A handheld crop leaf vegetation index detection method and device

Abstract

This invention provides a handheld method and device for detecting vegetation index of crop leaves, relating to the field of plant phenotyping technology. The method includes: controlling a contact linear array image sensor to scan the leaf of the crop to be tested line by line in the shaded leaf channel to obtain multi-channel reflectance grayscale digital quantities corresponding to each band; sequentially performing radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction on the multi-channel reflectance grayscale digital quantities to obtain multi-channel response data; determining the vegetation index of the leaf of the crop to be tested based on the multi-channel response data, and extracting the leaf phenotypic parameters of the leaf of the crop to be tested; determining the growth status of the leaf of the crop to be tested based on the vegetation index and leaf phenotypic parameters. This method can suppress the influence of ambient light interference, fluorescence crosstalk, and multi-band crosstalk on the measurement results, improve the accuracy of the vegetation index and the reliability of the leaf phenotypic parameters, and is beneficial for high-frequency, continuous, single-person handheld operation under field conditions.

Description

Technical Field

[0001] This invention relates to the field of plant phenotyping technology, and more particularly to a handheld method and apparatus for detecting crop leaf vegetation index. Background Technology

[0002] With the development of modern information technology and intelligent sensing, smart agriculture places higher demands on the rapid, sensitive, and non-destructive acquisition of crop nutrition and physiological status. This is an important foundation for precision fertilization, early warning of pests and diseases, and field management decisions. Chlorophyll is the main pigment in photosynthesis, and changes in its content can reflect crop photosynthetic capacity, growth potential, and stress response, thus becoming one of the key indicators for leaf phenotype and nutritional diagnosis.

[0003] Existing leaf-level detection technologies mainly include two methods: the first is a portable multi-band reflectance imaging device, which acquires two-dimensional reflectance images of leaves through multi-wavelength narrowband illumination and further calculates vegetation indices; the second is a handheld leaf spectrometer, which obtains leaf reflectance and transmission spectra through point measurement and outputs several indices. These two methods have advantages in spatial phenotypic information and broadband fine spectral information, respectively, but they also have drawbacks such as large size, low operational efficiency, sensitivity to ambient light, and contamination of reflectance signals by non-reflectance components, limiting their application in high-frequency, continuous, single-person handheld operation and real-time output under field conditions.

[0004] Developed by a team from the Department of Agro-Biological Engineering at Purdue University, LeafScope is used for whole-field multispectral reflectance imaging of crop leaves, such as soybeans. The system typically consists of a monochrome camera, a closed imaging chamber, and a multi-band LED light source. The imaging chamber reduces ambient light interference, and with the leaf laid flat on the imaging window, different wavelengths of LEDs illuminate it sequentially, acquiring multiple narrowband leaf reflectance images. In post-processing, vegetation indices such as NDVI are calculated for analysis of leaf nutrition and stress correlations. However, its limitations include: the system typically consists of a chamber, camera, and external computing terminal, limiting overall portability and continuous operation capabilities, making it more suitable for greenhouse or small-scale use; furthermore, its output focuses primarily on image acquisition and post-processing, lacking the ability to quickly output integrated data including geometric parameters, lesion percentage, and index results on-site.

[0005] The CI-710s is a handheld leaf spectrometer from CID Bio-Science. It can collect leaf transmission, absorption, and reflectance spectra in the visible to near-infrared range and calculate vegetation indices such as NDVI and pigment-related parameters. This type of device is compact, supports one-click acquisition, on-site display, and calibration, and is suitable for leaf spectral measurements in leaf color change monitoring, nitrogen nutrition assessment, and breeding experiments. However, its limitations include: spot measurement typically covers only a small local area of ​​the leaf, making it difficult to reflect the geometric characteristics and spatial distribution of lesions on the entire leaf; furthermore, the output is mainly spectral curves and indices, requiring some experience in parameter interpretation and not easily leading to intuitive conclusions that are user-friendly for agricultural production; and its limited ability to jointly characterize "geometric features + lesion phenotype + vegetation indices" restricts its application in large-scale, rapid leaf phenotypic acquisition and visualization management. Summary of the Invention

[0006] This invention provides a handheld crop leaf vegetation index detection method and device to solve the technical problem that the judgment of plant leaf growth status is not accurate enough in the prior art.

[0007] On one hand, the present invention provides a handheld method for detecting crop leaf vegetation index. The method is applied to a handheld device, which includes a contact linear array image sensor, a multi-band illumination source, and a light-shielding leaf channel. The method includes: The multi-band illumination source is controlled to light up sequentially according to a preset time sequence, and the contact linear array image sensor is controlled to scan the crop leaf under test line by line in the shading leaf channel to obtain the multi-channel reflectance grayscale digital quantity corresponding to each band. The multi-channel reflectance grayscale digital quantities are sequentially subjected to radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction to obtain multi-channel response data; Based on the multi-channel response data, the vegetation index of the leaves of the crop under test is determined, and the leaf phenotypic parameters of the leaves of the crop under test are extracted. Based on the vegetation index and the leaf phenotypic parameters, the growth status of the leaves of the crop under test is determined.

[0008] Optionally, the radiation baseline drift suppression includes: For each line of scan data scanned by the contact linear array image sensor, pixels are selected as reference bands according to preset rules, and the line-level light leakage baseline of the corresponding scan data is estimated based on the reference bands. A broadband fingerprint of ambient light is obtained, and the ambient light component is inverted based on the row-level leakage baseline and the broadband fingerprint. The ambient light component is then removed from the corresponding scan data. Scale back-scaling and gain alignment were performed on the multi-channel response data after removing the ambient light component to suppress radiation baseline drift.

[0009] Optionally, the fluorescence cross-channel crosstalk compensation includes: The digital values ​​of the first multi-channel reflectance grayscale under blue light excitation and the digital values ​​of the second multi-channel reflectance grayscale under green light excitation were obtained respectively. Based on the response values ​​of the preset fluorescence detection channels in the first multi-channel reflective gray digital quantity and the second multi-channel reflective gray digital quantity, fluorescence discrimination quantities under blue light excitation conditions and green light excitation conditions are constructed respectively, and the first fluorescence intensity under blue light excitation and the second fluorescence intensity under green light excitation are estimated according to the fluorescence discrimination quantities respectively. Based on the first fluorescence intensity, the second fluorescence intensity, and the multiple detection channels corresponding to the first multi-channel reflective grayscale digital quantity and the second multi-channel reflective grayscale digital quantity, the fluorescence crosstalk coefficient of each of the multiple detection channels is inverted under the joint constraint of the blade region. Based on the fluorescence crosstalk coefficient and the first fluorescence intensity, the fluorescence crosstalk component under blue light excitation is removed from the first multi-channel reflective grayscale digital quantity, and based on the fluorescence crosstalk coefficient and the second fluorescence intensity, the fluorescence crosstalk component under green light excitation is removed from the second multi-channel reflective grayscale digital quantity.

[0010] Optionally, the crosstalk bias correction includes: Based on the illumination sequence formed by multi-band illumination sources according to a preset time sequence, the proportion of charge remaining in the current frame after the previous frame is acquired by the contact linear array image sensor is estimated. Based on the charge ratio, differential processing is performed on the scan data of two adjacent scans in the illumination sequence to obtain differential data, and the cross-channel crosstalk coefficient is determined based on the ratio between the differential data of different detection channels. The bias inverse solution compensation is performed on the multi-channel reflective grayscale digital quantity based on the cross-channel crosstalk coefficient.

[0011] Optionally, before determining the growth status of the crop leaves to be tested, the method further includes: Based on the vegetation index, the growth period of the leaves of the crop to be tested is determined; Based on the growth period, growth status is diagnosed according to the vegetation index and the leaf phenotypic parameters to obtain the diagnostic results. Mechanism consistency testing was performed on the diagnostic results.

[0012] Optionally, the step of performing mechanistic consistency testing on the diagnostic results includes: Determine the feasible domain of the preset mechanism corresponding to the reproductive period; Determine whether the diagnostic result falls within the feasible region of the preset mechanism; If not, output an alarm flag.

[0013] Optionally, the leaf phenotypic parameters include geometric parameters and lesion percentage, and the extraction of leaf phenotypic parameters from the leaves of the crop to be tested includes: A two-dimensional multi-band image is constructed based on the multi-channel response data; Leaf region segmentation and lesion identification are performed on the two-dimensional multi-band image to obtain the geometric parameters and lesion regions of the crop leaf under test; wherein, the geometric parameters include length, width and area; Based on the geometric parameters and the lesion area, the area ratio of the lesion area in the leaf of the crop under test is determined as the lesion ratio.

[0014] Optionally, selecting pixels as reference bands according to preset rules includes: Calculate the brightness of the pixels in each row; Apply Gaussian smoothing to the pixel brightness; Calculate the average gray value of each pixel across multiple reference channels; When the brightness of a pixel after Gaussian smoothing is less than or equal to a preset texture threshold, and the average gray value is less than or equal to a preset reflection threshold, that pixel is selected as the reference band.

[0015] Optionally, the vegetation index includes: Normalized Difference Vegetation Index (NDVI); and / or Red-edged Normalized Difference Vegetation Index (NDVI).

[0016] On the other hand, the present invention provides a handheld device, including a contact linear array image sensor, a multi-band illumination source, a light-shielding blade channel, and a controller, wherein the controller includes: The scanning imaging module is used to control the multi-band illumination source to light up sequentially according to a preset time sequence, and to control the contact linear array image sensor to scan the crop leaf under test line by line in the shading leaf channel to obtain the multi-channel reflectance grayscale digital quantity corresponding to each band. The correction module is used to sequentially perform radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction on the multi-channel reflectance grayscale digital quantities to obtain multi-channel response data; The parameter extraction module is used to determine the vegetation index of the leaves of the crop under test based on the multi-channel response data, and to extract the leaf phenotypic parameters of the leaves of the crop under test. The state determination module is used to determine the growth state of the leaves of the crop under test based on the vegetation index and the leaf phenotypic parameters.

[0017] This invention provides a handheld method and device for detecting crop leaf vegetation index. It controls a multi-band illumination source to be lit sequentially according to a preset time sequence, and controls a contact linear array image sensor to scan the crop leaf under test line by line in the shaded leaf channel, obtaining multi-channel reflectance grayscale digital quantities corresponding to each band. The multi-channel reflectance grayscale digital quantities are then subjected to radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction to obtain multi-channel response data. Based on the multi-channel response data, the vegetation index of the crop leaf under test is determined, and the leaf phenotypic parameters are extracted. Based on the vegetation index and leaf phenotypic parameters, the growth status of the crop leaf under test is determined. This method can suppress the influence of ambient light interference, fluorescence crosstalk, and multi-band crosstalk on the measurement results, improving the accuracy of the vegetation index and the reliability of the leaf phenotypic parameters, thereby improving the accuracy of the growth status. This is beneficial for high-frequency, continuous, single-person handheld operation under field conditions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic flowchart of the handheld crop leaf vegetation index detection method provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the module structure of the handheld device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the overall architecture of the handheld device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the closed structure of the handheld device provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the open structure of the handheld device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] Figure 1This is a schematic flowchart of a handheld crop leaf vegetation index detection method provided in an embodiment of the present invention. The handheld crop leaf vegetation index detection method is applied to a handheld device, which includes a contact linear array image sensor, a multi-band illumination source, and a light-shielding leaf channel. The handheld device may include an upper cover and a lower cover; when closed, the upper and lower covers form a closed cavity, serving as the light-shielding leaf channel. The lower cover has a reference surface, which can be glass or acrylic sheet; the leaves can be laid flat on the reference surface.

[0022] Multi-band illumination sources can be positioned below or to the side of the reference plane. These sources are linearly arranged along the scanning direction, typically coaxially or symmetrically with the contact linear array image sensor. Multi-band illumination sources generally include red, green, blue, red-edge, and near-infrared light. The contact linear array image sensor is positioned below the reference plane, receiving light reflected from the blades through a window, which can be considered the reference plane itself.

[0023] See Figure 1 The handheld crop leaf vegetation index detection method includes the following steps: Step 101: Control the multi-band illumination source to light up sequentially according to the preset time sequence, and control the contact linear array image sensor to scan the leaf of the crop under test line by line in the shading leaf channel to obtain the multi-channel reflectance grayscale digital quantity corresponding to each band.

[0024] In this step, the leaves of the crop to be tested can be laid flat on a reference surface. Each time a band of illumination is lit, the contact linear array image sensor acquires one row of pixels, thus obtaining a row of reflected grayscale digital values. Controlling the multi-band illumination sources to light up sequentially according to a preset time sequence helps to accurately distinguish which band of illumination the reflected grayscale digital values ​​originate from.

[0025] The width of the shading leaf channel is set to allow only reflected light from the surface of the crop leaf to be acquired by the contact linear image sensor, thus blocking ambient light outside the leaf edge. Generally, the width of the shading leaf channel is slightly greater than or equal to the width of the leaf. Line-by-line scanning can be understood as acquiring information line by line along the length of the leaf.

[0026] Step 102: Perform radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction on the digital quantities of multi-channel reflectance grayscale sequentially to obtain multi-channel response data.

[0027] Step 103: Based on multi-channel response data, determine the vegetation index of the leaves of the crop to be tested, and extract the leaf phenotypic parameters of the leaves of the crop to be tested.

[0028] In this step, vegetation indices may include the Normalized Difference Vegetation Index (NDVI); and / or the Red Edge Normalized Difference Vegetation Index (RBVI).

[0029] Step 104: Determine the growth status of the leaves of the crop to be tested based on vegetation index and leaf phenotypic parameters.

[0030] In this embodiment, multi-band illumination sources are controlled to illuminate sequentially according to a preset time sequence, and a contact linear array image sensor is controlled to scan the leaf of the crop under test line by line in the shaded leaf channel to obtain the multi-channel reflectance grayscale digital quantity corresponding to each band. The multi-channel reflectance grayscale digital quantity is then subjected to radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction to obtain multi-channel response data. Based on the multi-channel response data, the vegetation index of the leaf of the crop under test is determined, and the leaf phenotypic parameters of the leaf of the crop under test are extracted. Based on the vegetation index and leaf phenotypic parameters, the growth status of the leaf of the crop under test is determined. This method can suppress the influence of ambient light interference, fluorescence crosstalk, and multi-band crosstalk on the measurement results, improve the accuracy of the vegetation index and the reliability of the leaf phenotypic parameters, thereby improving the accuracy of the growth status and facilitating high-frequency, continuous, single-person handheld operation under field conditions.

[0031] In one embodiment of this specification, radiation baseline drift suppression includes: Step 1: For each line of scan data scanned by the contact linear array image sensor, select pixels as reference bands according to preset rules, and estimate the line-level light leakage baseline of the corresponding scan data based on the reference bands. In this step, the preset rules generally refer to pixels with low texture and low reflection; specifically, low texture means that the smooth gradient is less than or equal to the preset texture threshold, and low reflection means that the average reflection value is less than or equal to the preset reflection threshold.

[0032] Specifically, the brightness of each pixel in each row can be calculated as shown in the following formula (1): (1); in, The brightness of the pixels within the row; Line number; This represents the pixel's position within the row; R indicates the red light channel; G indicates the green light channel; B indicates the blue light channel; This refers to a contact-type linear array image sensor. This refers to the digital quantity of the reflected grayscale. , , These are the digital values ​​of reflected grayscale for the red, green, and blue light channels, respectively. , , This is the brightness conversion factor.

[0033] The brightness of the pixels is smoothed using Gaussian, as shown in the following formula (2): (2); in, It is a Gaussian filter; The standard deviation of the Gaussian kernel; For gradient operators; This represents the smoothed gradient magnitude.

[0034] The average gray value of each pixel across multiple reference channels is calculated as shown in the following formula (3): (3); in, For the reference channel set; For channel The digital quantity of reflected grayscale; This represents the average grayscale value. This refers to the number of channels in the reference channel set. A reference channel is a set of channels selected from multiple spectral bands whose reflected signals are relatively stable and less affected by noise and interference. Reference channels typically include one or more visible light channels from near-infrared, red, green, blue, and red-edge wavelengths.

[0035] The preset rule is satisfied when the brightness of a pixel after Gaussian smoothing is less than or equal to the preset texture threshold, and the average gray value is less than or equal to the preset reflection threshold.

[0036] Based on the reference band, the row-level light leakage baseline of the corresponding scan data is estimated as shown in the following formula (4): (4); in, For the first Walking in the passage The light leakage baseline; For the first Reference band for the line; This is for median calculation.

[0037] Step 2: Obtain the broadband fingerprint of ambient light, invert the ambient light component based on the row-level leakage baseline and the broadband fingerprint, and remove the ambient light component from the corresponding scan data; In this step, ambient light refers to stray ambient light. The stray ambient light is considered as a broadband interpolation component formed after passing through the shielding cavity and optical channel; its cross-channel relative shape is determined by the stray light fingerprint vector of the device. Characterization: Stray light fingerprint vectors can be obtained through dark field slit calibration or ambient light injection calibration. This represents the total number of multi-band acquisition channels.

[0038] Construct the row-level baseline observation vector as shown in the following formula (5): (5); in, For the first The baseline vector of light leakage in the row; to Acquisition channels for all bands, .

[0039] The stray light intensity of this row is fitted with a robust scaling factor, as shown in Equation (6) or Equation (7) below: (6); in, A robust scaling factor; For the first The stray light intensity obtained by line fitting; Minimize the result .

[0040] (7); in, To prevent positive numbers with a denominator of zero, which are very small. Let be the stray light fingerprint vector of channel c.

[0041] Based on the stray light intensity and stray light fingerprint vector, the stray ingress amount is obtained as shown in the following formula (8): (8); in, This represents the stray interference.

[0042] Broadband ingress culling is performed on the entire row of pixels, as shown in the following formula (9): (9); in, Data to exclude ambient light components.

[0043] Step 3: Scale back-calibrate and gain align the multi-channel response data after removing the ambient light component to complete radiation baseline drift suppression; In this step, the main task is to perform scale backscaling and gain alignment on the red and near-infrared channels. The row-level scale factor of the red channel relative to the near-infrared channel is estimated within the reference band, as shown in formula (10): (10); in, For the first The indexing factor of the row; Near-infrared channel to eliminate ambient light component value; Red channel to eliminate ambient light component value.

[0044] Gain scaling is applied to the red channel (with the near-infrared channel maintained as the reference scale), as shown in the following formula (11): (11); in, For the enhanced red light channel value. It serves as the benchmark.

[0045] In one embodiment of this specification, fluorescence crosstalk compensation includes: Step 1: Obtain the digital values ​​of the first multi-channel reflectance grayscale under blue light excitation and the second multi-channel reflectance grayscale under green light excitation. In this step, the first multi-channel reflective grayscale digital quantity can refer to the grayscale digital quantity after removing the dark background under blue light excitation conditions, expressed as: The second multi-channel reflectance grayscale digital quantity can refer to the grayscale digital quantity after removing the dark background under green light excitation conditions, expressed as: Specifically, as shown in formulas (12) and (13) below: (12); (13); in, The first multi-channel reflective grayscale digital value without removing the dark background; A dark background under blue light excitation conditions; The second multi-channel reflective grayscale digital value without removing the dark background; The dark background is under green light excitation conditions. This indicates a row. The dark background can be black level or dark current, which can be measured in advance.

[0046] Step 2: Based on the response values ​​of the preset fluorescence detection channels in the first multi-channel reflective grayscale digital quantity and the second multi-channel reflective grayscale digital quantity, construct fluorescence discrimination quantities under blue light excitation conditions and green light excitation conditions respectively, and estimate the first fluorescence intensity under blue light excitation and the second fluorescence intensity under green light excitation according to the fluorescence discrimination quantities respectively. In this step, the sensitive responses of chlorophyll fluorescence radiation to 690nm and 730nm channels were utilized to construct fluorescence-sensitive discriminants under two excitation conditions. The detection channels were then used... For example, the blue-excited fluorescence discriminant is shown in the following formula (14): (14); The green excitation fluorescence discrimination quantification is shown in the following formula (15): (15); in, The fluorescence discrimination metric under blue light excitation conditions; and These are the weighting coefficients; Data for the 690nm channel under blue light excitation conditions; This is data for the 730nm channel under blue light excitation conditions. This represents the fluorescence discrimination metric under green light excitation conditions. Data for the 690nm channel under green light excitation conditions; Data for the 730nm channel under green light excitation conditions.

[0047] The first fluorescence intensity is shown in the following formula (16): (16); The second fluorescence intensity is shown in the following formula (17): (17); in, The first fluorescence intensity; The second fluorescence intensity; These are the calibration coefficients under blue light excitation conditions; These are the calibration coefficients under green light excitation conditions. The calibration coefficients can be obtained in advance through reference targets, multi-point calibration, or regression fitting.

[0048] Step 3: Based on the first fluorescence intensity, the second fluorescence intensity, and the multiple detection channels corresponding to the first multi-channel reflective grayscale digital quantity and the second multi-channel reflective grayscale digital quantity, the fluorescence crosstalk coefficient of each detection channel in the multiple detection channels is inverted under the joint constraint of the blade region. Specifically, fluorescence crosstalk is considered as an additional contamination term for each channel, and its cross-channel crosstalk coefficients are denoted as follows: Then for any channel Fluorescent crosstalk is shown in formulas (18) and (19) below: (18); (19); in, Under blue light excitation conditions, the channel The ideal pure reflection component, i.e. the reflection signal without fluorescence crosstalk. Under green light excitation conditions, the channel The ideal pure reflection component.

[0049] In low fluorescence reference areas or chlorophyll-free regions Under the joint constraint of the blade region, a robust inversion of the fluorescence crosstalk coefficient is performed. Taking blue light excitation as an example, the result is shown in the following formula (20): (20); in, Channel under blue light excitation The fluorescence crosstalk coefficient. Determine what makes the function reach its minimum value. . Channel under blue light excitation The ideal pure reflection component. For crosstalk coefficient variables; This refers to the leaf region. Similarly, the channel under green light excitation conditions can be obtained. fluorescence crosstalk coefficient In general, a low-fluorescence reference region can also be used. The combined constraints of (e.g., chlorophyll-free regions) and leaf regions are considered. In the inversion of fluorescence crosstalk, it is necessary to know the pure reflectance signal without fluorescence crosstalk. In the low-fluorescence reference region, since the fluorescence intensity is approximately zero, the fluorescence crosstalk model can be simplified; that is, the linearized signal observed in this region is approximately equal to the pure reflectance component. Therefore, data from the low-fluorescence reference region can be used to estimate the value of the pure reflectance signal.

[0050] Step 4: Based on the fluorescence crosstalk coefficient and the first fluorescence intensity, remove the fluorescence crosstalk component under blue light excitation from the first multi-channel reflective grayscale digital quantity, and based on the fluorescence crosstalk coefficient and the second fluorescence intensity, remove the fluorescence crosstalk component under green light excitation from the second multi-channel reflective grayscale digital quantity. Specifically, fluorescence crosstalk removal is performed on the affected channels, as shown in the following formulas (21) and (22): (twenty one); (twenty two); in, Data obtained under blue light excitation conditions after removing fluorescence crosstalk components. Data obtained by removing fluorescence crosstalk components under green light excitation conditions.

[0051] In one embodiment of this specification, crosstalk bias correction includes: Step 1: Based on the illumination sequence formed by multi-band illumination sources according to a preset time sequence, estimate the proportion of charge remaining in the current frame after the contact linear array image sensor was acquired in the previous frame; the charge proportion can also be understood as the tail residual coefficient. Specifically, the alternating lighting sequence is defined as , No. The frame excitation band is , No. The observed values ​​of the channel are (Pixel coordinates can be omitted to indicate pixel-by-pixel or ROI-by-ROI processing). The response tail is described using a first-order recursive model, as shown in formula (23). (twenty three); in, For the ideal response currently generated, The tail residue coefficient, for Frame number Observations of the channel, This represents the stray superposition and bias term. It can be estimated using empty fields or stable target sequences. .

[0052] Step 2: Based on the charge ratio, perform differential processing on the scan data of two adjacent scans in the illumination sequence to obtain differential data, and determine the cross-channel crosstalk coefficient based on the ratio between the differential data of different detection channels; Excite different times for two adjacent frames and The differential method is constructed to eliminate the slow-varying bias and highlight the tail crosstalk, as shown in the following formula (24): (twenty four); in, This is differential data.

[0053] The ratio is used to estimate the cross-channel crosstalk ratio (in terms of the target channel). Related to stimulation channels For example, as shown in the following formula (25): (25); in, For the channel To the passage The ratio of .

[0054] The crosstalk coefficient is obtained by taking robust statistics within the ROI, as shown in the following formula (26): (26); Among them, from the channel To the passage The crosstalk coefficient; It can be the entire image or a selected stable region.

[0055] Step 3: Perform bias inverse solution compensation on the digital quantity of multi-channel reflection grayscale based on the cross-channel crosstalk coefficient.

[0056] Specifically, the equivalent crosstalk is viewed as a hybrid matrix. For ideal signals The linear aliasing is shown in the following formula (27): (27); in, For the first The observation vectors for all channels of the frame; The ideal vector for all channels; The mixture matrix is ​​composed of crosstalk coefficients. Equal structure, For slow-changing bias.

[0057] The decoupling signal is obtained by inverse solution or iteration, as shown in the following formula (28): (28); Alternatively, a channel-by-channel elimination method can be used, as shown in the following formula (29): (29); in, This is the decoupling signal, which is the compensated signal; Estimated slow-varying bias. It is the inverse of the mixed matrix. For the compensated channel The signal. aisle The observed values.

[0058] In one embodiment of this specification, before determining the growth status of the crop leaves to be tested, the method further includes: Step 1: Determine the growth stage of the leaves of the crop to be tested based on vegetation indices; In this step, for each scan line or region First, the discriminant of the structural representation and the red edge representation is calculated to map the sample to the corresponding reproductive period layer or growth state layer, and then the matching calibration interval and parameter set are selected accordingly to avoid misjudgment caused by cross-stage mixing.

[0059] The stage discrimination feature is calculated as shown in formulas (30) and (31) below: (30); (31); in, The currently processed blade region is typically a scan line or a complete blade region. For pixels Normalized Difference Vegetation Index; For the region of median; For pixels The red-edge feature; For the region of median. It can take any red edge sensitive feature, such as red edge ratio, red edge slope, or red edge position.

[0060] The stage mapping is shown in the following formula (32): (32); in, It can be a threshold segmentation map or a lightweight classifier. For discrete stage labels or continuous stage variables. The mapping result, also known as the growth stage label, represents the growth stage layer or growth state layer in which the leaf is located. It can be a discrete stage label (such as the jointing stage, heading stage, and grain-filling stage) or a continuous stage variable (such as a value between 0 and 1, representing the growth process).

[0061] The selection of the calibration interval and parameter set corresponding to the stage can be expressed as: .

[0062] in, To coincide with the reproductive period The corresponding parameter set. (Through...) To mitigate systematic bias caused by stage differences, the same index can be interpreted using different scales across different reproductive stages. Nutritional grading thresholds serve as cutoff values ​​for classifying diagnostic results into different nutritional levels (e.g., deficiency, normal, excess). Regression coefficients are used to calculate model coefficients for nutritional content or stress risk from features, such as weights in linear regression. The mechanistic feasible domain boundary refers to the reasonable range of values ​​for each feature parameter within the reproductive stage, such as the normal range for NDVI and the reasonable range for fluorescence ratios, used for subsequent consistency testing.

[0063] Step 2: Based on the growth stage, perform growth status diagnosis according to vegetation index and leaf phenotypic parameters to obtain diagnostic results; Specifically, in phase constraints Under this approach, three complementary features—structural, biochemical, and physiological—are constructed and integrated to output nutrient classification and stress risk, thereby achieving decoupled expression of "structural influence—biochemical content—physiological state".

[0064] Structure item As shown in the following formula (33): (33); Biochemistry As shown in the following formula (34): (34); Overlayable red border parameters As shown in the following formula (35): (35); Among them, the structural parameters reflect the canopy structure, coverage, or greenness of the leaves. The biochemical parameters reflect the chlorophyll content or nitrogen nutrition status of the leaves. The red-bordered parameters enhance the characterization ability of biochemical information.

[0065] The physiological terms are composed of fluorescence-related information, as shown in the following formula (36): (36); in, This is a physiological item. For pixels Fluorescence intensity at 690 nm; For pixels Fluorescence intensity at 730nm; The fusion diagnostic output under stage constraints is shown in the following formula (37): (37); in, It can be output as chlorophyll content level, nitrogen nutrition level, or stress risk score; It can be a piecewise linear model, a rule set, or a lightweight regressor, and its threshold and coefficients are determined by... Constraints and choices are made at the stage level.

[0066] Step 3: Perform a mechanism consistency test on the diagnostic results.

[0067] In one embodiment of this specification, the consistency of diagnostic results is tested, including: Determine the feasible domain of the preset mechanism corresponding to the reproductive period; Determine whether the diagnostic results fall within the pre-defined feasible region of the mechanism; If not, output an alarm flag.

[0068] In this embodiment, to ensure that the output results are consistent with the plant nutrition and photosynthetic physiological mechanisms, the present invention sets stage-related mechanistic feasible domains and consistency constraints, performs quality checks on the diagnostic results, and provides a credibility score and alarm flags.

[0069] Relevant feasible domains in the definition phase This is used to constrain the reasonable range of key features and the boundaries of collaborative relationships, for example: In the stage The normal range below; and their ratio in the stage The reasonable range below; the boundary of the synergistic relationship between structural terms, biochemical terms, and physiological terms.

[0070] The out-of-bounds indication is calculated for each feature, as shown in the following formula (38): (38); in, For feature indexes, different feature items (such as NDVI, NDRE, RE, F690, F730, etc.) are represented. For the first The numerical values ​​of each feature; For the stage The The range of values ​​for each feature; For the first Outbound indication of a feature; Let be the distance function, calculate and The distance.

[0071] The consistency residual is constructed by combining the out-of-bounds amount with the stage reference level (or model residual), as shown in the following formula (39): (39); in, For the stage The expected center or reference level below, and As weight.

[0072] Credibility rating: Generally, the higher the rating, the more credible the product or service, as shown in the following formula (40): (40); in, Assess credibility. Specifically, if... Then output an alarm flag. The instructions suggest shading, bonding, exposure adjustment, or resampling, and also address... Reduce the weight or mark it as low trust; if Then output This results in nutritional grading and stress risk diagnosis with reliable quantitative information. Map the consistent residuals to a range of 0 to 1. For consistent residuals. This is a preset diagnostic threshold.

[0073] In one embodiment of this specification, leaf phenotypic parameters include geometric parameters and lesion percentage. Extracting leaf phenotypic parameters from the leaves of the crop under test includes: Two-dimensional multi-band images are constructed based on multi-channel response data; Leaf region segmentation and lesion identification are performed on two-dimensional multi-band images to obtain the geometric parameters and lesion regions of the leaves of the crop under test; wherein, the geometric parameters include at least one of length, width and area; Based on geometric parameters and lesion areas, the area ratio of lesion areas on the leaves of the crop under test is determined as the lesion ratio.

[0074] Based on the same general inventive concept, this invention also protects a handheld device. Figure 2 This is a schematic diagram of the modular structure of the handheld device provided in an embodiment of the present invention. The handheld device provided by the present invention will be described below, and the handheld device described below can be referred to in correspondence with the handheld crop leaf vegetation index detection method described above.

[0075] The handheld device includes a contact linear array image sensor, a multi-band illumination source, a light-shielding blade channel, and a controller. The controller includes a scanning imaging module 201, a calibration module 202, a parameter extraction module 203, and a status determination module 204.

[0076] The scanning imaging module 201 is used to control the multi-band illumination source to light up sequentially according to a preset time sequence, and to control the contact linear array image sensor to scan the crop leaf under test line by line in the shading leaf channel to obtain the multi-channel reflectance grayscale digital quantity corresponding to each band. The correction module 202 is used to sequentially perform radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction on the digital quantities of multi-channel reflectance grayscale to obtain multi-channel response data; The parameter extraction module 203 is used to determine the vegetation index of the leaves of the crop to be tested based on multi-channel response data, and to extract the leaf phenotypic parameters of the leaves of the crop to be tested. The state determination module 204 is used to determine the growth status of the leaves of the crop under test based on vegetation index and leaf phenotypic parameters.

[0077] Figure 3 This is a schematic diagram of the overall architecture of the handheld device provided in an embodiment of the present invention. See also... Figure 3 The handheld device mainly consists of a power supply module, a scanning imaging module, a main control processing module, a vegetation index and phenotypic analysis module, and a data storage and communication module.

[0078] The power supply module includes a lithium battery power supply module, a charging and power management module, and a power detection and display module, which realizes battery charging and discharging management, overvoltage and overcurrent protection, and visualization of remaining power.

[0079] The main control processing module is based on the OK3562 embedded processing platform and integrates a built-in TF or SD storage submodule, a human-machine interaction display submodule, and a key input submodule. It is responsible for core tasks such as power-on self-test, timing and light source control, data acquisition and caching, image processing and result display.

[0080] The scanning imaging module includes a multi-band LED light source driving module, a contact linear array image sensor module, and a blade narrow slit channel and clamping module. During operation, multiple narrow-band light sources such as red, green, blue, and red-edge or near-infrared light sources are driven in a preset sequence. The linear array sensor scans the blade attached to the reference surface line by line and stitches them together to form a multi-band two-dimensional reflection image.

[0081] The vegetation index and phenotypic analysis module completes algorithm processing on the main control end: First, leaf region segmentation and geometric feature extraction (length, width, area) are performed based on visible light images, followed by lesion identification and lesion proportion statistics; NDVI, NDRE and other vegetation indices are calculated on multi-band data after dark background removal and whiteboard normalization, and spatial distribution maps and overall leaf mean values ​​can also be output.

[0082] The data storage and communication module is used for result retention and external interconnection, and includes a local measurement data storage module, a serial communication module, and a wireless communication (4G) module. This data storage and communication module can write images, processing results, and metadata to a memory card according to the blade number, and supports exporting or uploading to the cloud via wired or wireless means.

[0083] Figure 4 and Figure 5 These correspond to the closed appearance and the open working position of the device, respectively. The handheld device features a slim, integrated housing. A cylindrical probe 401 is located at the front to house imaging or illumination components. A rectangular display screen 402 and a six-button human-machine interface (HMI) are arranged in the middle, including multiple buttons 403. The sides have reserved power interfaces 404 and data interfaces 406 (such as USB-C or serial ports) and status indicator lights 405 for easy power supply, data export, and working status indication. The outer shell has a wear-resistant matte coating, and the corners are chamfered to improve grip stability. The internal structure can integrate a main control processing board, TF or SD storage slots, and a battery compartment, forming a handheld integrated field measurement terminal.

[0084] Figure 5In this design, the upper cover 407, hinged open, forms a linear blade channel, with the blades resting flat on the reference glass or acrylic window of the lower cover 408. A multi-band LED linear light source (including visible light and red-edge or near-infrared channels) is arranged on one side of the channel near the window, coaxially with a contact linear image sensor (CIS) on the opposite side along the scanning direction. Limiting or clamping components are provided on both sides to ensure the blades adhere to the focal plane. The inner wall features light-absorbing and light-shielding structures to suppress stray light. When the upper cover is closed, a closed imaging cavity is formed, reducing ambient light interference. During operation, the light sources of each band are driven in a preset sequence, and the linear array is controlled to sample line by line to complete two-dimensional stitching. Simultaneously, within the same channel, leaf geometric features are extracted, lesion proportions are statistically analyzed, and indices such as NDVI and NDRE are calculated. This structure combines single-handed portability, rapid placement, and stable imaging, making it suitable for efficient measurement and batch operations under field conditions.

[0085] To facilitate understanding, the workflow of the handheld device is described below: Step 1, Power On and Initialization: Turn on the power, the main controller completes the peripheral self-test (CIS linear array, multi-band LED, display and buttons, storage and communication, etc.), loads the acquisition configuration and calibration parameters (dark field black level and row baseline parameters, light shielding cavity stray light fingerprint, red light and near-infrared back-labeling parameters, blue light and green light dual-excitation fluorescence calibration parameters, cross-channel crosstalk coefficient and threshold, mechanism feasible region and confidence threshold, etc.), and enters the standby interface.

[0086] Step 2, Environment and Channel Detection: Check whether the top cover and the light-shielding cavity are closed properly (optional detection of ambient light or light leakage indicator in the cavity). If not, prompt for closing the light-shielding or adjusting the fit; if satisfied, enter the parameter setting interface.

[0087] Step 3, Parameters and Sample Information: Users set crop type, number of measurements and sample number, band sequence and alternating illumination time sequence (including blue light excitation and green light excitation), exposure and gain strategy and output content, and the system establishes a traceability index (time, sample number, configuration version, etc.).

[0088] Step 4, blade placement and clamping: Attach the blade to the reference surface and close the contact-type light-blocking channel through the clamping and limiting mechanism to ensure consistent working distance and posture, and suppress external ambient light and geometric drift.

[0089] Step 5: Multi-band line scan acquisition: Obtain the original DN of multiple channels line by line according to the "single line single band lighting, linear array synchronous integration readout" method, and record abnormal markers such as missing lines, frame loss, saturation, and low signal-to-noise ratio (dark field lines or reference lines can be inserted for online correction).

[0090] Step 6, Exposure Adaptation and Quality Control: Adaptively adjust the LED current and integration time based on the reference band or target brightness, set saturation constraints and minimum signal-to-noise ratio constraints to ensure the sequence is stable and usable; if the quality control conditions are not met, prompt to adjust the shading bonding or re-examine.

[0091] Step 7, Light Leakage and Stray Light Suppression (Radiative Consistency): Intra-row low-texture, low-reflection reference bands are extracted to estimate row-level baseline terms. Combined with broadband fingerprint fitting of stray light from the light-shielding cavity and the removal of intrusive components, baseline drift is suppressed. At the same time, red light and near-infrared gain backscalars and scale alignment are performed to ensure that the exponential scales such as NDVI and NDRE are consistent.

[0092] Step 8: Fluorescence intrusion separation and compensation: Fluorescence-sensitive discriminant parameters are constructed using windows such as 690nm and 730nm for blue and green light excitation data, fluorescence intensity is estimated and contamination coefficient is inverted; fluorescence intrusion is removed from affected channels, and the key reflection channels are recalibrated and scaled to obtain a multi-channel response that is closer to the true reflection characteristics.

[0093] Step 9, Crosstalk and Integral Bias Correction (Optional): Model the tail residue or trailing of the response based on the alternating illumination sequence, estimate the crosstalk coefficient and compensate for it, and output a uniform multi-channel response.

[0094] Step 10, Index and Fusion Diagnosis: Calculate NDVI, NDRE, red edge characteristics and fluorescence-related characteristics within the leaf area, perform growth period stratification, and fuse output nutrient classification and stress risk under stage constraints.

[0095] Step 11, Reliability Assessment and Output: Perform a mechanism consistency quality check and calculate the reliability score; if the reliability is low, prompt for light-shielding bonding or resampling, and mark the result as low reliability or downweighted; if the reliability is high, output the diagnostic value, reliability, and alarm flag.

[0096] Step 12, Storage and Looping: Display the results and write the raw data, calibration data, index layer, statistics table and log to the TF card or SD card, or perform data export to move on to the next blade measurement or end the process.

[0097] Specific Implementation Example 1 (Nutritional Grading of Winter Wheat Leaves): Select winter wheat leaf samples, complete multi-band line scan acquisition and radiometric consistency correction, as well as fluorescence crosstalk separation correction according to the above process; calculate NDVI, NDRE, red edge features and fluorescence correlation features in the leaf area, perform fusion diagnosis based on growth period stratification constraints, output chlorophyll and nitrogen nutrient levels, and provide a confidence score and alarm marker to achieve rapid and traceable nutritional diagnosis in the field.

[0098] Specific Implementation Example 2 (Early Stress Risk in Maize Leaves): Leaf samples were selected under different management conditions in maize, and consistent multi-channel responses and fluorescence correlation quantities were obtained according to the above process; structural, biochemical, and physiological items were integrated to output a stress risk score, and the reliability was calculated through mechanism consistency quality inspection. Low reliability triggered an alarm and prompted to check the shading and bonding or re-sampling, while high reliability output the risk value and reliability indicator, thereby improving the reliability and interpretability of early stress identification.

[0099] In summary, the present invention has the following advantages: Integrating various types of modules into a compact body, the overall structure is lightweight and easy to operate, making it suitable for single-person long-term carrying and continuous measurement; the clamping and channel structure supports quick disassembly and replacement, and can be adapted to different crops and leaves of different widths and thicknesses, improving the equipment's versatility and ease of maintenance.

[0100] Integrating multiple narrowband reflection channels such as visible light, red edge band (e.g., 730nm), and near-infrared band (e.g., 810nm, 940nm), and employing active illumination and a shading cavity to work together to avoid ambient light interference, it can simultaneously acquire color, red edge, and near-infrared information at the leaf scale. This enables multidimensional joint characterization of chlorophyll and nitrogen nutritional status (NDVI, NDRE), water absorption characteristics (WBI), and yellowing and color change (EGI). Compared to detection methods that rely on a single index or a few bands of reflection, it provides more comprehensive information and is more sensitive and stable to early stress and slight nutrient changes.

[0101] To address the issues of uneven illumination, thermal drift, and light source attenuation affecting contact linear array scanning structures, a row-level flattening and reflectivity recovery strategy of "edge strip white board + dark field" is proposed. During the scanning process, dark field and reference reflectivity information are acquired simultaneously, and row-level darkening, flattening, and normalization corrections are performed on the data of each band. Combined with the nominal reflectivity of the white board, consistent calibration is achieved across time and devices, significantly reducing systematic deviations caused by hardware differences, aging, or environmental changes, and improving the comparability and long-term stability of measurement results.

[0102] This invention proposes a multi-band affine registration method based on leaf margin gradient features and channel geometric priors. This method achieves pixel-level alignment of different reflection bands at the end side, effectively suppressing geometric misalignment caused by scanning offset or optical path differences. It also avoids jagged edges, artifacts, and streaks in ratio-based vegetation indices such as NDVI, NDRE, and WBI at the edges and lesion areas, thereby improving the accuracy and repeatability of index maps, lesion maps, and spatial statistical results.

[0103] This invention simultaneously outputs geometric feature parameters (length, width, area), lesion proportion, and multiple vegetation indices (NDVI, NDRE, WBI, EGI, etc.) at the single-leaf scale. Based on these, a multivariate composite health index (MVHI) is constructed for leaf health grading evaluation, achieving integrated quantitative characterization of nutritional status, pathological status, and morphological characteristics. The detection results support end-side visualization, local storage, and data export, reducing dependence on user spectrum and model background, and are suitable for rapid judgment and decision support in agricultural production sites.

[0104] This invention operates at the leaf scale using a contact scanning structure. The measurement process is unaffected by factors such as soil background, weeds, changes in canopy structure, and fluctuations in observation distance. Compared with remote sensing or canopy reflectance measurement methods, it offers stable calibration conditions and high spatial resolution. It can be used for fine phenotypic analysis of single leaves as well as for rapid batch measurements, providing a portable, reliable, and information-dense technical means for monitoring crop growth, nutrient diagnosis, early disease identification, and precision management in the field.

[0105] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units.

[0106] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as OM / AM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A handheld method for detecting crop leaf vegetation index, characterized in that, The method is applied to a handheld device, the handheld device including a contact linear array image sensor, a multi-band illumination source, and a light-shielding blade channel, the method comprising: The multi-band illumination source is controlled to light up sequentially according to a preset time sequence, and the contact linear array image sensor is controlled to scan the crop leaf under test line by line in the shading leaf channel to obtain the multi-channel reflectance grayscale digital quantity corresponding to each band. The multi-channel reflectance grayscale digital quantities are sequentially subjected to radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction to obtain multi-channel response data; Based on the multi-channel response data, the vegetation index of the leaves of the crop under test is determined, and the leaf phenotypic parameters of the leaves of the crop under test are extracted. Based on the vegetation index and the leaf phenotypic parameters, the growth status of the leaves of the crop under test is determined.

2. The handheld crop leaf vegetation index detection method according to claim 1, characterized in that, The radiation baseline drift suppression includes: For each line of scan data scanned by the contact linear array image sensor, pixels are selected as reference bands according to preset rules, and the line-level light leakage baseline of the corresponding scan data is estimated based on the reference bands. A broadband fingerprint of ambient light is obtained, and the ambient light component is inverted based on the row-level leakage baseline and the broadband fingerprint. The ambient light component is then removed from the corresponding scan data. Scale back-scaling and gain alignment were performed on the multi-channel response data after removing the ambient light component to suppress radiation baseline drift.

3. The handheld crop leaf vegetation index detection method according to claim 2, characterized in that, The step of selecting pixels as reference bands according to preset rules includes: Calculate the brightness of the pixels in each row; Apply Gaussian smoothing to the pixel brightness; Calculate the average gray value of each pixel across multiple reference channels; When the brightness of a pixel after Gaussian smoothing is less than or equal to a preset texture threshold, and the average gray value is less than or equal to a preset reflection threshold, that pixel is selected as the reference band.

4. The handheld crop leaf vegetation index detection method according to claim 1, characterized in that, The fluorescence cross-channel crosstalk compensation includes: The digital values ​​of the first multi-channel reflectance grayscale under blue light excitation and the digital values ​​of the second multi-channel reflectance grayscale under green light excitation were obtained respectively. Based on the response values ​​of the preset fluorescence detection channels in the first multi-channel reflective gray digital quantity and the second multi-channel reflective gray digital quantity, fluorescence discrimination quantities under blue light excitation conditions and green light excitation conditions are constructed respectively, and the first fluorescence intensity under blue light excitation and the second fluorescence intensity under green light excitation are estimated according to the fluorescence discrimination quantities respectively. Based on the first fluorescence intensity, the second fluorescence intensity, and the multiple detection channels corresponding to the first multi-channel reflective grayscale digital quantity and the second multi-channel reflective grayscale digital quantity, the fluorescence crosstalk coefficient of each of the multiple detection channels is inverted under the constraint of the blade region. Based on the fluorescence crosstalk coefficient and the first fluorescence intensity, the fluorescence crosstalk component under blue light excitation is removed from the first multi-channel reflective grayscale digital quantity, and based on the fluorescence crosstalk coefficient and the second fluorescence intensity, the fluorescence crosstalk component under green light excitation is removed from the second multi-channel reflective grayscale digital quantity.

5. The handheld crop leaf vegetation index detection method according to claim 1, characterized in that, The crosstalk bias correction includes: Based on the illumination sequence formed by multi-band illumination sources according to a preset time sequence, the proportion of charge remaining in the current frame after the previous frame is acquired by the contact linear array image sensor is estimated. Based on the charge ratio, differential processing is performed on the scan data of two adjacent scans in the illumination sequence to obtain differential data, and the cross-channel crosstalk coefficient is determined based on the ratio between the differential data of different detection channels. The bias inverse solution compensation is performed on the multi-channel reflective grayscale digital quantity based on the cross-channel crosstalk coefficient.

6. The handheld crop leaf vegetation index detection method according to claim 1, characterized in that, Before determining the growth status of the crop leaves to be tested, the method further includes: Based on the vegetation index, the growth period of the leaves of the crop to be tested is determined; Based on the growth period, growth status is diagnosed according to the vegetation index and the leaf phenotypic parameters to obtain the diagnostic results. Mechanism consistency testing was performed on the diagnostic results.

7. The handheld crop leaf vegetation index detection method according to claim 6, characterized in that, The mechanism consistency test of the diagnostic results includes: Determine the feasible domain of the preset mechanism corresponding to the reproductive period; Determine whether the diagnostic result falls within the feasible region of the preset mechanism; If not, output an alarm flag.

8. The handheld crop leaf vegetation index detection method according to claim 1, characterized in that, The leaf phenotypic parameters include geometric parameters and the proportion of lesions. Extracting the leaf phenotypic parameters of the crop leaves to be tested includes: A two-dimensional multi-band image is constructed based on the multi-channel response data; Leaf region segmentation and lesion identification are performed on the two-dimensional multi-band image to obtain the geometric parameters and lesion regions of the crop leaf under test; wherein, the geometric parameters include length, width and area; Based on the geometric parameters and the lesion area, the area ratio of the lesion area in the leaf of the crop under test is determined as the lesion ratio.

9. The handheld crop leaf vegetation index detection method according to claim 1, characterized in that, The vegetation index includes: Normalized Difference Vegetation Index (NDVI); and / or Red-edged Normalized Difference Vegetation Index (NDVI).

10. A handheld device, characterized in that, It includes a contact linear array image sensor, a multi-band illumination source, a light-shielding blade channel, and a controller, wherein the controller includes: The scanning imaging module is used to control the multi-band illumination source to light up sequentially according to a preset time sequence, and to control the contact linear array image sensor to scan the crop leaf under test line by line in the shading leaf channel to obtain the multi-channel reflectance grayscale digital quantity corresponding to each band. The correction module is used to sequentially perform radiation baseline drift suppression, fluorescence cross-channel crosstalk compensation, and crosstalk bias correction on the multi-channel reflectance grayscale digital quantities to obtain multi-channel response data; The parameter extraction module is used to determine the vegetation index of the leaves of the crop under test based on the multi-channel response data, and to extract the leaf phenotypic parameters of the leaves of the crop under test. The state determination module is used to determine the growth state of the leaves of the crop under test based on the vegetation index and the leaf phenotypic parameters.

Images