Method and system for detecting leaves damaged by litchi stem borer based on deep learning

Abstract

The present invention discloses a method and system for detecting leaves damaged by litchi stalk borers based on deep learning, comprising the following steps: first, obtaining litchi leaf image data of different litchi planting areas in the target season, and extracting features to obtain litchi leaf feature data. Then, clustering operations are performed on these feature data based on a clustering algorithm to determine the detection representative area of ​​leaves damaged by litchi stalk borers, and real-time litchi leaf image data of the area is obtained in a preset time period. Next, the real-time image data is used to detect damage symptoms using a deep learning algorithm to obtain damage symptom change data within a preset time period. Based on these change data, the damage development trend of litchi leaves is predicted, and finally a prediction result of the damage degree of litchi leaves is formed, and corresponding prevention and control measures are formulated. This method improves the accuracy of detection of leaves damaged by litchi stalk borers and the prevention and control effect by combining deep learning with a clustering algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of litchi stem borer detection, and in particular to a method and system for detecting litchi stem borer-infested leaves based on deep learning. Background Art

[0002] At present, litchi production is often harmed by litchi stem borers. During the litchi fruiting period, the insect mainly bores into the fruit and the population is relatively high. In autumn, it feeds on the leaves of the autumn shoots of litchi, which are the fruiting branches of the next year. If they are damaged, the litchi yield will be seriously affected. In winter, it still feeds on the young leaves of litchi. Although the population is low, the number of insect sources in winter determines the severity of the damage caused by the litchi stem borer in the next year. During the fruiting period, the traditional prediction and forecast of litchi stem borers relies on picking up fallen fruits to cultivate stem borers and predicting the peak period of adult egg-laying. In autumn and winter, it relies entirely on manual observation and empirical judgment, which has problems such as low detection efficiency and low accuracy, and cannot meet the prediction needs of large litchi planting areas.

[0003] After the lychee stem borer infests the leaves, the early symptoms are not obvious, but in the middle and late stages of the damage, that is, before the larvae of the lychee stem borer are about to emerge, the tips and veins of the lychee leaves die, and the symptoms are more obvious and unique, which are suitable for automatic identification using image recognition technology.

[0004] With the continuous development of deep learning technology, it has made significant progress in the fields of image processing and pattern recognition and has been widely used in the agricultural field. The image recognition method based on deep learning can automatically extract feature information from images and perform efficient and accurate classification and recognition, providing new possibilities for intelligent detection of crop diseases and pests.

[0005] However, there are still some challenges in the detection of litchi leaves damaged by litchi stem borers. Due to the complexity of the litchi growth environment and the diversity of natural characteristics of leaves, the application of traditional image processing and machine learning methods in this field is limited. Therefore, it is necessary to propose a deep learning-based detection method for litchi leaves damaged by litchi stem borers to solve the problems of traditional methods and improve the efficiency and accuracy of automatic detection of litchi leaves damaged by litchi stem borers.

[0006] In view of the above background, the present invention proposes a method for detecting leaves infested by litchi stalk borers based on deep learning. By acquiring litchi leaf image data, extracting litchi leaf features and performing clustering operations, and combining deep learning algorithms, automatic detection of litchi leaves infested by litchi stalk borers is achieved, thereby providing a highly efficient and accurate detection technology for leaves infested by litchi stalk borers for litchi production. Summary of the invention

[0007] In order to solve at least one of the above technical problems, the present invention proposes a method and system for detecting leaves infested by litchi stem borers based on deep learning.

[0008] The first aspect of the present invention provides a method for detecting leaves damaged by litchi stem borers based on deep learning, comprising:

[0009] Acquire litchi leaf image data from different litchi planting areas in a target season, extract litchi leaf features from different litchi planting areas according to the litchi leaf image data, and obtain litchi leaf feature data;

[0010] Performing a clustering operation on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result;

[0011] Determine a detection representative area of ​​leaves damaged by litchi peduncle borers according to the clustering result, and obtain real-time litchi leaf image data of the detection representative area in a preset time period;

[0012] Based on the deep learning algorithm, the real-time litchi leaf image data is used to detect the symptoms of litchi pedicle borer damage, and the litchi pedicle borer damage symptom change data in a preset time period is obtained;

[0013] Predicting the development trend of litchi leaf damage based on the symptom change data of litchi peduncle borer damage, and obtaining a prediction result of litchi leaf damage;

[0014] The control measures for litchi stalk borer are formulated according to the prediction results of the damage degree of litchi leaves.

[0015] In this solution, the litchi leaf image data of different litchi planting areas in the target season are obtained, and the litchi leaf features of different litchi planting areas are extracted according to the litchi leaf image data to obtain litchi leaf feature data, specifically:

[0016] The target litchi planting area is divided into N sub-areas, and the litchi leaf image data of each sub-area in the target season is obtained by high-definition camera equipment;

[0017] The litchi leaf image data is converted from the RGB color space to the HSV color space based on an image recognition algorithm, a statistical analysis is performed on the pixels of each litchi leaf image in the HSV color space, and a color distribution of each litchi leaf image is calculated based on a density estimation method;

[0018] According to the color distribution, the litchi leaf image data of each sub-region is divided into color depths, and the color information of the litchi leaves of each sub-region is determined according to the divided color depths to obtain the color characteristics of the litchi leaves of each sub-region;

[0019] Extracting leaf contours from the litchi leaf image data based on a contour recognition algorithm, estimating the number of litchi leaves in each sub-region according to the leaf contours, and obtaining a quantitative feature of the litchi leaves;

[0020] The color characteristics and quantity characteristics of the litchi leaves in each sub-region are integrated to obtain litchi leaf characteristic data.

[0021] In this solution, the clustering operation is performed on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result, which is specifically:

[0022] Quantifying the color features and quantity features in the litchi leaf feature data, scoring the color features and quantity features of the litchi leaves in each sub-region, and obtaining standardized numerical data of the color scores and quantity scores;

[0023] Introducing the K-means clustering algorithm, calculating the sum of square errors of the standardized color features and quantity features at different K values ​​based on the elbow method, and drawing a relationship graph between the K value and the sum of square errors, and determining the number of clusters K according to the relationship graph;

[0024] Randomly select K data points in the standardized numerical data as initial cluster centers, assign each data point in the standardized numerical data to the nearest cluster center, calculate the mean of each cluster, use the mean of the cluster as the new cluster center, repeat the assignment of data points and the update of cluster centers until the cluster centers no longer change, stop the clustering operation, obtain the clustering results, and assign a clustering label to each clustering result.

[0025] In this solution, the detection representative area of ​​the leaves damaged by litchi pedicle borers is determined according to the clustering results, and the real-time litchi leaf image data of the detection representative area in a preset time period is obtained, specifically:

[0026] Mapping the clustering labels of the clustering results to the corresponding sub-regions, and drawing the sub-regions after the clustering labels are mapped as a clustering label-sub-region two-dimensional plane diagram;

[0027] The ant colony algorithm is introduced to select only one of the sub-regions with the same cluster label as a condition restriction;

[0028] Based on the ant colony algorithm and conditional restrictions, the shortest cluster label connection path search is performed on the cluster label-sub-region two-dimensional plane graph to obtain a search result;

[0029] The sub-region corresponding to the cluster label in the search result is used as a representative detection region of leaves damaged by litchi peduncle borers;

[0030] Real-time litchi leaf image data of the detection representative area in a preset time period is obtained.

[0031] In this solution, the real-time litchi leaf image data is subjected to litchi pedicle borer damage symptom detection based on the deep learning algorithm to obtain litchi pedicle borer damage symptom change data for a preset time period, specifically:

[0032] Collecting a preset amount of historical litchi leaf image data, wherein the historical litchi leaf image data includes litchi leaves at different growth stages, and images of different types and degrees of litchi pedicle borer damage symptoms;

[0033] The historical litchi leaf image data are annotated, wherein the annotated content includes information on whether there are symptoms of litchi pedicle borer infestation, infestation sites, and infestation degree, and the annotated historical litchi leaf image data are divided into a training set and a test set according to a preset ratio;

[0034] Constructing a symptom detection model for litchi pedicle borer infestation based on a convolutional neural network, configuring the loss function, optimizer and evaluation index of the model, importing the training set into the model for training, and evaluating the accuracy and adjusting the model parameters of the trained model through a test set to obtain a trained symptom detection model for litchi pedicle borer infestation;

[0035] Importing the real-time litchi leaf image data of the preset time period into the trained litchi pedicle borer damage symptom detection model to detect litchi pedicle borer damage symptoms, and obtaining damage symptom detection results, wherein the damage symptom detection results include whether damage symptoms exist, the type and degree of damage symptoms, and the damage location;

[0036] The damage symptom detection results are sorted in chronological order to obtain the litchi peduncle borer damage symptom change data in a preset time period.

[0037] In this scheme, the damage degree development trend of litchi leaves is predicted based on the symptom change data of litchi peduncle borer, and the prediction result of litchi leaf damage degree is obtained, specifically:

[0038] A prediction model for changes in symptom damage caused by litchi pedicle borers is constructed based on the LSTM algorithm, and an input layer, an LSTM layer, a fully connected layer, an output layer, a loss function, and an optimizer of the prediction model are set, and data on changes in symptom damage caused by litchi pedicle borers in the preset time period are imported into the prediction model for learning and training;

[0039] Real-time data on changes in symptom damage of lychee stalk borer in the representative area detected in the current monitoring time period are obtained, and the real-time data on changes in symptom damage of lychee stalk borer are imported into the prediction model for changes in symptom damage of lychee stalk borer, and the development trend of the damage degree of lychee leaves in a future preset time period is predicted to obtain a prediction result of the damage degree of lychee leaves, which is the change in the damage degree, damage location, and pest spread path of lychee leaves in a future preset time period.

[0040] In this scheme, the control measures for litchi stem borer are formulated according to the prediction results of the damage degree of litchi leaves, specifically:

[0041] According to the litchi leaf damage degree prediction result, the litchi leaf damage risk, damage time period, and damage degree of each sub-region in a preset time period in the future are identified;

[0042] The control priority of litchi stalk borer in each sub-area is determined according to the damage risk and the damage time period, and the control means of litchi stalk borer is determined according to the damage degree, so as to obtain the control measures of litchi stalk borer.

[0043] The second aspect of the present invention further provides a deep learning-based litchi stalk borer leaf damage detection system, the system comprising: a memory, a processor, the memory comprising a deep learning-based litchi stalk borer leaf damage detection method program, the deep learning-based litchi stalk borer leaf damage detection method program when executed by the processor, the following steps are implemented:

[0044] Acquire litchi leaf image data from different litchi planting areas in a target season, extract litchi leaf features from different litchi planting areas according to the litchi leaf image data, and obtain litchi leaf feature data;

[0045] Performing a clustering operation on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result;

[0046] Determine a detection representative area of ​​leaves damaged by litchi peduncle borers according to the clustering result, and obtain real-time litchi leaf image data of the detection representative area in a preset time period;

[0047] Based on the deep learning algorithm, the real-time litchi leaf image data is used to detect the symptoms of litchi pedicle borer damage, and the litchi pedicle borer damage symptom change data in a preset time period is obtained;

[0048] Predicting the development trend of litchi leaf damage based on the symptom change data of litchi peduncle borer damage, and obtaining a prediction result of litchi leaf damage;

[0049] The control measures for litchi stalk borer are formulated according to the prediction results of the damage degree of litchi leaves.

[0050] In this scheme, the damage degree development trend of litchi leaves is predicted based on the symptom change data of litchi peduncle borer, and the prediction result of litchi leaf damage degree is obtained, specifically:

[0051] A prediction model for changes in symptom damage caused by litchi pedicle borers is constructed based on the LSTM algorithm, and an input layer, an LSTM layer, a fully connected layer, an output layer, a loss function, and an optimizer of the prediction model are set, and data on changes in symptom damage caused by litchi pedicle borers in the preset time period are imported into the prediction model for learning and training;

[0052] Real-time data on changes in symptom damage of lychee stalk borer in the representative area detected in the current monitoring time period are obtained, and the real-time data on changes in symptom damage of lychee stalk borer are imported into the prediction model for changes in symptom damage of lychee stalk borer, and the development trend of the damage degree of lychee leaves in a future preset time period is predicted to obtain a prediction result of the damage degree of lychee leaves, which is the change in the damage degree, damage location, and pest spread path of lychee leaves in a future preset time period.

[0053] In this scheme, the control measures for litchi stem borer are formulated according to the prediction results of the damage degree of litchi leaves, specifically:

[0054] According to the litchi leaf damage degree prediction result, the litchi leaf damage risk, damage time period, and damage degree of each sub-region in a preset time period in the future are identified;

[0055] The control priority of litchi stalk borer in each sub-area is determined according to the damage risk and the damage time period, and the control means of litchi stalk borer is determined according to the damage degree, so as to obtain the control measures of litchi stalk borer.

[0056] The present invention discloses a method and system for detecting leaves damaged by litchi stalk borers based on deep learning, comprising the following steps: first, obtaining litchi leaf image data of different litchi planting areas in the target season, and extracting features to obtain litchi leaf feature data. Then, clustering operations are performed on these feature data based on a clustering algorithm to determine the detection representative area of ​​leaves damaged by litchi stalk borers, and real-time litchi leaf image data of the area is obtained in a preset time period. Next, the real-time image data is used to detect damage symptoms using a deep learning algorithm to obtain damage symptom change data within a preset time period. Based on these change data, the damage development trend of litchi leaves is predicted, and finally a prediction result of the damage degree of litchi leaves is formed, and corresponding prevention and control measures are formulated. This method improves the accuracy of detection of leaves damaged by litchi stalk borers and the prevention and control effect by combining deep learning with a clustering algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flow chart of a method for detecting leaves damaged by litchi stem borers based on deep learning is shown in the present invention;

[0058] Figure 2 A flow chart showing the prediction results of litchi leaf damage degree obtained by the present invention;

[0059] Figure 3 A flow chart showing the control measures for litchi stem borer formulated by the present invention;

[0060] Figure 4 A block diagram of a deep learning-based litchi stem borer leaf detection system of the present invention is shown. DETAILED DESCRIPTION

[0061] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0063] Figure 1 A flow chart of a method for detecting leaves infested by litchi stem borers based on deep learning according to the present invention is shown.

[0064] like Figure 1 As shown, the first aspect of the present invention provides a method for detecting leaves damaged by litchi stem borers based on deep learning, comprising:

[0065] S102, acquiring litchi leaf image data from different litchi planting areas in a target season, and extracting litchi leaf features from different litchi planting areas according to the litchi leaf image data to obtain litchi leaf feature data;

[0066] S104, performing a clustering operation on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result;

[0067] S106, determining a detection representative area of ​​leaves damaged by litchi peduncle borers according to the clustering result, and obtaining real-time litchi leaf image data of the detection representative area in a preset time period;

[0068] S108, performing litchi pedicle borer damage symptom detection on the real-time litchi leaf image data based on a deep learning algorithm to obtain litchi pedicle borer damage symptom change data in a preset time period;

[0069] S110, predicting the development trend of the damage degree of litchi leaves according to the symptom change data of litchi peduncle borer, and obtaining the prediction result of the damage degree of litchi leaves;

[0070] S112, formulating prevention and control measures for litchi stem borer according to the prediction result of the damage degree of litchi leaves.

[0071] It should be noted that by acquiring litchi leaf image data from different litchi planting areas in the target season and extracting litchi leaf feature data for clustering operations, the representative detection area of ​​leaves infested by litchi stalk borers is determined based on the clustering results, and areas with similar leaf features can be classified into one category, thereby accurately locating leaf areas that may be infested by stalk borers, which helps to accurately locate the detection range. By determining the representative area, the detection focus can be concentrated on areas that may be infested by stalk borers, avoiding comprehensive detection of the entire planting area, thereby saving detection manpower and time costs; obtaining real-time litchi leaf image data of the representative detection area The data can be used to predict the development trend of the degree of damage to litchi leaves, and then determine the prevention and control measures for litchi stem borers. By predicting the development trend of leaf damage, the trend changes of stem borers damaging leaves can be discovered in advance, and timely warnings of possible leaf damage can be given. By understanding the development trend of leaf damage, the demand and distribution of resources required for prevention and control can be better assessed, and prevention and control resources can be allocated in a targeted manner to avoid waste of resources and repeated investment, and improve resource utilization efficiency. According to the prediction results, the most appropriate prevention and control methods and agents can be selected, and corresponding prevention and control measures can be taken according to different periods and degrees of leaf damage to improve the prevention and control effect and reduce the impact on the environment.

[0072] According to an embodiment of the present invention, the lychee leaf image data of different lychee planting areas in the target season is obtained, and the lychee leaf features of different lychee planting areas are extracted according to the lychee leaf image data to obtain the lychee leaf feature data, specifically:

[0073] The target litchi planting area is divided into N sub-areas, and the litchi leaf image data of each sub-area in the target season is obtained by high-definition camera equipment;

[0074] The litchi leaf image data is converted from the RGB color space to the HSV color space based on an image recognition algorithm, a statistical analysis is performed on the pixels of each litchi leaf image in the HSV color space, and a color distribution of each litchi leaf image is calculated based on a density estimation method;

[0075] According to the color distribution, the litchi leaf image data of each sub-region is divided into color depths, and the color information of the litchi leaves of each sub-region is determined according to the divided color depths to obtain the color characteristics of the litchi leaves of each sub-region;

[0076] Extracting leaf contours from the litchi leaf image data based on a contour recognition algorithm, estimating the number of litchi leaves in each sub-region according to the leaf contours, and obtaining a quantitative feature of the litchi leaves;

[0077] The color characteristics and quantity characteristics of the litchi leaves in each sub-region are integrated to obtain litchi leaf characteristic data.

[0078] It should be noted that, by using image recognition algorithms, features such as color and quantity of litchi leaves can be accurately extracted, avoiding the subjectivity and uncertainty of traditional manual feature extraction, ensuring the accuracy and stability of feature extraction, saving manpower and time costs, and being suitable for leaf feature extraction in large-scale litchi planting areas; the target season is autumn and winter; the color feature is the color information of litchi leaves, and the quantity feature is the estimated leaf quantity information; the image recognition algorithm includes a target detection algorithm and an image segmentation algorithm.

[0079] According to an embodiment of the present invention, the clustering operation is performed on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result, which is specifically:

[0080] Quantifying the color features and quantity features in the litchi leaf feature data, scoring the color features and quantity features of the litchi leaves in each sub-region, and obtaining standardized numerical data of the color scores and quantity scores;

[0081] Introducing the K-means clustering algorithm, calculating the sum of square errors of the standardized color features and quantity features at different K values ​​based on the elbow method, and drawing a relationship graph between the K value and the sum of square errors, and determining the number of clusters K according to the relationship graph;

[0082] Randomly select K data points in the standardized numerical data as initial cluster centers, assign each data point in the standardized numerical data to the nearest cluster center, calculate the mean of each cluster, use the mean of the cluster as the new cluster center, repeat the assignment of data points and the update of cluster centers until the cluster centers no longer change, stop the clustering operation, obtain the clustering results, and assign a clustering label to each clustering result.

[0083] It should be noted that, since it is necessary to perform clustering operations on the color features and quantity features of the litchi leaves in each sub-region at the same time, but the numerical ranges of the number of leaves and the color value are different, it is necessary to standardize the data. The standardization here means expressing the color features and quantity features with scores in the same range. The closer the color is to green, the higher the score is, and the closer to yellow and gray, the lower the score is; the more leaves, the higher the score is, and the fewer leaves, the lower the score is, so that the value range of all features is the same, preventing a certain feature from having too much influence on the clustering results; the K-means clustering algorithm can effectively divide the litchi leaf feature data into different categories or clusters, so that similar leaf features are classified into one category, for example, by analyzing the features of different clusters, a group with a large number of leaves and green color, or a group with a small number of leaves and yellow color is identified, and the cluster label is a category label such as a large number of leaves and green color, a small number of leaves and yellow color; the distance is the Manhattan distance between the data point and the cluster center; the clustering result classifies sub-regions with similar litchi leaf color features and quantity features into one category.

[0084] According to an embodiment of the present invention, determining the detection representative area of ​​the leaves damaged by litchi pedicle borers according to the clustering result, and obtaining the real-time litchi leaf image data of the detection representative area in a preset time period, specifically:

[0085] Mapping the clustering labels of the clustering results to the corresponding sub-regions, and drawing the sub-regions after the clustering labels are mapped as a clustering label-sub-region two-dimensional plane diagram;

[0086] The ant colony algorithm is introduced to select only one of the sub-regions with the same cluster label as a condition restriction;

[0087] Based on the ant colony algorithm and conditional restrictions, the shortest cluster label connection path search is performed on the cluster label-sub-region two-dimensional plane graph to obtain a search result;

[0088] The sub-region corresponding to the cluster label in the search result is used as a representative detection region of leaves damaged by litchi peduncle borers;

[0089] Real-time litchi leaf image data of the detection representative area in a preset time period is obtained.

[0090] It should be noted that, since the clustering result is the result of clustering the color characteristics and quantity characteristics of the lychee leaves in each sub-region, and each color characteristic and quantity characteristic exists in different sub-regions, clustering the color characteristics and quantity characteristics is to cluster the sub-regions according to the color characteristics and quantity characteristics, and the clustering label is the clustering label of each sub-region, and the clustering label-sub-region two-dimensional plane map is a two-dimensional plane map formed by drawing the target lychee planting area as a two-dimensional plane map and dividing the sub-regions, and then marking the clustering label to each corresponding sub-region; since the planting range of the target lychee planting area is large, the lychees in many planting areas have similar leaf growth characteristics, and areas with similar leaf growth characteristics generally have similar lychee stalk borers. The degree of damage is similar. Therefore, if all sub-areas are inspected for leaves damaged by litchi stalk borers, a large number of redundant inspections will occur, wasting inspection resources. Therefore, by selecting only one of the sub-areas with the same clustering label as the inspection area, and then searching the inspection area with the shortest connection path in the target litchi planting area through the ant colony algorithm as the inspection representative area, the consumption of human resources is reduced, and the waste of human resources due to the long distance between the inspection areas is avoided. The shortest clustering label connection path is the shortest path formed in the clustering label-sub-area two-dimensional plane diagram by connecting only one of the sub-areas with the same clustering label. The shortest clustering label connection path includes all clustering class labels, and each clustering class label is not repeated in the connection path.

[0091] According to an embodiment of the present invention, the real-time litchi leaf image data is subjected to litchi pedicle borer damage symptom detection based on a deep learning algorithm to obtain litchi pedicle borer damage symptom change data for a preset time period, specifically:

[0092] Collecting a preset amount of historical litchi leaf image data, wherein the historical litchi leaf image data includes litchi leaves at different growth stages, and images of different types and degrees of litchi pedicle borer damage symptoms;

[0093] The historical litchi leaf image data are annotated, wherein the annotated content includes information on whether there are symptoms of litchi pedicle borer infestation, infestation sites, and infestation degree, and the annotated historical litchi leaf image data are divided into a training set and a test set according to a preset ratio;

[0094] Constructing a symptom detection model for litchi pedicle borer infestation based on a convolutional neural network, configuring the loss function, optimizer and evaluation index of the model, importing the training set into the model for training, and evaluating the accuracy and adjusting the model parameters of the trained model through a test set to obtain a trained symptom detection model for litchi pedicle borer infestation;

[0095] Importing the real-time litchi leaf image data of the preset time period into the trained litchi pedicle borer damage symptom detection model to detect litchi pedicle borer damage symptoms, and obtaining damage symptom detection results, wherein the damage symptom detection results include whether damage symptoms exist, the type and degree of damage symptoms, and the damage location;

[0096] The damage symptom detection results are sorted in chronological order to obtain the litchi peduncle borer damage symptom change data in a preset time period.

[0097] It should be noted that by constructing a litchi stalk borer infestation symptom detection model through a convolutional neural network and training the model based on historical litchi leaf image data, the deep learning model can be optimized and adjusted to improve its accuracy and stability in identifying borer infestation symptoms in litchi leaf image data; through an automated borer-infested leaf detection method, labor costs and time costs can be reduced, production efficiency can be improved, and borer infestation symptoms on litchi leaves can be detected in a timely and accurate manner.

[0098] Figure 2 The flowchart of the present invention for obtaining the prediction result of the damage degree of litchi leaves is shown.

[0099] According to an embodiment of the present invention, the prediction result of the damage degree of litchi leaves is obtained by predicting the development trend of the damage degree of litchi leaves according to the symptom change data of the litchi peduncle borer, specifically:

[0100] S202, constructing a prediction model for changes in symptom damage caused by litchi pedicle borer based on the LSTM algorithm, setting the input layer, LSTM layer, fully connected layer and output layer, loss function and optimizer of the prediction model, and importing the data on changes in symptom damage caused by litchi pedicle borer in the preset time period into the prediction model for learning and training;

[0101] S204, obtaining real-time data on changes in symptom damage of litchi stalk borer in the representative area detected in the current monitoring time period, importing the real-time data on changes in symptom damage of litchi stalk borer into the prediction model for changes in symptom damage of litchi stalk borer, predicting the development trend of the damage degree of litchi leaves in a future preset time period, and obtaining a prediction result of the damage degree of litchi leaves, which is the change in the damage degree, damage location, and pest spread path of litchi leaves in a future preset time period.

[0102] It should be noted that the prediction model for changes in symptom damage of litchi stalk borer constructed based on the LSTM algorithm can predict the development trend of the damage degree of litchi leaves. By learning and training the damage symptom change data of a preset time period, the model can accurately predict the changes in the damage degree of litchi leaves, the changes in the damage location and the changes in the pest spread path within the future preset time period; by predicting the development trend of the damage degree of litchi leaves, the planting structure can be adjusted according to the prediction results, and the agricultural production layout can be optimized, thereby improving the efficiency and economic benefits of agricultural production; the LSTM algorithm is a long short-term memory network (Long Short-Term Memory).

[0103] Figure 3 The flowchart of the present invention for formulating control measures for litchi stem borer is shown.

[0104] According to an embodiment of the present invention, the control measures for litchi peduncle borer are formulated according to the prediction result of the damage degree of litchi leaves, specifically:

[0105] S302, identifying the litchi leaf damage risk, damage time period, and damage degree of each sub-region in a preset time period in the future according to the litchi leaf damage degree prediction result;

[0106] S304, determining the control priority of litchi stalk borer in each sub-area according to the damage risk and damage time period, determining the control means of litchi stalk borer according to the damage degree, and obtaining control measures of litchi stalk borer.

[0107] It should be noted that based on the prediction results of the damage degree of litchi leaves, the risk of litchi leaf damage degree and the damage time period of each sub-region in the preset time period in the future are determined through the information obtained from the prediction model, including which areas may be affected by the stalk borer and when the damage is most serious. According to the damage risk and time period, the priority of litchi stalk borer control in each sub-region is determined. For those areas with higher risks in the future time period, priority should be given to the implementation of prevention and control measures to minimize losses; the means of prevention and control of litchi stalk borer are determined according to the degree of damage. Different degrees of damage may require different prevention and control methods. For example, for lightly damaged areas, biological control or cultural control methods can be adopted, while for severely damaged areas, more powerful means such as chemical control or mechanical control may be required.

[0108] Figure 4 A block diagram of a deep learning-based litchi stem borer leaf detection system of the present invention is shown.

[0109] The second aspect of the present invention further provides a deep learning-based litchi stalk borer leaf damage detection system 4, the system comprising: a memory 41, a processor 42, the memory comprising a deep learning-based litchi stalk borer leaf damage detection method program, the deep learning-based litchi stalk borer leaf damage detection method program when executed by the processor, the following steps are implemented:

[0110] Acquire litchi leaf image data from different litchi planting areas in a target season, extract litchi leaf features from different litchi planting areas according to the litchi leaf image data, and obtain litchi leaf feature data;

[0111] Performing a clustering operation on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result;

[0112] Determine a detection representative area of ​​leaves damaged by litchi peduncle borers according to the clustering result, and obtain real-time litchi leaf image data of the detection representative area in a preset time period;

[0113] Based on the deep learning algorithm, the real-time litchi leaf image data is used to detect the symptoms of litchi pedicle borer damage, and the litchi pedicle borer damage symptom change data in a preset time period is obtained;

[0114] Predicting the development trend of litchi leaf damage based on the symptom change data of litchi peduncle borer damage, and obtaining a prediction result of litchi leaf damage;

[0115] The control measures for litchi stalk borer are formulated according to the prediction results of the damage degree of litchi leaves.

[0116] According to an embodiment of the present invention, the lychee leaf image data of different lychee planting areas in the target season is obtained, and the lychee leaf features of different lychee planting areas are extracted according to the lychee leaf image data to obtain the lychee leaf feature data, specifically:

[0117] The target litchi planting area is divided into N sub-areas, and the litchi leaf image data of each sub-area in the target season is obtained by high-definition camera equipment;

[0118] The litchi leaf image data is converted from the RGB color space to the HSV color space based on an image recognition algorithm, a statistical analysis is performed on the pixels of each litchi leaf image in the HSV color space, and a color distribution of each litchi leaf image is calculated based on a density estimation method;

[0119] According to the color distribution, the litchi leaf image data of each sub-region is divided into color depths, and the color information of the litchi leaves of each sub-region is determined according to the divided color depths to obtain the color characteristics of the litchi leaves of each sub-region;

[0120] Extracting leaf contours from the litchi leaf image data based on a contour recognition algorithm, estimating the number of litchi leaves in each sub-region according to the leaf contours, and obtaining a quantitative feature of the litchi leaves;

[0121] The color characteristics and quantity characteristics of the litchi leaves in each sub-region are integrated to obtain litchi leaf characteristic data.

[0122] According to an embodiment of the present invention, the clustering operation is performed on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result, which is specifically:

[0123] Quantifying the color features and quantity features in the litchi leaf feature data, scoring the color features and quantity features of the litchi leaves in each sub-region, and obtaining standardized numerical data of the color scores and quantity scores;

[0124] Introducing the K-means clustering algorithm, calculating the sum of square errors of the standardized color features and quantity features at different K values ​​based on the elbow method, and drawing a relationship graph between the K value and the sum of square errors, and determining the number of clusters K according to the relationship graph;

[0125] Randomly select K data points in the standardized numerical data as initial cluster centers, assign each data point in the standardized numerical data to the nearest cluster center, calculate the mean of each cluster, use the mean of the cluster as the new cluster center, repeat the assignment of data points and the update of cluster centers until the cluster centers no longer change, stop the clustering operation, obtain the clustering results, and assign a clustering label to each clustering result.

[0126] According to an embodiment of the present invention, determining the detection representative area of ​​the leaves damaged by litchi pedicle borers according to the clustering result, and obtaining the real-time litchi leaf image data of the detection representative area in a preset time period, specifically:

[0127] Mapping the clustering labels of the clustering results to the corresponding sub-regions, and drawing the sub-regions after the clustering labels are mapped as a clustering label-sub-region two-dimensional plane diagram;

[0128] The ant colony algorithm is introduced to select only one of the sub-regions with the same cluster label as a condition restriction;

[0129] Based on the ant colony algorithm and conditional restrictions, the shortest cluster label connection path search is performed on the cluster label-sub-region two-dimensional plane graph to obtain a search result;

[0130] The sub-region corresponding to the cluster label in the search result is used as a representative detection region of leaves damaged by litchi peduncle borers;

[0131] Real-time litchi leaf image data of the detection representative area in a preset time period is obtained.

[0132] According to an embodiment of the present invention, the real-time litchi leaf image data is subjected to litchi pedicle borer damage symptom detection based on a deep learning algorithm to obtain litchi pedicle borer damage symptom change data for a preset time period, specifically:

[0133] Collecting a preset amount of historical litchi leaf image data, wherein the historical litchi leaf image data includes litchi leaves at different growth stages, and images of different types and degrees of litchi pedicle borer damage symptoms;

[0134] The historical litchi leaf image data are annotated, wherein the annotated content includes information on whether there are symptoms of litchi pedicle borer infestation, infestation sites, and infestation degree, and the annotated historical litchi leaf image data are divided into a training set and a test set according to a preset ratio;

[0135] Constructing a symptom detection model for litchi pedicle borer infestation based on a convolutional neural network, configuring the loss function, optimizer and evaluation index of the model, importing the training set into the model for training, and evaluating the accuracy and adjusting the model parameters of the trained model through a test set to obtain a trained symptom detection model for litchi pedicle borer infestation;

[0136] Importing the real-time litchi leaf image data of the preset time period into the trained litchi pedicle borer damage symptom detection model to detect litchi pedicle borer damage symptoms, and obtaining damage symptom detection results, wherein the damage symptom detection results include whether damage symptoms exist, the type and degree of damage symptoms, and the damage location;

[0137] The damage symptom detection results are sorted in chronological order to obtain the litchi peduncle borer damage symptom change data in a preset time period.

[0138] According to an embodiment of the present invention, the prediction result of the damage degree of litchi leaves is obtained by predicting the development trend of the damage degree of litchi leaves according to the symptom change data of the litchi peduncle borer, specifically:

[0139] A prediction model for changes in symptom damage caused by litchi pedicle borers is constructed based on the LSTM algorithm, and an input layer, an LSTM layer, a fully connected layer, an output layer, a loss function, and an optimizer of the prediction model are set, and data on changes in symptom damage caused by litchi pedicle borers in the preset time period are imported into the prediction model for learning and training;

[0140] Real-time data on changes in symptom damage of lychee stalk borer in the representative area detected in the current monitoring time period are obtained, and the real-time data on changes in symptom damage of lychee stalk borer are imported into the prediction model for changes in symptom damage of lychee stalk borer, and the development trend of the damage degree of lychee leaves in a future preset time period is predicted to obtain a prediction result of the damage degree of lychee leaves, which is the change in the damage degree, damage location, and pest spread path of lychee leaves in a future preset time period.

[0141] According to an embodiment of the present invention, the control measures for litchi peduncle borer are formulated according to the prediction result of the damage degree of litchi leaves, specifically:

[0142] According to the litchi leaf damage degree prediction result, the litchi leaf damage risk, damage time period, and damage degree of each sub-region in a preset time period in the future are identified;

[0143] The control priority of litchi stalk borer in each sub-area is determined according to the damage risk and the damage time period, and the control means of litchi stalk borer is determined according to the damage degree, so as to obtain the control measures of litchi stalk borer.

[0144] The present invention discloses a method and system for detecting leaves infested by litchi stalk borers based on deep learning, comprising the following steps: first, acquiring litchi leaf image data of different litchi planting areas in the target season, and performing feature extraction to obtain litchi leaf feature data. Then, clustering operations are performed on these feature data based on a clustering algorithm to determine the representative detection area of ​​leaves infested by litchi stalk borers, and real-time litchi leaf image data of the area is obtained in a preset time period. Next, damage symptoms are detected on the real-time image data using a deep learning algorithm to obtain damage symptom change data within a preset time period. Based on these change data, the damage development trend of litchi leaves is predicted, and finally a prediction result of the damage degree of litchi leaves is formed, and corresponding prevention and control measures are formulated. This method improves the accuracy of detection of leaves infested by litchi stalk borers and the prevention and control effect by combining deep learning with a clustering algorithm.

[0145] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0146] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0147] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0148] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks or optical disks, and other media that can store program codes.

[0149] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0150] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting leaves damaged by litchi stem borers based on deep learning, characterized in that: The following steps are involved: Acquire litchi leaf image data from different litchi planting areas in a target season, extract litchi leaf features from different litchi planting areas according to the litchi leaf image data, and obtain litchi leaf feature data; Performing a clustering operation on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result; Determine a detection representative area of ​​leaves damaged by litchi peduncle borers according to the clustering result, and obtain real-time litchi leaf image data of the detection representative area in a preset time period; Based on the deep learning algorithm, the real-time litchi leaf image data is used to detect the symptoms of litchi pedicle borer damage, and the litchi pedicle borer damage symptom change data in a preset time period is obtained; Predicting the development trend of litchi leaf damage based on the symptom change data of litchi peduncle borer damage, and obtaining a prediction result of litchi leaf damage; Formulate prevention and control measures for litchi stem borer according to the prediction results of the damage degree of litchi leaves; The acquiring of litchi leaf image data from different litchi planting areas in the target season, and extracting litchi leaf features from different litchi planting areas according to the litchi leaf image data to obtain litchi leaf feature data, specifically comprises: The target litchi planting area is divided into N sub-areas, and the litchi leaf image data of each sub-area in the target season is obtained by high-definition camera equipment; The litchi leaf image data is converted from the RGB color space to the HSV color space based on an image recognition algorithm, a statistical analysis is performed on the pixels of each litchi leaf image in the HSV color space, and a color distribution of each litchi leaf image is calculated based on a density estimation method; According to the color distribution, the litchi leaf image data of each sub-region is divided into color depths, and the color information of the litchi leaves of each sub-region is determined according to the divided color depths to obtain the color characteristics of the litchi leaves of each sub-region; Extracting leaf contours from the litchi leaf image data based on a contour recognition algorithm, estimating the number of litchi leaves in each sub-region according to the leaf contours, and obtaining a quantitative feature of the litchi leaves; Integrating the color characteristics and quantity characteristics of the litchi leaves in each sub-region to obtain litchi leaf characteristic data; The clustering operation is performed on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result, which is specifically: Quantifying the color features and quantity features in the litchi leaf feature data, scoring the color features and quantity features of the litchi leaves in each sub-region, and obtaining standardized numerical data of the color scores and quantity scores; The K-means clustering algorithm is introduced, and the sum of square errors of standardized color features and quantity features under different K values ​​is calculated based on the elbow method, and a relationship diagram between the K value and the sum of square errors is drawn, and the number of clusters K is determined according to the relationship diagram; Randomly select K data points in the standardized numerical data as initial cluster centers, assign each data point in the standardized numerical data to the nearest cluster center, calculate the mean of each cluster, use the mean of the cluster as the new cluster center, repeatedly assign data points and update cluster centers until the cluster centers no longer change, stop clustering, obtain clustering results, and assign cluster labels to each cluster result; The prediction result of the damage degree of litchi leaves is obtained by predicting the development trend of the damage degree of litchi leaves according to the change data of the symptoms of litchi peduncle borer, specifically: A prediction model for changes in symptom damage caused by litchi pedicle borers is constructed based on the LSTM algorithm, and an input layer, an LSTM layer, a fully connected layer, an output layer, a loss function, and an optimizer of the prediction model are set, and data on changes in symptom damage caused by litchi pedicle borers in the preset time period are imported into the prediction model for learning and training; Acquire real-time lychee stalk borer infestation symptom change data of the detection representative area in the current monitoring time period, import the real-time lychee stalk borer infestation symptom change data into the lychee stalk borer infestation symptom change prediction model, predict the development trend of lychee leaf damage degree in a future preset time period, and obtain a lychee leaf damage degree prediction result, wherein the lychee leaf damage degree prediction result is a change in the damage degree of the lychee leaves, a change in the damage location, and a change in the pest diffusion path in the future preset time period; The method of determining the detection representative area of ​​the leaves damaged by litchi peduncle borers according to the clustering result and obtaining the real-time litchi leaf image data of the detection representative area in a preset time period is specifically as follows: Mapping the clustering labels of the clustering results to the corresponding sub-regions, and drawing the sub-regions after the clustering labels are mapped as a clustering label-sub-region two-dimensional plane diagram; The ant colony algorithm is introduced to select only one of the sub-regions with the same cluster label as a condition restriction; Based on the ant colony algorithm and conditional restrictions, the shortest cluster label connection path search is performed on the cluster label-sub-region two-dimensional plane graph to obtain a search result; The sub-region corresponding to the cluster label in the search result is used as a representative detection region of leaves damaged by litchi peduncle borers; Real-time litchi leaf image data of the detection representative area in a preset time period is obtained.

2. A method for detecting leaves damaged by litchi peduncle borers based on deep learning according to claim 1, characterized in that: The real-time litchi leaf image data is subjected to litchi pedicle borer damage symptom detection based on a deep learning algorithm to obtain litchi pedicle borer damage symptom change data for a preset time period, specifically: Collecting a preset amount of historical litchi leaf image data, wherein the historical litchi leaf image data includes litchi leaves at different growth stages, and images of different types and degrees of litchi pedicle borer damage symptoms; The historical litchi leaf image data are annotated, wherein the annotated content includes information on whether there are symptoms of litchi pedicle borer infestation, infestation sites, and infestation degree, and the annotated historical litchi leaf image data are divided into a training set and a test set according to a preset ratio; Constructing a symptom detection model for litchi pedicle borer infestation based on a convolutional neural network, configuring the loss function, optimizer and evaluation index of the model, importing the training set into the model for training, and evaluating the accuracy and adjusting the model parameters of the trained model through a test set to obtain a trained symptom detection model for litchi pedicle borer infestation; Importing the real-time litchi leaf image data of the preset time period into the trained litchi pedicle borer damage symptom detection model to detect litchi pedicle borer damage symptoms, and obtaining damage symptom detection results, wherein the damage symptom detection results include whether damage symptoms exist, the type and degree of damage symptoms, and the damage location; The damage symptom detection results are sorted in chronological order to obtain the litchi peduncle borer damage symptom change data in a preset time period.

3. A method for detecting leaves damaged by litchi peduncle borers based on deep learning according to claim 1, characterized in that: The control measures for litchi peduncle borer are formulated according to the prediction results of the damage degree of litchi leaves, specifically: According to the litchi leaf damage degree prediction result, the litchi leaf damage risk, damage time period, and damage degree of each sub-region in a preset time period in the future are identified; The control priority of litchi stalk borer in each sub-area is determined according to the damage risk and the damage time period, and the control means of litchi stalk borer is determined according to the damage degree, so as to obtain the control measures of litchi stalk borer.

4. A deep learning-based detection system for litchi pedicle borer-damaged leaves, characterized in that: The deep learning-based litchi stalk borer-damaged leaf detection system comprises a storage device and a processor, wherein the storage device comprises a deep learning-based litchi stalk borer-damaged leaf detection method program, and when the deep learning-based litchi stalk borer-damaged leaf detection method program is executed by the processor, the following steps are implemented: Acquire litchi leaf image data from different litchi planting areas in a target season, extract litchi leaf features from different litchi planting areas according to the litchi leaf image data, and obtain litchi leaf feature data; Performing a clustering operation on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result; Determine a detection representative area of ​​leaves damaged by litchi peduncle borers according to the clustering result, and obtain real-time litchi leaf image data of the detection representative area in a preset time period; Based on the deep learning algorithm, the real-time litchi leaf image data is used to detect the symptoms of litchi pedicle borer damage, and the litchi pedicle borer damage symptom change data in a preset time period is obtained; Predicting the development trend of litchi leaf damage based on the symptom change data of litchi peduncle borer damage, and obtaining a prediction result of litchi leaf damage; Formulate prevention and control measures for litchi stem borer according to the prediction results of the damage degree of litchi leaves; The acquiring of litchi leaf image data from different litchi planting areas in the target season, and extracting litchi leaf features from different litchi planting areas according to the litchi leaf image data to obtain litchi leaf feature data, specifically comprises: The target litchi planting area is divided into N sub-areas, and the litchi leaf image data of each sub-area in the target season is obtained by high-definition camera equipment; The litchi leaf image data is converted from the RGB color space to the HSV color space based on an image recognition algorithm, a statistical analysis is performed on the pixels of each litchi leaf image in the HSV color space, and a color distribution of each litchi leaf image is calculated based on a density estimation method; According to the color distribution, the litchi leaf image data of each sub-region is divided into color depths, and the color information of the litchi leaves of each sub-region is determined according to the divided color depths to obtain the color characteristics of the litchi leaves of each sub-region; Extracting leaf contours from the litchi leaf image data based on a contour recognition algorithm, estimating the number of litchi leaves in each sub-region according to the leaf contours, and obtaining a quantitative feature of the litchi leaves; Integrating the color characteristics and quantity characteristics of the litchi leaves in each sub-region to obtain litchi leaf characteristic data; The clustering operation is performed on the litchi leaf feature data based on a clustering algorithm to obtain a clustering result, which is specifically: Quantifying the color features and quantity features in the litchi leaf feature data, scoring the color features and quantity features of the litchi leaves in each sub-region, and obtaining standardized numerical data of the color scores and quantity scores; The K-means clustering algorithm is introduced, and the sum of square errors of standardized color features and quantity features under different K values ​​is calculated based on the elbow method, and a relationship diagram between the K value and the sum of square errors is drawn, and the number of clusters K is determined according to the relationship diagram; Randomly select K data points in the standardized numerical data as initial cluster centers, assign each data point in the standardized numerical data to the nearest cluster center, calculate the mean of each cluster, use the mean of the cluster as the new cluster center, repeatedly assign data points and update cluster centers until the cluster centers no longer change, stop clustering, obtain clustering results, and assign cluster labels to each cluster result; The prediction result of the damage degree of litchi leaves is obtained by predicting the development trend of the damage degree of litchi leaves according to the change data of the symptoms of litchi peduncle borer, specifically: A prediction model for changes in symptom damage caused by litchi pedicle borers is constructed based on the LSTM algorithm, and an input layer, an LSTM layer, a fully connected layer, an output layer, a loss function, and an optimizer of the prediction model are set, and data on changes in symptom damage caused by litchi pedicle borers in the preset time period are imported into the prediction model for learning and training; Acquire real-time lychee stalk borer infestation symptom change data of the detection representative area in the current monitoring time period, import the real-time lychee stalk borer infestation symptom change data into the lychee stalk borer infestation symptom change prediction model, predict the development trend of lychee leaf damage degree in a future preset time period, and obtain a lychee leaf damage degree prediction result, wherein the lychee leaf damage degree prediction result is a change in the damage degree of the lychee leaves, a change in the damage location, and a change in the pest diffusion path in the future preset time period; The method of determining the detection representative area of ​​the leaves damaged by litchi peduncle borers according to the clustering result and obtaining the real-time litchi leaf image data of the detection representative area in a preset time period is specifically as follows: Mapping the clustering labels of the clustering results to the corresponding sub-regions, and drawing the sub-regions after the clustering labels are mapped as a clustering label-sub-region two-dimensional plane diagram; The ant colony algorithm is introduced to select only one of the sub-regions with the same cluster label as a condition restriction; Based on the ant colony algorithm and conditional restrictions, the shortest cluster label connection path search is performed on the cluster label-sub-region two-dimensional plane graph to obtain a search result; The sub-region corresponding to the cluster label in the search result is used as a representative detection region of leaves damaged by litchi peduncle borers; Real-time litchi leaf image data of the detection representative area in a preset time period is obtained.

5. A deep learning-based detection system for litchi stem borer-damaged leaves according to claim 4, characterized in that: The control measures for litchi peduncle borer are formulated according to the prediction results of the damage degree of litchi leaves, specifically: According to the litchi leaf damage degree prediction result, the litchi leaf damage risk, damage time period, and damage degree of each sub-region in a preset time period in the future are identified; The control priority of litchi stalk borer in each sub-area is determined according to the damage risk and the damage time period, and the control means of litchi stalk borer is determined according to the damage degree, so as to obtain the control measures of litchi stalk borer.

Images