Method and system for evaluating damage level of grassland spodoptera frugiperda based on machine vision

Abstract

The invention discloses a method and system for evaluating the damage level of grassland spodoptera frugiperda based on machine vision, and the method comprises the following steps: obtaining and preprocessing a corn leaf image, and obtaining a preprocessed image; segmenting adhered wormholes in the preprocessed image and removing small-area noise to obtain a wormhole image; extracting wormhole features in the wormhole image, wherein the wormhole features comprise diameters, areas and number features of wormholes; according to the wormhole features, carrying out pest level labeling on the pre-processed image to obtain a labeled image, and training a support vector machine through the labeled image to obtain a recognition model; and identifying the to-be-identified corn leaves through the identification model to obtain the damage level of the grassland spodoptera exigua. Through a machine vision method, each feature of the spodoptera exigua hole can be accurately extracted, then the damage level is judged through the features, and finally the accurate grassland spodoptera exigua damage level is obtained.

Description

Technical Field

[0001] The present invention belongs to the technical field of damage level judgment, and in particular relates to a method and system for evaluating the damage level of fall armyworm based on machine vision. Background Art

[0002] In the early stage of corn damage, the larvae of the fall armyworm mainly feed on the leaves of the seedlings, causing pinhole-shaped insect holes or transparent window holes. As the larvae grow and develop, the leaves are fed to form long or larger irregular insect holes. The evaluation of the resistance of the fall armyworm is usually carried out by investigating the damage of single leaves and the leaves of the whole corn plant, mainly the size, length and number of holes. According to the damage, the average damage level of the leaves of the variety is calculated to determine its resistance level in the heart leaf stage. Through measurement, statistics and calculation, it is divided into different levels such as high resistance, resistance, medium resistance, sensitivity and high sensitivity according to certain levels. This is crucial for evaluating the characteristic values ​​of the variety. At present, the method of field detection of the degree of damage caused by the fall armyworm is mainly based on the detection personnel using calipers to measure the number and size of insect holes on each corn leaf affected by the insect. Because the leaves are fed by larvae to form insect holes of different shapes, this method has extremely high requirements for detection personnel, is time-consuming and labor-intensive, and has low efficiency and accuracy. Moreover, this method also requires the use of Excel software to perform data statistics and calculate the average damage level. Moreover, the statistical data results are easily affected by subjective factors, with large errors, which will cause deviations from the actual resistance situation. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a method and system for evaluating the damage level of fall armyworm based on machine vision to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above object, the present invention provides a method for evaluating the damage level of fall armyworm based on machine vision, comprising the following steps:

[0005] Acquire a corn leaf image and perform preprocessing to obtain a preprocessed image;

[0006] Segmenting the contiguous wormholes in the preprocessed image and removing small area noise to obtain a wormhole image;

[0007] Extracting wormhole features in the wormhole image, wherein the wormhole features include diameter, area and quantity features of the wormholes;

[0008] Annotating the preprocessed image for pest level according to the wormhole features to obtain an annotated image, and training a support vector machine with the annotated image to obtain a recognition model;

[0009] The corn leaves to be identified are identified through the recognition model to obtain the damage level of the fall armyworm.

[0010] Preferably, preprocessing the corn leaf image includes:

[0011] The corn leaf image is grayed, and Gaussian filtering is used to remove image noise. The image after noise removal is binarized to obtain a binarized image. A morphological opening operation is performed on the binarized image to obtain a preprocessed image.

[0012] Preferably, obtaining the wormhole image comprises:

[0013] The binary image is subjected to distance transformation to obtain a distance map of the wormhole area. The distance map is segmented into single wormholes by a watershed algorithm. If the area of ​​the segmented wormhole is smaller than a judgment threshold, it is judged as small area noise. The small area noise is removed to obtain a wormhole image.

[0014] Preferably, the extracting wormhole features from the wormhole image comprises:

[0015] The edge contour of a single wormhole is extracted, the length and width of the minimum circumscribed rectangle of the wormhole are calculated according to the edge contour, and the average value of the length and width of the minimum circumscribed rectangle is used as the diameter of the wormhole; the area and number of the wormholes are calculated according to the edge contour.

[0016] Preferably, the pest level marking includes:

[0017] The damage level of the preprocessed image is calculated and labeled according to the average leaf-eating level algorithm and the pest level classification standard.

[0018] The present invention also proposes a system for evaluating the damage level of fall armyworm based on machine vision, comprising:

[0019] A preprocessing module is used to obtain a corn leaf image and perform preprocessing to obtain a preprocessed image;

[0020] A wormhole segmentation module is used to segment the contiguous wormholes in the preprocessed image and remove small area noise to obtain a wormhole image;

[0021] A feature extraction module is used to extract wormhole features in the wormhole image, wherein the wormhole features include diameter, area and quantity features of the wormholes;

[0022] A model building module is used to annotate the pre-processed image with the pest level according to the wormhole features to obtain an annotated image, and train a support vector machine with the annotated image to obtain a recognition model;

[0023] The recognition module is used to identify the corn leaves to be identified through the recognition model to obtain the damage level of the fall armyworm.

[0024] Preferably, the preprocessing module comprises:

[0025] A collection unit, used for collecting corn leaf images;

[0026] The preprocessing unit is used to grayscale the corn leaf image, remove image noise by Gaussian filtering, binarize the image after noise removal to obtain a binarized image, and perform morphological opening operation on the binarized image to obtain a preprocessed image.

[0027] Preferably, the wormhole segmentation module comprises:

[0028] A distance transformation unit is used to perform distance transformation on the binary image to obtain a distance map of the wormhole area;

[0029] A segmentation unit, used for segmenting the connected wormholes into individual wormholes by using a watershed algorithm on the distance map;

[0030] The denoising unit is used to determine whether the segmented wormhole area is small area noise, and remove it if it is small area noise to obtain a wormhole image.

[0031] Preferably, the feature extraction module comprises:

[0032] A contour extraction unit, used to extract the edge contour of a single wormhole;

[0033] A calculation unit is used to calculate the diameter, area and number of wormholes according to the edge contour; the diameter is calculated by obtaining the length and width of the minimum circumscribed rectangle of the wormhole according to the edge contour, and taking the average value of the length and width of the minimum circumscribed rectangle as the diameter of the wormhole.

[0034] The present invention also proposes a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] The present invention discloses a method and system for evaluating the damage level of fall armyworm based on machine vision, comprising the following steps: obtaining a corn leaf image and preprocessing it to obtain a preprocessed image; segmenting the sticky wormholes in the preprocessed image and removing small area noise to obtain a wormhole image; extracting wormhole features in the wormhole image, the wormhole features including the diameter, area and quantity features of the wormholes; annotating the preprocessed image with the wormhole features to obtain an annotated image, training a support vector machine through the annotated image to obtain a recognition model; identifying the corn leaf to be identified through the recognition model to obtain the damage level of fall armyworm. The present application can accurately extract the various features of the wormholes through the machine vision method, and finally obtain the accurate damage level of fall armyworm by judging the damage level through the features. The present invention realizes the recognition of the damage level of fall armyworm through standardized image acquisition and image processing methods, and improves the efficiency and accuracy of pest identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0040] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0041] Embodiment 1

[0042] like Figure 1 As shown, this embodiment provides a method for evaluating the damage level of fall armyworm based on machine vision, including the following steps:

[0043] Acquire a corn leaf image and perform preprocessing to obtain a preprocessed image;

[0044] Segment the connected wormholes in the preprocessed image and remove small area noise to obtain the wormhole image;

[0045] Extracting wormhole features from wormhole images, the wormhole features include diameter, area and quantity features of the wormholes;

[0046] The pre-processed image is labeled with the pest level according to the worm hole characteristics to obtain a labeled image, and a support vector machine is trained through the labeled image to obtain a recognition model;

[0047] The corn leaves to be identified are identified through the recognition model to obtain the damage level of the fall armyworm.

[0048] Further, obtaining the corn leaf image includes:

[0049] A device with a uniform light source and a fixed shooting angle was built to stabilize the lighting conditions and shooting angle, eliminating the impact of field ambient lighting and corn leaf posture on image acquisition quality. Black light-absorbing material was used as the background inside the device to reduce background interference.

[0050] The collected corn leaves were placed in a standardized image acquisition device to obtain corn leaf images with stable lighting, fixed angles, and no background interference. By fixing the angle and position of the corn leaves in a standardized system and using preset lighting conditions, the stability of the lighting during image acquisition was ensured, thereby obtaining preliminary image data without background interference.

[0051] In this embodiment, a box with a black light-absorbing material coated inside is built, and a leaf holder with adjustable height and angle and an LED light source array are provided inside the box. The leaf holder can ensure that the leaf position and angle are consistent each time the image is taken, the black light-absorbing material can reduce the interference of background stray light, and the LED light source array can provide stable and uniform lighting, for example, a color temperature of 6500K and an illumination of 10000 lux. In this way, a corn leaf image with a simple background and uniform lighting can be obtained, for example, an RGB image with a shooting resolution of 4096*3072 pixels.

[0052] Further, preprocessing the corn leaf image includes:

[0053] The corn leaf image is grayed and Gaussian filtering is used to remove image noise. The image after noise removal is binarized to obtain a binarized image. The binarized image is morphologically opened to obtain a preprocessed image.

[0054] Specifically: grayscale the original color image. The weighted average method is used to calculate the gray value of each pixel. The formula is as follows: Gray = 299R + 587G + 114B, where R, G, and B represent the values ​​of the red, green, and blue channels of each pixel in the original color image, respectively, and Gray represents the calculated gray value. A grayscale image is obtained. Gaussian filtering is performed on the grayscale image. The grayscale image is denoised using a Gaussian filter. The formula of the Gaussian filter is as follows: G(x, y) = (1 / (2pisigma^2))exp(-(x^2+y^2) / (2sigma^2)), where x and y represent pixel coordinates, and sigma represents the standard deviation of the Gaussian distribution. The Gaussian filter is applied to the grayscale image through a convolution operation to obtain a smoothed grayscale image. Image enhancement is performed. According to the histogram distribution of the smoothed grayscale image, the histogram equalization method is used to enhance the image contrast, highlight the leaf features, and obtain an enhanced grayscale image. Binarize the enhanced grayscale image. Use the Otsu algorithm to automatically calculate the binarization threshold. If the pixel grayscale value is greater than the threshold, the pixel value is set to 255, otherwise it is set to 0. Get the binary image. Perform morphological processing on the binary image. Use morphological opening operation to remove small noise and burrs in the binary image. The opening operation is a combination of corrosion and expansion, which can eliminate small objects and smooth the outline of the object without changing its overall size. Get the processed binary image.

[0055] Further, obtaining the wormhole image includes:

[0056] The binary image is transformed to obtain the distance map of the wormhole area. The watershed algorithm is used to segment the connected wormholes into single wormholes. If the area of ​​the segmented wormhole is smaller than the judgment threshold, it is judged as small area noise. The small area noise is removed to obtain the wormhole image.

[0057] Specifically, the distance transform calculates the distance from each pixel in the image to the nearest background pixel. Through the distance transform, a binary image can be converted into a grayscale image, where the grayscale value represents the distance from the pixel to the background. For example, if the coordinates of a pixel are (x, y), its value is 1 (indicating a wormhole), and the coordinates of the nearest background pixel are (x-2, y-1), then the value of the pixel in the distance map is The distance map can help find the center area of ​​wormholes and distinguish the boundaries between different wormholes. The watershed algorithm is an image segmentation method based on the concept of topography. It regards the image as a topographic map, where the grayscale value represents the height of the terrain. In the distance map, each local maximum point is considered to be the center of a catchment basin. The watershed algorithm simulates the process of water flowing from high to low, and finally divides the image into multiple catchments, each of which represents a potential wormhole area. For example, if the distance between two wormholes is close, they may form a connected area in the distance map, containing two local maximum points. The watershed algorithm can separate the two wormholes to avoid misidentifying them as a large wormhole. After the watershed algorithm is segmented, the area of ​​each segmented area needs to be calculated to distinguish between real wormholes and noise. For example, suppose one segmented area contains 10 pixels and the other segmented area contains 1000 pixels. If the preset area threshold is 50 pixels, the first segmented area will be considered as noise, and the second segmented area will be considered as a real wormhole. The size of the preset area threshold needs to be determined according to the actual size of the wormhole. For example, if the wormholes caused by a small insect are being studied, the area threshold should be set smaller. If the wormholes caused by a large insect are being studied, the area threshold should be set larger. An area threshold that is too small will cause some small wormholes to be misidentified as noise, while an area threshold that is too large will cause some noise to be misidentified as wormholes. The purpose of judging whether the area of ​​each segmented region is smaller than the preset area threshold is to remove the noise region. For example, assuming that the area of ​​a segmented region is 30 pixels and the preset area threshold is 50 pixels, then the segmented region will be considered as noise. This is because regions with too small an area are likely to be caused by noise rather than real wormholes. After removing the noise region, the final segmentation result image can be obtained, in which each connected region represents a single wormhole. For example, assuming that the initial segmentation result image contains 5 segmented regions, of which the areas of 2 regions are smaller than the preset area threshold. After removing these two noise regions, the final segmentation result image will contain 3 connected regions, which represent 3 different wormholes. By screening the area threshold, noise can be effectively removed and the accuracy of wormhole identification can be improved.

[0058] Furthermore, extracting wormhole features from the wormhole image includes:

[0059] The edge contour of a single wormhole is extracted, and the length and width of the minimum circumscribed rectangle of the wormhole are calculated according to the edge contour, and the average value of the length and width of the minimum circumscribed rectangle is taken as the diameter of the wormhole; the area and number of the wormholes are calculated according to the edge contour.

[0060] Specifically, the Canny edge detection algorithm can effectively identify edge information in an image. The algorithm implements edge detection through multiple steps. First, the image is Gaussian filtered to reduce the impact of noise on edge detection. For example, using a 5x5 Gaussian kernel to smooth the image can effectively remove high-frequency noise in the image and make the edge clearer. Next, the gradient amplitude and direction of the image are calculated to determine the position and strength of the edge. For example, the Sobel operator can be used to calculate the gradient in the horizontal and vertical directions, and then the gradient amplitude is calculated based on the square root of the sum of the squares of the two gradients. Then non-maximum suppression is performed to refine the edge and only retain the local maximum pixel points in the gradient direction. Finally, double threshold processing is used to mark the pixels with gradient amplitude greater than the high threshold as strong edge pixels, and the pixels with gradient amplitude between the high threshold and the low threshold as weak edge pixels, and the complete edge is formed by connecting the strong edge pixels and the weak edge pixels. The Canny algorithm adopts these steps to detect edges more accurately and reduce the interference of noise and pseudo edges. The image segmentation method based on distance transform and watershed algorithm is an effective method for segmenting adhesion objects. For the calculation of the wormhole diameter, the original image needs to be preprocessed and segmented first to obtain a binary image of a single wormhole. Then the Canny edge detection algorithm is used to extract the edge contour of the wormhole. Assuming that the edge of a wormhole is missing, it can be repaired using morphological closing operations. Closing operations are operations that first expand and then erode, which can fill small gaps in the edge and smooth the edge contour. For example, using a 3x3 structure element to close the edge image can effectively repair the missing part of the edge and make the edge contour more complete. Next, the minimum bounding rectangle of the repaired edge contour is calculated. For example, the length of the minimum bounding rectangle of a wormhole is 10 pixels and the width is 8 pixels. Finally, the diameter of the wormhole is calculated based on the length L and width W of the minimum bounding rectangle, and the formula is diameter = (L + W) / 2. In this example, the diameter of the wormhole is (10 + 8) / 2 = 9 pixels. The method of using the minimum bounding rectangle to calculate the diameter is simple and easy, and can better reflect the size of the wormhole. If the area of ​​the wormhole region is directly calculated and then converted into an equivalent diameter, it may be affected by the irregular shape of the wormhole. The minimum circumscribed rectangle can better fit the overall shape of the wormhole, thereby obtaining a more accurate diameter estimate.

[0061] Furthermore, by combining the diameter and area of ​​each wormhole calculated previously, as well as the total number of wormholes, a pest image feature vector, such as [average diameter, average area, number of wormholes], can be constructed for subsequent pest level judgment. The use of a support vector machine model for pest level judgment can effectively classify the degree of pests based on image features. A support vector machine is a machine learning algorithm that can establish a mapping relationship between pest levels and image features based on training data. For example, a large number of pest images can be collected in advance and their pest levels (mild, moderate, severe) can be manually labeled. Then, the feature vectors of these images are extracted, and a support vector machine model is trained with these data. The trained model can predict the corresponding pest level based on the new pest image feature vector. The acquired pest image feature vector is input into the trained support vector machine model, and the model will output the corresponding pest level based on the learned mapping relationship. For example, a new pest image feature vector is [3 mm, 8 mm2, 20]. When it is input into the support vector machine model, the model may output "moderate" pest. The output of the support vector machine model can be used to determine the level of pests, thereby guiding agricultural production. For example, if the model determines that the pest level is "severe", it is recommended to take corresponding prevention and control measures, such as spraying pesticides. If the model determines that the pest level is "mild", it can continue to observe and not take intervention measures for the time being. This can effectively control pests and increase crop yields.

[0062] Furthermore, pest level labeling includes:

[0063] The damage level of the preprocessed image is calculated and labeled according to the average leaf-eating level algorithm and the pest level classification standard.

[0064] Embodiment 2

[0065] This embodiment also provides a system for evaluating the damage level of fall armyworm based on machine vision, including:

[0066] A preprocessing module is used to obtain a corn leaf image and perform preprocessing to obtain a preprocessed image;

[0067] The wormhole segmentation module is used to segment the contiguous wormholes in the preprocessed image and remove small area noise to obtain the wormhole image;

[0068] A feature extraction module is used to extract wormhole features in the wormhole image, where the wormhole features include the diameter, area and quantity of the wormholes;

[0069] A model building module is used to annotate the pre-processed image with the pest level according to the wormhole features to obtain an annotated image, and train a support vector machine through the annotated image to obtain a recognition model;

[0070] The recognition module is used to identify the corn leaves to be identified through the recognition model to obtain the damage level of the fall armyworm.

[0071] Furthermore, the preprocessing module includes:

[0072] A collection unit, used for collecting corn leaf images;

[0073] The preprocessing unit is used to grayscale the corn leaf image, remove image noise by Gaussian filtering, binarize the image after noise removal to obtain a binarized image, and perform morphological opening operation on the binarized image to obtain a preprocessed image.

[0074] Furthermore, the wormhole segmentation module includes:

[0075] A distance transformation unit is used to perform distance transformation on the binary image to obtain a distance map of the wormhole area;

[0076] A segmentation unit, used for segmenting the connected wormholes into individual wormholes by using a watershed algorithm on the distance map;

[0077] The denoising unit is used to determine whether the segmented wormhole area is small area noise, and remove it if it is small area noise to obtain a wormhole image.

[0078] Furthermore, the feature extraction module includes:

[0079] A contour extraction unit, used to extract the edge contour of a single wormhole;

[0080] The calculation unit is used to calculate the diameter, area and number of the wormhole according to the edge contour; the diameter is calculated by obtaining the length and width of the minimum circumscribed rectangle of the wormhole according to the edge contour, and taking the average value of the length and width of the minimum circumscribed rectangle as the diameter of the wormhole.

[0081] Embodiment 3

[0082] This embodiment further provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0083] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for evaluating the damage level of fall armyworm based on machine vision, characterized in that: The following steps are involved: Acquire a corn leaf image and perform preprocessing to obtain a preprocessed image; Segmenting the contiguous wormholes in the preprocessed image and removing small area noise to obtain a wormhole image; Extracting wormhole features in the wormhole image, wherein the wormhole features include diameter, area and quantity features of the wormholes; Annotating the preprocessed image for pest level according to the wormhole features to obtain an annotated image, and training a support vector machine with the annotated image to obtain a recognition model; The corn leaves to be identified are identified through the recognition model to obtain the damage level of the fall armyworm.

2. The method according to claim 1, characterized in that The preprocessing of corn leaf images includes: The corn leaf image is grayed, and Gaussian filtering is used to remove image noise. The image after noise removal is binarized to obtain a binarized image. A morphological opening operation is performed on the binarized image to obtain a preprocessed image.

3. The method according to claim 1, characterized in that The obtaining of the wormhole image comprises: The binary image is subjected to distance transformation to obtain a distance map of the wormhole area. The distance map is segmented into single wormholes by a watershed algorithm. If the area of ​​the segmented wormhole is smaller than a judgment threshold, it is judged as small area noise. The small area noise is removed to obtain a wormhole image.

4. The method according to claim 1, characterized in that: The extracting of wormhole features in the wormhole image comprises: The edge contour of a single wormhole is extracted, the length and width of the minimum circumscribed rectangle of the wormhole are calculated according to the edge contour, and the average value of the length and width of the minimum circumscribed rectangle is used as the diameter of the wormhole; the area and number of the wormholes are calculated according to the edge contour.

5. The method according to claim 1, characterized in that The pest level marking includes: The damage level of the preprocessed image is calculated and labeled according to the average leaf-eating level algorithm and the pest level classification standard.

6. A system for evaluating the damage level of fall armyworm based on machine vision, characterized in that: include: A preprocessing module is used to obtain a corn leaf image and perform preprocessing to obtain a preprocessed image; A wormhole segmentation module is used to segment the contiguous wormholes in the preprocessed image and remove small area noise to obtain a wormhole image; A feature extraction module is used to extract wormhole features in the wormhole image, wherein the wormhole features include diameter, area and quantity features of the wormholes; A model building module is used to annotate the pre-processed image with the pest level according to the wormhole features to obtain an annotated image, and train a support vector machine with the annotated image to obtain a recognition model; The recognition module is used to identify the corn leaves to be identified through the recognition model to obtain the damage level of the fall armyworm.

7. The system according to claim 6, characterized in that The preprocessing module comprises: A collection unit, used for collecting corn leaf images; The preprocessing unit is used to grayscale the corn leaf image, remove image noise by Gaussian filtering, binarize the image after noise removal to obtain a binarized image, and perform morphological opening operation on the binarized image to obtain a preprocessed image.

8. The system according to claim 6, characterized in that The wormhole segmentation module comprises: A distance transformation unit is used to perform distance transformation on the binary image to obtain a distance map of the wormhole area; A segmentation unit, used for segmenting the connected wormholes into individual wormholes by using a watershed algorithm on the distance map; The denoising unit is used to determine whether the segmented wormhole area is small area noise, and remove it if it is small area noise to obtain a wormhole image.

9. The system according to claim 6, characterized in that The feature extraction module comprises: A contour extraction unit, used to extract the edge contour of a single wormhole; A calculation unit is used to calculate the diameter, area and number of wormholes according to the edge contour; the diameter is calculated by obtaining the length and width of the minimum circumscribed rectangle of the wormhole according to the edge contour, and taking the average value of the length and width of the minimum circumscribed rectangle as the diameter of the wormhole.

10. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 5.

Images