Pepper leaf powdery mildew condition diagnosis method and system, storage medium and equipment

Abstract

The invention discloses a pepper leaf powdery mildew disease diagnosis method and system, a storage medium and equipment. The method comprises the following steps: acquiring pepper leaf images corresponding to different time nodes, angles and illumination conditions through an industrial camera; carrying out region segmentation on the pepper leaf image by adopting an image segmentation algorithm to obtain a plurality of leaf disease regions; based on evaluation indexes of energy, entropy, contrast and correlation, the influence of each leaf disease area on the accuracy of the disease diagnosis model is evaluated; a neural network mode recognition system and a quantitative conjugate gradient training algorithm are combined, and a fusion algorithm is obtained and evaluated; pepper leaf images corresponding to different disease degrees are collected in a linear array scanning mode, and the disease diagnosis model is used for recognizing the pepper leaf images so as to diagnose the disease grade of the pepper leaves. The problems that in the prior art, the detection level is not uniform and the early stage of the disease cannot be recognized due to the mode of judging the disease through a single feature are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of plant disease diagnosis, and in particular to a method, system, storage medium and equipment for diagnosing powdery mildew on pepper leaves. Background Art

[0002] Pepper is a vegetable with the largest sales volume in China and is widely used in food processing, chemical raw material extraction and other industries. With the continuous industrialization from planting to processing, disease prevention and control of pepper in the field planting process has become an important link in the entire industrial chain.

[0003] At present, the diagnosis method of pepper diseases mainly relies on the color characteristics of pepper leaves to assess the severity of the disease. Under manual identification conditions, this method of judging pepper leaf diseases by a single feature can easily lead to uneven detection levels and fail to identify the early stages of pepper leaf diseases, thus leading to the misuse and abuse of pesticides. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method, system, storage medium and equipment for diagnosing powdery mildew on pepper leaves, aiming to solve the problem that the prior art method of judging pepper leaf diseases by a single feature easily leads to uneven detection levels and cannot identify the early stages of pepper leaf diseases.

[0005] A first aspect of the present invention is to provide a method for diagnosing powdery mildew on pepper leaves, the method comprising: The industrial camera is used to collect pepper leaf images corresponding to different time points, angles and lighting conditions; Using a four-dimensional adaptive clustering image segmentation algorithm to perform regional segmentation on the pepper leaf image to obtain multiple leaf disease regions; Based on the evaluation indicators of energy, entropy, contrast and correlation, the influence of each leaf disease area on the accuracy of the disease diagnosis model was evaluated; Combining a neural network pattern recognition system with a quantitative conjugate gradient training algorithm to obtain a fusion algorithm, and evaluating the fusion algorithm on the dataset of pepper leaf images; A linear array scanning method is used to collect pepper leaf images corresponding to different disease degrees, and the pepper leaf images are identified using the disease diagnosis model to diagnose the disease level of the pepper leaves.

[0006] According to one aspect of the above technical solution, the step of using a four-dimensional adaptive clustering image segmentation algorithm to segment the pepper leaf image to obtain multiple leaf disease areas includes: According to the visual system and feature parameter extraction method, saturation, intensity and color channels are obtained and used as feature parameters for input; The characteristic parameters are used to perform four-dimensional segmentation and two-dimensional segmentation on the pepper leaf image, thereby achieving regional segmentation and obtaining multiple leaf disease regions.

[0007] According to one aspect of the above technical solution, the saturation, intensity and color channel are respectively represented as saturation S ,strength I , Color Channel a , b , the corresponding models are: ; ; ; ; in, R , G , B Are the three primary colors of the image.

[0008] According to one aspect of the above technical solution, the step of evaluating the influence of each leaf disease area on the accuracy of the disease diagnosis model based on the evaluation indicators of energy, entropy, contrast and correlation includes: Selecting measurement points and deviation points in each of the leaf disease areas; Calculating the occurrence number, normalized probability and gray-level co-occurrence matrix of any gray value in the leaf disease area; Identify matrix features in the gray-level co-occurrence matrix, and derive energy, entropy, contrast, and correlation corresponding to identification parameters based on the matrix features; The identification parameters are used to describe the texture states corresponding to pepper leaves with different diseases; The pepper leaves are evenly divided into multiple equal-width areas from the root to the top, and the influencing parameters of each leaf disease area in the pepper leaf image on the disease diagnosis model are identified; According to the influencing parameters, one of the target leaf disease areas is selected to replace the pepper leaf image to train the disease diagnosis model.

[0009] According to one aspect of the above technical solution, the energy, entropy, contrast and correlation are respectively expressed as energy ,entropy , Contrast Relevance , the corresponding models are: ; ; ; ; In the formula, and , and They are and The mean and standard deviation of in, and The models are , .

[0010] According to one aspect of the above technical solution, a fusion algorithm is obtained by combining a neural network pattern recognition system with a quantitative conjugate gradient training algorithm, and the step of evaluating the fusion algorithm on the data set of the pepper leaf image includes: Multiple pepper leaf images were randomly divided into training set, validation set and test set with a division ratio of 70:15:15; Combined with the quantitative conjugate gradient training algorithm, the gradient is 6.62×10 -7 Disease diagnosis model.

[0011] According to one aspect of the above technical solution, the steps of collecting pepper leaf images corresponding to different disease degrees by linear array scanning, and using the disease diagnosis model to identify the pepper leaf images to diagnose the disease level of the pepper leaves include: Collect panoramic images of pepper plants with different disease severity levels; Each panoramic image is divided into multiple observation points, and the disease matching degree of each observation point is measured by linear array scanning; Measuring the recognition rate of the whole pepper plant using the disease diagnosis model; According to the measurement results, the degree of powdery mildew disease on pepper leaves is identified and the disease grade is output.

[0012] The second aspect of the present invention is to provide a pepper leaf powdery mildew disease diagnosis system, which is applied to the method described in the above technical solution, and the system comprises: An image acquisition module is used to collect pepper leaf images corresponding to different time nodes, angles and lighting conditions through an industrial camera; An image segmentation module is used to perform regional segmentation on the pepper leaf image using a four-dimensional adaptive clustering image segmentation algorithm to obtain multiple leaf disease regions; The first evaluation module is used to evaluate the impact of each leaf disease area on the accuracy of the disease diagnosis model based on the evaluation indicators of energy, entropy, contrast and correlation; A second evaluation module is used to combine the neural network pattern recognition system with the quantitative conjugate gradient training algorithm to obtain a fusion algorithm, and evaluate the fusion algorithm on the data set of the pepper leaf image; The disease diagnosis module is used to collect pepper leaf images corresponding to different disease degrees by using a linear array scanning method, and use the disease diagnosis model to identify the pepper leaf images to diagnose the disease level of the pepper leaves.

[0013] A third aspect of the present invention is to provide a readable storage medium having a computer program stored thereon, which implements the method described in the above technical solution when executed by a processor.

[0014] The fourth aspect of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the method described in the above technical solution when executing the computer program.

[0015] Compared with the prior art, the pepper leaf powdery mildew diagnosis method, system, storage medium and device shown in the present invention have the following beneficial effects: By using the method shown in the present invention, the disease features of pepper leaves can be automatically extracted by combining feature transfer learning and image processing technology, which can reduce the randomness of traditional disease identification methods and overcome the limitations of detection methods. An adaptive clustering algorithm and color space parameters are used to construct a pepper leaf powdery mildew diagnosis system for identifying diseased leaves in complex backgrounds. Based on texture feature parameters, the degree of regional disease is quantified, which can improve the computational efficiency of the disease diagnosis model while maintaining detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which: Figure 1 It is a flow chart of a method for diagnosing powdery mildew on pepper leaves according to an embodiment of the present invention; Figure 2 Schematic diagram of a pepper leaf image acquisition device used in one embodiment of the present invention.

[0017] Figure 3 Four-dimensional and two-dimensional segmented images of pepper leaves are collected in one embodiment of the present invention.

[0018] Figure 4 These are the values ​​of energy, entropy, contrast and correlation of pepper leaves with 12 different disease degrees in one embodiment of the present invention.

[0019] Figure 5 These are the values ​​of energy, entropy, contrast and correlation of ten regions of a diseased pepper leaf in one embodiment of the present invention.

[0020] Figure 6A technical roadmap of a method for diagnosing powdery mildew on pepper leaves in one embodiment of the present invention; Figure 7 The present invention is a block diagram of a pepper leaf powdery mildew disease diagnosis system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. Several embodiments of the present invention are shown in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0022] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0024] Embodiment 1 See also Figure 1 The first embodiment of the present invention provides a method for diagnosing powdery mildew on pepper leaves, the method comprising steps S10 to S50: Step S10, collecting pepper leaf images corresponding to different time nodes, angles and lighting conditions through an industrial camera.

[0025] In this embodiment, the industrial camera belongs to a pre-built pepper leaf collection system, such as Figure 2 As shown, the pepper leaf acquisition system includes a parallel light source 1, a controllable lifting platform 2, a camera lens 3, an industrial camera 4, a computer 5, a first network cable 6, a second network cable 7, a tripod 8 and a camera 9, so as to construct an acquisition system for obtaining pepper leaf image data, thereby acquiring the image data of the pepper leaves through the system.

[0026] The pepper leaf collection system shown in this embodiment works as follows: The pepper leaf is placed flat on the center of the lifting platform 2, and a parallel light source 1 is used to simulate sunlight shining on the leaf surface. The height of the lifting platform 2 is adjusted to obtain the best definition and unexposed pepper leaf image as experimental data. The image data of the pepper leaf image with the best definition is sequentially transmitted to the computer 5 through the camera lens 3, the industrial camera 4, and the first network cable 6, while the image data of the unexposed pepper leaf image is sequentially transmitted to the computer 5 through the camera 9 and the second network cable 7. According to the moldy area on the pepper leaf recorded in the pepper leaf image, the disease degree of the same variety of pepper plants in the early stage of powdery mildew is pre-divided into 12 categories, and 10 pepper leaves with similar mold areas are selected from each category, and the pepper leaf image of each pepper leaf is collected 10 times under the same measurement conditions, and a total of 1200 pepper leaf images are collected.

[0027] Step S20, using a four-dimensional adaptive clustering image segmentation algorithm to perform region segmentation on the pepper leaf image to obtain a plurality of leaf disease regions.

[0028] In this embodiment, the step of using a four-dimensional adaptive clustering image segmentation algorithm to segment the pepper leaf image to obtain multiple leaf disease areas includes: According to the visual system and feature parameter extraction method, saturation, intensity and color channels are obtained and used as feature parameters for input; The characteristic parameters are used to perform four-dimensional segmentation and two-dimensional segmentation on the pepper leaf image, thereby achieving regional segmentation and obtaining multiple leaf disease regions.

[0029] The saturation, intensity and color channel are respectively represented as saturation S ,strength I , Color Channel a , b , the corresponding models are: ; ; ; ; Among them, R, G, and B are the three primary colors of the image. , , The values ​​of are set to 95.047, 100 and 108.883 respectively, and it is required .

[0030] The schematic diagram of four-dimensional segmentation and two-dimensional segmentation of pepper leaf images is as follows: Figure 3 shown.

[0031] Step S30, based on the evaluation indicators of energy, entropy, contrast and correlation, the influence of each leaf disease area on the accuracy of the disease diagnosis model is evaluated.

[0032] In this embodiment, the step of evaluating the influence of each leaf disease area on the accuracy of the disease diagnosis model based on the evaluation indicators of energy, entropy, contrast and correlation includes: Selecting measurement points and deviation points in each of the leaf disease areas; Calculating the occurrence number, normalized probability and gray-level co-occurrence matrix of any gray value in the leaf disease area; Identify matrix features in the gray-level co-occurrence matrix, and derive energy, entropy, contrast, and correlation corresponding to identification parameters based on the matrix features; The identification parameters are used to describe the texture states corresponding to pepper leaves with different diseases; The pepper leaves are evenly divided into multiple equal-width areas from the root to the top, and the influencing parameters of each leaf disease area in the pepper leaf image on the disease diagnosis model are identified; According to the influencing parameters, one of the target leaf disease areas is selected to replace the pepper leaf image to train the disease diagnosis model.

[0033] Among them, the energy, entropy, contrast and correlation are respectively expressed as energy ,entropy , Contrast Relevance , the corresponding models are: ; ; ; ; In the formula, and , and They are and The mean and standard deviation of in, and The models are , .

[0034] Specifically, in this embodiment, measurement points are selected in the image of each leaf disease area obtained by four-dimensional segmentation and two-dimensional segmentation. With deviation point , where the gray value of the measurement point is defined as , the gray value series is set toL Furthermore, the occurrence frequency and normalized probability of any target gray value in the image of the entire leaf disease area are calculated. Gray-level co-occurrence matrix . Further, based on the gray level co-occurrence matrix The matrix characteristics of the energy corresponding to the identification parameters are derived ,entropy , Contrast Relevance . Further, the texture state of pepper leaves with different diseases is described by using identification parameters. The energy of pepper leaves with 12 different disease degrees shown in this embodiment ,entropy , Contrast Relevance The corresponding values ​​are Figure 4 Further, the pepper leaves are evenly divided into 10 equal-width regions from the root to the top, and the influence of each region in the pepper leaves on the disease diagnosis model is calculated. The energy of 10 regions of pepper leaves with one disease shown in this embodiment ,entropy , Contrast Relevance The corresponding values ​​are Figure 5 Finally, according to the influencing parameters, the leaf disease area 5 is selected to replace the entire pepper leaf image to train the disease diagnosis model.

[0035] Step S40, combining the neural network pattern recognition system with the quantitative conjugate gradient training algorithm to obtain a fusion algorithm, and evaluating the fusion algorithm on the data set of the pepper leaf image.

[0036] In this embodiment, a fusion algorithm is obtained by combining a neural network pattern recognition system with a quantitative conjugate gradient training algorithm. The step of evaluating the fusion algorithm on the data set of pepper leaf images includes: Multiple pepper leaf images were randomly divided into training set, validation set and test set with a division ratio of 70:15:15; Combined with the quantitative conjugate gradient training algorithm, the gradient is 6.62×10 -7 Disease diagnosis model.

[0037] Specifically, in this embodiment, 35577 pepper leaf images collected in advance are randomly divided into training set, validation set and test set with a division ratio of 70:15:15. Then, the disease diagnosis model is trained by combining the quantitative conjugate gradient training algorithm, and the gradient is 6.62×10 -7 Disease diagnosis model.

[0038] Step S50, using a linear array scanning method to collect pepper leaf images corresponding to different disease degrees, and using the disease diagnosis model to identify the pepper leaf images to diagnose the disease level of the pepper leaves.

[0039] In this embodiment, the steps of collecting pepper leaf images corresponding to different disease degrees by linear array scanning, and using the disease diagnosis model to identify the pepper leaf images to diagnose the disease level of the pepper leaves include: Collect panoramic images of pepper plants with different disease severity levels; Each panoramic image is divided into multiple observation points, and the disease matching degree of each observation point is measured by linear array scanning; Measuring the recognition rate of the whole pepper plant using the disease diagnosis model; According to the measurement results, the degree of powdery mildew disease on pepper leaves is identified and the disease grade is output.

[0040] Specifically, panoramic images of pepper plants with different disease degrees in the field were collected, each image was divided into 560 observation points, the disease matching degree of each observation point was measured using the linear scanning method, and the recognition rate of the whole pepper plant was measured using the disease diagnosis model. Finally, the disease degree of powdery mildew on pepper leaves was identified based on the measurement results, and the disease level of powdery mildew on pepper leaves was displayed.

[0041] like Figure 6 The diagram shown is a disease grade diagnosis diagram of two pepper leaves obtained according to step S5, as shown in FIG. Figure 6 As reflected, after the linear array scanning method is used to measure the leaves of two pepper plants, the pepper leaf images obtained can be recognized by the disease diagnosis model shown in this embodiment, and the specific disease degree of the diseased pepper leaves can be determined.

[0042] The pepper leaf powdery mildew disease diagnosis system based on the feature transfer learning model shown in this embodiment improves the capture rate of the powdery mildew data set by combining feature transfer learning and a neural network algorithm, and solves the problem of identifying insufficiently labeled pepper leaf images under the interference of complex background information; a four-dimensional adaptive clustering image segmentation algorithm is used to extract pepper leaf images under complex backgrounds, a fusion algorithm is used to analyze the RGB value of each pixel in the pepper leaf image, and an effectiveness index is calculated based on the image of each pepper leaf, and the optimal number of clusters is automatically determined to obtain a better segmentation result; in addition, the present invention also selects the leaf disease area 5 with the best linearity as the main target of texture feature extraction based on parameters such as energy, entropy, contrast and correlation, uses a neural network pattern recognition system to train and process pepper leaf images, and combines a quantized conjugate gradient training algorithm to construct a diagnosis model for pepper leaf powdery mildew.

[0043] In summary, compared with the prior art, the method for diagnosing powdery mildew on pepper leaves shown in this embodiment has the following beneficial effects: By using the method shown in this embodiment, the disease features of pepper leaves are automatically extracted by combining feature transfer learning and image processing technology, which can reduce the randomness of traditional disease identification methods and overcome the limitations of detection methods. An adaptive clustering algorithm and color space parameters are used to construct a pepper leaf powdery mildew diagnosis system for identifying diseased leaves in complex backgrounds. Based on texture feature parameters, the regional disease degree is quantified, which can improve the computational efficiency of the disease diagnosis model while maintaining the detection accuracy.

[0044] Combined with feature transfer learning and image processing technology, the disease features of pepper leaves are automatically extracted. Adaptive clustering algorithm and color space parameters are used to build a pepper leaf powdery mildew diagnosis system to identify diseased leaves in complex backgrounds. Based on texture feature parameters, the regional disease degree is quantified, which can improve the computational efficiency of the recognition model while maintaining the detection accuracy. Leaves with 12 disease degrees are used to evaluate the feature transfer learning system, and a diagnosis system for early powdery mildew on pepper leaves is obtained.

[0045] Embodiment 2 See also Figure 2 The second embodiment of the present invention provides a pepper leaf powdery mildew disease diagnosis system, the system comprising: The image acquisition module 10 is used to collect pepper leaf images corresponding to different time nodes, angles and lighting conditions through an industrial camera.

[0046] In this embodiment, the industrial camera belongs to a pre-built pepper leaf collection system, such as Figure 2 As shown, the pepper leaf acquisition system includes a parallel light source 1, a controllable lifting platform 2, a camera lens 3, an industrial camera 4, a computer 5, a first network cable 6, a second network cable 7, a tripod 8 and a camera 9, so as to construct an acquisition system for obtaining pepper leaf image data, thereby acquiring the image data of the pepper leaves through the system.

[0047] The pepper leaf collection system shown in this embodiment works as follows: The pepper leaf is placed flat on the center of the lifting platform 2, and a parallel light source 1 is used to simulate sunlight shining on the leaf surface. The height of the lifting platform 2 is adjusted to obtain the best definition and unexposed pepper leaf image as experimental data. The image data of the pepper leaf image with the best definition is sequentially transmitted to the computer 5 through the camera lens 3, the industrial camera 4, and the first network cable 6, while the image data of the unexposed pepper leaf image is sequentially transmitted to the computer 5 through the camera 9 and the second network cable 7. According to the moldy area on the pepper leaf recorded in the pepper leaf image, the disease degree of the same variety of pepper plants in the early stage of powdery mildew is pre-divided into 12 categories, and 10 pepper leaves with similar mold areas are selected from each category, and the pepper leaf image of each pepper leaf is collected 10 times under the same measurement conditions, and a total of 1200 pepper leaf images are collected.

[0048] The image segmentation module 20 is used to perform region segmentation on the pepper leaf image using a four-dimensional adaptive clustering image segmentation algorithm to obtain a plurality of leaf disease regions.

[0049] In this embodiment, the step of using a four-dimensional adaptive clustering image segmentation algorithm to segment the pepper leaf image to obtain multiple leaf disease areas includes: According to the visual system and feature parameter extraction method, saturation, intensity and color channels are obtained and used as feature parameters for input; The characteristic parameters are used to perform four-dimensional segmentation and two-dimensional segmentation on the pepper leaf image, thereby achieving regional segmentation and obtaining multiple leaf disease regions.

[0050] The saturation, intensity and color channel are respectively represented as saturation S ,strength I , Color Channel a , b , the corresponding models are: ; ; ; ; Among them, R, G, and B are the three primary colors of the image. , , The values ​​of are set to 95.047, 100 and 108.883 respectively, and it is required .

[0051] The schematic diagram of four-dimensional segmentation and two-dimensional segmentation of pepper leaf images is as follows: Figure 3 shown.

[0052] The first evaluation module 30 is used to evaluate the influence of each leaf disease area on the accuracy of the disease diagnosis model based on the evaluation indicators of energy, entropy, contrast and correlation.

[0053] In this embodiment, the step of evaluating the influence of each leaf disease area on the accuracy of the disease diagnosis model based on the evaluation indicators of energy, entropy, contrast and correlation includes: Selecting measurement points and deviation points in each of the leaf disease areas; Calculating the occurrence number, normalized probability and gray-level co-occurrence matrix of any gray value in the leaf disease area; Identify matrix features in the gray-level co-occurrence matrix, and derive energy, entropy, contrast, and correlation corresponding to identification parameters based on the matrix features; The identification parameters are used to describe the texture states corresponding to pepper leaves with different diseases; The pepper leaves are evenly divided into multiple equal-width areas from the root to the top, and the influencing parameters of each leaf disease area in the pepper leaf image on the disease diagnosis model are identified; According to the influencing parameters, one of the target leaf disease areas is selected to replace the pepper leaf image to train the disease diagnosis model.

[0054] Among them, the energy, entropy, contrast and correlation are respectively expressed as energy ,entropy , Contrast Relevance , the corresponding models are: ; ; ; ; In the formula, and , and They are and The mean and standard deviation of in, and The models are , .

[0055] Specifically, in this embodiment, measurement points are selected in the image of each leaf disease area obtained by four-dimensional segmentation and two-dimensional segmentation. With deviation point , where the gray value of the measurement point is defined as , the gray value series is set to L Furthermore, the occurrence frequency and normalized probability of any target gray value in the image of the entire leaf disease area are calculated. Gray-level co-occurrence matrix . Further, based on the gray level co-occurrence matrix The matrix characteristics of the energy corresponding to the identification parameters are derived ,entropy , Contrast Relevance . Further, the texture state of pepper leaves with different diseases is described by using identification parameters. The energy of pepper leaves with 12 different disease degrees shown in this embodiment ,entropy , Contrast Relevance The corresponding values ​​are Figure 4 Further, the pepper leaves are evenly divided into 10 equal-width regions from the root to the top, and the influence of each region in the pepper leaves on the disease diagnosis model is calculated. The energy of 10 regions of pepper leaves with one disease shown in this embodiment ,entropy , Contrast Relevance The corresponding values ​​are Figure 5 Finally, according to the influencing parameters, the leaf disease area 5 is selected to replace the entire pepper leaf image to train the disease diagnosis model.

[0056] The second evaluation module 40 is used to combine the neural network pattern recognition system with the quantitative conjugate gradient training algorithm to obtain a fusion algorithm, and evaluate the fusion algorithm on the data set of the pepper leaf image.

[0057] In this embodiment, a fusion algorithm is obtained by combining a neural network pattern recognition system with a quantitative conjugate gradient training algorithm. The step of evaluating the fusion algorithm on the data set of pepper leaf images includes: Multiple pepper leaf images were randomly divided into training set, validation set and test set with a division ratio of 70:15:15; Combined with the quantitative conjugate gradient training algorithm, the gradient is 6.62×10 -7 Disease diagnosis model.

[0058] Specifically, in this embodiment, 35577 pepper leaf images collected in advance are randomly divided into training set, validation set and test set with a division ratio of 70:15:15. Then, the disease diagnosis model is trained by combining the quantitative conjugate gradient training algorithm, and the gradient is 6.62×10 -7 Disease diagnosis model.

[0059] The disease diagnosis module 50 is used to collect pepper leaf images corresponding to different disease degrees by using a linear array scanning method, and use the disease diagnosis model to identify the pepper leaf images to diagnose the disease level of the pepper leaves.

[0060] In this embodiment, the steps of collecting pepper leaf images corresponding to different disease degrees by linear array scanning, and using the disease diagnosis model to identify the pepper leaf images to diagnose the disease level of the pepper leaves include: Collect panoramic images of pepper plants with different disease severity levels; Each panoramic image is divided into multiple observation points, and the disease matching degree of each observation point is measured by linear array scanning; Measuring the recognition rate of the whole pepper plant using the disease diagnosis model; According to the measurement results, the degree of powdery mildew disease on pepper leaves is identified and the disease grade is output.

[0061] Specifically, panoramic images of pepper plants with different disease degrees in the field were collected, each image was divided into 560 observation points, the disease matching degree of each observation point was measured using the linear scanning method, and the recognition rate of the whole pepper plant was measured using the disease diagnosis model. Finally, the disease degree of powdery mildew on pepper leaves was identified based on the measurement results, and the disease level of powdery mildew on pepper leaves was displayed.

[0062] like Figure 6 The diagram shows the disease grade diagnosis diagram of two pepper leaves diagnosed by the disease diagnosis module 50. Figure 6 As reflected, after the linear array scanning method is used to measure the leaves of two pepper plants, the pepper leaf images obtained can be recognized by the disease diagnosis model shown in this embodiment, and the specific disease degree of the diseased pepper leaves can be determined.

[0063] The pepper leaf powdery mildew disease diagnosis system based on the feature transfer learning model shown in this embodiment improves the capture rate of the powdery mildew data set by combining feature transfer learning and a neural network algorithm, and solves the problem of identifying insufficiently labeled pepper leaf images under the interference of complex background information; a four-dimensional adaptive clustering image segmentation algorithm is used to extract pepper leaf images under complex backgrounds, a fusion algorithm is used to analyze the RGB value of each pixel in the pepper leaf image, and an effectiveness index is calculated based on the image of each pepper leaf, and the optimal number of clusters is automatically determined to obtain a better segmentation result; in addition, the present invention also selects the leaf disease area 5 with the best linearity as the main target of texture feature extraction based on parameters such as energy, entropy, contrast and correlation, uses a neural network pattern recognition system to train and process pepper leaf images, and combines a quantized conjugate gradient training algorithm to construct a diagnosis model for pepper leaf powdery mildew.

[0064] In summary, compared with the prior art, the pepper leaf powdery mildew disease diagnosis system shown in this embodiment has the following beneficial effects: By using the system shown in this embodiment, the disease features of pepper leaves can be automatically extracted by combining feature transfer learning and image processing technology, which can reduce the randomness of traditional disease identification methods and overcome the limitations of detection methods. An adaptive clustering algorithm and color space parameters are used to construct a pepper leaf powdery mildew diagnosis system for identifying diseased leaves in complex backgrounds. Based on texture feature parameters, the degree of regional disease is quantified, which can improve the computational efficiency of the disease diagnosis model while maintaining detection accuracy.

[0065] Embodiment 3 A third embodiment of the present invention provides a readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method described in the above embodiment is implemented.

[0066] Embodiment 4 A fourth embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the method described in the above embodiment when executing the computer program.

[0067] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0068] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for diagnosing powdery mildew on pepper leaves, characterized in that: The method comprises: The industrial camera is used to collect pepper leaf images corresponding to different time points, angles and lighting conditions; Using a four-dimensional adaptive clustering image segmentation algorithm to perform regional segmentation on the pepper leaf image to obtain multiple leaf disease regions; Based on the evaluation indicators of energy, entropy, contrast and correlation, the influence of each leaf disease area on the accuracy of the disease diagnosis model was evaluated; Combining a neural network pattern recognition system with a quantitative conjugate gradient training algorithm to obtain a fusion algorithm, and evaluating the fusion algorithm on the dataset of pepper leaf images; A linear array scanning method is used to collect pepper leaf images corresponding to different disease degrees, and the pepper leaf images are identified using the disease diagnosis model to diagnose the disease level of the pepper leaves.

2. The method for diagnosing powdery mildew on pepper leaves according to claim 1, characterized in that: The step of performing regional segmentation on the pepper leaf image using a four-dimensional adaptive clustering image segmentation algorithm to obtain multiple leaf disease regions comprises: According to the visual system and feature parameter extraction method, saturation, intensity and color channels are obtained and used as feature parameters for input; The characteristic parameters are used to perform four-dimensional segmentation and two-dimensional segmentation on the pepper leaf image, thereby achieving regional segmentation and obtaining multiple leaf disease regions.

3. The method for diagnosing powdery mildew on pepper leaves according to claim 2, characterized in that: The saturation, intensity and color channels are respectively represented as saturation S ,strength I , Color Channel a , b , the corresponding models are: ； ； ； ； in, R , G , B Are the three primary colors of the image.

4. The method for diagnosing powdery mildew on pepper leaves according to claim 1, characterized in that: The steps of evaluating the impact of each leaf disease area on the accuracy of the disease diagnosis model based on the evaluation indicators of energy, entropy, contrast and correlation include: Selecting measurement points and deviation points in each of the leaf disease areas; Calculating the occurrence number, normalized probability and gray-level co-occurrence matrix of any gray value in the leaf disease area; Identify matrix features in the gray-level co-occurrence matrix, and derive energy, entropy, contrast, and correlation corresponding to identification parameters based on the matrix features; The identification parameters are used to describe the texture states corresponding to pepper leaves with different diseases; The pepper leaves are evenly divided into multiple equal-width areas from the root to the top, and the influencing parameters of each leaf disease area in the pepper leaf image on the disease diagnosis model are identified; According to the influencing parameters, one of the target leaf disease areas is selected to replace the pepper leaf image to train the disease diagnosis model.

5. The method for diagnosing powdery mildew on pepper leaves according to claim 4, characterized in that: The energy, entropy, contrast and correlation are respectively expressed as energy ,entropy , Contrast Relevance , the corresponding models are: ； ； ； ； In the formula, and , and They are and The mean and standard deviation of in, and The models are , .

6. The method for diagnosing powdery mildew on pepper leaves according to claim 1, characterized in that: Combining the neural network pattern recognition system with the quantitative conjugate gradient training algorithm to obtain a fusion algorithm, the step of evaluating the fusion algorithm on the data set of the pepper leaf image comprises: Multiple pepper leaf images were randomly divided into training set, validation set and test set with a division ratio of 70:15:15; Combined with the quantitative conjugate gradient training algorithm, the gradient is 6.62×10 -7 Disease diagnosis model.

7. The method for diagnosing powdery mildew on pepper leaves according to claim 1, characterized in that: The steps of collecting pepper leaf images corresponding to different disease degrees by using a linear array scanning method, and identifying the pepper leaf images by using the disease diagnosis model to diagnose the disease level of the pepper leaves include: Collect panoramic images of pepper plants with different disease severity levels; Each panoramic image is divided into multiple observation points, and the disease matching degree of each observation point is measured by linear array scanning; Measuring the recognition rate of the whole pepper plant using the disease diagnosis model; According to the measurement results, the degree of powdery mildew disease on pepper leaves is identified and the disease grade is output.

8. A pepper leaf powdery mildew disease diagnosis system, characterized in that: The method applied to any one of claims 1 to 7, wherein the system comprises: An image acquisition module is used to collect pepper leaf images corresponding to different time nodes, angles and lighting conditions through an industrial camera; An image segmentation module is used to perform regional segmentation on the pepper leaf image using a four-dimensional adaptive clustering image segmentation algorithm to obtain multiple leaf disease regions; The first evaluation module is used to evaluate the impact of each leaf disease area on the accuracy of the disease diagnosis model based on the evaluation indicators of energy, entropy, contrast and correlation; A second evaluation module is used to combine the neural network pattern recognition system with the quantitative conjugate gradient training algorithm to obtain a fusion algorithm, and evaluate the fusion algorithm on the data set of the pepper leaf image; The disease diagnosis module is used to collect pepper leaf images corresponding to different disease degrees by using a linear array scanning method, and use the disease diagnosis model to identify the pepper leaf images to diagnose the disease level of the pepper leaves.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and implemented to run on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images