Forestry disease and pest monitoring method and system based on machine learning

Abstract

The invention provides a forestry pest monitoring method and system based on machine learning, and the method comprises the steps: carrying out the principal component analysis of a forestry pest image, obtaining a plurality of recognition indexes of the pest of a target forestry region, determining an effective feature vector when the image recognition of the forestry pest is carried out according to all recognition indexes, and carrying out the recognition of the effective feature vector. Determining a first-level correlation feature of the pest image in the target forestry region through the effective feature vector; extracting a plurality of environmental influence factors from the monitoring image, and determining secondary correlation features of the pest image in the target forestry area according to all the environmental influence factors; carrying out interactive fusion on the first-level correlation features and the second-level correlation features to obtain multi-level distribution features for carrying out image recognition on the diseases and insect pests; and identifying the current pest and disease distribution of the target forestry area based on the multi-level distribution characteristics through a pest and disease identification model. By adopting the scheme of the application, diversified monitoring of the forestry diseases and insect pests can be realized under a complex background, so that the false detection rate of the forestry diseases and insect pests is reduced.

Description

Technical Field

[0001] The present application relates to the field of machine learning technology, and more specifically, to a forestry pest and disease monitoring method and system based on machine learning. Background Art

[0002] Forest pests and diseases have a significant impact on the health and productivity of forest ecosystems. They can lead to tree death, biodiversity decline and economic losses. Timely detection and control of pests and diseases are the key to protecting forest resources and ecological balance. Machine learning can not only accurately identify and predict pests and diseases in forests, but also provide efficient and accurate auxiliary decision-making in large-scale forest management, thereby significantly improving the management efficiency of forestry resources and reducing the economic losses caused by pests and diseases.

[0003] Monitoring of forest pests and diseases based on machine learning mainly collects and analyzes data in forest ecosystems (such as image data, sensor data, environmental data, etc.), and then uses machine learning models to automatically identify and predict the occurrence of pests and diseases. Although monitoring based on machine learning has improved the efficiency of pest and disease identification to a certain extent, it still faces challenges such as data quality and model generalization ability. In existing technologies, the types and appearance characteristics of forest pests and diseases are diverse and variable. For example, a pest will show different forms in different growth stages or seasons, and the emergence of foreign pests and diseases will introduce new types, which are not included in existing models. The diversity and variability of characteristic data of different types of pests and diseases place high generalization requirements on machine learning models, making it difficult for existing models to effectively adapt to new types and changes of pests and diseases, and prone to missed detections or false detections. In addition, due to the complex backgrounds, diverse types of pests and diseases, and unstable quality of monitoring images faced by forestry areas, the accuracy and comprehensiveness of identifying different types of pests and diseases through monitoring images are insufficient, making it difficult to fully identify the various types of pests and diseases in forestry. Therefore, how to achieve diversified monitoring of forest pests and diseases under complex backgrounds, thereby reducing the false detection rate of forest pests and diseases, has become a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides a forestry pest and disease monitoring method and system based on machine learning, which can realize diversified monitoring of forestry pests and diseases under complex backgrounds, thereby reducing the false detection rate of forestry pests and diseases.

[0005] In a first aspect, the present application provides a forestry pest monitoring method based on machine learning, comprising the following steps: Obtain historical forest pest and disease images of the target forestry area; Based on the machine learning model, principal component analysis is performed on the historical forestry pest and disease images to obtain multiple identification indicators of pests and diseases in the target forestry area, and effective feature vectors are determined when performing image recognition on forestry pests and diseases according to all the identification indicators, and then the primary correlation features of the pest and disease images in the target forestry area are determined through the effective feature vectors; Conduct real-time monitoring of the target forestry area, collect current monitoring images of the target forestry area, extract multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area based on the perception scene of the monitoring image, and then determine the secondary correlation features of the pest and disease images in the target forestry area from all the environmental influencing factors; The primary correlation features and the secondary correlation features are interactively integrated by using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in a target forestry area; A pest and disease identification model is pre-trained through machine learning, and the pest and disease identification model identifies the current pest and disease distribution in the target forestry area based on the multi-level distribution characteristics.

[0006] In some embodiments, principal component analysis is performed on the historical forestry pest and disease images based on a machine learning model to obtain multiple identification indicators of pests and diseases in the target forestry area, specifically including: For each pest and disease image in the historical forestry pest and disease images; Preprocessing the pest and disease image to obtain a preprocessed pest and disease image; Determine the multi-dimensional features of forestry pests and diseases in the pest and disease images based on the preprocessed pest and disease images through a machine learning model, and then obtain the multi-dimensional features of forestry pests and diseases in each pest and disease image; A plurality of principal components of the historical forestry pest and disease images are extracted from all the multi-dimensional features, and the extracted principal components are then used as identification indicators of the pest and disease in the target forestry area.

[0007] In some embodiments, determining effective feature vectors for image recognition of forest pests and diseases based on all recognition indicators specifically includes: Get the variance of the principal component corresponding to each identification index; Determine the variance contribution corresponding to each identification indicator through the variance of the principal component corresponding to each identification indicator; The effective feature vector for image recognition of forest pests and diseases is determined based on all variance contributions.

[0008] In some embodiments, determining the primary correlation features of the pest and disease images in the target forestry area through the effective feature vector specifically includes: Determine the correlation between each identification index in the effective feature vector and the distribution of forest pests and diseases; The first-level correlation features of pest and disease images in the target forestry area are determined based on all correlations.

[0009] In some embodiments, the multiple environmental influencing factors extracted during the pest and disease monitoring process in the target forestry area based on the perceived scene of the monitoring image specifically include: Segmenting the monitoring image into different perception areas based on the perception scene of the monitoring image; Extract pixel-level features of each perception area; By combining all pixel-level features with external environmental data, multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area are determined.

[0010] In some embodiments, the first-level association features and the second-level association features are interactively integrated using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in a target forestry area, specifically including: Determine the interaction rate between the first-level association feature and the second-level association feature by using a tree model; The primary association feature and the secondary association feature are fused based on the interaction rate to obtain a multi-level distribution feature for image recognition of pests and diseases in a target forestry area.

[0011] In some embodiments, the pest identification model identifies the current pest distribution in the target forestry area based on the multi-level distribution characteristics, specifically including: Using the current monitoring image of the target forestry area as the input of the pest and disease identification model; The pest and disease identification model is used to confidently identify the pests and diseases in the target forestry area based on the multi-level distribution characteristics to obtain the current pest and disease distribution results of the target forestry area.

[0012] In a second aspect, the present application provides a forestry pest monitoring system based on machine learning, comprising: An acquisition module is used to acquire historical forestry pest and disease images of the target forestry area; A processing module, for performing principal component analysis on the historical forestry pest and disease images based on a machine learning model to obtain multiple identification indicators of pests and diseases in a target forestry area, determining effective feature vectors for image recognition of forestry pests and diseases based on all the identification indicators, and further determining primary correlation features of the pest and disease images in the target forestry area through the effective feature vectors; The processing module is also used to monitor the target forestry area in real time, collect the current monitoring image of the target forestry area, extract multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area based on the perception scene of the monitoring image, and then determine the secondary correlation features of the pest and disease image in the target forestry area from all the environmental influencing factors; The processing module is further used to interactively fuse the primary association features and the secondary association features using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in the target forestry area; An execution module is used to pre-train a pest and disease monitoring model through machine learning, and the pest and disease monitoring model identifies the current pest and disease distribution in the target forestry area based on the multi-level distribution characteristics.

[0013] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the above-mentioned forestry pest and disease monitoring method based on machine learning.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium, in which instructions or codes are stored. When the instructions or codes are run on a computer, the computer implements the above-mentioned forestry pest and disease monitoring method based on machine learning when executing.

[0015] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects: In the present application, historical forestry pest and disease images of a target forestry area are obtained; principal component analysis is performed on the historical forestry pest and disease images based on a machine learning model to obtain multiple identification indicators of pests and diseases in the target forestry area, and effective feature vectors for image recognition of forestry pests and diseases are determined based on all the identification indicators, and then the primary correlation features of the pest and disease images in the target forestry area are determined through the effective feature vectors; the target forestry area is monitored in real time, and the current monitoring images of the target forestry area are collected, and multiple environmental influencing factors in the pest and disease monitoring process of the target forestry area are extracted based on the perceived scenes of the monitoring images, and then the secondary correlation features of the pest and disease images in the target forestry area are determined from all the ring influencing factors; the primary correlation features and the secondary correlation features are interactively fused using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in the target forestry area; a pest and disease recognition model is pre-trained through machine learning, and the pest and disease recognition model identifies the current distribution of pests and diseases in the target forestry area based on the multi-level distribution features.

[0016] It can be seen that in this application, firstly, the primary correlation features of the pest and disease images in the target forestry area are determined by the effective feature vectors when performing image recognition on forestry pests and diseases, which can help the pest and disease recognition model to better classify and identify forestry pests and diseases, and improve the accuracy of comprehensive recognition of forestry pests and diseases; secondly, the secondary correlation features of the pest and disease images in the target forestry area are determined by the environmental influencing factors in the pest and disease monitoring process in the target forestry area, and the feature information of the pest and disease images can be effectively supplemented by the environmental influencing factors, so as to better understand the influence of external factors on the distribution of pests and diseases; then, the primary correlation features and the secondary correlation features are interactively fused, the primary correlation features mainly focus on the morphological and behavioral characteristics of the pests and diseases themselves, and the secondary correlation features are mainly used to identify the pests and diseases themselves. The features take into account the impact of the external environment on pests and diseases. Through interactive fusion, the image features of forest pests and diseases and environmental information can be combined to provide more dimensional clues for forest pest and disease identification. The multi-level distribution features obtained by interactive fusion can provide more comprehensive information on the distribution of forest pests and diseases, thereby improving the ability of the pest and disease identification model to identify various pests and diseases, so as to realize diversified monitoring of forest pests and diseases under complex backgrounds; finally, the pest and disease identification model pre-trained by machine learning can confidently identify pests and diseases in the target forestry area based on the multi-level distribution features, which can reduce the false detection rate of pests and diseases in the target forestry area; in summary, this scheme can realize diversified monitoring of forest pests and diseases under complex backgrounds, thereby reducing the false detection rate of forest pests and diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0018] Figure 1 is an exemplary flow chart of a forestry pest and disease monitoring method based on machine learning according to some embodiments of the present application; Figure 2 is an exemplary flow chart of determining a valid feature vector according to some embodiments of the present application; Figure 3 is an exemplary flow chart of determining primary association features according to some embodiments of the present application; Figure 4 It is a structural schematic diagram of a forestry pest and disease monitoring system based on machine learning according to some embodiments of the present application; Figure 5It is a structural schematic diagram of a computer device for implementing a forestry pest and disease monitoring method based on machine learning as shown in some embodiments of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] refer to Figure 1 , which is an exemplary flow chart of a forestry pest monitoring method based on machine learning according to some embodiments of the present application. The forestry pest monitoring method based on machine learning 100 mainly includes the following steps: In step 101, historical forestry pest and disease images of a target forestry area are obtained.

[0021] In specific implementation, all pest and disease images of the target forestry area in the past year can be obtained from the forestry pest and disease image resource library, and all the obtained pest and disease images are combined into historical forestry pest and disease images of the target forestry area. In other embodiments, other methods can also be used for acquisition, which is not specifically limited here.

[0022] It should be noted that the historical forestry pests and diseases images in this application are images of all forestry pests and diseases collected in the past in the target forestry area. In addition, the forestry pests and diseases image resource library in this application includes various types of pests and diseases in the target forestry area collected in the past by equipment such as drones, satellite images or fixed cameras. The type, location and severity of the pests and diseases in each pest and disease image are marked, which will not be repeated here.

[0023] In step 102, principal component analysis is performed on the historical forestry pest and disease images based on the machine learning model to obtain multiple identification indicators of pests and diseases in the target forestry area, and an effective feature vector is determined for image recognition of forestry pests and diseases based on all the identification indicators, and then the primary correlation features of the pest and disease images in the target forestry area are determined through the effective feature vector.

[0024] In some embodiments, principal component analysis is performed on the historical forestry pest and disease images based on a machine learning model to obtain multiple identification indicators of pests and diseases in the target forestry area, which can be achieved by the following steps: For each pest and disease image in the historical forestry pest and disease images; Preprocessing the pest and disease image to obtain a preprocessed pest and disease image; Determine the multi-dimensional features of forestry pests and diseases in the pest and disease images based on the preprocessed pest and disease images through a machine learning model, and then obtain the multi-dimensional features of forestry pests and diseases in each pest and disease image; A plurality of principal components of the historical forestry pest and disease images are extracted from all the multi-dimensional features, and the extracted principal components are then used as identification indicators of the pest and disease in the target forestry area.

[0025] In specific implementation, the pest and disease image is preprocessed to obtain the preprocessed pest and disease image, which can be implemented in the following manner, namely: the pest and disease image can be denoised (for example, smoothed by Gaussian filtering), image size adjusted (for example, the image size is adjusted to a fixed 224x224 pixels), color space converted (for example, the pest and disease image is converted from RGB space to other color spaces (such as HSV)) and other preprocessing to obtain the preprocessed pest and disease image. Other methods can also be used to implement it in other embodiments, which are not limited here. In addition, it should be noted that the goal of preprocessing in the present application is to clearly display the key features of pests and diseases in forestry pest and disease images, remove noise, and reduce interference from irrelevant information, thereby preparing for subsequent feature extraction and principal component analysis.

[0026] In specific implementation, determining the multi-dimensional features of forestry pests and diseases in pest images based on the preprocessed pest images through a machine learning model can be achieved in the following manner, namely: first, the color change features in the forestry pest images can be captured using a color histogram or the distribution of other color spaces (such as HSV). Secondly, the texture features in the forestry pest images are extracted through gray-level co-occurrence matrices (GLCMs), local binary patterns (LBPs), etc. Then, the geometric shape features of the pests and diseases in the forestry pest images are extracted through morphological operations (such as edge detection and contour extraction), for example: the area, perimeter, shape moment, etc. of the contour of the pests and diseases. Finally, different types of features (i.e., the color change features, texture features, and geometric shape features obtained above) are merged into a feature vector through a machine learning model, and the obtained results are used as the multi-dimensional features of forestry pests and diseases in the pest images, thereby obtaining the multi-dimensional features of the pests and diseases in each pest image. In other embodiments, other methods may also be used for determination, which is not limited here.

[0027] It should be noted that the multi-dimensional features in the present application are feature vectors that describe various information of forestry pests and diseases in pest and disease images, including feature dimensions such as color, texture, and shape.

[0028] In specific implementation, multiple principal components of the historical forestry pest and disease images are extracted from all multi-dimensional features, and then the extracted principal components are used as identification indicators of pests and diseases in the target forestry area. This can be achieved in the following manner, namely: first, the multi-dimensional features corresponding to all forestry pest and disease images are integrated into a feature matrix, wherein each column in the feature matrix represents a feature dimension, and the value of each row is the eigenvalue of each pest and disease image in the historical forestry pest and disease images under the corresponding feature dimension, and then, the principal component analysis method is used to reduce the dimension of the feature matrix to obtain multiple principal components of the historical forestry pest and disease images (each principal component has a corresponding variance, and the variance represents the contribution of the corresponding principal component to the data variability), and then the extracted principal components are used as identification indicators of pests and diseases in the target forestry area. In other embodiments, other methods may also be used for determination, which is not limited here.

[0029] It should be noted that the recognition indicators in this application characterize the distinguishing features of pests and diseases in historical forestry pest and disease images. Therefore, the recognition indicators of pests and diseases in the target forestry area can provide effective input data for subsequent pest and disease classification, identification and other tasks.

[0030] In some embodiments, reference Figure 2 As shown in FIG. 1 , this figure is an exemplary flow chart for determining effective feature vectors in some embodiments of the present application. In this embodiment, determining effective feature vectors for image recognition of forestry pests and diseases based on all recognition indicators can be implemented by the following steps: First, in step 1021, the variance of the principal component corresponding to each identification index is obtained; Next, in step 1022, the variance contribution corresponding to each identification indicator is determined by the variance of the principal component corresponding to each identification indicator; Finally, in step 1023, effective feature vectors for image recognition of forest pests and diseases are determined based on all variance contributions.

[0031] In specific implementation, the variance contribution in this application reflects the importance of the principal component in the pest and disease index in the overall data. The larger the variance contribution, the higher the importance of the principal component in the pest and disease index in the overall data, and the smaller the variance contribution, the lower the importance of the principal component in the pest and disease index in the overall data; the variance contribution corresponding to each identification indicator can be determined by the variance of the principal component corresponding to each identification indicator, which can be implemented in the following way, namely: output the principal component corresponding to each identification indicator through the explained_variance_ratio in the principal component analysis The ratio of the variance of the component to the variance of all principal components, and the obtained ratio is used as the variance contribution of each recognition index. In other embodiments, other methods can also be used for determination, which is not limited here; determining the effective feature vector for image recognition of forestry pests and diseases according to all variance contributions can be achieved in the following way, namely: the first few recognition indicators with larger variance contributions (i.e., the recognition indicators corresponding to the principal components that can explain more than 95% of the variance) can be selected to form the effective feature vector for image recognition of forestry pests and diseases. In other embodiments, other methods can also be used for determination, which is not limited here.

[0032] It should be noted that the effective feature vector in this application is a feature that is valuable for the task of forestry pest and disease image recognition, which includes a combination of the most significant features (such as color, texture, shape, etc.) in the forestry pest and disease image. The effective feature vector can improve the efficiency and accuracy of forestry pest and disease recognition.

[0033] In some embodiments, determining the primary correlation features of the pest and disease images in the target forestry area by using the effective feature vector can be achieved by using the following steps: Determine the correlation between each identification index in the effective feature vector and the distribution of forest pests and diseases; The first-level correlation features of pest and disease images in the target forestry area are determined based on all correlations.

[0034] In specific implementation, the correlation of the present application reflects the degree of relationship between the identification indicators in the effective feature vector and the distribution of forestry pests and diseases. The higher the correlation, the closer the relationship between the identification indicators in the effective feature vector and the distribution of forestry pests and diseases. On the contrary, the lower the correlation, the more distant the relationship between the identification indicators in the effective feature vector and the distribution of forestry pests and diseases. The correlation between each identification indicator in the effective feature vector and the distribution of forestry pests and diseases can be determined in the following way, namely: the degree of association between each identification indicator in the effective feature vector and the distribution of forestry pests and diseases can be evaluated by an evaluation algorithm, and the result of the evaluation is used as the correlation between each identification indicator in the effective feature vector and the distribution of forestry pests and diseases. The evaluation algorithm is, for example, genetic algorithm, ensemble learning and reinforcement learning, etc. Other methods can also be used in other embodiments, which are not limited here.

[0035] In specific implementation, the first-level correlation features of the pest and disease images in the target forestry area based on all the correlations can be determined in the following manner, namely: first, a correlation threshold is set based on prior experience and historical experimental data. The correlation threshold is used to select important features (i.e., identification indicators) with higher correlations in the effective feature vector. Usually, the correlation threshold is set to 0.5. Then, all the correlations are compared with the correlation threshold, and then, the identification indicators whose absolute values ​​of the correlations in the effective feature vectors are greater than the correlation threshold are extracted. Finally, all the extracted identification indicators constitute the first-level correlation features of the pest and disease images in the target forestry area. Figure 3 As shown, this figure is an exemplary flow chart for determining the primary association features in some embodiments of the present application. Other methods may also be used for determination in other embodiments, which is not limited here.

[0036] It should be noted that the primary correlation features in the present application are image features that are representative of the characteristics of pests and diseases when performing image recognition on diversified forestry pests and diseases. The accuracy of comprehensive recognition of forestry pests and diseases can be improved through the primary correlation features.

[0037] In step 103, the target forestry area is monitored in real time, and the current monitoring image of the target forestry area is collected. Based on the perception scene of the monitoring image, multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area are extracted, and then the secondary correlation features of the pest and disease images in the target forestry area are determined by all the environmental influencing factors.

[0038] In specific implementation, the target forestry area is monitored in real time. The current image of the target forestry area can be collected in real time by image acquisition devices such as cameras or drones as the current monitoring image of the target forestry area. Other methods can also be used for collection in other embodiments, which are not limited here.

[0039] It should be noted that the monitoring images in the present application are images used to monitor and identify pests and diseases in the target forestry area.

[0040] In some embodiments, the extraction of multiple environmental impact factors in the process of pest and disease monitoring in the target forestry area based on the perceived scene of the monitoring image can be achieved by the following steps: Segmenting the monitoring image into different perception areas based on the perception scene of the monitoring image; Extract pixel-level features of each perception area; By combining all pixel-level features with external environmental data, multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area are determined.

[0041] In specific implementation, segmenting the monitoring image into different perception areas based on the perception scene of the monitoring image can be achieved in the following manner, namely, segmenting the monitoring image into different perception areas based on the perception scene of the monitoring image through an image segmentation algorithm (such as U-Net, Mask R-CNN), wherein the perception areas are, for example, healthy vegetation, damaged vegetation, soil, shadow and other areas; extracting pixel-level features of each perception area can be achieved in the following manner, namely, image analysis can be performed on each perception area through a convolutional neural network in deep learning to extract pixel-level features of each perception area, wherein the pixel-level features include: color (such as RGB value, HSV component, etc., which reflects the visual information of the perception area), texture (such as contrast, correlation, energy, etc., which is used to describe the surface texture of the perception area), spectral reflectance (the reflectance of each pixel in a specific band, which reflects the health status of vegetation in the perception area), etc. In other embodiments, other methods may also be used for determination, which is not limited here.

[0042] It should be noted that the perception area in the present application reflects the regionalized structural information in the monitoring image, such as vegetation distribution, bare land and shadows in the monitoring image; in addition, the pixel-level features characterize the detail information in the perception area, that is, the attribute value of a single pixel in the perception area (such as color, texture or spectral value).

[0043] In specific implementation, the following method can be used to determine multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area by combining all pixel-level features with external environmental data, namely: first, use the weather API (WeatherForecast API) (such as OpenWeatherMap, WeatherStack) to obtain external environmental data consistent with the collection time corresponding to the current monitoring image of the target forestry area, the external environmental data including temperature, humidity, precipitation, slope, slope direction and other external environmental factors, then, through correlation analysis based on all pixel-level features and the external environmental data, calculate the weight value of each external environmental factor in the external environmental data affecting the distribution (i.e., growth and spread) of forestry pests and diseases in the target forestry area, and finally, the obtained results are combined into multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area, and other methods can also be used for determination in other embodiments, which are not limited here.

[0044] It should be noted that the environmental impact factors in this application represent the key environmental conditions that affect the distribution of forestry pests and diseases in the target forestry area.

[0045] In some embodiments, determining the secondary correlation features of the pest and disease images in the target forestry area by all ring impact factors can be achieved by the following steps: Determine the impact amount corresponding to each environmental impact factor of each external environmental factor in the target forestry area when identifying the characteristics of forestry pests and diseases through machine learning and each environmental impact factor; The secondary correlation features of pest and disease images in the target forestry area are determined through all influencing quantities.

[0046] In specific implementation, the influence amount corresponding to each environmental influence factor when identifying the characteristics of forestry pests and diseases for each external environmental factor of the target forestry area through machine learning and each environmental influence factor can be achieved in the following manner, namely: a model of the mapping relationship between each environmental influence factor and the identification characteristics of forestry pests and diseases can be trained based on historical experience data and each environmental influence factor through machine learning methods (such as multivariate regression, random forest, XGBoost). The model is used to describe the degree of influence of each external environment in the target forestry area on the identification of forestry pests and diseases, and then each environmental influence factor is used as the input of the model. The degree of influence of the forestry pest and disease identification characteristics under each environmental influence factor output by the model is used as the influence amount corresponding to each environmental influence factor when each external environmental factor in the target forestry area identifies the characteristics of forestry pests and diseases. In other embodiments, other methods can also be used for determination, which is not limited here.

[0047] It should be noted that the impact value in this application reflects the degree of influence of external environmental factors on the identification characteristics of forestry pests and diseases. The larger the impact value, the greater the degree of influence of the external environmental factors on the identification characteristics of forestry pests and diseases. The smaller the impact value, the smaller the degree of influence of the external environmental factors on the identification characteristics of forestry pests and diseases.

[0048] In specific implementation, determining the secondary correlation features of the pest and disease images in the target forestry area through all the influence quantities can be achieved in the following manner, namely: based on all the influence quantities, selecting environmental influence factors with significant influence (i.e., influence quantities higher than 80%), and then using the selected environmental influence factors and the associated image features corresponding to the forestry pests and diseases as the secondary correlation features of the pest and disease images in the target forestry area. For example, for each external environmental factor, the associated image features corresponding to the external environmental factors corresponding to all environmental influence factors higher than 80% and the forestry pest and disease features (i.e., the vector formed by the color features of the forestry pest and disease images when the external environmental factors change) are used as the secondary correlation features of the pest and disease images in the target forestry area. In other embodiments, other methods may also be used for determination, which is not limited here.

[0049] It should be noted that the secondary correlation features in the present application represent image features in which pests and diseases in forestry pest images are significantly correlated with the external environment, which can reflect the distribution pattern of forestry pests and diseases affected by the external environment. In addition, the secondary correlation features can enhance the robustness of the forestry pest and disease recognition model, reflecting the indirect impact of the external environment on the distribution of forestry pests and diseases.

[0050] In step 104, the primary association features and the secondary association features are interactively fused using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in the target forestry area.

[0051] In some embodiments, the interactive fusion of the primary association features and the secondary association features using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in a target forestry area can be achieved by the following steps: Determine the interaction rate between the first-level association feature and the second-level association feature by using a tree model; The primary association feature and the secondary association feature are fused based on the interaction rate to obtain a multi-level distribution feature for image recognition of pests and diseases in a target forestry area.

[0052] It should be noted that the interaction rate in the present application represents the degree of complementarity between the fusion of the primary association features and the secondary association features. As a preferred embodiment, the interaction rate between the primary association features and the secondary association features can be determined using a tree model in the following manner, namely: a tree model (such as a random forest) can be used to evaluate the strength of the relationship between the primary association features and the secondary association features, and the obtained evaluation result is used as the interaction rate between the primary association features and the secondary association features. In other embodiments, other methods can also be used for determination, which is not specifically limited here.

[0053] In specific implementation, the first-level association feature and the second-level association feature are fused based on the interaction rate to obtain the multi-level distribution feature for image recognition of pests and diseases in the target forestry area. This can be achieved in the following manner, namely: the first-level association feature and the second-level association feature can be automatically fused based on the interaction rate through a feature fusion method, and then the fused feature vector is reshaped into a matrix form (such as a two-dimensional array) suitable for convolution operation, so that the obtained matrix form (such as a two-dimensional array) is used as the multi-level distribution feature for image recognition of pests and diseases in the target forestry area, wherein the feature fusion method is, for example, a deep learning feature fusion method, and other methods can also be used for determination in other embodiments, which are not limited here.

[0054] It should be noted that the multi-level distribution features in the present application are matrices that characterize the characteristic laws of the distribution of forest pests and diseases at different image spatial scales. The multi-level distribution features can improve the ability of the forest pest and disease recognition model to capture complex features.

[0055] In step 105, a pest and disease identification model is pre-trained through machine learning, and the pest and disease identification model identifies the current pest and disease distribution in the target forestry area based on the multi-level distribution characteristics.

[0056] It should be noted that the pest and disease identification model in the present application is a model for monitoring and identifying the distribution of pests and diseases in the target forestry area. It has a strong ability to identify pest and disease characteristics. As a preferred embodiment, the pest and disease identification model can be pre-trained based on a general large-scale forestry pest and disease monitoring data set through a machine learning model (such as a convolutional neural network), and then the pest and disease identification model can be made to learn to identify forestry pest and disease characteristics through an optimization algorithm, thereby completing the pre-training of the pest and disease identification model. In other embodiments, other methods can also be used for pre-training, which is not limited here.

[0057] In some embodiments, the pest identification model can identify the current pest distribution in the target forestry area based on the multi-level distribution characteristics by using the following steps: Using the current monitoring image of the target forestry area as the input of the pest and disease identification model; The pest and disease identification model is used to confidently identify the pests and diseases in the target forestry area based on the multi-level distribution characteristics to obtain the current pest and disease distribution results of the target forestry area.

[0058] In specific implementation, the pests and diseases in the target forestry area are confidently identified based on the multi-level distribution characteristics through the pest and disease identification model, and the current pest and disease distribution result of the target forestry area is obtained in the following manner, namely: the current monitoring image of the target forestry area is input into the pest and disease identification model, and then the confidence of the pest and disease category in each area in the current monitoring image of the target forestry area is determined based on the multi-level distribution characteristics through the pest and disease identification model, and a pest and disease distribution map (i.e., the forestry pest and disease classification result of each area) is generated, so as to obtain the current pest and disease distribution result of the target forestry area, and finally, the obtained pest and disease distribution result is sent to forestry management personnel or a forestry pest and disease monitoring platform.

[0059] In addition, in another aspect of the present application, in some embodiments, the present application provides a forestry pest monitoring system based on machine learning, referring to Figure 4 , which is a structural diagram of a forestry pest monitoring system based on machine learning according to some embodiments of the present application. The forestry pest monitoring system based on machine learning 400 includes: an acquisition module 401, a processing module 402 and an execution module 403, which are respectively described as follows: Acquisition module 401, in this application, acquisition module 401 is mainly used to acquire historical forestry pest images of the target forestry area; Processing module 402, in the present application, processing module 402 is mainly used to perform principal component analysis on the historical forestry pest and disease images based on a machine learning model to obtain multiple identification indicators of pests and diseases in the target forestry area, determine the effective feature vector for image recognition of forestry pests and diseases based on all the identification indicators, and then determine the primary correlation features of the pest and disease images in the target forestry area through the effective feature vector; The processing module 402 in the present application is also used to monitor the target forestry area in real time, collect the current monitoring image of the target forestry area, extract multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area based on the perception scene of the monitoring image, and then determine the secondary correlation features of the pest and disease image in the target forestry area from all the environmental influencing factors; The processing module 402 in the present application is also used to interactively fuse the primary association features and the secondary association features using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in the target forestry area; Execution module 403, in the present application, execution module 403 is mainly used to pre-train a pest and disease identification model through machine learning, and the pest and disease identification model identifies the current pest and disease distribution in the target forestry area based on the multi-level distribution characteristics.

[0060] The above describes in detail the examples of the forestry pest monitoring method and system based on machine learning provided by the embodiments of the present application. It can be understood that the corresponding device includes a hardware structure and / or software module corresponding to the execution of each function in order to realize the above functions. It should be easily appreciated by those skilled in the art that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0061] In some embodiments, the present application also provides a computer device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the above-mentioned forestry pest and disease monitoring method based on machine learning.

[0062] In some embodiments, reference Figure 5 , the dotted line in the figure indicates that the unit or the module is optional, and the figure is a schematic diagram of the structure of a computer device for implementing the forestry pest monitoring method based on machine learning in the present application. The forestry pest monitoring method based on machine learning in the above embodiment can be Figure 5 The computer device 500 shown in the figure is implemented, and the computer device 500 includes at least one processor 501, a memory 502 and at least one communication unit 505. The computer device 500 can be a terminal device, a server or a chip.

[0063] The processor 501 may be a general-purpose processor or a special-purpose processor. For example, the processor 501 may be a central processing unit (CPU), which may be used to control the computer device 500, execute software programs, and process data of the software programs. The computer device 500 may also include a communication unit 505 to implement signal input (reception) and output (transmission).

[0064] For example, the computer device 500 may be a chip, the communication unit 505 may be an input and / or output circuit of the chip, or the communication unit 505 may be a communication interface of the chip, and the chip may be a component of a terminal device, a network device, or other devices.

[0065] For another example, the computer device 500 may be a terminal device or a server, and the communication unit 505 may be a transceiver of the terminal device or the server, or the communication unit 505 may be a transceiver circuit of the terminal device or the server.

[0066] The computer device 500 may include one or more memories 502, on which a program 504 is stored. The program 504 can be executed by the processor 501 to generate instructions 503, so that the processor 501 performs the method described in the above method embodiment according to the instructions 503. Optionally, data (such as a target audit model) can also be stored in the memory 502. Optionally, the processor 501 can also read the data stored in the memory 502, and the data can be stored at the same storage address as the program 504, or the data can be stored at a different storage address from the program 504.

[0067] The processor 501 and the memory 502 may be provided separately or integrated together, for example, integrated on a system on chip (SOC) of the terminal device.

[0068] It should be understood that each step of the above method embodiment can be completed by a hardware-based logic circuit or software-based instructions in the processor 501. The processor 501 can be a CPU, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, such as discrete gates, transistor logic devices, or discrete hardware components.

[0069] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0070] For example, in some embodiments, the present application also provides a computer-readable storage medium, in which instructions or codes are stored. When the instructions or codes are run on a computer, the computer implements the above-mentioned forestry pest and disease monitoring method based on machine learning when executing.

[0071] In summary, in the forestry pest and disease monitoring method and system based on machine learning disclosed in the embodiments of the present application, historical forestry pest and disease images of the target forestry area are obtained; principal component analysis is performed on the historical forestry pest and disease images based on a machine learning model to obtain multiple recognition indicators of pests and diseases in the target forestry area, and an effective feature vector is determined when performing image recognition on forestry pests and diseases according to all the recognition indicators, and then the primary correlation features of the pest and disease images in the target forestry area are determined through the effective feature vector; the target forestry area is monitored in real time, and the current monitoring images of the target forestry area are collected, and based on the perception of the monitoring images The scene extracts multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area, and then determines the secondary correlation features of the pest and disease images in the target forestry area by all the ring influencing factors; uses the tree model to interactively fuse the primary correlation features and the secondary correlation features to obtain multi-level distribution features for image recognition of pests and diseases in the target forestry area; pre-trains the pest and disease recognition model through machine learning, and the pest and disease recognition model identifies the current distribution of pests and diseases in the target forestry area based on the multi-level distribution features; can realize diversified monitoring of forest pests and diseases under complex backgrounds, thereby reducing the false detection rate of forest pests and diseases.

[0072] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0073] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A forestry pest monitoring method based on machine learning, characterized in that: The steps include: Obtain historical forest pest and disease images of the target forestry area; Based on the machine learning model, principal component analysis is performed on the historical forestry pest and disease images to obtain multiple identification indicators of pests and diseases in the target forestry area, and effective feature vectors are determined when performing image recognition on forestry pests and diseases according to all the identification indicators, and then the primary correlation features of the pest and disease images in the target forestry area are determined through the effective feature vectors; Conduct real-time monitoring of the target forestry area, collect current monitoring images of the target forestry area, extract multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area based on the perception scene of the monitoring image, and then determine the secondary correlation features of the pest and disease images in the target forestry area from all the environmental influencing factors; The primary correlation features and the secondary correlation features are interactively integrated by using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in a target forestry area; A pest and disease identification model is pre-trained through machine learning, and the pest and disease identification model identifies the current pest and disease distribution in the target forestry area based on the multi-level distribution characteristics.

2. The method according to claim 1, characterized in that Based on the machine learning model, principal component analysis is performed on the historical forestry pest and disease images to obtain multiple identification indicators of pests and diseases in the target forestry area, including: For each pest and disease image in the historical forestry pest and disease images; Preprocessing the pest and disease image to obtain a preprocessed pest and disease image; Determine the multi-dimensional features of forestry pests and diseases in the pest and disease images based on the preprocessed pest and disease images through a machine learning model, and then obtain the multi-dimensional features of forestry pests and diseases in each pest and disease image; A plurality of principal components of the historical forestry pest and disease images are extracted from all the multi-dimensional features, and the extracted principal components are then used as identification indicators of the pest and disease in the target forestry area.

3. The method according to claim 1, characterized in that According to all the identification indicators, the effective feature vectors for image recognition of forest pests and diseases include: Get the variance of the principal component corresponding to each identification index; Determine the variance contribution corresponding to each identification indicator through the variance of the principal component corresponding to each identification indicator; The effective feature vector for image recognition of forest pests and diseases is determined based on all variance contributions.

4. The method according to claim 1, characterized in that Determining the primary correlation features of the pest and disease image in the target forestry area through the effective feature vector specifically includes: Determine the correlation between each identification index in the effective feature vector and the distribution of forest pests and diseases; The first-level correlation features of pest and disease images in the target forestry area are determined based on all correlations.

5. The method according to claim 1, characterized in that The multiple environmental influencing factors extracted based on the perception scene of the monitoring image in the process of pest and disease monitoring in the target forestry area specifically include: Segmenting the monitoring image into different perception areas based on the perception scene of the monitoring image; Extract pixel-level features of each perception area; By combining all pixel-level features with external environmental data, multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area are determined.

6. The method according to claim 1, characterized in that The tree model is used to interactively fuse the primary correlation features and the secondary correlation features to obtain multi-level distribution features for image recognition of pests and diseases in the target forestry area, specifically including: Determine the interaction rate between the first-level association feature and the second-level association feature by using a tree model; The primary association feature and the secondary association feature are fused based on the interaction rate to obtain a multi-level distribution feature for image recognition of pests and diseases in a target forestry area.

7. The method according to claim 1, characterized in that The pest identification model identifies the current pest distribution in the target forestry area based on the multi-level distribution characteristics, specifically including: Using the current monitoring image of the target forestry area as the input of the pest and disease identification model; The pest and disease identification model is used to confidently identify the pests and diseases in the target forestry area based on the multi-level distribution characteristics to obtain the current pest and disease distribution results of the target forestry area.

8. A forestry pest monitoring system based on machine learning, characterized in that: include: An acquisition module is used to acquire historical forestry pest and disease images of the target forestry area; A processing module, for performing principal component analysis on the historical forestry pest and disease images based on a machine learning model to obtain multiple identification indicators of pests and diseases in a target forestry area, determining effective feature vectors for image recognition of forestry pests and diseases based on all the identification indicators, and further determining primary correlation features of the pest and disease images in the target forestry area through the effective feature vectors; The processing module is also used to monitor the target forestry area in real time, collect the current monitoring image of the target forestry area, extract multiple environmental influencing factors in the process of pest and disease monitoring in the target forestry area based on the perception scene of the monitoring image, and then determine the secondary correlation features of the pest and disease image in the target forestry area from all the environmental influencing factors; The processing module is further used to interactively fuse the primary association features and the secondary association features using a tree model to obtain multi-level distribution features for image recognition of pests and diseases in the target forestry area; An execution module is used to pre-train a pest and disease monitoring model through machine learning, and the pest and disease monitoring model identifies the current pest and disease distribution in the target forestry area based on the multi-level distribution characteristics.

9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the forestry pest and disease monitoring method based on machine learning as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions or codes, and when the instructions or codes are executed on a computer, the computer implements the forestry pest monitoring method based on machine learning as described in any one of claims 1 to 7.

Images