A method and system for measuring phenotypic parameters of rapeseed siliques

Abstract

The present invention relates to the technical field of visual image recognition, and discloses a method for measuring phenotypic parameters of rapeseed siliques, comprising the following steps: step 1, constructing a rapeseed silique phenotypic parameter measurement system to obtain rapeseed silique image information; step 2, constructing a rapeseed silique structure measurement model based on a UNet network and a rapeseed silique seed counting model based on a CrowdCounting‑P2PNet network; step 3, performing multi-scale fusion processing on the image information to obtain a multi-scale fused image; step 4, inputting the multi-scale fused image into the rapeseed silique structure measurement model and the rapeseed silique seed counting model; step 5, outputting the precise phenotypic parameters of the rapeseed siliques. The present invention can efficiently collect rapeseed silique images, effectively fuse rapeseed silique long-exposure and short-exposure images, comprehensively measure the phenotypic parameters of rapeseed siliques, and can construct a rapeseed silique phenotypic parameter detection model using the multi-scale fused image, and has the advantages of low cost, high speed, high precision, simple system, etc.

Description

Technical Field

[0001] The present invention relates to the technical field of visual image recognition, and in particular to a method and system for measuring phenotypic parameters of rapeseed siliques. Background Art

[0002] As a multifunctional plant, rapeseed is not only a nutritious vegetable, but also an important source of vegetable oils and fats. It is a typical representative of oil crops in my country. The quality of rapeseed siliques has an important impact on yield. Measuring its phenotypic parameters can provide key data for studying genetic influences. Therefore, accurate detection of the phenotypic parameters of rapeseed siliques is of great practical significance. The phenotypic parameters of rapeseed siliques include the number of seeds in the valver, the length of the beak, valver and pedicel, and the geometric angles between the key structures. Traditional manual measurement methods have problems such as high cost, low efficiency, easy fatigue, and large errors.

[0003] Chinese patent document, publication number CN108300801A, discloses a molecular marker closely related to rapeseed grain weight and silique length and its application, which uses cotton thread to measure the length of rapeseed siliques. Analyzing the distribution and shape of rapeseed siliques by computer software to draw paths, combined with the angle tool in the software to achieve automatic angle detection, has become the mainstream method in the industry. For example, the document (Chinese Journal of Oil Crops, 2022, 44(1):25-34) uses the straight line tool and angle tool in the computer Photoshop software to measure the angle between the main stem and branches of rapeseed in the photo; the document (Transactions of the Chinese Society of Agricultural Engineering, 2021, 37(23):153-160) proposes a method for testing the number of grains per angle based on rapeseed silique length image recognition.

[0004] However, there is no effective method for measuring the phenotypic parameters of rapeseed siliques. Therefore, it is still a challenging task to effectively measure the phenotypic parameters of rapeseed siliques, and it is urgent to propose a method for measuring the phenotypic parameters of rapeseed siliques that can realize automated and batch detection. Summary of the invention

[0005] In order to solve the technical problems raised in the background technology, the present invention provides a method and system for measuring phenotypic parameters of rapeseed siliques.

[0006] The present invention is implemented by the following technical scheme: a method for measuring phenotypic parameters of rapeseed siliques, comprising the following steps:

[0007] Step 1, constructing a rapeseed silique phenotypic parameter measurement system to obtain rapeseed silique image information;

[0008] Step 2: Construct a rapeseed silique structure measurement model based on the UNet network and a rapeseed silique seed counting model based on the CrowdCounting-P2PNet network;

[0009] Step 3: Perform multi-scale fusion processing on the image information to obtain a multi-scale fused image;

[0010] Step 4, inputting the multi-scale fusion image into the rapeseed silique structure measurement model and the rapeseed silique seed counting model;

[0011] Step 5: Output the precise phenotypic parameters of rapeseed siliques.

[0012] Preferably, in step 1, the rapeseed silique image information includes two groups of rapeseed silique images under long exposure and short exposure conditions.

[0013] Preferably, in step 2, after obtaining two groups of rapeseed silique images under long exposure and short exposure conditions, LabelMe software is used to annotate the multi-scale fusion images, specifically:

[0014] Accurately mark each rapeseed kernel in the rapeseed silique image and assign a label category;

[0015] The beak, valve, stalk and stem of rapeseed silique images are annotated with polygons and given different label categories.

[0016] Finally, an image dataset of rapeseed siliques was constructed;

[0017] Based on the multi-scale fusion image, the image information collected under long exposure and short exposure conditions is integrated, which effectively balances the clarity of the rapeseed silique outline and the visibility of the internal grains.

[0018] Preferably, in step 3, the multi-scale fusion image is an RGB three-channel image.

[0019] Preferably, in step 2, the rapeseed silique image dataset is divided into a training set and a test set according to a certain ratio;

[0020] Among them: the training set is used for the training and optimization of the rapeseed silique structure measurement model based on the UNet network and the rapeseed silique seed counting model based on the CrowdCounting-P2PNet network;

[0021] The test set is used to evaluate the performance and generalization ability of the constructed model.

[0022] Preferably, in the UNet network model, regarding length measurement, the steps are as follows:

[0023] After the rapeseed silique image is segmented, the boundary contour of the segmentation mask needs to be extracted, and the center line of the rapeseed silique needs to be generated through the skeletonization algorithm. The checkerboard calibration plate is placed on a transparent acrylic plate, and a complete image with the checkerboard calibration plate as the background is captured. Based on the captured checkerboard calibration plate image, the sub-pixel corner point detection algorithm is used to extract the corner points of the checkerboard, and the pixel distance between adjacent corner points is calculated. Combined with the physical size of the checkerboard, the proportional coefficient between the pixel and the actual physical length is determined;

[0024] By traversing the pixel points on the center line and calculating their spacing, the total pixel length of the center line is determined. Using the proportional coefficient, the pixel length of the center line is converted into the actual physical length, thereby accurately measuring the length of the fruit beak, the length of the fruit petal, and the length of the fruit stalk.

[0025] Preferably, in the UNet network model, regarding angle measurement, the steps are as follows:

[0026] The least squares method was used to fit the generated center lines to calculate the slope differences of the fitted center lines, and the inverse tangent function was used to calculate the angles between these fitted center lines to determine the valve-stalk angle, the stalk-branch angle, and the valve-branch angle.

[0027] Preferably, in a CrowdCounting-P2PNet network model;

[0028] When counting the seeds of rapeseed siliques, in order to ensure the accuracy of the seed counting, only the seeds in the masked area of ​​the fruit valve segmented by the UNet network were counted, that is, only the number of seeds in the fruit valve was counted, and the counting results of the seeds in the rest of the area were considered invalid.

[0029] The present invention also proposes a rapeseed silique phenotypic parameter measurement system, which is used in the above-mentioned measurement method, and comprises a fixed bracket, a flat light source, a transparent acrylic plate, a checkerboard calibration plate and a color camera;

[0030] The flat light source is installed at the bottom of the fixed bracket to provide uniform background light upward to enhance the visibility of the seeds inside the rapeseed silique;

[0031] The transparent acrylic plate is installed in the middle of the fixed bracket through a clamping mechanism to place the rapeseed siliques to be tested;

[0032] The color camera is installed on the top of the fixed bracket and is used to capture the backlit image of the rapeseed silique to be tested;

[0033] Among them: after adjusting the position of the transparent acrylic plate according to the exposure time and fixing it, the checkerboard calibration plate is placed on the transparent acrylic plate, and its complete image is taken to obtain the scale information of the image and ensure that the pixel coordinates in the image can be accurately converted into actual physical sizes; then the checkerboard calibration plate is removed and the rapeseed silique is placed for data collection.

[0034] Preferably, the clamping mechanism includes a slider and two clamping parts respectively fixed on opposite sides of the slider, the slider is provided with a through hole for the fixed bracket to move through, and the slider is also spirally mounted with a knob, which is pressed against the outer side of the fixed bracket to fix the position of its clamping mechanism, and the transparent acrylic plate is installed between the two clamping parts.

[0035] In order to obtain the best imaging effect, the position of the transparent acrylic plate needs to be adjusted appropriately according to the exposure time of the color camera. After determining the optimal position, it should be fixed and the checkerboard calibration plate image should be taken. The color camera collects images under long exposure conditions to enhance the visibility of the internal kernels and images under short exposure conditions to retain the clear outline of the rapeseed silique. The two sets of images are effectively fused through the multi-scale exposure fusion algorithm to maximize the retention of key information in the image. In the fusion process, through multi-scale analysis and feature extraction, and optimizing the weight distribution of each scale, the images under different exposure conditions are weighted to achieve a balance between details and outlines, ensuring that the fused image retains the details of the silique outline while enhancing the visibility of the internal kernels, and the fused image has the same resolution as the original image.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention can efficiently collect rapeseed silique images, effectively fuse rapeseed silique long-exposure and short-exposure images, comprehensively measure rapeseed silique phenotypic parameters, and construct a rapeseed silique phenotypic parameter detection model using multi-scale fused images. The invention has the advantages of low cost, high speed, high precision, and simple system. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of the measurement method proposed by the present invention;

[0039] Figure 2 It is a schematic diagram of the overall structure of the measurement system proposed by the present invention;

[0040] Figure 3 This is a schematic diagram of the multi-scale image fusion principle proposed by the present invention;

[0041] Figure 4 This is a schematic diagram of the principle of point marking and polygon marking of rapeseed siliques proposed by the present invention;

[0042] Figure 5 This is a schematic diagram of the principle of angle marking of rapeseed siliques proposed by the present invention;

[0043] Figure 6 A schematic diagram of the checkerboard calibration plate proposed by the present invention;

[0044] Figure 7 This is a schematic structural diagram of the clamping mechanism proposed in the present invention.

[0045] Description of main symbols:

[0046] 1. Color camera; 2. Clamping mechanism; 3. Fixed bracket; 4. Transparent acrylic plate; 5. Flat light source; 7. Knob; 8. Slider; 9. Clamping part. DETAILED DESCRIPTION

[0047] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form a new embodiment.

[0048] Example 1

[0049] Reference Figure 1 ,This scheme proposes a method for measuring phenotypic parameters of rapeseed siliques, including

[0050] Step 1, constructing a rapeseed silique phenotypic parameter measurement system to obtain rapeseed silique image information;

[0051] Step 2: Construct a rapeseed silique structure measurement model based on the UNet network and a rapeseed silique seed counting model based on the CrowdCounting-P2PNet network;

[0052] Step 3: Perform multi-scale fusion processing on the image information to obtain a multi-scale fused image;

[0053] Step 4, inputting the multi-scale fusion image into the rapeseed silique structure measurement model and the rapeseed silique seed counting model;

[0054] Step 5: Output the precise phenotypic parameters of rapeseed siliques.

[0055] As an optional implementation manner of the present invention, the rapeseed silique image information includes two groups of rapeseed silique images under long exposure and short exposure conditions.

[0056] As an optional implementation of the present invention, in step 2, after obtaining two groups of rapeseed silique images under long exposure and short exposure conditions, LabelMe software is used to annotate the multi-scale fusion image, specifically:

[0057] Accurately mark each rapeseed kernel in the rapeseed silique image and assign a label category;

[0058] The beak, valve, stalk and stem of rapeseed silique images are annotated with polygons and given different label categories.

[0059] Finally, an image dataset of rapeseed siliques was constructed.

[0060] Reference Figure 3 Based on the multi-scale fusion image, the image information collected under long exposure and short exposure conditions is integrated, which effectively balances the clarity of the rapeseed silique outline and the visibility of the internal grains.

[0061] As an optional implementation manner of the present invention, in step 3, the multi-scale fusion image is an RGB three-channel image.

[0062] As an optional implementation mode of the present invention, in step 2, the rapeseed silique image dataset is divided into a training set and a test set according to a certain ratio.

[0063] Among them: the training set is used for the training and optimization of the rapeseed silique structure measurement model based on the UNet network and the rapeseed silique seed counting model based on the CrowdCounting-P2PNet network;

[0064] The test set is used to evaluate the performance and generalization ability of the constructed model.

[0065] Optionally, in the UNet network model, regarding length measurement, the steps are as follows:

[0066] Reference Figure 4 After the rapeseed silique image is segmented, the boundary contour of the segmentation mask needs to be extracted, and the center line of the rapeseed silique needs to be generated through the skeletonization algorithm. The checkerboard calibration plate is placed on a transparent acrylic plate, and a complete image with the checkerboard calibration plate as the background is captured. Based on the captured checkerboard calibration plate image, the sub-pixel corner point detection algorithm is used to extract the corner points of the checkerboard, and the pixel distance between adjacent corner points is calculated. Combined with the physical size of the checkerboard, the proportional coefficient between the pixel and the actual physical length is determined.

[0067] By traversing the pixel points on the center line and calculating their spacing, the total pixel length of the center line is determined. Using the proportional coefficient, the pixel length of the center line is converted into the actual physical length, thereby accurately measuring the length of the fruit beak, the length of the fruit petal, and the length of the fruit stalk.

[0068] Reference Figure 5 As an optional implementation of the present invention, in the UNet network model, regarding angle measurement, the steps are as follows:

[0069] The least squares method was used to fit the generated center lines to calculate the slope differences of the fitted center lines, and the inverse tangent function was used to calculate the angles between these fitted center lines to determine the valve-stalk angle, the stalk-branch angle, and the valve-branch angle.

[0070] As an optional implementation of the present invention, in the CrowdCounting-P2PNet network model;

[0071] When counting the seeds of rapeseed siliques, in order to ensure the accuracy of the seed counting, only the seeds in the masked area of ​​the fruit valve segmented by the UNet network were counted, that is, only the number of seeds in the fruit valve was counted, and the counting results of the seeds in the rest of the area were considered invalid.

[0072] Example 2

[0073] Reference Figure 2 , Figure 7 The present invention also proposes a rapeseed silique phenotypic parameter measurement system, comprising a fixed support 3, a flat light source 5, a transparent acrylic plate 4, a checkerboard calibration plate and a color camera 1;

[0074] The flat light source 5 is installed at the bottom of the fixed bracket 3 to provide uniform background light upward to enhance the visibility of the seeds inside the rapeseed silique;

[0075] The transparent acrylic plate is installed in the middle of the fixed bracket through the clamping mechanism 2 to place the rapeseed siliques to be tested;

[0076] The color camera 1 is installed on the top of the fixed bracket 3 and is used to capture the backlit image of the rapeseed silique to be tested;

[0077] Among them: after adjusting the position of the transparent acrylic plate according to the exposure time and fixing it, the checkerboard calibration plate is placed on the transparent acrylic plate 4, and its complete image is taken to obtain the scale information of the image to ensure that the pixel coordinates in the image can be accurately converted into actual physical sizes; then the checkerboard calibration plate is removed and the rapeseed silique is placed for data collection.

[0078] Reference Figure 7 In this solution, the clamping mechanism 2 includes a slider 8 and two clamping parts 9 respectively fixed on opposite sides of the slider 8, wherein the slider begins to have a through hole for the fixed bracket to move through, and the slider 8 is also spirally mounted with a knob 7, which is pressed against the outer side of the fixed bracket to fix the position of the clamping mechanism 2, and the transparent acrylic plate 4 is installed between the two clamping parts 9.

[0079] In order to obtain the best imaging effect, the position of the transparent acrylic plate needs to be adjusted appropriately according to the exposure time of the color camera. After determining the optimal position, it should be fixed and the checkerboard calibration plate image should be taken. The color camera collects images under long exposure conditions to enhance the visibility of the internal kernels and images under short exposure conditions to retain the clear outline of the rapeseed silique. The two sets of images are effectively fused through the multi-scale exposure fusion algorithm to maximize the retention of key information in the image. In the fusion process, through multi-scale analysis and feature extraction, and optimizing the weight distribution of each scale, the images under different exposure conditions are weighted to achieve a balance between details and outlines, ensuring that the fused image retains the details of the silique outline while enhancing the visibility of the internal kernels, and the fused image has the same resolution as the original image.

[0080] The above-mentioned embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and substitutions made by technicians in this field on the basis of the present invention shall fall within the scope of protection required by the present invention.

Claims

1. A method for measuring phenotypic parameters of rapeseed siliques, characterized in that: include Step 1, constructing a rapeseed silique phenotypic parameter measurement system to obtain rapeseed silique image information; Step 2: Construct a rapeseed silique structure measurement model based on the UNet network and a rapeseed silique seed counting model based on the CrowdCounting-P2PNet network; Step 3: Perform multi-scale fusion processing on the image information to obtain a multi-scale fused image; Step 4, inputting the multi-scale fusion image into the rapeseed silique structure measurement model and the rapeseed silique seed counting model; Step 5: Output the precise phenotypic parameters of rapeseed siliques.

2. A method for measuring phenotypic parameters of rapeseed siliques according to claim 1, characterized in that: In the step 1, the rapeseed silique image information includes two groups of rapeseed silique images under long exposure and short exposure conditions.

3. A method for measuring phenotypic parameters of rapeseed siliques according to claim 2, characterized in that: In step 2, after obtaining two groups of rapeseed silique images under long exposure and short exposure conditions, LabelMe software is used to annotate the multi-scale fusion images, specifically: Accurately mark each rapeseed kernel in the rapeseed silique image and assign a label category; The beak, valve, stalk and stem of rapeseed silique images are annotated with polygons and given different label categories. Finally, an image dataset of rapeseed siliques was constructed.

4. A method for measuring phenotypic parameters of rapeseed siliques according to claim 3, characterized in that: In step 3, the multi-scale fusion image is an RGB three-channel image.

5. The method for measuring phenotypic parameters of rapeseed siliques according to claim 1, characterized in that: In step 2, the rapeseed silique image data set is divided into a training set and a test set according to a certain ratio; Among them: the training set is used for the training and optimization of the rapeseed silique structure measurement model based on the UNet network and the rapeseed silique seed counting model based on the CrowdCounting-P2PNet network; The test set is used to evaluate the performance and generalization ability of the constructed model.

6. A method for measuring phenotypic parameters of rapeseed siliques according to claim 5, characterized in that: In the UNet network model, the steps for length measurement are as follows: After the rapeseed silique image is segmented, the boundary contour of the segmentation mask needs to be extracted, and the center line of the rapeseed silique needs to be generated through the skeletonization algorithm. The checkerboard calibration plate is placed on a transparent acrylic plate, and a complete image with the checkerboard calibration plate as the background is captured. Based on the captured checkerboard calibration plate image, the sub-pixel corner point detection algorithm is used to extract the corner points of the checkerboard, and the pixel distance between adjacent corner points is calculated. Combined with the physical size of the checkerboard, the proportional coefficient between the pixel and the actual physical length is determined; By traversing the pixel points on the center line and calculating their spacing, the total pixel length of the center line is determined. Using the proportional coefficient, the pixel length of the center line is converted into the actual physical length, thereby accurately measuring the length of the fruit beak, the length of the fruit petal, and the length of the fruit stalk.

7. A method for measuring phenotypic parameters of rapeseed siliques according to claim 6, characterized in that: In the UNet network model, the steps for angle measurement are as follows: The least squares method was used to fit the generated center lines to calculate the slope differences of the fitted center lines, and the inverse tangent function was used to calculate the angles between these fitted center lines to determine the valve-stalk angle, the stalk-branch angle, and the valve-branch angle.

8. The method for measuring phenotypic parameters of rapeseed siliques according to claim 5, characterized in that: In the CrowdCounting-P2PNet network model; When counting the seeds of rapeseed siliques, only the seeds within the masked area of ​​the fruit valve segmented by the UNet network are counted, that is, only the number of seeds in the fruit valve is counted, and the counting results of the seeds in the rest of the area are considered invalid.

9. A rapeseed silique phenotypic parameter measurement system, which adopts the measurement method according to any one of claims 1 to 8, characterized in that: It includes a fixing bracket, a flat light source, a transparent acrylic plate, a checkerboard calibration plate and a color camera; The flat light source is installed at the bottom of the fixed bracket to provide uniform background light upwards; The transparent acrylic plate is installed in the middle of the fixed bracket through a clamping mechanism to place the rapeseed siliques to be tested; The color camera is installed on the top of the fixed bracket and is used to capture the backlit image of the rapeseed silique to be tested; Among them: after adjusting the position of the transparent acrylic plate according to the exposure time and fixing it, the checkerboard calibration plate is placed on the transparent acrylic plate, and its complete image is taken to obtain the scale information of the image; then the checkerboard calibration plate is removed, and the rapeseed silique is placed for data collection.

10. The rapeseed silique phenotypic parameter measurement system according to claim 9, characterized in that: The clamping mechanism includes a slider and two clamping parts respectively fixed on opposite sides of the slider. The slider is provided with a through hole for the fixed bracket to move through. The slider is also spirally mounted with a knob, which is pressed against the outer side of the fixed bracket to fix the position of the clamping mechanism. The transparent acrylic plate is installed between the two clamping parts.

Images