Prediction of Growth and Quality of Chinese Cabbage Seedlings Cultivated in Different Plug Cell Sizes via Analysis of Image Data Using Multispectral Camera

: In recent times, there has been an increasing demand for the development of rapid and non-destructive assessment of the growth and quality of seedlings before transplanting. This study was conducted to examine the growth and quality of Chinese cabbage seedlings that can be determined via the image data acquired using a multispectral camera. Chinese cabbage seedlings were cultivated in ﬁve different plug trays (72, 105, 128, 162, and 200 cells/tray) for 30 days after sowing (DAS). The growth of seedlings had no signiﬁcant difference in the early stage of cultivation; however, it decreased with increasing the number of cells in the plug tray due to the restricted root zone volume in the mid to late stages. Individual leaf area was predicted by analyzing of image data with high accuracy (R 2 > 0.8) after 15 DAS; however, the accuracy of leaf area prediction per tray decreased due to overlapping and twisting leaves. Among six different vegetation indices, mrNDVI showed a high correlation (R 2 > 0.6) with the dry weight of seedlings at 25 and 30 DAS. We conﬁrmed that the leaf area of seedlings can be predicted non-destructively by analyzing the acquired image data per seedling and tray and suggested the applicability of vegetation indices for predicting the growth and quality of vegetable seedlings.


Introduction
Chinese cabbage (Brassica rapa L. ssp.pekinensis) is a very important vegetable crop in Korea, and it accounts for 23 and 30% of domestic vegetable production and consumption, respectively [1,2].In the vegetable nursery industry, Chinese cabbage is the second most produced vegetable crop in Korea.Chinese cabbages are cultivated in the fields throughout the year, and the cropping systems and cultivars are decided depending on the season.
Seedling cultivation is a process of cultivation in which young high-quality plants are grown uniformly under appropriate conditions before transplanting.The benefits of seedling cultivation include seed savings, faster growth, and higher quality and uniformity compared to direct seeding [3].In Korea, a large number of vegetable farmers purchase seedlings produced in plug trays from nurseries rather than self-raising the seedlings, and commercial nurseries producing vegetable seedlings have increased in number since the 1990s [4].Seedling production using plug trays is efficient in terms of growing more seedlings in less space, automation, and labor saving in the cultivation process [5].Commercial nurseries produce the seedlings using plug trays by controlling the environmental conditions in greenhouses, and their goal is to produce healthy and uniform seedlings for fast establishment and growth after transplanting.As vegetable seedlings are young and vulnerable, even short-term exposure to inappropriate environments can cause a rapid deterioration in the growth and quality; however, most nurseries rely on the experience of their managers to determine whether the seedlings are in a healthy condition.
In Korea, Chinese cabbage is cultivated in all seasons by employing various types of cropping.Plug trays have various cell numbers ranging from 50 to 288 according to the cropping type.The plug trays with a smaller number of cells are used for the production of Chinese cabbage seedlings in winter and spring cropping types due to a longer period of seedling cultivation, and those with a larger number of cells are used for the seedling production in summer and autumn cropping types due to a shorter period of seedling cultivation [6].The cell size decreased with increasing the number of cells in a plug tray, and the growth of seedlings was affected by the cell size of the plug tray.In many crops, using a plug tray with a larger cell size induced better growth of the seedlings due to the increased root zone volume and amount of light received per seedling [7][8][9].However, it does not mean that the efficiency of seedling production is promoted by using plug trays with larger cell sizes in a commercial nursery.A smaller number of seedlings were planted in the same area, and more resources (commercial medium, water, plug tray, etc.) were required to produce one seedling [10].
Improper conditions during the seedling cultivation period can reduce the growth and quality of seedlings [11][12][13].In vegetable crops, the status of seedling growth and quality can be decided by measuring plant height, number of leaves, leaf area, chlorophyll content, and fresh and dry weight of the sample seedlings [14][15][16].These parameters conventionally employ destructive measurement, and it takes a lot of time and labor for measuring.Therefore, researchers try to develop a method for quickly and non-destructively analyzing the growth and quality of plants.Min et al. [17] detected the nitrogen content in leaves using VIS-NIR spectroscopy and reported that predicting the nitrogen content of the seedlings using the stepwise multiple linear regression (SMLR) and partial least squares (PLS) regression procedures was highly correlated with the actual nitrogen measurements (R 2 > 0.7).However, hyperspectral imaging using VIS-NIR spectroscopy is difficult in terms of data processing, which makes it hard to apply in the agricultural industry.
Multispectral imaging, which is a simplified version of hyperspectral imaging, is a leading non-destructive measurement system in agriculture [18].The reflectance at different wavelengths changes with the chlorophyll content, nutrient, and water content of a plant.Analyzing the reflectance of a plant using images can provide a non-destructive and quick assessment of the current growth, nutrient, and water status of the plant [19][20][21].Recently, the PLS method, which is a linear regression method for multivariate calibration and is widely used for data processing in hyperspectral imaging, was applied to analyze multispectral images in order to develop a powerful model between the extracted spectra and growth parameters [22].Multispectral image analysis was conducted in crop cultivation, ranging from large-scale fields to leaves [23][24][25].Open-field imaging studies were mainly conducted on food crops such as maize, wheat, and rice [26][27][28].Among the vegetables grown in greenhouses, the technologies of image processing were applied in the case of fruit vegetables [29][30][31].Previous research has focused on the application of multispectral image analysis for crop cultivation after transplanting, and there is still a lack of studies on the applicability to seedling production.Wang et al. [32] found that multispectral imaging in individual leaf areas can detect nitrogen and moisture stress in tomato seedlings.
The objective of this study is to confirm the applicability of image analysis for detecting the change in the seedling growth and quality as affected by different cell sizes of plug trays during the cultivation period of Chinese cabbage seedlings.

Plant Materials and Cultivations
In this study, Chinese cabbage 'Chun-gwang' (SAKATA KOREA Co., Ltd., Seoul, Republic of Korea) was used as the plant material.The experiment was conducted in the glass greenhouse at the University of Seoul, Seoul, Republic of Korea (37 • 34 57.1 N, 127 • 03 37.7 E) from 8 March to 7 April 2023.The seeds were sown in the plug trays with 72, 105, 128, 162, and 200 cells filled with the commercial medium for seedling production (Biosangto, Nongwoobio Co., Ltd., Incheon, Gyeonggi-do, Republic of Korea).The seeds were germinated in the dark room maintained at a 25 • C air temperature for two days.Then, they were transferred to the greenhouse and cultivated for 28 days with sub-irrigation every day.The air temperature, relative humidity, and light intensity in the greenhouse were recorded using the agricultural environmental measuring instrument (aM-31, WISE Sensing Inc., Suwon, Gyeonggi-do, Republic of Korea).During the cultivation period, the average air temperature, relative humidity, and daily light integral in the greenhouse were 23.1 ± 2.5 • C, 33.4 ± 14.2%, and 16.4 ± 6.4 mol•m −2 , respectively.

Growth Measurements
Plant height, number of leaves, SPAD value, leaf area, and fresh and dry weight of shoot in Chinese cabbage seedlings were measured at 10, 15, 20, 25, and 30 DAS.The leaf area meter (LI-3100, LI-COR Inc., Lincoln, NE, USA) and chlorophyll meter (SPAD-502PLUS, KONICA MINOLTA, Inc., Tokyo, Japan) were used for measuring the leaf area and SPAD value, respectively.The shoot fresh weight was obtained using a scale (KERN EWJ 300-3, Kern & Sohn GmbH, Balingen, Germany), and the shoot dry weight was measured after drying in an oven at 70 • C for 7 days.The leaf area index (LAI) was calculated based on the measured leaf area and the area of one cell in a plug tray.The areas of one cell were 24.10, 17.18, 13.60, 11.07, and 8.89 cm 2 in the plug trays with 72, 105, 128, 162, and 200 cells, respectively.

Image-Based Measurements and Analysis
The images of seedlings were obtained using the Plant Image Measurement System (PIMS) shown in Figure 1.PIMS consists of a multispectral camera (FS-3200T-10GE-NNC, JAI, Copenhagen, Denmark) and LED lighting.The LED provides light optimized for acquiring visible (VIS)-near-infrared (NIR) spectral information.

Plant Materials and Cultivations
In this study, Chinese cabbage 'Chun-gwang' (SAKATA KOREA Co., Ltd., Seou public of Korea) was used as the plant material.The experiment was conducted i glass greenhouse at the University of Seoul, Seoul, Republic of Korea (37°34′57.1 127°03′37.7″E) from 8 March to 7 April 2023.The seeds were sown in the plug trays 72, 105, 128, 162, and 200 cells filled with the commercial medium for seedling produ (Biosangto, Nongwoobio Co., Ltd., Incheon, Gyeonggi-do, Republic of Korea).The s were germinated in the dark room maintained at a 25 °C air temperature for two d Then, they were transferred to the greenhouse and cultivated for 28 days with sub-ir tion every day.The air temperature, relative humidity, and light intensity in the g house were recorded using the agricultural environmental measuring instrument (aM WISE Sensing Inc., Suwon, Gyeonggi-do, Republic of Korea).During the cultivatio riod, the average air temperature, relative humidity, and daily light integral in the g house were 23.1 ± 2.5 °C, 33.4 ± 14.2%, and 16.4 ± 6.4 mol•m −2 , respectively.

Growth Measurements
Plant height, number of leaves, SPAD value, leaf area, and fresh and dry weig shoot in Chinese cabbage seedlings were measured at 10, 15, 20, 25, and 30 DAS.The area meter (LI-3100, LI-COR Inc., Lincoln, NE, USA) and chlorophyll meter (SP 502PLUS, KONICA MINOLTA, Inc., Tokyo, Japan) were used for measuring the leaf and SPAD value, respectively.The shoot fresh weight was obtained using a scale (K EWJ 300-3, Kern & Sohn GmbH, Balingen, Germany), and the shoot dry weight was m ured after drying in an oven at 70 °C for 7 days.The leaf area index (LAI) was calcu based on the measured leaf area and the area of one cell in a plug tray.The areas o cell were 24.10, 17.18, 13.60, 11.07, and 8.89 cm 2 in the plug trays with 72, 105, 128 and 200 cells, respectively.

Image-Based Measurements and Analysis
The images of seedlings were obtained using the Plant Image Measurement Sy (PIMS) shown in Figure 1.PIMS consists of a multispectral camera (FS-3200T-10GE-N JAI, Copenhagen, Denmark) and LED lighting.The LED provides light optimized fo quiring visible (VIS)-near-infrared (NIR) spectral information.The images of seedlings and trays were obtained before measuring the seedling growth destructively at 10, 15, 20, 25, and 30 DAS.The images of seedlings and trays were captured at RGB (450 nm, 550 nm, and 650 nm), NIR1 (750 nm), and NIR2 (830 nm).The predicted leaf area (PLA) and spectral reflectance at each wavelength band were analyzed using the ENVI program (ENVI 5.3, L3Harris Geospatial, Broomfield, CO, USA) (Figure 2).using the ENVI program (ENVI 5.3, L3Harris Geospatial, Broomfield, CO, USA) (Figure 2).
The predicted changes in the leaf area per plug tray with 72, 105, 128, 162, and 200 cells were fitted using a sigmoidal equation as a function of DAS.
where a, b, and c are regression parameters and x is DAS.

Figure 2. Seedling image analysis using ENVI program.
To investigate the feasibility of using a vegetation index to detect the changes in seedling growth as affected by the cell size in plug trays, the vegetation indices were calculated with the spectral reflectance, as shown in Table 1.
Table 1.Vegetation indices examined in this study.R represents the reflectance values at the indicated wavelengths in nm.RNIR is the average of R750 and R830.

Index Equation Reference
Normalized Difference Vegetation Index (NDVI) (R NIR − R RED )/(R NIR + R RED ) [33] Green Normalized Difference Vegetation Index (GNDVI) where a, b, and c are regression parameters and x is DAS.
To investigate the feasibility of using a vegetation index to detect the changes in seedling growth as affected by the cell size in plug trays, the vegetation indices were calculated with the spectral reflectance, as shown in Table 1.
Table 1.Vegetation indices examined in this study.R represents the reflectance values at the indicated wavelengths in nm.R NIR is the average of R 750 and R 830 .

Image-Based Measurements and Analysis
The experimental data were analyzed using SAS 9.4 statistical software (Enterprise Guide 8.3, SAS Institute Inc., Cary, NC, USA), and Duncan's multiple range tests were performed to determine the significant difference at p < 0.05.Linear regression analysis was conducted using Microsoft ® Excel ® (Microsoft 365 MSO Version 16.0.16501.20074,Redmond, WA, USA), and SigmaPlot statistical software (Version 15, Systat Software Inc., Palo Alto, CA, USA) was used for logistic regression analysis.

Seedling Growth in Chinese Cabbage as Affected by Cell Size of Plug Tray
The seedling growth was not affected significantly by the cell size of the plug tray at 10 DAS; however, the growth of seedlings in the plug tray with 72 and 105 cells was higher after 15 DAS (Table 2).In all treatments, the seedling growth tended to increase rapidly until 20 DAS and then slowed down.The SPAD value did not show significant differences among different cell plug trays in the cultivation period.Figure 3 shows the change in the leaf area of seedlings by the cell size of the plug tray.At 10 DAS, there were no significant leaf area differences between treatments; however, the leaf area of seedlings increased by decreasing the number of cells in a plug tray at 15 DAS.The increase in leaf area tended to slow over time after 20 DAS, and the leaf area of the seedlings grown in the plug tray with 72 and 105 cells was approximately three times higher than that grown in the plug tray with 200 cells.
The results are consistent with previous studies that show that the leaf area, shoot fresh weight, and shoot dry weight of seedlings increased with increasing plug tray cell size [10].Cabbage seedlings grown in plug trays with larger cell sizes had relatively better growth, and the seedlings with better growth led to increased productivity [40].However, Kim et al. [41] reported that zucchini seedlings grown in a 32-cell plug tray showed lower quality of the first harvested fruit than those grown in plug trays with 50, 105, and 163 cells.The optimal cell size in the plug tray should be determined carefully considering the seedling growth characteristics, cultivation area, period, and resources used during the nursery process.Figure 3 shows the change in the leaf area of seedlings by the cell size of the p tray.At 10 DAS, there were no significant leaf area differences between treatments; ho ever, the leaf area of seedlings increased by decreasing the number of cells in a plug t at 15 DAS.The increase in leaf area tended to slow over time after 20 DAS, and the area of the seedlings grown in the plug tray with 72 and 105 cells was approximately th times higher than that grown in the plug tray with 200 cells.The results are consistent with previous studies that show that the leaf area, sh fresh weight, and shoot dry weight of seedlings increased with increasing plug tray size [10].Cabbage seedlings grown in plug trays with larger cell sizes had relatively be growth, and the seedlings with better growth led to increased productivity [40].Howe Kim et al. [41] reported that zucchini seedlings grown in a 32-cell plug tray showed lo quality of the first harvested fruit than those grown in plug trays with 50, 105, and cells.The optimal cell size in the plug tray should be determined carefully considering seedling growth characteristics, cultivation area, period, and resources used during nursery process.

Correlation Analysis between Measured and Predicted Leaf Area
We correlated the leaf area predicted using multispectral imaging with the measured leaf area (Figure 4).Except at 10 DAS, the coefficients of determination of the linear regression between the predicted and measured values were very high (R 2 > 0.8) (Figure 4B-E).As shown in the results for the entire experimental period (Figure 4F), the predicted leaf area correlated with the measured leaf area with a high coefficient of determination (R 2 = 0.927).It was confirmed that the individual leaf area can be predicted with high accuracy using multispectral imaging.However, the negative correlation between the predicted and measured leaf areas is likely due to leaf twisting at the time of imaging, resulting in a smaller predicted leaf area.

Correlation Analysis between Measured and Predicted Leaf Area
We correlated the leaf area predicted using multispectral imaging with the measured leaf area (Figure 4).Except at 10 DAS, the coefficients of determination of the linear regression between the predicted and measured values were very high (R 2 > 0.8) (Figure 4B-E).As shown in the results for the entire experimental period (Figure 4F), the predicted leaf area correlated with the measured leaf area with a high coefficient of determination (R 2 = 0.927).It was confirmed that the individual leaf area can be predicted with high accuracy using multispectral imaging.However, the negative correlation between the predicted and measured leaf areas is likely due to leaf twisting at the time of imaging, resulting in a smaller predicted leaf area.The leaf area of the plug tray unit was also predicted well through the use of multispectral imaging in all treatments; however, the coefficient of determination was lower compared to the correlation analysis of individual leaf area prediction (Figure 5).The decreased accuracy of leaf area prediction in a plug tray was due to the overlapping and wrapping area of leaves within the plug tray not being considered.The leaf area of the plug tray unit was also predicted well through the use of multispectral imaging in all treatments; however, the coefficient of determination was lower compared to the correlation analysis of individual leaf area prediction (Figure 5).The decreased accuracy of leaf area prediction in a plug tray was due to the overlapping and wrapping area of leaves within the plug tray not being considered.In Figure 6 and Table 3, the predicted leaf area of the plug tray unit was a strong fit to a sigmoidal curve with DAS as independent variables (R 2 > 0.95).Sigmoidal regression curves can be used to estimate the predicted leaf area of a plug tray unit over time according to the different cell sizes of plug trays.In Figure 6 and Table 3, the predicted leaf area of the plug tray unit was a strong fit to a sigmoidal curve with DAS as independent variables (R 2 > 0.95).Sigmoidal regression curves can be used to estimate the predicted leaf area of a plug tray unit over time according to the different cell sizes of plug trays.Leaf area is one of the important parameters for determining the growth and quality of Chinese cabbage seedlings.The Chinese cabbage is cultivated year round by employing different cropping types in Korea; therefore, Chinese cabbage seedlings are produced using different cultivation managements (different cell plug trays, cultivation period, fertilization, etc.) depending on the nursery throughout the year.As Chinese cabbage seedlings are cultivated using different cell plug trays and different growing days, it is important to analyze the leaf area of the seedling non-destructively in different cell plug trays during the cultivation period.Many researchers conducted the prediction of individual leaf area using image analysis and reported that the leaf area of the plant unit was predicted well [42][43][44][45].Our results also showed a high level of accuracy in the prediction of individual leaf area.
In commercial nurseries, the management of cultivation and shipment is handled on a plug tray basis and not on a seedling basis.Therefore, the prediction of leaf area per plug tray using image analysis has less accuracy but is more meaningful than the prediction of individual leaf area.Tong et al. [46] found that the leaf area of tomato, cucumber, aubergine, and pepper seedlings in a plug tray was predicted through the segmentation of seedling images and reported that the prediction accuracy was higher than 95% in the four different seedlings.Leaf overlapping, which was estimated to be a source of error in the Leaf area is one of the important parameters for determining the growth and quality of Chinese cabbage seedlings.The Chinese cabbage is cultivated year round by employing different cropping types in Korea; therefore, Chinese cabbage seedlings are produced using different cultivation managements (different cell plug trays, cultivation period, fertilization, etc.) depending on the nursery throughout the year.As Chinese cabbage seedlings are cultivated using different cell plug trays and different growing days, it is important to analyze the leaf area of the seedling non-destructively in different cell plug trays during the cultivation period.Many researchers conducted the prediction of individual leaf area using image analysis and reported that the leaf area of the plant unit was predicted well [42][43][44][45].Our results also showed a high level of accuracy in the prediction of individual leaf area.
In commercial nurseries, the management of cultivation and shipment is handled on a plug tray basis and not on a seedling basis.Therefore, the prediction of leaf area per plug tray using image analysis has less accuracy but is more meaningful than the prediction of individual leaf area.Tong et al. [46] found that the leaf area of tomato, cucumber, aubergine, and pepper seedlings in a plug tray was predicted through the segmentation of seedling images and reported that the prediction accuracy was higher than 95% in the four different seedlings.Leaf overlapping, which was estimated to be a source of error in the leaf area analysis of the plug tray unit, was smaller than that in our study because the seedlings were cultivated for 10-15 days in plug trays with 50 and 72 cells in the study by Tong et al. [46].As the predicted leaf area of a plug tray unit can be used to characterize the changes in leaf area among different cell plug trays, we obtained a sigmoid regression curve of predicted leaf area of a plug tray unit using DAS with a high fitness.A previous study showed that the predicted leaf area of individual lettuce plants fitted to a sigmoidal curve using days after germination, and the models were used to calculate the total incident light throughout the cultivation period [47].Through further research under various environmental conditions, we can improve the model of predicted leaf area in different cell plug trays using cumulative temperature or light integral during the cultivation period.

Correlation Analysis between Vegetation Indices and Growth Parameters
Linear regression analysis was conducted to determine the correlation between the measured growth parameters and the vegetation indices calculated using spectral reflectance.At 25 and 30 DAS, shoot dry weight was significantly correlated with several vegetation indices; however, LAI and SPAD values were less correlated (Tables 4 and 5).Among the vegetation indices, mrNDVI was highly correlated with shoot dry weight at both 25 and 30 DAS.The vegetation indices used in this study were NDVI, GNDVI, CI green , TVI, mrNDVI, and RDVI.The vegetation indices were applied widely to estimate crop growth parameters such as biomass, yield, LAI, etc. NDVI and RDVI are the indices that use the NIR and RED spectrums to estimate the green biomass of plants, and RDVI is the index that modifies NDVI to consider the density of the canopy [39].GNDVI and CI green are the indices that use the NIR and GREEN spectrums and are more sensitive to chlorophyll concentration than NDVI [48,49].TVI is the index that calculates the area of a triangle consisting of the green peak, the chlorophyll absorption minimum, and the NIR shoulder based on changes in the spectral reflectance due to chlorophyll absorption and leaf tissue abundance.mrNDVI is the index modified from NDVI to correct for increased reflectance across the entire visible spectrum due to high leaf surface reflectance.Sims and Gamon [38] chose R 445 as a measure of surface reflectance in mrNDVI because it has a stable value under changes in chlorophyll content.In this study, we replaced R 445 with R 450 , which can be obtained from a multispectral camera.
Our results showed that the correlation between vegetation indices and growth parameters was not clear until 20 DAS, and some vegetation indices correlated with shoot dry weight at 25 and 30 DAS.Previous research reported that NDVI increased with increasing LAI until LAI reached 3 [50][51][52][53].It was found that GNDVI and CI green were more sensitive to chlorophyll concentration than NDVI [34,35,54].RDVI and TVI showed a high correlation with biomass in wheat and sorghum, respectively [55,56].The vegetation indices were mainly applied to estimate the growth status in an open field or a forest, and there was some research in greenhouses [57].mrNDVI, which was highly correlated with water stress [57], showed a high correlation with the shoot dry weight of Chinese cabbage seedlings at 25 and 30 DAS.
NDVI, the most commonly used vegetation index for crop growth detection, showed a lower correlation with growth parameters than the other vegetation indices in our study.In addition, LAI and SPAD values show a low correlation with vegetable indices.As our research is intended to build basic data for the non-destructive detection of Chinese cabbage seedling growth and quality using vegetation indices, we conducted linear regression analysis.However, some vegetation indices applied in this study have been reported to have a non-linear relationship with growth parameters such as LAI and chlorophyll content in previous studies [34,38,58].NDVI saturates asymptotically when the LAI exceeds 2 and becomes nearly constant in value when the LAI is above 3 [50,58].At 25 DAS, the LAI of seedlings was higher than 2, which would have reduced the correlation between NDVI and growth parameters.
The application of vegetation indices to the seedling production in a greenhouse is a first attempt, and we suggested that mrNDVI can be used to analyze the seedling growth non-destructively and quickly at the last stage of cultivation.However, early determination of seedling growth using vegetation indices was not possible in this study.In the early stage of seedling production, the reflectance analyzed from the image was strongly influenced by the commercial medium due to the small leaf area.The noise in reflectance prevents the vegetation indices from reflecting seedling growth.Xue et al. [59] also reported that the interference of soil reflectance was one of the factors limiting the application of vegetation indices.The use of vegetation indices can be a powerful tool to save time, money, and crop losses for growth and quality measurements in agriculture; however, vegetation indices can have different sensitivities to different varieties, which means further research is needed to find the optimal vegetation index for each variety [60].Therefore, the proper vegetation index should be found by measuring the reflectance at a wider range of wavelengths for detecting the growth and quality of seedlings throughout the cultivation period.

Conclusions
In this study, we investigated the changes in the growth of Chinese cabbage seedlings affected by plug tray cell size and analyzed the growth and quality of the seedlings nondestructively using multispectral imaging.The growth of seedlings was increased by increasing the cell size in a plug tray.The leaf area, which is one of the important growth parameters in Chinese cabbage seedlings, was predicted using the image data of the seedlings and plug tray units, and individual leaf area was predicted with a high level of accuracy.The accuracy of predicting the leaf area of the plug tray unit was lower than that of the seedling unit due to the increased overlapping and twisting areas according to DAS.Among several vegetation indices, the applicability of mrNDVI was confirmed in terms of evaluating the growth of the Chinese cabbage seedlings at the last stage of cultivation.In this study, we can confirm the applicability of non-destructive analysis using multispectral imaging to detect the growth and quality of seedlings; however, further studies are needed for the field application of non-destructive analysis in commercial nurseries, including improving the accuracy of leaf area prediction of plug tray unit and developing an appropriate vegetation index for early detection of seedling growth.

Figure 1 .
Figure 1.Plant image measurement system (PIMS).An LED with a wavelength in the range of 400 to 1000 nm and a multispectral camera are installed in this system.

Figure 2 .
Figure 2. Seedling image analysis using ENVI program.The predicted changes in the leaf area per plug tray with 72, 105, 128, 162, and 200 cells were fitted using a sigmoidal equation as a function of DAS.Predicted leaf area = a/(1 + e −(x−c)/b ) (1)

z
Days after sowing.y Means in columns followed by different letters are significantly ferent when considering Duncan's multiple range test at p < 0.05.

Figure 5 .
Figure 5. Correlation between measured and predicted leaf areas of plug tray unit in Chinese cabbage 'Chun-gwang'.

Figure 5 .
Figure 5. Correlation between measured and predicted leaf areas of plug tray unit in Chinese cabbage 'Chun-gwang'.
z Days after sowing.y Means in columns followed by different letters are significantly different when considering Duncan's multiple range test at p < 0.05.

Table 4 .
Coefficient of determination (R 2 ) of linear regression analysis between growth parameter and vegetation index in the seedlings of Chinese cabbage 'Chun-gwang' at 25 days after sowing.

Table 5 .
Coefficient of determination (R 2 ) of linear regression analysis between growth parameter and vegetation index in the seedlings of Chinese cabbage 'Chun-gwang' at 30 days after sowing.