Design of Device for Optical Luminescent Diagnostic of the Seeds Infected by Fusarium

Abstract

The development and application of optical luminescent methods and devices will help obtain information quickly and objectively about the level of Fusarium infection of agricultural plants. For the previously obtained ranges, the spectral characteristics of excitation and luminescence of wheat, barley, and oats of various degrees of infection were measured. The obtained dependences of flows on infection were approximated by linear regression models and relative sensitivities were determined. For wheat and barley, it is advisable to determine the degree of infection by the ratio of flows Φλ1/Φλ2, which makes it possible to calibrate the measuring device in relative units and increase its sensitivity. A method for determining the degree of infected seeds with Fusarium was developed. After the seeds are placed in a light-tight chamber, they are excited by radiation, and photoluminescence is recorded. The electrical signal from the radiation receiver is amplified and processed accounting for previously obtained calibration curves. In the universal device that measures the infection of wheat, barley, and oats seeds, it is necessary to have three radiation sources: 362 nm, 424 nm, and 485 nm. Based on the energy efficiency criteria, optimal LEDs and photodiodes, as well as a microcontroller, switches, operational amplifiers, a display, and other components of the device, were selected.

1. Introduction

One of the main factors which affects yield losses are plant diseases. Plant diseases are dangerous because they are difficult to detect and identify in the early stages. This factor affects crop yields and it can seriously affect the sustainability of the agricultural sector. It is important for agricultural enterprises to detect diseases early in order to control their spread.

The acceptance of operational management decisions depends on the availability of information about the diseases. Traditionally, plant diseases are detected by interpretation of visual symptoms followed by laboratory evaluation [1]. However, these methods require skills and experience in the relevant plant pathology, significant time to complete the diagnosis, and expensive chemicals and equipment. These disadvantages of traditional methods have prompted the development of modern technologies such as machine vision and remote sensing for the detection and identification of plant diseases. These technologies make it possible to assess the disease with greater reliability, accuracy, and speed [2]. These technologies are based on the determination of the optical properties of plants in various spectral ranges. Optical methods such as RGB imaging, multi- and hyperspectral sensors, thermography, or chlorophyll fluorescence show potential for conscious, objective, and reproducible detections for disease detection and quantification of epidemic diseases [3,4,5].

Using infrared thermography, chlorophyll fluorescence imaging and hyperspectral imaging, Fusarium damage to wheat on spikelet scales was monitored. The method made it possible to visualize the temperature difference inside the infected spikelets starting from 5 days. At the same time, the violation of photosynthetic activity was confirmed by the maximum fluorescence yield of spikelets over 5 days. NIR spectroscopy with a reflectance prefix was used to categorize weedy rice from cultivated rice [6]. As a means of assessing microbial contamination and shelf life of leafy green vegetables, spectral analysis of near-IR reflectance (NIR) and transmission (absorption) in the visible region was used [7]. The effectiveness of early detection of three European endemic diseases, wheat septoria, rust, and blotch, has also been analyzed [8].

Monitoring of Fusarium damage to wheat on spikelet scales was carried out using infrared thermography, chlorophyll fluorescence imaging, and hyperspectral imaging. The method made it possible to visualize the temperature difference inside the infected spikelets starting from 5 days.

Hyperspectral imaging is used to detect rust blistering in southwestern white pine seedlings [9,10]. The application of hyperspectral imaging for detection of Fusarium was evaluated in these study [11]. Hyperspectral images were obtained in the wavelength range of 400–1000 nm. Another study [12] provided the basis for the development of automated and monitoring systems of the myrtle rust detection based on reflectance spectrum sensors. An algorithm for recognizing cucumber diseases based on leaf images using sparse representation classification has been developed [13].

In a coffee leaf rust study [14], researchers obtained data using a Sequoia camera, which produced images with a spatial resolution of 10.6 cm in four spectral bands: green (530–570 nm), red (640–680 nm), far red (730–740 nm), and near infrared (770–810 nm). Researchers used conventional cameras such as the Nikon D80 to image leaf diseases in the study of rice and wheat leaf diseases [15]. The use of low-cost drones equipped with digital cameras as a field phenotyping tool to determine the severity of foliar diseases in a wheat breeding program is being explored [16].

Based on the application of chlorophyll fluorescence and hyperspectral imaging, it is possible to detect Fusarium species infection in wheat [17]. To detect brown rust disease in winter wheat, an optical sensor device was developed that excites chlorophyll fluorescence at discrete wavelengths and detects induced emissions [18].

Increasingly, remote sensors are being used to monitor plant health, offering non-destructive spatial detection and quantification of plant diseases at various levels of measurement. When applied on various platforms, these optoelectronic sensors open new possibilities for predicting and responding to stress and plant diseases [19].

Portable NIR reflectance instruments have been used to evaluate feed on farms, including predicting crude proteins, acidic detergent fiber, and others [20,21,22].

The Leaf Scanner device was developed to analyze the distribution of chlorophyll content in whole leaves based on visible and near-infrared LEDs for visual and near-infrared imaging of leaves [23]. For the same purposes, a crop chlorophyll detector based on an optical sensor with an interference filter was developed in the spectral range of 400–1000 nm [24].

A portable near infrared spectroscopy system has been developed for the rapid measurement of water content in rapeseed leaves. The spectra were collected using an integrated spectrometer from 900 to 1700 nm [25]. Sensors capable of automatically measuring canopy temperature using a thermal imaging camera [26] have been made to non-destructively measure the area of individual leaves outdoors in daylight using an RGB-D sensor, Kinect v2 as part of a handheld device [27], or image processing techniques [28].

A system for estimating biomass from measurements of the vegetation cover of leafy vegetables based on laser sensors has been developed [29]. A portable system has been developed that can take 3D measurements and classify objects based on color and depth images obtained from multiple RGB sensors [30].

A diagnostic tool was proposed to assess the degree of ripeness of oilseed olive in the visible and near infrared ranges, calibrated using image analysis. An RGB image has been received. Spectroscopic analyses were performed using a benchtop FT-NIR and a portable vis/NIR instrument. The desktop device was equipped with a fiber optic probe, and the spectra were recorded in the range of 800–2500 nm [31]. Yang et al. [32] presented a portable touch detector for the degree of sweetness and firmness of kiwi fruit. The detector consisted of a control/processing unit, an LED panel, a driver unit, and a unit for detecting and amplifying the light signal. LEDs of 1000 and 1100 nm were used as the light source.

A reflectance spectrometer using NIR spectra (850–1040 nm) has been proposed to measure protein concentration directly on the combine [33]. At the same time, insufficient attention has been paid to luminescent methods and devices for determining plant infection, which also make it possible to diagnose plant pathologies with high sensitivity but at a lower cost.

Most of the existing methods for assessing biochemical characteristics use destructive chemical analyses, which require time and a lot of labor. In addition, these methods use highly potent chemicals that require special handling and disposal. Currently, various spectroscopic methods are used: reflection spectroscopy, chlorophyll fluorescence spectroscopy, IR thermal imaging, terahertz spectroscopy in the time domain, and hyperspectral imaging have all been used for biochemical studies by focusing on changes in the ratio of chlorophyll content, physical changes, or the aquatic status of plants. The development of reliable, non-invasive, accurate, and effective methods for evaluating breeding material based on physiological, morphological, biochemical, and other indicators that closely correlate with the productivity and stability of plants is relevant.

However, reflective infrared, thermal imaging, terahertz spectroscopy requires expensive high-precision equipment, so luminescent spectroscopy of the ultraviolet and visible range may be an effective alternative. During the excitation and emission of luminescence, a deeper and longer interaction of radiation and biological tissue occurs than during reflection.

The effectiveness of the use of spectral methods has been revealed to assess the contamination during key periods of growth of collected seeds when they are laid for storage.

The spectroscopic diagnostics included FTIR in the mid-IR region, Raman, and luminescence methods. Combination of chemometric tools with FTIR and Raman spectroscopy allowed obtaining approaches based on identified characteristic spectral features which may be used as infection markers. These approaches make it possible to detect the infection on the grain husk. The carotenoid type fungi pigment was identified within the resonance conditions of Raman scattering excitation [34].

The aim of this study is to design a device for optical photoluminescent diagnostics of Fusarium infection of seeds of cereal plants. On the basis of previously obtained spectral characteristics, it is necessary to establish the dependence of luminescent fluxes on infection and develop a method for determining the proportion of infected seeds in a sample. For practical implementation of the methodology, based on energy efficiency criteria, components and parts of the device were selected.

2. Materials and Methods

The infection of seeds with Fusarium was investigated. Winter wheat “Irishka No. 172”, barley “Moskovsky 86”, and oats “Salp” were used as seed samples.

The degree of infection of seeds was determined by external signs. Additionally, the chromatography method was used to determine the T-2 toxin by fluorescence in long-wavelength ultraviolet light.

The spectral characteristics of excitation and luminescence were measured using a previously developed technique [35,36]. Statistical processing was carried out, averaged over 250 spectra. The integral absorption capacity N and photoluminescence flux Φ were calculated using the Panorama Pro software package.

3. Results and Discussion

3.1. Obtainment and Analysis of the Spectra

The luminescence spectra φ_l(λ) were measured based on previously obtained results [35] for wavelengths of excitation maxima λ_e equal to 232, 362, 424, 485, and 528 nm. Figure 1 shows the spectral characteristics of the luminescence of wheat seeds at various λ_e. Researchers [34,37] have suggested that the glow is caused by the luminescence of chlorophyll α. Its presence in infected plants may be associated with infection during the flowering period.

Figure 1 shows that all photoluminescence spectra of infected seeds are higher than those of healthy seeds at λ_e,1 = 232 nm, λ_e,2 = 362 nm, λ_e,5 = 528 nm. When excited at λ_e,3 = 424 nm, the spectrum of infected seeds is located lower in the range of 450–542 nm, and at λ_e,4 = 485 nm, the spectra repeatedly intersect.

The calculation of integral flows and the construction of infestation dependencies on the β(Φ) flow are presented in previously published articles [35,36].

3.2. Calculation of Regression Models

The obtained dependencies were approximated by linear regression models.

The relative sensitivity of a change in flux with a change in the degree of infection can be determined by the formula:

$$S_{\mathsf{\Phi }}=100\times \left|\frac{{\mathsf{\Delta }\mathsf{\Phi }}_{\mathsf{\lambda }}}{∆\mathsf{\beta }\times {\Phi }_{\beta =0}}\right|$$

(1)

The coefficients of determination and sensitivity are presented in

Table 1. Parameters of linear regression models of flow dependences on the degree of infection.

Parameter	λe = 232 nm	λe = 362 nm	λe = 424 nm	λe = 485 nm
Wheat
R2	0.94	0.84	0.83	0.91
SΦ	1.50	0.40	0.13	0.09
Barley
R2	0.94	0.95	0.63	0.29
SΦ	2.06	1.44	0.29	0.11
Oats
R2	0.50	0.62	0.88	0.97
SΦ	1.33	0.19	0.31	0.17

.

For wheat, all dependencies are statistically significant. Approximations for barley at λ_e = 424 nm and λ_e = 485 nm and for oats at λ_e = 232 nm and λ_e = 362 nm are not statistically significant.

The highest sensitivity of wheat during luminescence excitation was at the wavelength of λ_e = 232 nm, similarly for barley. Lower sensitivities were observed when excited by 362 nm wavelength radiation. For oat seeds, the highest sensitivity was observed at λ_e = 424 nm.

It is advisable to determine the degree of infection by the ratio of fluxes Φ_λ1/Φ_λ2, which makes it possible to calibrate the measuring device in relative units. It will also help to increase the sensitivity if the wavelength λ₂ is chosen for the falling dependence Φ_λ(β). For barley, the highest sensitivity (S_Φ = 2.48) was achieved at the flux ratio Φ₂₃₂/Φ₄₈₅, and for wheat (Φ₂₃₂/Φ₄₂₄) it was 1.85. The coefficients of determination of the obtained dependencies were 0.99 and 0.98, respectively. For practical problems, inverse dependences of the degree of infection of seeds on their photoluminescence flux β(Φ_λ) were obtained.

Using 232 nm radiation to excite photoluminescence has disadvantages. First, there is a relatively low radiation flux (5.8–10.8 times less than at λ_e = 362 nm). Second, there are serious problems with the radiation source: narrow spectrum emitters (for example, LEDs) are available only with a wavelength of at least 250 nm and they are very expensive (from USD 500). In addition, in the shortwave spectrum, UV radiation produces ozone gas and this gas is toxic to humans from the air (sustainable ozone generation occurs when air is irradiated with a wavelength of less than 242 nm). There is not yet sufficient evidence to support widespread use where direct human exposure is expected [38].

Therefore, it is proposed to use excitation by radiation at the wavelength of 362 nm for the practical application of the method for wheat and barley.

Then the calibration equations are as follows, for wheat

$$\mathsf{\beta }=287\frac{{\mathsf{\Phi }}_{362}}{{\mathsf{\Phi }}_{424}}-137$$

(2)

for barley

$$\mathsf{\beta }=245\frac{{\mathsf{\Phi }}_{362}}{{\mathsf{\Phi }}_{485}}-57$$

(3)

and for oats

$$\mathsf{\beta }=0.22 {\mathsf{\Phi }}_{424}-289$$

(4)

The determination coefficients for Equations (2)–(4) are 0.88, 0.96, and 0.88, respectively.

3.3. Device of Photoluminescent Diagnostics of Seeds

Based on the obtained results [36], the method for determining the degree of infection of seeds with Fusarium was developed. To implement the method, it is necessary to develop a device. The block diagram of the device is shown in Figure 2.

The algorithm of the sensor is as follows:

• The selected seeds (D) are placed in a dark, light-tight chamber (B).
• Photoluminescence is excited for 20 microseconds by two (for wheat or barley) or one radiation source (for oats).
• Luminescence is detected 0.75–1.0 microseconds after the radiation source is turned off by two photodetectors (for wheat or barley) or one (for oats).
• The received analog electrical signal from the receivers (photodiodes E) is amplified by an amplifier (F), converted into a digital signal, and fed to the microcontroller (G).
• On the microcontroller, the degree of infection is calculated taking into account the photo signal and a priori information (calibration Equations (2)–(4)).
• The received and processed signal is fed to the indicator (I). Then a decision is made on subsequent operations with seeds.

3.4. Justification of the Choice of the Element Base of the Device

Requirements for an optical device for express diagnostics of grain seeds based on spectral luminescent parameters of grain.

(1)

Easy operation and minimal labor intensity.

(2)

Minimum permissible measurement error.

(3)

The weight of the device is not more than 3.5 kg to ensure maximum operator mobility.

(4)

Safety during operation.

(5)

Battery life of 5 h.

(6)

A light-tight chamber when measuring the seed material should ensure that there is no radiation from the external light background, as well as minimal reflections of the flow from the walls when the LEDs are working.

(7)

Convenience of cleaning the light-tight chamber from agricultural crops.

(8)

The result of measuring the diagnostic parameter, namely the degree of infection, should be displayed on the indicator for the operator.

(9)

Low cost of an express diagnostic device for maximum accessibility to users.

In a universal device that measures the infection of wheat, barley, and oats, it is necessary to have three radiation sources: 362 nm, 424 nm, and 485 nm. The most preferred is the use of LEDs having a narrow spectrum, excellent speed, making it possible to switch sources. The list and parameters of LEDs that can be used to create a device are shown in

Table 2. List of radiation sources.

LED Name [Data Source]	Radiation Wavelength λ, nm	Forward Current I, mA	Forward Voltage U, V	Angle of Radiation φ, ˚	Flow of Radiation Φe, W
VLMU3510-365-130	362	700	4.0	160	0.95
NICHIA NCSU276A	365	700	4.0	160	0.9
NICHIA NVSU233B	365	1400	3.85	160	1.4
NICHIA NCSU033C	365	700	4.0	160	0.75
CREELED424	424	700	4.0	150	0.8
LHUV-0420-0550	420	1000	3.5	150	1.0
LZ4-00UA00-00U6	400	1000	3.9	120	5.45
XPEBBL-L1-0000-00201	485	1000	3.5	140	3.5
MLEBLU-A1-0000-000T01	480	350	3.5	150	0.8
L1CUBLU100000000	485	350	3.5	150	0.034

.

The numerical criterion for choosing an LED is its effective output during excitation.

$$k_{ls}=\frac{{\Phi }_{eff}}{{\Phi }_{full}}=\frac{{\displaystyle \underset{0}{\overset{\infty }{\int }}\mathsf{\phi }{\left(\mathsf{\lambda }\right)}_{\mathrm{LED}}S\left(\mathsf{\lambda }\right)\mathrm{d}\mathsf{\lambda }}}{{\displaystyle \underset{0}{\overset{\infty }{\int }}\mathsf{\phi }{\left(\mathsf{\lambda }\right)}_{\mathrm{LED}}\mathrm{d}\mathsf{\lambda }}},$$

(5)

гд е Φ_eff—efficient flux;

Φ_full—full flux of exciting radiation;

S(λ)—spectral sensitivity of seeds—excitation spectrum;

φ(λ)_LED—radiation spectrum of the LED.

The calculation results are presented in

Table 3. Results of calculating the effective recoil of radiation sources.

Sources of Radiation	Kls
VLMU3510-365-130	0.98
NICHIA NCSU276A	0.97
NICHIA NVSU233B	0.98
NICHIA NCSU033C	0.98
CREELED424	0.95
LHUV-0420-0550	0.94
LZ4-00UA00-00U6	0.73
XPEBBL-L1-0000-00201	0.90
MLEBLU-A1-0000-000T01	0.93
L1CUBLU100000000	0.92

.

Secondary selection criterion are the radiation angle, the minimum value of the electric power and the maximum value of the radiation flux. From the analysis of the results of Table 3 and

Table 4. Radiation receiver.

Photodiode Name [Data Source]	Spectral Range λ, nm	Sensitivity S, A/W	Dark Current Id, fA
VEMD5510CF	440–620	0.2	2 × 105
FGAP71	150–550	0.12	4 × 104
Hamamatsu S8265	340–720	0.3	2 × 104
Hamamatsu S1133	320–730	0.4	1 × 104
Hamamatsu S7686	480–660	0.38	2 × 104
BWR21R	420–675	0.009	2 × 107
VBPW34S	430–1100	0.004	3 × 107
VEMD5510CF	440–620	0.2	2 × 105
SFH 2711	470–670	0.001	2 × 105
SLD-70 BG2A	400–700	0.001	1 × 108

, it follows that the most optimal sources of radiation will be VLMU3510-365-130 LEDs (for λ_e = 362 nm), CREELED424 (for λ_e = 424 nm), and XPEBBL-L1-0000-00201 (for λ_e = 485 nm).

As a radiation receiver, photodiodes are the most optimal due to their effective speed and small overall dimensions. The parameters of some photodiodes are presented in Table 4.

The main criterion for choosing a radiation receiver is matching the sensitivity spectrum of the receiver with the photoluminescence spectrum of seeds.

The effective luminescence output for photodiodes in the short-wavelength region of the spectrum is calculated using the formula:

$$k_{ld}=\frac{{\Phi }_{\mathrm{eff}}}{{\Phi }_{\mathrm{full}}}=\frac{{{\displaystyle \int }}_0^{\infty }\mathsf{\phi }\left(\mathsf{\lambda }\right)S_{\mathrm{ph}}{\left(\mathsf{\lambda }\right)}_{}\mathrm{d}\mathsf{\lambda }}{{{\displaystyle \int }}_0^{\infty }\mathsf{\phi }\left(\mathsf{\lambda }\right)\mathrm{d}\mathsf{\lambda }},$$

(6)

где S_ph(λ)—spectral sensitivity of the photodetector;

φ(λ)—photoluminescence spectrum of seeds.

The calculation results are presented in

Table 5. Results of calculating the effective luminescence recoil for radiation detectors.

Radiation Receiver	kld
VEMD5510CF	0.68
FGAP71	0.66
Hamamatsu S8265	0.75
Hamamatsu S1133	0.79
Hamamatsu S7686	0.49
BWR21R	0.87
VBPW34S	0.26
VEMD5510CF	0.76
SFH 2711	0.79
SLD-70 BG2A	0.89

.

Secondary selection criteria are maximum sensitivity and minimum dark current.

Thus, the optimal receivers are photodiodes VEMD5510CF (for λ_e = 362 nm and λ_e = 485 nm) and BWR21R (for λ_e = 485 nm). Turning on and off the LEDs, photodiodes for the corresponding excitation wavelengths is carried out using the ATmega328P control microcontroller. This microcontroller has a low cost (USD 8), high prevalence, but a low input supply voltage range (1.8–5.0 V). In this regard, it is necessary to use a high-precision voltage converter. The ATmega328P has 14 digital inputs/outputs and 6 analog inputs, which is enough to control selected sources and receivers.

To control the turning on and off of the LEDs, it is necessary to use three analog switches with a high current and voltage load capacity, since the selected LEDs, the parameters of which are presented in Table 3, have an input current of 700–1000 mA and a forward voltage of 3.5–4 V. A single-channel RDC1-S2 N power MOSFET power switch was chosen for such analog switches, which is designed to operate with high voltage, namely with 100 V and a current of 5.6 A. This switch has a low gate voltage threshold of 2 V and is compatible with the selected ATmega328P controller.

It is necessary to use operational amplifiers with high speed and gain, as well as low noise to amplify the electrical signal received from photodiodes. The AD820ANZ was chosen as such an operational amplifier. The AD820ANZ has an output signal slew rate of 3 V/µs, which is sufficient to detect luminescence after the light source is turned off. The operational amplifier is capable of operating with a single supply voltage in the range from 5 V to 36 V.

The output of information on the degree of infection of crops was carried out on the LCD 2004 display with I2C. This display has the 5 V supply voltage, as well as an I2C adapter that provides data exchange between two buses: a parallel LCD bus and an I2C bus. The I2C adapter also has an image contrast adjustment resistor and its own voltage regulator. LCD 2004 with I2С has a sufficient number of familiarity spaces, which allows you to correctly display the name of the culture, calibration Equations (2)–(4), and the degree of infection.

The power source was three NCR18650B batteries with a capacity of 3350 mA·h and a voltage of 3.7 V with the ability to replenish the capacity. The batteries were placed in the battery compartment and connected in series. The ATmega328P microcontroller was designed for power supply up to 5 V; in this regard, it is advisable to use the DC–DC converter MT3608, which will reduce the voltage from the batteries from 11 V to 5 V, and if the voltage drops below 5 V, it will increase to the required values (DC–DC).

The ATmega328P controller has a low output voltage of 3–5 V and a low output current of 50 mA, therefore you must use an additional current converter up to 2 A and voltage from 5 V to power radiation sources, single-channel power switches, operational amplifiers. LM2596 converter, which has an input voltage of 3 to 40 V, an output voltage of 1.5–35 V and an output current of up to 3 A.

The following areas of future research are expected:

-

laboratory and field tests of the developed device.

-

extension of the application of the developed method to other agricultural plants and other diseases.

A similar spectral method, but based on reflectivity data, was used to detect Fusarium pepper disease [39]. Unfortunately, the creation of a device implementing the method was not reported.

The device proposed in this study, unlike analogs, does not need to analyze the structure of volatile organic compounds [40] and does not require the construction of an image [41,42].

4. Conclusions

The method for determining infestation involves the excitation of seed luminescence and its registration in a light-protective chamber, as well as amplification and processing of the received electrical signal. For instrumental implementation of the method, it is energy efficient to use VLMU3510-365-130, CREELED424, and XPEBBL-L1-0000-00201 LEDs, as well as VEMD5510CF and BWR21R photodiodes.