Next Article in Journal
The Effects of Reducing Nitrogen and Increasing Density in the Main Crop on Yield and Cadmium Accumulation of Ratoon Rice
Previous Article in Journal
Effect of Doses, Application Dates, Adjuvants and Zinc Sulfate on Bromoxynil + Terbuthylazine + Nicosulfuron Weed Control Efficacy and Maize Grain Yield in Mid-West Poland
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Agronomy
Volume 15
Issue 2
10.3390/agronomy15020484
agronomy-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Dual-Channel Co-Spectroscopy–Based Non-Destructive Detection Method for Fruit Quality and Its Application to Fuji Apples

by
Xin Liang
1,†,
Tian Jiang
2,†,
Wanli Dai
3 and
Sai Xu
1,*
1
Institute of Facility Agriculture, Guangdong Academy of Agricultural Sciences, Guangzhou 510640, China
2
College of Mathematics and Informatics, South China Agricultural University, Guangzhou 510642, China
3
Faculty of Mechanical and Electrical Engineering, Zhongkai University of Agriculture and Engineering, Guangzhou 510225, China
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this work.
Agronomy 2025, 15(2), 484; https://doi.org/10.3390/agronomy15020484
Submission received: 24 January 2025 / Revised: 13 February 2025 / Accepted: 14 February 2025 / Published: 17 February 2025
(This article belongs to the Section Precision and Digital Agriculture)
Browse Figures
Review Reports Versions Notes

Abstract

:
Visible/near-infrared spectroscopy is widely used for non-destructive fruit quality detection, but the high cost of spectrometers (400–1100 nm range) in sorting equipment limits its accessibility. This study proposes a dual-channel co-spectroscopy method to address this issue. Using apples’ soluble solids content as the research target, a dual-channel platform was constructed to optimize parameters for full-transmission spectral signal acquisition. Spectral data were collected using dual channels (400–700 nm and 700–1100 nm bands, separated by filters) and a single channel (400–1100 nm range). Preprocessing methods (MSC, SNV, FD, SD, SG) and feature extraction algorithms (CARS, SPA, UVE) were applied, followed by PLSR modeling. The dual-channel method with Raw spectrum + FD + CARS + PLSR achieved optimal results, with R2v = 0.88, RMSEP = 0.39 for the 400–700 nm band, and R2v = 0.94, RMSEP = 0.33 for the 700–1100 nm band. The single-channel method with Raw spectrum + MSC + CARS + PLSR achieved R2v = 0.90, RMSEP = 0.36. These findings validate dual-channel co-spectroscopy as a cost-effective, accurate solution for non-destructive fruit quality detection, providing a practical approach to reduce spectrometer costs and enhance sorting system efficiency.

1. Introduction

China is a major producer and consumer of fruit, with a production volume of 312.96 million tons in 2022, representing a 4.4% increase compared to 2021 and ranking among the highest globally [1]. Fruits are rich in dietary fiber, vitamins, minerals, polyphenols, and organic acids, offering significant nutritional value. Moderate consumption is crucial in maintaining health and preventing foodborne diseases [2]. Research has shown a positive correlation between increased fruit consumption, improved psychological well-being, and subjective happiness [3]. As consumer standards rise, the demand for higher-quality fruit increases. Fruits intended for the fresh market undergo rigorous quality inspection and grading to ensure superior standards, improve economic value, and enhance competitiveness. Soluble solid content (SSC) is a key indicator of fruit flavor [4]. High production volumes and stricter requirements have driven growing demand for non-destructive fruit quality inspection and grading equipment. Visible/near-infrared spectroscopy technologies used for non-destructive fruit quality inspection are currently implemented in single-line systems with one spectrometer, dual-line systems with two spectrometers, or multiline systems with multiple spectrometers. However, the high cost of spectrometers, as critical components of these systems, has significantly increased the overall expense of non-destructive inspection. The imbalance between large production volumes and relatively low export quantities underscores the urgency of finding cost-effective solutions. Dual-channel co-spectroscopy technology offers a promising approach to improving non-destructive fruit quality inspection and grading systems, reducing costs, and enhancing the global competitiveness of Chinese fruits.
Non-destructive testing technologies offer advantages such as high detection efficiency, strong accuracy, and reduced damage. Compared to traditional manual inspection and physicochemical testing methods, they not only improve detection speed and precision but also minimize damage to the fruit, thereby enhancing fruit quality and market competitiveness [5,6]. Visible/near-infrared (VIS/NIR) spectroscopy has demonstrated significant potential and broad application prospects for non-destructive detection of fruit internal quality [7,8,9,10,11,12,13,14,15,16]. With characteristics such as being non-destructive, rapid, and intelligent, this technology is widely utilized in automated production lines for large-scale fruit quality inspection and sorting [17,18,19]. In scientific research, current VIS/NIR spectroscopy methods for non-destructive detection of fruit internal quality are categorized into three types: full-transmission, partial-transmission, and diffuse reflection. The VIS/NIR full-transmission spectroscopy method [20] involves light passing through the entire diameter of the fruit, entering on one side and exiting on the opposite side. The transmitted light carries information about the fruit’s internal quality, used for quality modeling and identification. This approach effectively captures comprehensive internal quality information with a strong representation of data integrity. It has been applied to the non-destructive internal quality inspection of various fruits, including pineapple [21], pomelo [22], apple [23], citrus [24], peach [25], and pear [26]. The VIS/NIR partial-transmission spectroscopy method captures light entering from two lateral sides of the fruit and exiting through the bottom, allowing spectral information from a portion of the fruit’s flesh to represent the overall quality. This method has achieved favorable results in the non-destructive detection of TSSC in fruits such as watermelon [27] and melon [28]. The VIS/NIR diffuse reflection spectroscopy method [29] collects light reflected from the fruit surface and analyzes the spectral signals containing quality-related information. This method is typically limited to surface-level quality assessment and is particularly effective for TSSC detection in thin-skinned fruits such as apple [30], pear [31], peach [32], and kiwifruit [33]. However, partial-transmission and diffuse reflection methods are less effective at acquiring complete and accurate information on internal quality, especially for deep internal regions of larger fruits.
The development of fruit internal quality inspection and sorting equipment has advanced significantly on domestic and international fronts. VIS/NIR spectroscopy has become a widely adopted method for non-destructive quality assessment and sorting based on diverse grading criteria. Sorting technologies have transitioned from basic mechanical systems to advanced designs incorporating electromechanical control, automation, and intelligent features [34,35]. Prominent international manufacturers include Aweta (Netherlands) [36], COMPAC (New Zealand) [37], TOMRA (Norway) [38], MAF-RODA (France) [39], and Greefa (Netherlands) [40]. Key domestic manufacturers include Jiangxi Reemoon [41], Taiwan Jinong [42], Hangzhou Kepler [43], Shenzhen Deltron [44], Shandong Jereh [45], Longkou Kaixiang [46], and Hefei Taiho [47]. Existing domestic or imported equipment primarily employs single-line systems with one spectrometer, dual-line systems with two spectrometers, or multiline systems with multiple spectrometers. Although domestically produced equipment is generally more cost-effective than imported alternatives, the high cost of spectrometers continues to pose a significant barrier. Consequently, only large-scale fruit enterprises can afford such systems, leaving smaller businesses unable to manage the high acquisition expenses.
This study introduces a novel non-destructive fruit quality detection method based on dual-channel co-spectroscopy, addressing key limitations in current technologies. As no prior research or reports on dual-channel co-spectroscopy exist, this approach represents an innovative contribution to the field. Its core principle lies in the combination of “spectrometer sharing” and “band segmentation”, which is an innovative method that achieves multiband spectral acquisition and analysis through spectroscopic design and data collaborative processing. The method involves three key concepts. First, the dual-channel system consists of two channels. Channel 1 collects spectral data in the 400–700 nm range using an optical filter to block other wavelengths from the halogen light source, ensuring only light within this range reaches the detection sample. Channel 2 collects spectral data in the 700–1100 nm range by similarly filtering out other wavelengths, leaving only the specified range of light to illuminate the sample in Channel 2. Second, co-spectroscopy refers to the shared use of a single VIS/NIR spectrometer operating across the 400–1100 nm range for both channels. Third, data collected by Channels 1 and 2 is transmitted to a computer, where recognition models process the spectral information for the 400–700 nm and 700–1100 nm ranges separately. Based on the results, sorting commands are issued to the respective channels. The primary research objectives are as follows: (1) establishing a dual-channel co-spectroscopy experimental platform to optimize parameters for acquiring VIS/NIR full-transmission spectral signals; (2) developing PLSR models for non-destructive detection of soluble solids in apples using single-channel (400–1100 nm) and dual-channel co-spectroscopy (400–700 nm and 700–1100 nm) approaches; and (3) comparing the spectral effects of filtered and unfiltered light, as well as the modeling accuracy of single-channel and dual-channel systems, to validate the feasibility of the dual-channel co-spectroscopy method for non-destructive fruit quality detection.

2. Materials and Methods

2.1. Experimental Apples

The Fuji apples selected for this experiment were harvested on 29 October 2024, from multiple commercial orchards in Yantai City, Shandong Province, China. A total of 150 samples were collected, all of which were uniform in color, smooth, uniform in size, and free from visible damage. After harvesting, the samples were shipped via courier and arrived at the Guangzhou laboratory on 31 October 2024. The experiment was completed within 24 h of their arrival. Prior to data collection, the apples were placed in a laboratory environment at 20 ± 5 °C for 2 h to ensure thermal equilibrium with the surrounding environment.

2.2. Spectral Platform Setup and Data Collection

2.2.1. Spectral Platform Construction

A self-constructed dual-channel co-spectroscopy platform was used for the non-destructive detection of soluble solids in apples (Figure 1). Apples were placed on silicone trays in Channels 1 and 2 during sampling. The trays ensured complete contact and airtight sealing between the apple and the tray surface, stabilizing the apple’s position and providing valuable reference data for dynamic detection in conveyor systems. To minimize light scattering and energy loss, light emitted by halogen lamps was directed through light source tubes, transmitted through the apple sample, and passed through a central aperture in the tray. The transmitted light was then received by a Y-shaped optical fiber (wavelength range: 350–2000 nm; core diameter: 600 μm; connector type: SMA905x2). A T-shaped adjustable bracket linked the dual light sources, ensuring consistent and uniform illumination across both channels. The light sources, each consisting of a light source tube and halogen lamp (Osram MR16, 24 V/30 W, Munich, Germany), maintained equal distances from the apple samples. The bifurcated ends of the Y-shaped optical fiber were connected to flange receivers at the bottom of Trays 1 and 2, while the converged end was linked to a spectrometer (QE Pro, Ocean Optics, Shanghai, China) that transmitted the collected spectral data to a computer via a data cable. Spectral data collection was conducted within a dual-channel dark box, where apples were centrally positioned on the tray and supported upright by the tray’s fruit holder. The light source, emission aperture, near-focus aperture, apple, and optical fiber were aligned along the same central axis. Optimal platform parameters were determined through testing, including a spectrometer integration time of 1500 ms, a 6 mm distance between the receiving fiber and the tray’s bottom surface, and a 105 mm distance between the lower edge of the light source and the tray’s upper surface. The spectral acquisition software used was BiaoQi Specsuite (developed by BiaoQi Optoelectronics Co., Ltd. in Guangzhou, China). The spectral data processing software used was MATLAB R2017a.
In the dual-channel setup, Light Source 1 used a quartz optical filter (50 mm diameter, 1 mm thickness, high-temperature resistant) to allow only wavelengths in the 400–700 nm range (“Filter 1”), while Light Source 2 used a similar filter for the 700–1100 nm range (“Filter 2”). These filters ensured precise segmentation of spectral data for each channel, enabling accurate and efficient non-destructive detection of soluble solids in apples.

2.2.2. Spectral Data Collection

(1)
Single-channel (400–1100 nm range) apple spectral data collection
Before the experiment, 150 apple samples were equilibrated to the ambient temperature by leaving them in the testing environment for a sufficient period. Each sample was then assigned a unique identifier. The Y-shaped optical fiber end connected to Channel 2 was covered with a fiber cap to block light completely. In Channel 1, a circular quartz glass filter (diameter: 50 mm, thickness: 1 mm, high-temperature resistant, transparent) was embedded in Light Source 1, allowing light within the 400–1100 nm wavelength range to pass through. During the experiment, apple samples were sequentially placed on Tray 1 in Channel 1, following the assigned numbering order (Figure 2). Spectral data for each apple in the 400–1100 nm range were collected using the platform and saved in Folder 1 for further analysis.
(2)
Dual-channel co-spectroscopy (400–700 nm and 700–1100 nm range) apple spectral data collection
Before data collection, the 150 apples tested in Section 2.2.2 (1), along with one reference apple (referred to as the “R apple”), were left in the testing environment until their temperature equilibrated with the surroundings. The dual-channel co-spectroscopy platform was prepared by embedding Filter 1 in Light Source 1 of Channel 1 and Filter 2 in Light Source 2 of Channel 2. The specific data collection steps were as follows:
Step 1: Place Apple 1 on Tray 1 in Channel 1 as described in Section 2.2.2 (1). Simultaneously, position the R apple on Tray 2 in Channel 2. Collect spectral data for both apples to obtain the full spectral range of 400–1100 nm. The dataset includes 400–700 nm spectral data for Apple 1 and 700–1100 nm spectral data for the R apple. Remove the R apple’s 700–1100 nm data, retaining only Apple 1’s 400–700 nm spectral data. Save the data in Folder 2.
Step 2: Switch the positions of the apples. Place the R apple on Tray 1 in Channel 1 and Apple 1 on Tray 2 in Channel 2. Collect spectral data for both apples to obtain the full spectral range of 400–1100 nm. The dataset includes 700–1100 nm spectral data for Apple 1 and 400–700 nm spectral data for the R apple. Remove the R apple’s 400–700 nm data, retaining only Apple 1’s 700–1100 nm spectral data. Save the data in Folder 3.
Step 3: Repeat Steps 1 and 2 for the remaining apples (apples 2, 3, 4 … 150). Save the 400–700 nm spectral data for each apple in Folder 2 and the 700–1100 nm spectral data in Folder 3, using the corresponding apple numbering sequence.

2.3. Determination of Soluble Solids Content

Following the collection of apple spectral data, the soluble solids content of each apple sample was measured immediately. The measurement process followed the refractometer method specified in NY/T2637-2014 [48]. For the analysis, the edible portion of the tested apple was placed into a piece of gauze and juiced using a juicer. The extracted juice was transferred into a disposable aviation cup. A handheld sugar content meter (PAL-BX/ACID5, designed for apples, ATAGO, Tokyo, Japan) was used to measure the soluble solids content. Before measurement, the sugar content meter was calibrated using distilled water [49]. A disposable pipette was used to collect an appropriate amount of juice from the cup, which was then dropped onto the detection window of the meter. Each sample was measured three times, and the average value was recorded as the soluble solids content of the apple.

2.4. Data Analysis and Modeling

Noise interference in the detection signals, caused by factors such as scattered light and the relatively large optical path space during apple transmission spectral signal collection [50], was addressed through structural and algorithmic optimization. Structural adjustments included the use of a silicone base to ensure complete contact between the apple and the tray, minimizing scattering noise. Additionally, five noise reduction algorithms were compared for their effectiveness: Multiplicative Scatter Correction (MSC) [51], Standard Normal Variate (SNV) [52], First Derivative (FD), Second Derivative (SD) [53,54], and Savitzky–Golay polynomial smoothing (SG) [55]. Since not all spectral bands contribute meaningfully to modeling and identification, removing redundant features improves model accuracy and computational efficiency [56]. Three feature selection methods were evaluated: Competitive Adaptive Reweighted Sampling (CARS) [57,58,59], Successive Projections Algorithm (SPA) [60], and Uninformative Variable Elimination (UVE) [61]. Finally, the traditional modeling method most widely used in non-destructive fruit quality detection, partial least squares regression (PLSR) [62], was selected. PLSR is also the most commonly used and practical mainstream method in current industrial applications [63]. Compared to other modeling methods, PLSR offers several advantages, such as strong identification ability, high identification accuracy, robust predictive analysis of high-dimensional data, and strong model stability. In this study, PLSR was utilized for modeling, with latent variables (LV) optimized through repeated trials. PLSR model performance was assessed using the coefficient of determination (R2) and the root-mean-square error (RMSE). R2, ranging from 0 to 1, represents the correlation between predicted and actual values, with values closer to 1 indicating stronger linear correlation and better model performance. RMSE measures the deviation between predicted and actual values, where lower values indicate higher accuracy. To ensure reliable results, the 150 samples were randomly divided into 10 equal subsets. Each subset was split into calibration and validation sets in an 8:2 ratio, followed by 10-fold cross-validation. This method minimized overfitting and ensured a robust evaluation of the model’s performance.

3. Results

3.1. PLSR Modeling for Soluble Solids Content Using Single-Channel Spectra (400–1100 nm Range)

PLSR models for detecting soluble solids content in apples were developed using spectral data in the 400–1100 nm range, as shown in Table 1. Initially, raw spectral data were analyzed directly, with the number of LVs set to 20. Feature wavelengths were extracted using three methods: CARS, SPA, and UVE, resulting in 111, 24, and 263 selected feature points, respectively. Among these, the combination of Raw spectrum + CARS + PLSR yielded the best model, with R2c = 0.98 and R2v = 0.73, indicating optimal performance for non-destructive detection of soluble solids content. Raw spectra were further preprocessed using five methods: MSC, SNV, FD, SD, and SG. LVs were set between 10 and 30, and feature wavelengths were again extracted using CARS, SPA, and UVE. Results showed that the combination of Raw spectrum + MSC + CARS + PLSR achieved the best performance, with R2c = 0.99, R2v = 0.90, RMSEC = 0.05, and RMSEP = 0.36. This study demonstrates that preprocessing raw spectral data and combining it with feature wavelength extraction methods significantly improves the predictive performance of PLSR models for soluble solids content in apples.

3.2. Dual-Channel Co-Spectroscopy (400–700 nm and 700–1100 nm Range) PLSR Modeling for Soluble Solids Content

3.2.1. PLSR Modeling for the 400–700 nm Range

The PLSR model for detecting soluble solids content in apples using the 400–700 nm spectral range was developed, as shown in Table 2. Raw spectral data were analyzed directly with the number of LVs set between 10 and 30. Feature wavelengths were extracted using CARS, SPA, and UVE methods, resulting in 111, 24, and 263 selected points, respectively. Among these, the Raw spectrum + CARS + PLSR model achieved the best performance, with R2c = 0.99 and R2v = 0.61, indicating optimal results for non-destructive detection of soluble solids content in apples. Further analysis involved preprocessing raw spectra using MSC, SNV, FD, SD, and SG methods, with LVs again set between 10 and 30, followed by feature extraction using CARS, SPA, and UVE. The results showed that the combination of Raw spectrum + FD + CARS + PLSR provided the best fitting performance, achieving R2c = 0.99 and R2v = 0.88, RMSEC = 0.02, and RMSEP = 0.39. The findings indicate that preprocessing raw spectra combined with feature wavelength extraction significantly improves PLSR model accuracy for the 400–700 nm range. The combination of Raw spectrum + FD + CARS + PLSR produced the most reliable model for non-destructive detection of soluble solids content in apples, demonstrating superior fitting performance.

3.2.2. PLSR Modeling for the 700–1100 nm Range

The PLSR model for detecting soluble solids content in apples using the 700–1100 nm spectral range was developed, as shown in Table 3. Raw spectral data were analyzed directly, with the number of LVs set between 10 and 30. Feature wavelengths were extracted using CARS, SPA, and UVE methods, yielding 110, 24, and 22 selected points, respectively. Among these, the Raw spectrum + CARS + PLSR model achieved optimal performance, with R2c = 0.99 and R2v = 0.86, indicating superior non-destructive detection accuracy for soluble solids in apples. Further analysis involved preprocessing raw spectra using MSC, SNV, FD, SD, and SG methods, followed by feature extraction with CARS, SPA, and UVE. Results showed that the combination of Raw spectrum + FD + CARS + PLSR provided the best fitting performance, achieving R2c = 0.99 and R2v = 0.94, RMSEC = 0.02, and RMSEP = 0.33. The findings demonstrate that preprocessing raw spectra, combined with feature wavelength extraction, significantly enhances the predictive performance of PLSR models for the 700–1100 nm range. The combination of Raw spectrum + FD + CARS + PLSR was the most effective for non-destructive detection of soluble solids content in apples, providing superior fitting and prediction accuracy.

4. Discussion

4.1. Comparison of Spectral Effects After Filter Splitting

Spectral data were collected using a single-channel method covering the 400–1100 nm range and a dual-channel co-spectroscopy method, which split the spectral range into 400–700 nm and 700–1100 nm bands. The original spectral plots are shown in Figure 3, Figure 4 and Figure 5. A comparison of Figure 3 and Figure 4 shows that spectral data from both methods in the 400–550 nm range consist primarily of noise. At 625 nm, both methods display a similar peak feature. In the 400–700 nm range, the spectral waveforms exhibit comparable trends between the dual-channel and single-channel methods. A comparison of Figure 3 and Figure 5 indicates that in the 700–720 nm range, the spectral curves from both methods have positive slopes, trending upwards to the right. At 750 nm, both methods exhibit a similar trough feature, while at 820 nm, a similar peak is observed. Beyond 900 nm, the spectral data from both methods are dominated by noise. The analysis confirms that the spectral trends from the dual-channel co-spectroscopy method (400–700 nm and 700–1100 nm bands) are consistent with those from the single-channel method (400–1100 nm band). Since single-channel (400–1100 nm wavelength range) is widely used in detecting soluble solids content in Fuji apples, and a large body of literature has proven its feasibility [64,65], these results demonstrate that the spectral trends and features of apples collected using single-channel and dual-channel methods are similar. This similarity preliminarily verifies the feasibility of dual-channel co-spectral technology for detecting soluble solids content in Fuji apples.

4.2. Comparison of PLSR Modeling Performance Between Single and Dual Channels

As discussed in Section 2.1 and Section 2.2, the performance of PLSR models for non-destructive detection of soluble solids content in apples under various preprocessing and feature extraction methods is summarized in Table 4. For the single-channel method (400–1100 nm range), the combination Raw spectrum + MSC + CARS + PLSR provided the best performance, with R2c = 0.99 and R2v = 0.90, RMSEC = 0.05, and RMSEP = 0.36. For the dual-channel co-spectroscopy method (400–700 nm and 700–1100 nm ranges), the combination Raw spectrum + FD + CARS + PLSR yielded the best results. In the 400–700 nm range, the model achieved R2c = 0.99, R2v = 0.88, RMSEC = 0.02, and RMSEP = 0.39. In the 700–1100 nm range, the model achieved R2c = 0.99, R2v = 0.94, RMSEC = 0.02, and RMSEP = 0.33. Both single-channel and dual-channel methods demonstrated excellent modeling performance, with R2c values reaching 0.99 and R2v exceeding 0.88. These results validate the feasibility of the proposed dual-channel co-spectroscopy technique for non-destructive detection of soluble solids content in apples, providing a reliable and effective solution.

4.3. Prospects for Non-Destructive Fruit Quality Detection Based on Dual-Channel Co-Spectroscopy

This study explores the shared mechanism of dual-channel co-spectroscopy for the 400–700 nm and 700–1100 nm ranges using a 400–1100 nm spectrometer to detect soluble solids content in apples. A novel non-destructive fruit quality detection method based on dual-channel co-spectroscopy was proposed. Despite significant progress and widespread application of non-destructive fruit quality detection technologies worldwide [66], there are still shortcomings in terms of technological maturity, cost, and detection efficiency [67,68]. Increasing consumer demand for higher fruit quality has led to a surge in the need for such equipment. The rising demand has also increased the cost of spectrometers, a core component of detection systems. For instance, the market price of the Ocean Optics QE Pro spectrometer (400–1100 nm range) rose from CNY 130,000 in 2021 to CNY 159,000 in 2024, an increase of 22.30%, as shown in Table 5 [21,69]. In the dual-line sorting system, the detection efficiency of dual-channel co- spectral technology is the same as that of dual-channel detection technology. However, the spectrometer used in dual-channel co-spectral technology costs only half as much as that used in dual-channel detection technology. Furthermore, compared to single-line single-channel detection technology, although the spectrometer cost of dual-line dual-channel co-spectral technology is the same, its detection efficiency is doubled. Specifically, the detection efficiency of single-line single-channel detection technology is 3 to 5 items per second [70,71], while that of dual-line dual-channel co-spectral technology can reach 6 to 10 items per second. Future research should focus on applying dual-channel co-spectroscopy, exploring its potential to enhance practicality and expand its applications in fruit quality detection.
This study also designed intelligent sorting equipment for non-destructive fruit quality detection based on dual-channel co-spectroscopy (Figure 6). The detection process is as follows. Fruits are fed into the system through the loading area and conveyed along V-shaped belts (Belt 1 and Belt 2) to Trays 1 and 2, corresponding to Channels 1 and 2 of the detection system. As the trays move through the dual-channel co-spectroscopy detection chamber, the spectrometer is triggered via a switch. The spectrometer, connected to a Y-shaped optical fiber, simultaneously collects spectral data from Channels 1 and 2. The collected spectral data are transmitted via data cables to the intelligent sorting equipment’s controller. The embedded fruit recognition model analyzes the spectral data for Channels 1 and 2, displays the results on a user interface, and issues grading commands to sorting outlets based on the analysis. This system maintains production efficiency while significantly reducing equipment costs. The technology facilitates high-quality development in China’s fruit industry by ensuring reliable quality detection and intelligent sorting at a lower cost.

4.4. Selection of Research Subject and Model Application

This study primarily focuses on dual-channel co-spectral technology. During the selection of the research subject, we also conducted relevant investigations. Fuji apples, as a key pillar of China’s fruit industry, offer advantages such as high yield, good storage resistance, and high market recognition. They are an important tool for agricultural efficiency improvement and rural revitalization [72]. At the same time, non-destructive detection of soluble solids content in Fuji apples has been validated in numerous studies, making Fuji apples an ideal subject for verifying the feasibility of dual-channel co-spectral technology in this research. Furthermore, the research results on dual-channel co-spectral technology applied to Fuji apples can also be extended to quality detection of other fruits. Regarding model application, due to the small sample size and limited signals available in this study, the current model meets the basic requirements for industrial sorting. However, when higher sorting accuracy is required, the results of this study will need further improvement and optimization.

5. Conclusions

This study addresses the issue of high spectrometer costs in existing non-destructive fruit quality sorting line equipment by proposing a dual-channel co-spectroscopy-based method for non-destructive fruit quality detection. The feasibility of this method is verified by using Fuji apples as a case study for non-destructive detection of soluble solids content (SSC). An experimental platform was established to collect apple spectral data, which was then analyzed using machine learning methods.
This study conducted a comparative analysis of the spectral effects after splitting the optical filter and found that the spectral trends and features of apples collected by single-channel and dual-channel methods were similar, thereby preliminarily verifying the feasibility of dual-channel co-spectroscopy technology. Additionally, partial least squares regression (PLSR) modeling was performed for SSC detection in apples using single-channel (400–1100 nm) and dual-channel co-spectroscopy (400–700 nm and 700–1100 nm) data, and the modeling results were compared. For the single-channel (400–1100 nm) data, the best model for SSC detection was Raw spectrum + MSC + CARS + PLSR, achieving an R2c of 0.99 and an R2v of 0.90, with RMSEC and RMSEP values of 0.05 and 0.36, respectively. For the dual-channel co-spectroscopy (400–700 nm and 700–1100 nm) data, the optimal model was Raw spectrum + FD + CARS + PLSR. Specifically, in the 400–700 nm range, the model achieved an R2c of 0.99 and an R2v of 0.88, with RMSEC and RMSEP values of 0.02 and 0.39, respectively. In the 700–1100 nm range, the model achieved an R2c of 0.99 and an R2v of 0.94, with RMSEC and RMSEP values of 0.02 and 0.33, respectively. These results demonstrate that dual-channel co-spectroscopy technology is feasible for SSC detection in apples and further validate the reliability of this technology. This study also explores the application prospects of dual-channel co-spectroscopy for non-destructive fruit quality detection. It analyzes the fluctuations in spectrometer prices over the past five years and the cost differences between single- and dual-channel spectrometers in different sorting line configurations. The research found that the price of spectrometers has steadily increased, with a 22.30% rise by 2025. In dual-line sorting systems, dual-channel co-spectroscopy and dual-channel detection technologies achieve the same detection efficiency. However, the spectrometer cost for dual-channel co-spectroscopy is only half that of the conventional dual-channel detection technology. Furthermore, compared to single-line single-channel detection, dual-line dual-channel co-spectroscopy achieves twice the detection efficiency while maintaining the same spectrometer cost. Specifically, the detection efficiency of single-line single-channel detection is 3–5 items per second, while that of dual-line dual-channel co-spectroscopy can reach 6–10 items per second. Through the analysis of spectrometer costs and detection efficiency, the results indicate that dual-channel co-spectroscopy technology contributes to cost reduction and efficiency improvement in the industry, playing a significant role in promoting high-quality industrial development. In conclusion, the dual-channel co-spectroscopy-based non-destructive fruit detection method proposed in this study is feasible and has been successfully validated for SSC detection in Fuji apples. This technology not only meets the industry’s demand for accurate, rapid, and intelligent non-destructive fruit quality detection but also provides new insights and solutions for the non-destructive internal quality detection of other fruits.

Author Contributions

Resources, W.D. and T.J.; data curation, X.L., T.J. and W.D.; writing—original draft preparation, X.L.; writing—review and editing, S.X., X.L. and T.J.; project administration, W.D.; funding acquisition, S.X. and X.L. All authors have read and agreed to the published version of the manuscript.

Funding

The National Natural Science Foundation of China (32471998); International Science and Technology Cooperation Project of Guangdong Province (no. 2023A0505050129); the Innovation Fund Industry Special Project of Guangdong Academy of Agricultural Science (202306); the Special Training Project for Science and Technology Innovation Strategy of Guangdong Academy of Agricultural Sciences (Construction of Main Agricultural Research Force) (R2023PY-QN002); Rural Revitalization of Guangdong Province (no. 2024TS-1-2); and the National Key Research and Development Program of China (no. 2022YFD2002203).

Data Availability Statement

The data used to support the findings of this study can be made available by the corresponding author upon request.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Zhang, F. In 2022, China’s fruit production surpassed 300 million tons. China Fruit News 2023, 40, 51. [Google Scholar]
  2. Chen, Q.; Yang, Z.; Zhang, Q.; Liu, Y.; Yin, P.; Guo, Y. Research progress on nutrition and functional components in fruit. Hubei Agric. Sci. 2019, 58, 9–13+21. [Google Scholar]
  3. Ocean, N.; Howley, P.; Ensor, J. Lettuce be happy: A longitudinal UK study on the relationship between fruit and vegetable consumption and well-being. Soc. Sci. Med. 2019, 222, 335–345. [Google Scholar] [CrossRef]
  4. Ncama, K.; Opara, U.L.; Tesfay, S.Z.; Fawole, O.A.; Magwaza, L.S. Application of Vis/NIR spectroscopy for predicting sweetness and flavour parameters of ‘Valencia’orange (Citrus sinensis) and ‘Star Ruby’grapefruit (Citrus x paradisi Macfad). J. Food Eng. 2017, 193, 86–94. [Google Scholar] [CrossRef]
  5. Liu, M.; Liu, Y.; Guo, J. Research and Application on Fruit Nondestructive Detection Technology. Agric. Sci. Technol. Equip. 2017, 6, 51–55. [Google Scholar]
  6. Chen, L.; Duan, D.; Wang, F.; Meng, X.; Zhao, C.; Qian, Y. Research Progress of Non-destructive Testing of Fruit Quality by Hyperspectral Imaging Technology. Guangdong Agric. Sci. 2023, 50, 83–94. [Google Scholar]
  7. Altieri, G.; Genovese, F.; Tauriello, A.; Di Renzo, G.C. Models to improve the non-destructive analysis of persimmon fruit properties by VIS/NIR spectrometry. J. Sci. Food Agric. 2017, 97, 5302–5310. [Google Scholar] [CrossRef] [PubMed]
  8. Huang, Y.; Lu, R.; Chen, K. Detection of internal defect of apples by a multichannel Vis/NIR spectroscopic system. Postharvest Biol. Technol. 2020, 161, 111065. [Google Scholar] [CrossRef]
  9. Li, L.; Huang, W.; Wang, Z.; Liu, S.; He, X.; Fan, S. Calibration transfer between developed portable Vis/NIR devices for detection of soluble solids contents in apple. Postharvest Biol. Technol. 2022, 183, 111720. [Google Scholar] [CrossRef]
  10. Scalisi, A.; O’Connell, M.G. Application of visible/NIR spectroscopy for the estimation of soluble solids, dry matter and flesh firmness in stone fruits. J. Sci. Food Agric. 2021, 101, 2100–2107. [Google Scholar] [CrossRef]
  11. Soltani Nazarloo, A.; Rasooli Sharabiani, V.; Abbaspour Gilandeh, Y.; Taghinezhad, E.; Szymanek, M.; Sprawka, M. Feasibility of using VIS/NIR spectroscopy and multivariate analysis for pesticide residue detection in tomatoes. Processes 2021, 9, 196. [Google Scholar] [CrossRef]
  12. Tilahun, S.; Seo, M.H.; Hwang, I.G.; Kim, S.H.; Choi, H.R.; Jeong, C.S. Prediction of lycopene and β-carotene in tomatoes by portable chroma-meter and VIS/NIR spectra. Postharvest Biol. Technol. 2018, 136, 50–56. [Google Scholar] [CrossRef]
  13. Wang, J.; Guo, Z.; Zou, C.; Jiang, S.; El-Seedi, H.R.; Zou, X. General model of multi-quality detection for apple from different origins by Vis/NIR transmittance spectroscopy. J. Food Meas. Charact. 2022, 16, 2582–2595. [Google Scholar] [CrossRef]
  14. Wang, T.; Zhang, Y.; Liu, Y.; Zhang, Z.; Yan, T. Intelligent evaluation of stone cell content of korla fragrant pears by vis/NIR reflection spectroscopy. Foods 2022, 11, 2391. [Google Scholar] [CrossRef] [PubMed]
  15. Zhang, L. Research on Some Problems of On-Line Nondestructive Detection of Soluble Solids Contents in Crown Pear by VIS/NIR Spectroscopy. Master’s Thesis, Zhejiang University, Zhejiang, China, 2015. [Google Scholar]
  16. Xu, H. Development and Application of Optimal Model for Nondestructive Evaluation of Fruits Sugar Content Using Visible/Near Infrared Spectroscopy. Ph.D. Thesis, Zhejiang University, Zhejiang, China, 2010. [Google Scholar]
  17. Ma, J.J.; Wang, K.Q. Research progress of optical non-destructive testing technology for fruit quality. Sci. Technol. Food Ind. 2021, 42, 427–437. [Google Scholar]
  18. Liu, Y.; Xu, H.; Sun, X.; Jiang, X.; Rao, Y.; Xu, J.; Wang, J. On-line detection of soluble solids content of apples from different origins by visible and near-infrared spectroscopy. Chin. Opt. 2020, 13, 482–491. [Google Scholar]
  19. Liu, H.; Gong, Z.; Wen, T.; Wang, Z.; Dai, X. Design and experiment of an online detection and sorting system for citrus sugar content based on visible/near infrared spectroscopy. J. Hunan Agric. Univ. (Nat. Sci.) 2023, 49, 497–502. [Google Scholar]
  20. Huang, Y.; Dong, W.; Chen, Y.; Wang, X.; Luo, W.; Zhan, B.; Liu, X.; Zhang, H. Online detection of soluble solids content and maturity of tomatoes using Vis/NIR full transmittance spectra. Chemom. Intell. Lab. Syst. 2021, 210, 104243. [Google Scholar] [CrossRef]
  21. Xu, X.; Lu, H.; Wang, X.; Qiu, G.; Wang, C.; Liang, X. Nondestructive detection method for pineapple water core based on visible/near infrared spectroscopy. Trans. Chin. Soc. Agric. Eng. 2021, 37, 287–294. [Google Scholar]
  22. Xu, S.; Lu, H.; He, Z.; Liang, X. Non-destructive determination of internal soluble solid content in pomelo using visible/near infrared full-transmission spectroscopy. Postharvest Biol. Technol. 2024, 214, 112990. [Google Scholar] [CrossRef]
  23. Wang, F.; Zhao, C.; Yang, H.; Jiang, H.; Li, L.; Yang, G. Non-destructive and in-site estimation of apple quality and maturity by hyperspectral imaging. Comput. Electron. Agric. 2022, 195, 106843. [Google Scholar] [CrossRef]
  24. Li, Y.; Jin, G.; Jiang, X.; Yi, S.; Tian, X. Non-destructive determination of soluble solids content using a multi-region combination model in hybrid citrus. Infrared Phys. Technol. 2020, 104, 103138. [Google Scholar] [CrossRef]
  25. Liu, S.; Huang, W.; Lin, L.; Fan, S. Effects of orientations and regions on performance of online soluble solids content prediction models based on near-infrared spectroscopy for peaches. Foods 2022, 11, 1502. [Google Scholar] [CrossRef] [PubMed]
  26. Sun, X.; Liu, Y.; Li, Y.; Wu, M.; Zhu, D. Simultaneous measurement of brown core and soluble solids content in pear by on-line visible and near infrared spectroscopy. Postharvest Biol. Technol. 2016, 116, 80–87. [Google Scholar] [CrossRef]
  27. Ibrahim, A.; Daood, H.G.; Égei, M.; Takács, S.; Helyes, L. A comparative study between Vis/NIR spectroradiometer and NIR spectroscopy for the non-destructive quality assay of different watermelon cultivars. Horticulturae 2022, 8, 509. [Google Scholar] [CrossRef]
  28. Jie, D.; Xie, L.; Rao, X.; Ying, Y. Using visible and near infrared diffuse transmittance technique to predict soluble solids content of watermelon in an on-line detection system. Postharvest Biol. Technol. 2014, 90, 1–6. [Google Scholar] [CrossRef]
  29. Khodabakhshian, R.; Emadi, B.; Khojastehpour, M.; Golzarian, M.R. A comparative study of reflectance and transmittance modes of Vis/NIR spectroscopy used in determining internal quality attributes in pomegranate fruits. J. Food Meas. Charact. 2019, 13, 3130–3139. [Google Scholar] [CrossRef]
  30. Cortés, V.; Cubero, S.; Blasco, J.; Aleixos, N.; Talens, P. In-line application of visible and near-infrared diffuse reflectance spectroscopy to identify apple varieties. Food Bioprocess Technol. 2019, 12, 1021–1030. [Google Scholar] [CrossRef]
  31. Fu, X.; Ying, Y.; Lu, H.; Xu, H. Comparison of diffuse reflectance and transmission mode of visible-near infrared spectroscopy for detecting brown heart of pear. J. Food Eng. 2007, 83, 317–323. [Google Scholar] [CrossRef]
  32. Cortés, V.; Blasco, J.; Aleixos, N.; Cubero, S.; Talens, P. Visible and near-infrared diffuse reflectance spectroscopy for fast qualitative and quantitative assessment of nectarine quality. Food Bioprocess Technol. 2017, 10, 1755–1766. [Google Scholar] [CrossRef]
  33. Guo, W.; Zhao, F.; Dong, J. Nondestructive measurement of soluble solids content of kiwifruits using near-infrared hyperspectral imaging. Food Anal. Methods 2016, 9, 38–47. [Google Scholar] [CrossRef]
  34. Guo, L.; Tang, X.; Mao, X.; Wan, L. Problems and Countermeasures of Agricultural Products Quality Standard in China. Food Nutr. China 2017, 23, 11–14. [Google Scholar]
  35. Chu, Q.; Jiang, B.; Yang, L.; Zheng, J.; Dai, Z. Study on Current Situation of Quality Grading System and Chinese Standards for Agricultural Product. Stand. Sci. 2015, 65–69. [Google Scholar]
  36. Aweta[EB/OL]. Available online: http://www.aweta.nl/ (accessed on 10 April 2024).
  37. Compac[EB/OL]. Available online: https://www.tomra.com/food (accessed on 10 April 2024).
  38. TOMRA[EB/OL]. Available online: https://www.tomra.com/en (accessed on 10 April 2024).
  39. MAF RODA[EB/OL]. Available online: http://www.maf-roda.com/en/ (accessed on 10 April 2024).
  40. Greefa[EB/OL]. Available online: https://www.greefa.com/ (accessed on 10 April 2024).
  41. Reemoon[EB/OL]. Available online: https://www.reemoon.com/en (accessed on 15 April 2024).
  42. Jinong[EB/OL]. Available online: http://www.taiwanjinong.com/ (accessed on 15 April 2024).
  43. Kepler[EB/OL]. Available online: http://www.keplertech.com.cn/ (accessed on 5 June 2024).
  44. Deltron[EB/OL]. Available online: https://deltron.tech/ (accessed on 9 July 2024).
  45. Jereh[EB/OL]. Available online: https://www.jereh.com/en/ (accessed on 9 July 2024).
  46. Longkou Kaixiang[EB/OL]. Available online: http://www.chinakaixiang.com/ (accessed on 9 July 2024).
  47. Taiho[EB/OL]. Available online: https://www.chinataiho.com/ (accessed on 2 August 2024).
  48. NY/T 2637-2014; Determination of Soluble Solids Content in Fruits and Vegetables: Refractometer Method. Ministry of Agriculture and Rural Affairs of the People’s Republic of China: Beijing, China, 2015.
  49. Guo, Z.; Wang, J.; Song, Y.; Yin, X.; Zou, C.; Zou, X. Design and experiment of the handheld visible-near infrared nondestructive detecting system for apple quality. Trans. Chin. Soc. Agric. Eng. 2021, 37, 271–277. [Google Scholar]
  50. Yang, A.; Ding, J. Comparative assessment of two methods for estimation of soil organic carbon content by Vis-NIR spectra in Xinjiang Ebinur Lake Wetland. Trans. Chin. Soc. Agric. Eng. 2015, 31, 162–168. [Google Scholar]
  51. Isaksson, T.; Næs, T. The effect of multiplicative scatter correction (MSC) and linearity improvement in NIR spectroscopy. Appl. Spectrosc. 1988, 42, 1273–1284. [Google Scholar] [CrossRef]
  52. Bi, Y.; Yuan, K.; Xiao, W.; Wu, J.; Shi, C.; Xia, J.; Chu, G.; Zhang, G.; Zhou, G. A local pre-processing method for near-infrared spectra, combined with spectral segmentation and standard normal variate transformation. Anal. Chim. Acta 2016, 909, 30–40. [Google Scholar] [CrossRef] [PubMed]
  53. Alcin, O.F.; Sengur, A.; Ghofrani, S.; Ince, M.C. GA-SELM: Greedy algorithms for sparse extreme learning machine. Measurement 2014, 55, 126–132. [Google Scholar] [CrossRef]
  54. Feng, G.; Huang, G.-B.; Lin, Q.; Gay, R. Error minimized extreme learning machine with growth of hidden nodes and incremental learning. IEEE Trans. Neural Netw. 2009, 20, 1352–1357. [Google Scholar] [CrossRef] [PubMed]
  55. Jahani, S.; Setarehdan, S.K.; Boas, D.A.; Yücel, M.A. Motion artifact detection and correction in functional near-infrared spectroscopy: A new hybrid method based on spline interpolation method and Savitzky–Golay filtering. Neurophotonics 2018, 5, 015003. [Google Scholar] [CrossRef]
  56. Solorio-Fernandez, S.; Martínez-Trinidad, J.F.; Carrasco-Ochoa, J.A. A new unsupervised spectral feature selection method for mixed data: A filter approach. Pattern Recognit. 2017, 72, 314–326. [Google Scholar] [CrossRef]
  57. Sun, T.; Xu, W.-L.; Lin, J.-L.; Liu, M.-H.; He, X.-W. Determination of soluble solids content in navel oranges by Vis/NIR diffuse transmission spectra combined with CARS method. Spectrosc. Spectr. Anal. 2012, 32, 3229–3233. [Google Scholar]
  58. Liu, Y.; Jiang, X.; Xiong, S.; Liu, D. Using CARS to predict the content of soil organic matter based on Vis-NIR spectra. J. Chin. Agric. Mech. 2016, 37, 240–244. [Google Scholar]
  59. Liu, Y.; Shi, Y.; Cai, L.; Zhou, T. On-line NIR Detection Model Optimization of Soluble Solids Content in Navel Orange Based on CARS. Trans. Chin. Soc. Agric. Mach. 2013, 44, 138–144. [Google Scholar]
  60. Krepper, G.; Romeo, F.; de Sousa Fernandes, D.D.; Diniz, P.H.G.D.; de Araújo, M.C.U.; Di Nezio, M.S.; Pistonesi, M.F.; Centurión, M.E. Determination of fat content in chicken hamburgers using NIR spectroscopy and the Successive Projections Algorithm for interval selection in PLS regression (iSPA-PLS). Spectrochim. Acta Part A Mol. Biomol. Spectrosc. 2018, 189, 300–306. [Google Scholar] [CrossRef]
  61. Centner, V.; Massart, D.-L.; de Noord, O.E.; de Jong, S.; Vandeginste, B.M.; Sterna, C. Elimination of uninformative variables for multivariate calibration. Anal. Chem. 1996, 68, 3851–3858. [Google Scholar] [CrossRef]
  62. Liu, Y.; Sun, X.; Ouyang, A. Nondestructive measurement of soluble solid content of navel orange fruit by visible–NIR spectrometric technique with PLSR and PCA-BPNN. LWT-Food Sci. Technol. 2010, 43, 602–607. [Google Scholar] [CrossRef]
  63. Wolda, S.; Sjöstróma, M.; Eriksson, L. PLS-regression: A basic tool of chemometrics. Chemom. Intell. Lab. Syst. 2001, 50, 109–130. [Google Scholar] [CrossRef]
  64. Meng, Q.; Shang, J.; Huang, R.; Chen, L.; Zhang, Y. Nondestructive detection of soluble solids content in apple by visible-near infrared spectroscopy. Food Ferment. Ind. 2020, 46, 205–209. [Google Scholar]
  65. Xu, W.; Yao, L.; Sun, T.; Hu, T.; Hu, T.; Liu, M. CARS-SPA baesd Visble/near Infraed spectroscopy on-line detection of apple soluble solids content. Sci. Technol. Food Ind. 2014, 35, 61–64. [Google Scholar]
  66. Zhang, M. Design and Implementation of Apple Sugar Content Detection Based on Infrared Spectroscopy. Heilongjiang University: Harbin, China, 2017. [Google Scholar]
  67. Liu, Z.; Wang, L.; GXi, G.; Peng, C.; Jiao, Y. The Current Status and Development of Fruit Maturity Detection Technology. Agric. Technol. 2020, 40, 17–21. [Google Scholar]
  68. Hu, Y.; Jiang, H.; Zhou, H.; Wang, Y. Research Progress on Nondestructive Detection of Fruit Maturity by Near Infrared Spectroscopy and Hyperspectral Imaging. Sci. Technol. Food Ind. 2021, 42, 377–383. [Google Scholar]
  69. Guangzhou Biaoqi Optoelectronics [EB/OL]. Available online: http://www.gzbiaoqi.com/ (accessed on 22 August 2024).
  70. Yang Li, Y.; Peng, Y.; Lü, D.; Li, Y.; Liu, L.; Zhu, Y. Development of Mobile Orchard Local Grading System of Apple Internal Quality. Smart Agric. 2022, 4, 132–142. [Google Scholar]
  71. Guo, Z.; Huang, W.; Chen, Q.; Peng, Y.; Zhao, J. Model Optimization for the On-line Inspection of Internal Apple Quality by Shortwave Near-infrared Spectroscopy. Mod. Food Sci. Technol. 2016, 32, 147–153. [Google Scholar]
  72. 2024 China Apple Industry Data Analysis Report. [EB/OL]. Available online: https://www.thepaper.cn/newsDetail (accessed on 26 June 2024).
Agronomy 15 00484 g001
Figure 1. Experimental platform for non-destructive detection of soluble solids in apples based on dual-channel co-spectroscopy. Key components: 1. Computer; 2. Data transmission cable; 3. Spectrometer; 4. Y-shaped optical fiber; 5. Channel 1; 6. Light Source 1 (light source tube and halogen lamp); 7. Sample 1; 8. Tray 1; 9. Feet (×4); 10. Tray 2; 11. Sample 2; 12. Light Source 2 (light source tube and halogen lamp); 13. Channel 2; 14. Dual-channel co-spectroscopy dark box; 15. T-shaped adjustable bracket.
Figure 1. Experimental platform for non-destructive detection of soluble solids in apples based on dual-channel co-spectroscopy. Key components: 1. Computer; 2. Data transmission cable; 3. Spectrometer; 4. Y-shaped optical fiber; 5. Channel 1; 6. Light Source 1 (light source tube and halogen lamp); 7. Sample 1; 8. Tray 1; 9. Feet (×4); 10. Tray 2; 11. Sample 2; 12. Light Source 2 (light source tube and halogen lamp); 13. Channel 2; 14. Dual-channel co-spectroscopy dark box; 15. T-shaped adjustable bracket.
Agronomy 15 00484 g001
Agronomy 15 00484 g002
Figure 2. Schematic diagram of apple placement on the tray. Key components: 1. Apple samples; 2. Tray; 3. Tray fixing plate; 4. Optical fiber; 5. Fruit pedicel.
Figure 2. Schematic diagram of apple placement on the tray. Key components: 1. Apple samples; 2. Tray; 3. Tray fixing plate; 4. Optical fiber; 5. Fruit pedicel.
Agronomy 15 00484 g002
Agronomy 15 00484 g003
Figure 3. Original spectra of apples in the 400–1100 nm range.
Figure 3. Original spectra of apples in the 400–1100 nm range.
Agronomy 15 00484 g003
Agronomy 15 00484 g004
Figure 4. Original spectra of apples in the 400–700 nm range.
Figure 4. Original spectra of apples in the 400–700 nm range.
Agronomy 15 00484 g004
Agronomy 15 00484 g005
Figure 5. Original spectra of apples in the 700–1100 nm range.
Figure 5. Original spectra of apples in the 700–1100 nm range.
Agronomy 15 00484 g005
Agronomy 15 00484 g006
Figure 6. Schematic diagram of intelligent sorting equipment for non-destructive fruit quality detection based on dual-channel co-spectroscopy. Key components: 1. Loading area; 2. V-shaped belt 1; 3. V-shaped belt 2; 4. Channel 1; 5. Channel 2; 6. Dual-channel co-spectroscopy detection chamber; 7. Y-shaped optical fiber; 8. Spectrometer; 9. Data transmission cable; 10. Intelligent sorting equipment’s controller; 11. Sorting outlets.
Figure 6. Schematic diagram of intelligent sorting equipment for non-destructive fruit quality detection based on dual-channel co-spectroscopy. Key components: 1. Loading area; 2. V-shaped belt 1; 3. V-shaped belt 2; 4. Channel 1; 5. Channel 2; 6. Dual-channel co-spectroscopy detection chamber; 7. Y-shaped optical fiber; 8. Spectrometer; 9. Data transmission cable; 10. Intelligent sorting equipment’s controller; 11. Sorting outlets.
Agronomy 15 00484 g006
Table 1. PLSR modeling performance for soluble solids content in apples using spectra in the 400–1100 nm range.
Table 1. PLSR modeling performance for soluble solids content in apples using spectra in the 400–1100 nm range.
Preprocessing MethodFeature SelectionFeature WavelengthsLVsTraining SetTest Set
R2cRMSECR2vRMSEV
Raw spectrumCARS111200.980.190.730.58
Raw spectrum + MSC184300.990.050.900.36
Raw spectrum + SNV236280.990.060.880.37
Raw spectrum + FD162270.990.040.870.41
Raw spectrum + SD162290.990.040.830.48
Raw spectrum + SG184300.990.110.700.67
Raw spectrumSPA24200.570.750.530.65
Raw spectrum + MSC24240.500.790.480.68
Raw spectrum + SNV24200.560.750.470.68
Raw spectrum + FD24220.420.830.490.67
Raw spectrum + SD24230.410.820.500.68
Raw spectrum + SG24200.510.780.370.72
Raw spectrumUVE263200.940.300.430.92
Raw spectrum + MSC222200.930.340.381.08
Raw spectrum + SNV219250.960.250.311.08
Raw spectrum + FD78300.930.330.321.35
Raw spectrum + SD57250.900.390.391.06
Raw spectrum + SG145300.890.410.520.88
Section Summary: In the PLSR modeling analysis of apple soluble solids in the 400–1100 nm band, we collected the apples’ original spectra and combined five preprocessing methods (MSC, SNV, FD, SD, SG) with three feature extraction methods (CARS, SPA, UVE). Through multidimensional and multigradient processing, analysis, and comparison, we obtained the following results for the Raw spectrum + MSC + CARS + PLSR model: For R2c and R2v, the values were 0.99 and 0.90, respectively, while RMSEC and RMSEP were 0.05 and 0.36. These results indicate that this modeling method has the best fitting effect.
Table 2. PLSR modeling performance for soluble solids content in apples using spectra in the 400–700 nm range.
Table 2. PLSR modeling performance for soluble solids content in apples using spectra in the 400–700 nm range.
Preprocessing MethodFeature SelectionFeature WavelengthsLVsTraining SetTest Set
R2cRMSECR2vRMSEV
Raw spectrumCARS120300.990.090.610.87
Raw spectrum + MSC108300.990.110.640.92
Raw spectrum + SNV120300.990.010.710.65
Raw spectrum + FD133290.990.020.880.39
Raw spectrum + SD133280.990.030.640.69
Raw spectrum + SG149160.790.560.201.43
Raw spectrumSPA24240.550.760.360.76
Raw spectrum + MSC24240.540.760.300.76
Raw spectrum + SNV24240.570.740.480.69
Raw spectrum + FD24210.430.810.310.75
Raw spectrum + SD24230.470.790.260.77
Raw spectrum + SG24200.550.760.240.80
Raw spectrumUVE24190.710.640.200.94
Raw spectrum + MSC30180.710.640.211.05
Raw spectrum + SNV29240.620.710.170.97
Raw spectrum + FD14140.500.780.160.96
Raw spectrum + SD24240.660.670.171.03
Raw spectrum + SG29200.540.760.170.91
Section Summary: In the PLSR modeling analysis of apple soluble solids in the 400–700 nm band, we collected the apples’ original spectra and combined five preprocessing methods (MSC, SNV, FD, SD, SG) with three feature extraction methods (CARS, SPA, UVE). Through multidimensional and multigradient processing, analysis, and comparison, we obtained the following results for the Raw spectrum + FD + CARS + PLSR model: For R2c and R2v, the values were 0.99 and 0.88, respectively, while RMSEC and RMSEP were 0.02 and 0.39. These results indicate that this modeling method has the best fitting effect.
Table 3. PLSR modeling performance for soluble solids content in apples using spectra in the 700–1100 nm range.
Table 3. PLSR modeling performance for soluble solids content in apples using spectra in the 700–1100 nm range.
Preprocessing MethodFeature SelectionFeature WavelengthsLVsTraining SetTest Set
R2cRMSECR2vRMSEV
Raw spectrumCARS110300.990.060.860.53
Raw spectrum + MSC123300.990.040.800.62
Raw spectrum + SNV155300.990.030.850.52
Raw spectrum + FD123300.990.020.940.33
Raw spectrum + SD123300.990.020.880.46
Raw spectrum + SG138300.960.220.181.68
Raw spectrumSPA24200.540.720.310.91
Raw spectrum + MSC24240.520.730.320.90
Raw spectrum + SNV24240.420.770.370.89
Raw spectrum + FD24240.470.750.410.79
Raw spectrum + SD24240.490.740.470.83
Raw spectrum + SG24240.490.730.290.91
Raw spectrum + RAWUVE22180.650.640.270.93
Raw spectrum + MSC21190.640.640.320.91
Raw spectrum + SNV20200.620.660.260.96
Raw spectrum + FD28200.740.570.280.96
Raw spectrum + SD44200.870.420.181.11
Raw spectrum + SG20180.600.670.200.99
Section Summary: In the PLSR modeling analysis of apple soluble solids in the 700–1100 nm band, we collected the apples’ original spectra and combined five preprocessing methods (MSC, SNV, FD, SD, SG) with three feature extraction methods (CARS, SPA, UVE). Through multidimensional and multigradient processing, analysis, and comparison, we obtained the following results for the Raw spectrum + FD + CARS + PLSR model: For R2c and R2v, the values were 0.99 and 0.94, respectively, while RMSEC and RMSEP were 0.02 and 0.33. These results indicate that this modeling method has the best fitting effect.
Table 4. Comparison of PLSR modeling performance between single and dual channels.
Table 4. Comparison of PLSR modeling performance between single and dual channels.
Detection MethodWavelength Range (nm)PLSR
Training Set R2cRMSECTest Set
R2v		RMSEP
Single channel400–11000.990.050.900.36
Dual-channel co-spectroscopy400–7000.990.020.880.39
700–11000.990.020.940.33
Table 5. Spectrometer prices and price changes for different sorting line categories.
Table 5. Spectrometer prices and price changes for different sorting line categories.
Sorting Line CategoryDetection MethodWavelength Range (nm)Detection Efficiency (Items per Second)Number of Spectrometers2025 Market Price (CNY)2021 Market Price (CNY)Price Difference (CNY)
Single lineSingle channel400–11003–51159,000130,00029,000
Dual lineDual channel400–11006–102318,000260,00058,000
Dual channel co-spectroscopy400–7006–101159,000130,00029,000
700–1100
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Liang, X.; Jiang, T.; Dai, W.; Xu, S. Dual-Channel Co-Spectroscopy–Based Non-Destructive Detection Method for Fruit Quality and Its Application to Fuji Apples. Agronomy 2025, 15, 484. https://doi.org/10.3390/agronomy15020484

AMA Style

Liang X, Jiang T, Dai W, Xu S. Dual-Channel Co-Spectroscopy–Based Non-Destructive Detection Method for Fruit Quality and Its Application to Fuji Apples. Agronomy. 2025; 15(2):484. https://doi.org/10.3390/agronomy15020484

Chicago/Turabian Style

Liang, Xin, Tian Jiang, Wanli Dai, and Sai Xu. 2025. "Dual-Channel Co-Spectroscopy–Based Non-Destructive Detection Method for Fruit Quality and Its Application to Fuji Apples" Agronomy 15, no. 2: 484. https://doi.org/10.3390/agronomy15020484

APA Style

Liang, X., Jiang, T., Dai, W., & Xu, S. (2025). Dual-Channel Co-Spectroscopy–Based Non-Destructive Detection Method for Fruit Quality and Its Application to Fuji Apples. Agronomy, 15(2), 484. https://doi.org/10.3390/agronomy15020484

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop