A cotton leaf chlorophyll and nitrogen content detection instrument and detection method

Abstract

The present invention relates to an instrument and method for detecting the chlorophyll and nitrogen content of cotton leaves, belonging to the technical fields of spectral analysis and plant phenotyping. An RGB camera is used to obtain an RGB image of a cotton leaf to be tested that is illuminated by a light source, and a spectral probe is used to obtain spectral data of the cotton leaf to be tested that is illuminated by the light source. The color features, morphological features, and texture features of the RGB image are then extracted, and the spectral features of the spectral data are extracted. The color features, morphological features, texture features, and spectral features are then fused to obtain fused features. Finally, the fused features are used as input to calculate the chlorophyll content and nitrogen content of the cotton leaf to be tested using a prediction model. This method can achieve rapid and automatic acquisition of phenotypic data on the chlorophyll and nitrogen content of the cotton leaf, and can overcome the problem that existing plant leaf phenotyping platforms are difficult to carry and the chlorophyll and nitrogen content of cotton leaves in the field cannot be quickly and on-site obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of spectral analysis and plant phenotyping, and in particular to a closed portable detection instrument and a detection method thereof that can be used for detecting chlorophyll and nitrogen content in cotton leaves. Background Art

[0002] Cotton, with its excellent natural properties, is one of my country's important economic crops. Chlorophyll and nitrogen content in cotton leaves is crucial for characterizing its physiological and nutritional status. Traditional methods for measuring chlorophyll and nitrogen content in cotton leaves primarily rely on UV-Vis spectrophotometry and the Kjeldahl method. While these methods offer feasibility, reproducibility, and high accuracy, they are time-consuming, tedious, and subject to irreversible sample damage, limiting their practical application. In recent years, advances in automation, machine vision, and spectral analysis have enabled high-throughput, precise, and efficient plant leaf phenotyping technologies. These technologies use sensors to measure chlorophyll and nitrogen phenotypic data, enabling the prediction of chlorophyll and nitrogen content based on these data, with promising results. However, high-throughput plant phenotyping platforms are often limited in practice due to their bulk, high cost, low efficiency, and difficulty in storing and processing information. These platforms are difficult to transport to the field for field data collection.

[0003] Based on this, there is an urgent need for a detection instrument and a detection method with small size and high detection efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide a detection instrument and method for the chlorophyll and nitrogen content of cotton leaves, which can realize the rapid and automatic acquisition of phenotypic data of chlorophyll and nitrogen content of cotton leaves, and can overcome the problems that the existing plant leaf phenotyping platform is difficult to carry and the chlorophyll and nitrogen content of cotton leaves in the field cannot be quickly obtained on the spot.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A cotton leaf chlorophyll and nitrogen content detection instrument, comprising: a light-shielding housing, a control chip, and a reflectivity plate, a light source, an RGB camera, and a spectrum probe located within the light-shielding housing; the reflectivity plate is located on the bottom surface of the light-shielding housing; the control chip is communicatively connected to the RGB camera and the spectrum probe, respectively;

[0007] The cotton leaf to be tested is located on the reflectivity plate; the light source is used to illuminate the cotton leaf to be tested; the RGB camera is used to obtain an RGB image of the cotton leaf to be tested illuminated by the light source; and the spectrum probe is used to obtain spectral data of the cotton leaf to be tested illuminated by the light source;

[0008] The control chip is used to calculate the chlorophyll content and nitrogen content of the cotton leaves to be tested based on the RGB image and the spectral data.

[0009] A method for detecting chlorophyll and nitrogen content in cotton leaves, the method comprising:

[0010] Receive RGB images acquired by the RGB camera and spectral data acquired by the spectral probe;

[0011] Extracting color features, morphological features, and texture features of the RGB image, and extracting first spectral features and second spectral features of the spectral data;

[0012] fusing the color feature, the morphological feature, the texture feature, and the first spectral feature to obtain a first fused feature; fusing the color feature, the morphological feature, the texture feature, and the second spectral feature to obtain a second fused feature;

[0013] The first fusion feature is used as input, and the chlorophyll content of the cotton leaf to be tested is calculated using a first prediction model; the second fusion feature is used as input, and the nitrogen content of the cotton leaf to be tested is calculated using a second prediction model.

[0014] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0015] The present invention is used to provide an instrument and method for detecting the chlorophyll and nitrogen content of cotton leaves. The instrument utilizes an RGB camera to obtain an RGB image of the cotton leaf to be tested that is illuminated by a light source, utilizes a spectral probe to obtain spectral data of the cotton leaf to be tested that is illuminated by the light source, then extracts color features, morphological features, and texture features of the RGB image, extracts spectral features of the spectral data, and then fuses the color features, morphological features, texture features, and spectral features to obtain fused features. Finally, the fused features are used as input and a prediction model is used to calculate the chlorophyll content and nitrogen content of the cotton leaf to be tested. The rapid and automatic acquisition of phenotypic data of the chlorophyll and nitrogen content of the cotton leaf can be achieved, and the problems that the existing plant leaf phenotyping platform is difficult to carry and the chlorophyll and nitrogen content of cotton leaves in the field cannot be quickly and on-site obtained can be overcome. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 This is a schematic diagram of the overall structure of the detection instrument provided in Example 1 of the present invention;

[0018] Figure 2 A flow chart of the detection method provided in Example 2 of the present invention;

[0019] Figure 3 This is an operation flow chart of the detection instrument provided in Example 2 of the present invention;

[0020] Figure 4 This is a schematic diagram of the network structure of the prediction model provided in Example 2 of the present invention.

[0021] Explanation of symbols:

[0022] 1-Power supply; 2-Power plug; 3-Control chip; 4-Control lever; 5-Light source; 6-Spectral probe; 7-RGB camera; 8-Fan; 9-Light-shielding shell; 10-Start button; 11-Cotton leaves to be tested; 12-Bottom sample drawer; 13-Reflectivity plate; 14-USB output interface. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] The purpose of the present invention is to provide a detection instrument and method for the chlorophyll and nitrogen content of cotton leaves, which can realize the rapid and automatic acquisition of phenotypic data of chlorophyll and nitrogen content of cotton leaves, and can overcome the problems that the existing plant leaf phenotyping platform is difficult to carry and the chlorophyll and nitrogen content of cotton leaves in the field cannot be quickly obtained on the spot.

[0025] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] Example 1:

[0027] This embodiment is used to provide a cotton leaf chlorophyll and nitrogen content detection instrument, the detection instrument is a closed portable instrument. Figure 1As shown, the detection instrument includes: a light-shielding housing 9, a control chip 3, a reflectivity plate 13 located inside the light-shielding housing 9, a light source 5, an RGB camera 7, and a spectrum probe 6. The control chip 3 can be an AI control chip and peripheral circuits. The control chip 3 can be located inside the light-shielding housing 9, specifically, can be installed on the top of the light-shielding housing 9. The light source 5 can be a group of array halogen lamps, which are a light source with a long service life and a stable spectrum. The light source 5 can be installed on the top of the light-shielding housing 9, preferably at the top center of the light-shielding housing 9. The RGB camera 7 can be a detachable RGB camera, and the spectrum probe 6 can be a rectangular spectrum probe.

[0028] A reflectivity plate 13 is located on the bottom surface of the light-shielding housing 9, and the cotton leaf 11 to be tested is located on the reflectivity plate 13. The reflectivity of the reflectivity plate 13 can be 30%. A light source 5 is used to illuminate the cotton leaf 11 to be tested. The RGB camera 7 is used to obtain an RGB image of the cotton leaf 11 to be tested illuminated by the light source 5. The spectrum probe 6 is used to obtain spectral data of the cotton leaf 11 to be tested illuminated by the light source 5.

[0029] Since the cotton leaf 11 to be tested is in the light-shielding housing 9, and the bottom is a reflectivity plate 13 with a fixed reflectivity, and a light source 5 is provided to illuminate the cotton leaf 11 to be tested, the influence of external light can be avoided. In this shooting environment, there is no need to perform complex background elimination on the acquired RGB image and spectral data, which can reduce the processing volume.

[0030] The control chip 3 is respectively connected to the RGB camera 7 and the spectrum probe 6 for communication. The control chip 3 is used to calculate the chlorophyll content and nitrogen content of the cotton leaf 11 to be tested based on the RGB image and the spectrum data.

[0031] As an optional embodiment, the bottom of the light-shielding shell 9 adopts a drawer design, and a bottom sample drawer 12 is provided at the bottom of the inner side thereof. The bottom sample drawer 12 is slidably connected to the light-shielding shell 9 via a slide rail. The light-shielding shell 9 and the bottom sample drawer 12 form a closed environment, so that the detection process is free from interference from the external environment. The bottom surface of the bottom sample drawer 12 is black, and the reflectivity plate 13 is located on the bottom surface of the bottom sample drawer 12. The cotton leaf 11 to be tested is still located on the reflectivity plate 13. When it is necessary to place the cotton leaf 11 to be tested in the light-shielding shell 9, the bottom sample drawer 12 is pulled out, and after the cotton leaf 11 to be tested is placed in the bottom sample drawer 12, the bottom sample drawer 12 is closed to more conveniently place the cotton leaf 11 to be tested, and the placement position of the cotton leaf 11 to be tested can be easily adjusted.

[0032] The detection instrument of this embodiment also includes a moving part, which is located inside the light-shielding shell 9 and can be specifically installed on the top of the light-shielding shell 9. The RGB camera 7 and the spectrum probe 6 are installed on the moving part, and the moving part is used to drive the RGB camera 7 and the spectrum probe 6 to move horizontally and lift. Specifically, the moving part includes a driving member and a control rod 4, and the driving member is connected to the control rod 4. The driving member can be a stepping motor, and the stepping motor can be fixedly connected to a side wall inside the light-shielding shell 9. The control rod 4 can be a stepping control rod. The control rod 4 includes a horizontal axis and a vertical axis, and the RGB camera 7 and the spectrum probe 6 are installed on the vertical axis. Under the drive of the stepping motor, the lifting and horizontal movement of the spectrum probe 6 and the RGB camera 7 installed thereon are completed. By lifting and lowering, the imaging height can be easily set. For example, when the cotton leaf 11 to be measured is small, the lenses of the spectrum probe 6 and the RGB camera 7 can be lowered to obtain a more accurate spectrum and an RGB image of appropriate size.

[0033] Because the light source 5, RGB camera 7, and spectral probe 6 generate heat during long-term operation, the detection instrument of this embodiment further includes a fan 8 to dissipate heat. The fan 8 is located inside the light-shielding housing 9 and can be mounted on the sidewalls of the light-shielding housing 9. There can be multiple fans 8. The fan 8 is used to dissipate heat.

[0034] The control chip 3 of this embodiment can be controlled and connected with the light source 5, stepper motor, spectral probe 6, RGB camera 7 and fan 8, and is used to control the start and shutdown of the light source 5, stepper motor, spectral probe 6, RGB camera 7 and fan 8. The control chip 3 is also used to store the collected data, and the collected data includes RGB images and spectral data.

[0035] To provide power, the instrument in this embodiment also includes a power supply 1 located outside the light-shielding housing 9. This power supply 1 is an external mobile power source for easy portability. Power supply 1 connects to a power plug 2 extending from the outside of the light-shielding housing 9 and is used to power the light source 5, stepper motor, RGB camera 7, spectral probe 6, and fan 8 via peripheral circuitry.

[0036] The light shielding housing 9 of this embodiment is further provided with a start button 10 and a USB output interface 14 . The start button 10 is used to start the detection instrument, and the USB output interface 14 is used to export the collected data and the detection results of the control chip 3 .

[0037] The detection instrument of this embodiment is easy to carry to the field and can be powered by a mobile power supply. It collects spectral and image information at the same time, extracts the color, morphology and texture characteristics of the leaves, as well as the spectral characteristics at characteristic wavelengths, fuses the image characteristics and spectral characteristics, and uses the control chip 3 to process the collected data in real time. According to the model integrated in the control chip 3, the fusion information of the spectrum and image is used to predict, calculate and output the chlorophyll and nitrogen content parameters of the leaves. The detection instrument provided by this embodiment is a closed portable instrument that can be used for field plant leaf phenotypic detection. It has a simple structure, is easy to carry, easy to operate and has a low cost. It can provide the environment required for algorithm implementation. By using the device of this embodiment to obtain RGB images and spectral data of leaves, it is possible to directly realize the rapid acquisition of chlorophyll content and nitrogen content phenotypic data of cotton leaves. Data collection, calculation and output are all implemented on the hardware platform of the detection instrument, and there is no need to copy or transmit the data to a remote server for calculation and analysis via the Internet.

[0038] Example 2:

[0039] This embodiment is used to provide a method for detecting the chlorophyll and nitrogen content of cotton leaves, and the detection instrument described in Example 1 is controlled to work, such as Figure 1 and Figure 2 As shown, the detection method includes:

[0040] S1: receiving the RGB image acquired by the RGB camera 7 and the spectral data acquired by the spectral probe 6;

[0041] like Figure 3 As shown, the method of using the detection instrument to obtain RGB images and spectral data is as follows: after starting the power supply 1, the bottom sample drawer 12 is opened by the slide rail, the cotton leaf 11 to be tested is placed, and the bottom sample drawer 12 is closed; the control chip 3 controls the light source 5, the stepping motor, the RGB camera 7, the spectrum probe 6 and the fan 8 to be turned on; after waiting for the light source 5 to stabilize, the stepping control rod is controlled to move, driving the RGB camera 7 and the spectrum probe 6 carried on the stepping control rod to move, and judging whether the cotton leaf 11 to be tested is within the field of view based on the size of the imaging or spectral reflectance. Specifically, the position of the cotton leaf 11 to be tested is automatically identified based on the color and reflectance difference between the cotton leaf 11 to be tested and the 30% reflectance plate 13, and the position coordinates of the cotton leaf 11 to be tested are determined. Mainly, whether the complete leaf sample is within the field of view is judged based on the RGB imaging. The movement is continuously stopped until the cotton leaf 11 to be tested is within the data acquisition range of the RGB camera 7 and the spectrum probe 6, and the RGB image is collected by the RGB camera 7, and the spectrum data within the coverage range of the spectrum probe 6 is collected by the spectrum probe 6.

[0042] The bottom sample drawer 12 can also be opened automatically: the control chip 3 outputs a driving signal to make the bottom sample drawer 12 pop out.

[0043] It should be noted that if the cotton leaf 11 to be tested is too large, no matter how the RGB camera 7 and the spectral probe 6 are moved, the cotton leaf 11 to be tested cannot be completely located within the data acquisition range of the RGB camera 7 and the spectral probe 6. That is, when the cotton leaf 11 to be tested is large and the RGB camera 7 and the spectral probe 6 cannot collect the complete leaf image and spectral data at one time, the RGB camera 7 and the spectral probe 6 can be continuously moved to take multiple shots and perform data splicing to obtain the complete RGB image and spectral data of the cotton leaf 11 to be tested, which can better solve the problem that the cotton leaf 11 to be tested is too large to directly acquire data.

[0044] S2: extracting color features, morphological features, and texture features of the RGB image, and extracting first spectral features and second spectral features of the spectral data;

[0045] Specifically, in S2, extracting the color features, morphological features, and texture features of the RGB image may include:

[0046] (1) performing threshold segmentation on the RGB image to obtain a binary image; the binary image includes leaf pixels and background pixels;

[0047] More specifically, the threshold segmentation method for RGB images is as follows: convert the RGB image to HSV space. Since the S component values ​​of leaves and background are quite different, a fixed S (saturation) value is set to form a mask. The RGB image is processed using the mask to form a binary image. For each pixel in the RGB image, if the S component value of the pixel is greater than the fixed S value, it is a leaf pixel, and the corresponding pixel value is 1; otherwise, the pixel is a background pixel, and the corresponding pixel value is 0.

[0048] (2) Convert the RGB image to HSV color space and LAB color space respectively, and determine the R, G, B, H, S, V, L, A, and B component values ​​of each leaf pixel. R, G, B, H, S, V, L, A, and B are different color spaces. That is, the R, G, B, H, S, V, L, A, and B component values ​​of the connected area of ​​the leaf are extracted based on the binary image; then, the average value of each component is calculated based on the R, G, B, H, S, V, L, A, and B component values ​​of all leaf pixels to obtain the color features of the RGB image. The color features include the average values ​​of the R, G, B, H, S, V, L, A, and B components.

[0049] (3) Calculate the morphological features of the RGB image based on the number of leaf pixels; the morphological features include leaf area, perimeter, width, height, MajorAxisLength, MinorAxisLength, Eccentricity, and EquivDiameter;

[0050] More specifically, a leaf image can be viewed as a collection of leaf pixels. The pixels of the entire leaf are interconnected, and the leaf region in a binary image is a connected set. The leaf area is obtained by analyzing the number of pixels in the connected region and calculating the ratio of pixels in the leaf connected region to the total pixels. Multiplying this ratio by the image field of view yields the leaf area. The perimeter is the number of pixels at the junction of the background and the leaf connected region, i.e., the outermost circle of the leaf connected region. The width is the number of pixels along the X-axis of the smallest rectangle containing the leaf connected region (i.e., the smallest circumscribed rectangle of the leaf connected region). Correspondingly, the height is the number of pixels along the Y-axis of the smallest rectangle containing the leaf connected region. MajorAxisLength is the length of the major axis of an ellipse with the same standard second-order central moment as the leaf connected region, i.e., the number of pixels along the major axis of the ellipse. MinorAxisLength is the length of the minor axis of an ellipse with the same standard second-order central moment as the leaf connected region, i.e., the number of pixels along the minor axis of the ellipse. Eccentricity is the eccentricity of an ellipse with the same standard second-order central moment as the leaf-connected region. It is the ratio of the distance between the ellipse's foci to its major axis. EquivDiameter is the diameter of a circle with the same area, or number of pixels, as the leaf-connected region. These metrics can be calculated using the Regionprops function in Matlab.

[0051] (4) Calculating the gray level co-occurrence matrix of the binary image, and calculating the texture features of the RGB image based on the gray level co-occurrence matrix; the texture features include maximum probability, correlation, contrast, energy, homogeneity and entropy.

[0052] The co-occurrence matrix is ​​defined by the joint probability density of pixels at two locations. It reflects not only the distribution of brightness but also the positional distribution of pixels with the same or similar brightness. It is a second-order statistical characteristic of image brightness variations and the basis for defining a set of texture features. The gray-level co-occurrence matrix of an image is a matrix function of pixel distance and angle. It calculates the correlation between the gray-level values ​​of two points at a certain distance and orientation in the image to reflect comprehensive information about the direction, spacing, amplitude, and speed of change. Maximum probability refers to the probability of the largest pixel pair. Correlation measures the similarity of the elements of the spatial gray-level co-occurrence matrix in the row or column direction. Contrast reflects image clarity and the depth of texture grooves. Energy is the sum of the squares of the gray-level co-occurrence matrix elements and reflects the uniformity of the image's gray-level distribution and the coarseness of the texture. Homogeneity (also known as inverse disparity) reflects the homogeneity of the image texture and measures the extent of local variation in the image texture. Entropy is a measure of the amount of information contained in the image, indicating the degree of texture non-uniformity or complexity.

[0053] In S2, extracting the first spectral feature and the second spectral feature of the spectral data may include:

[0054] (1) Perform pixel-level correspondence between the RGB image and the spectral data to obtain the spectral data of each pixel in the RGB image;

[0055] (2) Threshold segmentation is performed on the RGB image to obtain a binary image, which includes leaf pixels and background pixels;

[0056] (3) The average spectrum of the leaves is calculated based on the spectral data of all leaf pixels, and the average spectrum of the leaves corresponding to the first characteristic wavelength is selected as the first spectral feature of the spectral data, and the average spectrum of the leaves corresponding to the second characteristic wavelength is selected as the second spectral feature of the spectral data.

[0057] Calculate the average of the spectral data of all leaf pixels to obtain the average leaf spectrum. The characteristic wavelengths can be: 380-450, 1630-2000, 2280-2495nm.

[0058] It should be noted that the characteristic wavelength is determined during the training of the prediction model. When training the prediction model, a training set needs to be constructed, and RGB images and spectral data of multiple leaf samples are collected. Feature extraction of the RGB images is performed to obtain the color, texture, and morphological characteristics of each leaf sample. The spectral data is processed to calculate the leaf average spectrum of each leaf sample. The leaf average spectrum is then preprocessed using a standard normal variate change and a first-order derivative. The characteristic wavelength is selected using the competitive adaptive weight sampling method (CARS) characteristic wavelength selection method. Specifically, the CARS characteristic wavelength selection method substitutes the leaf average spectrum of each leaf sample and the actual chlorophyll content value into the calculation to calculate the first characteristic wavelength. The CARS characteristic wavelength selection method substitutes the leaf average spectrum of each leaf sample and the actual nitrogen content value into the calculation to calculate the second characteristic wavelength. The leaf average spectrum corresponding to the first characteristic wavelength in the leaf sample is selected to form the first spectral feature of each leaf sample, and the leaf average spectrum corresponding to the second characteristic wavelength is selected to form the second spectral feature of each leaf sample. The color, texture, morphology, and first spectral features were fused to obtain the first fused feature of each leaf sample. The first fused feature of each leaf sample was used as the sample input data, and the actual chlorophyll content of each leaf sample was used as the label data of the sample to form the first training data set. The initial model was trained to obtain the first prediction model. The color, texture, morphology, and second spectral features were fused to obtain the second fused feature of each leaf sample. The second fused feature of each leaf sample was used as the sample input data, and the actual nitrogen content of each leaf sample was used as the label data of the sample to form the second training data set. The initial model was trained to obtain the second prediction model.

[0059] S3: fusing the color feature, the morphological feature, the texture feature, and the first spectral feature to obtain a first fused feature; fusing the color feature, the morphological feature, the texture feature, and the second spectral feature to obtain a second fused feature;

[0060] S4: Using the first fusion feature as input, the chlorophyll content of the cotton leaf to be tested is calculated using a first prediction model; using the second fusion feature as input, the nitrogen content of the cotton leaf to be tested is calculated using a second prediction model.

[0061] The first prediction model and the second prediction model of this embodiment both use a deep convolutional neural network model for calculating chlorophyll and nitrogen content, such as Figure 4As shown in the figure, the deep convolutional neural network model includes an input layer, a feature extraction layer, a fully connected network layer, and an output layer connected in sequence. The feature extraction layer includes multiple convolution blocks connected in sequence, the number of which can be three. Each convolution block includes a convolution layer, a ReLU activation function, and a maximum pooling layer. The convolution kernel size of the convolution layer is 3*3 with a stride of 1, and the size of the maximum pooling layer is 3*3 with a stride of 2. The fully connected network layer includes multiple fully connected layers connected in sequence, the number of which can be two, with 128 and 32 neurons respectively.

[0062] The deep convolutional neural network model is trained using a first training data set to learn and establish a mapping relationship between the first fusion feature and the chlorophyll content, thereby constructing a first prediction model. The deep convolutional neural network model is trained using a second training data set to learn and establish a mapping relationship between the second fusion feature and the nitrogen content, thereby constructing a second prediction model. The first prediction model and the second prediction model are integrated into the control chip 3. The first fusion feature obtained by the S3 process is used as the model input of the first prediction model integrated into the control chip 3, and the second fusion feature obtained by the S3 process is used as the model input of the second prediction model integrated into the control chip 3. The chlorophyll and nitrogen contents can be calculated and output according to the first prediction model and the second prediction model solidified in the control chip 3.

[0063] The control chip 3 of this embodiment combines the acquired spectral data with the RGB image at the pixel level, completing the correspondence between image information and spectral information, synthesizing a multi-channel spectral dot matrix image. The spectral image processing algorithm integrated within the control chip 3 then processes the data to calculate leaf area phenotypic indicators (including morphological, color, texture, and spectral characteristics). Based on the prediction model integrated within the control chip 3, the fusion characteristics of the spectrum and image are used as input to predict chlorophyll content and nitrogen content. After the calculation is completed, the control chip 3 packages and stores the calculated data for a single measurement, making the calculated data accessible to external devices via the USB output interface 14.

[0064] Each embodiment in this specification focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referenced to each other.

[0065] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. An instrument for detecting chlorophyll and nitrogen content in cotton leaves, characterized in that: The detection instrument includes: a light-shielding housing, a control chip, and a reflectivity plate, a light source, an RGB camera, and a spectrum probe located inside the light-shielding housing; the reflectivity plate is located on the bottom surface of the light-shielding housing; the control chip is respectively connected to the RGB camera and the spectrum probe for communication; The cotton leaf to be tested is located on the reflectivity plate; the light source is used to illuminate the cotton leaf to be tested; the RGB camera is used to obtain an RGB image of the cotton leaf to be tested illuminated by the light source; and the spectrum probe is used to obtain spectral data of the cotton leaf to be tested illuminated by the light source; The control chip is used to calculate the chlorophyll content and nitrogen content of the cotton leaves to be tested based on the RGB image and the spectral data; Extracting color features, morphological features, and texture features of the RGB image, and extracting first spectral features and second spectral features of the spectral data; fusing the color features, the morphological features, the texture features, and the first spectral features to obtain a first fused feature; fusing the color features, the morphological features, the texture features, and the second spectral features to obtain a second fused feature; using the first fused feature as input, calculating the chlorophyll content of the cotton leaf to be tested using a first prediction model; using the second fused feature as input, calculating the nitrogen content of the cotton leaf to be tested using a second prediction model; Extracting the color features, morphological features and texture features of the RGB image specifically includes: performing threshold segmentation on the RGB image to obtain a binary image; the binary image includes leaf pixels and background pixels; converting the RGB image into HSV color space and LAB color space respectively, determining the R, G, B, H, S, V, L, A, B component values ​​of each leaf pixel; calculating the average value of each component based on the R, G, B, H, S, V, L, A, B component values ​​of all the leaf pixels, and obtaining the color features of the RGB image; the color features include R , G, B, H, S, V, L, A, B components; calculating the morphological features of the RGB image based on the number of leaf pixels; the morphological features include the area, perimeter, width, height, MajorAxisLength, MinorAxisLength, Eccentricity and EquivDiameter of the leaf; calculating the gray level co-occurrence matrix of the binary image, and calculating the texture features of the RGB image based on the gray level co-occurrence matrix; the texture features include maximum probability, correlation, contrast, energy, homogeneity and entropy; Extracting the first spectral feature and the second spectral feature of the spectral data specifically includes: performing pixel-level correspondence between the RGB image and the spectral data to obtain spectral data of each pixel in the RGB image; performing threshold segmentation on the RGB image to obtain a binary image; the binary image includes leaf pixels and background pixels; calculating the average spectrum of the leaves based on the spectral data of all the leaf pixels, and selecting the average spectrum of the leaves corresponding to the first characteristic wavelength as the first spectral feature of the spectral data, and selecting the average spectrum of the leaves corresponding to the second characteristic wavelength as the second spectral feature of the spectral data.

2. The detection instrument according to claim 1, characterized in that: A bottom sample drawer is provided at the bottom inner side of the light-shielding shell; the bottom sample drawer is slidably connected to the light-shielding shell; the bottom surface of the bottom sample drawer is black, and the reflectivity plate is located on the bottom surface of the bottom sample drawer.

3. The detection instrument according to claim 1, characterized in that: The detection instrument also includes a moving component, which is located inside the light-shielding housing; the RGB camera and the spectrum probe are installed on the moving component; and the moving component is used to drive the RGB camera and the spectrum probe to move horizontally and rise and fall.

4. The detection instrument according to claim 3, characterized in that: The movable component includes a driving member and a control rod; the driving member is drivingly connected to the control rod; the control rod includes a horizontal axis and a vertical axis; the RGB camera and the spectral probe are installed on the vertical axis.

5. The detection instrument according to claim 1, characterized in that: The detection instrument further comprises a fan, which is located inside the light-shielding housing; the fan is used for heat dissipation.

6. The detection instrument according to claim 1, characterized in that: The detection instrument also includes a power supply; the power supply is used to supply power to the light source, the RGB camera and the spectral probe.

7. A method for detecting chlorophyll and nitrogen content in cotton leaves, comprising controlling the detection instrument according to any one of claims 1 to 6 to operate, characterized in that: The detection method comprises: Receive RGB images acquired by the RGB camera and spectral data acquired by the spectral probe; Extracting color features, morphological features, and texture features of the RGB image, and extracting first spectral features and second spectral features of the spectral data; fusing the color feature, the morphological feature, the texture feature, and the first spectral feature to obtain a first fused feature; fusing the color feature, the morphological feature, the texture feature, and the second spectral feature to obtain a second fused feature; The first fusion feature is used as input, and the chlorophyll content of the cotton leaf to be tested is calculated using a first prediction model; the second fusion feature is used as input, and the nitrogen content of the cotton leaf to be tested is calculated using a second prediction model; The extracting of the color features, morphological features and texture features of the RGB image specifically includes: Performing threshold segmentation on the RGB image to obtain a binary image; the binary image includes leaf pixels and background pixels; Convert the RGB image to HSV color space and LAB color space respectively, determine the R, G, B, H, S, V, L, A, and B component values ​​of each leaf pixel; calculate the average value of each component based on the R, G, B, H, S, V, L, A, and B component values ​​of all leaf pixels to obtain the color features of the RGB image; the color features include the average values ​​of the R, G, B, H, S, V, L, A, and B components; Calculating morphological features of the RGB image based on the number of leaf pixels; the morphological features include area, perimeter, width, height, MajorAxisLength, MinorAxisLength, Eccentricity, and EquivDiameter of the leaf; Calculating a gray level co-occurrence matrix of the binary image, and calculating texture features of the RGB image based on the gray level co-occurrence matrix; the texture features include maximum probability, correlation, contrast, energy, homogeneity, and entropy; The extracting of the first spectral feature and the second spectral feature of the spectral data specifically includes: Performing pixel-level correspondence between the RGB image and the spectral data to obtain spectral data of each pixel in the RGB image; Performing threshold segmentation on the RGB image to obtain a binary image; the binary image includes leaf pixels and background pixels; The leaf average spectrum is calculated based on the spectral data of all the leaf pixel points, and the leaf average spectrum corresponding to the first characteristic wavelength is selected as the first spectral feature of the spectral data, and the leaf average spectrum corresponding to the second characteristic wavelength is selected as the second spectral feature of the spectral data.

8. The detection method according to claim 7, characterized in that Both the first prediction model and the second prediction model adopt a deep convolutional neural network model; the deep convolutional neural network model includes an input layer, a feature extraction layer, a fully connected network layer and an output layer connected in sequence; The feature extraction layer includes multiple convolution blocks connected in sequence; each of the convolution blocks includes a convolution layer, a ReLU activation function and a maximum pooling layer; the fully connected network layer includes multiple fully connected layers connected in sequence.

Images