Crop anomaly identification method and system based on deep learning

Abstract

The invention relates to a crop anomaly recognition method and system based on deep learning, and relates to the technical field of intelligent agriculture, and the method comprises the steps: determining image features with crop degree anomaly and crop anomaly information corresponding to the image features based on precision image information, binding the crop anomaly information to a crop image, and obtaining a crop anomaly recognition result; performing sampling point analysis on the crop abnormal image to obtain an abnormal sampling point position; performing crop image acquisition on the abnormal sampling point position according to the abnormal sampling point position to obtain a current crop image; inputting image features and crop abnormal information as training samples into an abnormal identification model for training; and obtaining a trained anomaly recognition model, extracting crop image features and position coordinate information in the current crop image, and inputting the crop image features into the trained anomaly recognition model for recognition to obtain an image recognition result. The crop abnormity identification efficiency is improved.

Description

Technical Field

[0001] This application relates to the technical field of intelligent agriculture, and in particular, to a method and system for crop anomaly recognition based on deep learning. Background Art

[0002] Agriculture, as the foundation of the world economy, is crucial for human survival. However, crops often face various abnormal conditions during the production process, such as pests, diseases, and nutrient deficiencies. These abnormal conditions directly affect the growth and yield of crops, thereby affecting agricultural production efficiency and economic benefits. Traditional crop anomaly recognition mainly relies on manual observation and empirical judgment. However, this method is not only time-consuming and laborious but also limited by the lack of professional knowledge and experience, easily leading to inaccurate judgments, missing the best prevention and control opportunities, and causing yield reduction and economic losses.

[0003] In recent years, with the rapid development of artificial intelligence technology, deep learning, as a powerful tool, has achieved remarkable results in fields such as image recognition and natural language processing. Deep learning algorithms, especially convolutional neural networks (CNNs), generative adversarial networks (GANs), Transformers, etc., have the ability to automatically learn and recognize complex patterns, which provides a new solution for crop anomaly recognition. By training a deep learning model, it can automatically identify and analyze abnormal conditions in crop images. This method not only improves the accuracy and efficiency of recognition but also reduces the dependence on professional knowledge and experience, helps to timely detect and control crop anomalies, and reduces losses.

[0004] Before applying deep learning algorithms to recognize crop images, various shooting devices are often used to shoot crops. The shooting methods used are all general shooting methods, that is, images of all crops are taken. However, not all crops in the obtained crop images have abnormal conditions. Therefore, this increases the amount of ineffective processing for subsequent crop image recognition. Moreover, due to the existence of downward viewing angles during the shooting of crops, there are likely to be certain occlusions in the obtained crop images, which also increases the difficulty of subsequent anomaly recognition of crop images, thereby reducing the efficiency of crop anomaly recognition. Summary of the Invention

[0005] To solve at least one of the above technical problems, this application provides a method and system for crop anomaly recognition based on deep learning.

[0006] In a first aspect, a method for crop anomaly recognition based on deep learning provided by this application adopts the following technical solutions:

[0007] A method for crop anomaly recognition based on deep learning includes:

[0008] Obtain regional image information, where the regional image information is crop images and growth condition information of different types of crops in different growth stages in the target area during the historical period;

[0009] Preprocess the regional image information to obtain precision image information;

[0010] Based on the precision image information, determine the image features with abnormal farming degree and the corresponding crop abnormal information corresponding to the image features;

[0011] According to the image features, bind the corresponding crop abnormal information to the crop image to obtain a crop abnormal image;

[0012] Perform sampling point analysis on the crop abnormal image to obtain abnormal sampling points;

[0013] Generate a crop collection instruction according to the abnormal sampling points, and control the image collection device to collect crop images at the abnormal sampling points to obtain the current crop image;

[0014] Create an abnormal recognition model, and use the image features and crop abnormal information as training samples to input into the abnormal recognition model for training to obtain a trained abnormal recognition model;

[0015] Extract the crop image features and the corresponding position coordinate information in the current crop image, and input the crop image features into the trained abnormal recognition model for recognition to obtain an image recognition result.

[0016] By adopting the above technical solutions, images and growth condition information of different types of crops in the target area during different growth stages in the historical period are obtained, providing a comprehensive data basis for subsequent anomaly detection. Then, the regional image information is preprocessed to improve the image accuracy, making the anomaly features clearer and more distinguishable. Based on the accurate image information, it is possible to accurately determine the image features with anomalies and their corresponding crop anomaly information, providing accurate target positioning for subsequent operations. By corresponding and binding the crop anomaly information to the crop image, a crop anomaly image is obtained. This step makes the anomaly information more intuitive and facilitates subsequent analysis and processing. By performing sampling point analysis on the crop anomaly image, the anomaly sampling points can be accurately found, thereby guiding the image acquisition device to perform targeted acquisition on the anomaly area, improving the acquisition efficiency and accuracy. By creating an anomaly recognition model and inputting the image features and crop anomaly information as training samples, the model can learn the anomaly features, improving the accuracy of anomaly recognition. In the subsequent recognition process, by inputting the current crop image features into the trained anomaly recognition model, the image recognition result can be quickly obtained, realizing the rapid response and accurate recognition of crop anomalies, thereby improving the recognition efficiency of crop anomalies.

[0017] In a preferred example, the present application can be further configured as follows: The preprocessing of the regional image information to obtain accurate image information includes:

[0018] Denoise and perform image enhancement processing on the crop images in the regional image information to obtain a primary processed image;

[0019] Determine the target crops with crop anomalies according to the growth condition information in the regional image information, and determine the target images with the target crops in the primary processed image;

[0020] Adjust the size of the target image according to the proportional distribution of the target crops in the target image to obtain a secondary processed image;

[0021] Update and replace the primary processed image according to the secondary processed image to obtain an updated primary processed image;

[0022] Update and replace the crop image according to the updated primary processed image to obtain an updated crop image;

[0023] Associate and summarize the updated crop image and the growth condition information to obtain accurate image information.

[0024] In a preferred example, the present application can be further configured as follows: The sampling point analysis of the crop anomaly image to obtain anomaly sampling points includes:

[0025] Construct a regional three-dimensional model within the target area based on the abnormal crop image;

[0026] Determine a concentrated abnormal area that meets the preset conditions and a scattered abnormal area that does not meet the preset conditions according to the regional three-dimensional model, where the preset condition is that the number of abnormal crops satisfying a specific distance exceeds 3 plants;

[0027] Respectively use the concentrated abnormal area and the scattered abnormal area as abnormal area points, and connect the abnormal area points in the order of the abnormal occurrence time nodes to obtain an abnormal trend curve;

[0028] Conduct a regular extension analysis on the abnormal trend curve to obtain a future abnormal trend curve with different abnormal occurrence probabilities within the future period;

[0029] Determine the future area points corresponding to the future abnormal trend curve, and generate refined sampling points according to the concentration attribute of the future area points;

[0030] Summarize the refined sampling points to obtain abnormal sampling points.

[0031] In a preferred example, the present application can be further configured as: the conducting a regular extension analysis on the abnormal trend curve to obtain a future abnormal trend curve with different abnormal occurrence probabilities within the future period includes:

[0032] Determine the point extension vectors for different abnormal area points to extend to the next abnormal area point and the data feature set corresponding to each abnormal area point according to the abnormal trend curve;

[0033] Determine the extension vector features based on the point extension vectors, and add the extension vector features to the data feature set to obtain a point feature set;

[0034] Arrange each point feature set in a matrix according to the time sequence to obtain a point feature matrix corresponding to each point feature set;

[0035] Deduce the point feature sets according to the time period of the point feature matrix to obtain multiple predicted point matrices within the future period;

[0036] Determine a future point feature set according to the multiple predicted point matrices, and import the future point feature set into the regional three-dimensional model to obtain a future abnormal trend curve with different abnormal occurrence probabilities within the future period.

[0037] In a preferred example, the present application can be further configured as: the constructing a regional three-dimensional model within the target area based on the abnormal crop image includes:

[0038] Extract the crop distribution information in the abnormal crop image, and convert the crop distribution information into crop feature points;

[0039] Convert the crop feature points into a crop raster object group, and use the first raster function to obtain the source of the maximum elevation value at all planar spatial positions in the crop raster object group;

[0040] Record the source of the maximum elevation value at all planar spatial positions in the crop raster object group as a position raster in the form of integer raster data;

[0041] Based on the position raster, select the raster pixels at the corresponding positions in the crop raster object group to obtain a crop layer raster data composed of single or multiple bottom elevation rasters, and convert the crop layer raster data into crop occupied area data;

[0042] Based on different depth feature types in the crop raster object group, using the crop occupied area data as the basis and taking the distribution range of the crop layer corresponding to the depth feature type in the updated vector surface data as the condition, generate the spatial multi-dimensional data corresponding to the crop layer raster data of different depth feature types, and summarize the spatial multi-dimensional data corresponding to each crop layer raster data to obtain a regional three-dimensional model.

[0043] In a preferred example of the present application, it can be further configured that: determining the future regional point positions corresponding to the future abnormal trend curve, and generating refined sampling points according to the concentration attributes of the future regional point positions, including:

[0044] Determine the future point position feature set according to the future abnormal trend curve;

[0045] Retrieve the future extension vector feature and the future point position feature set in the future point position feature set;

[0046] Determine the future regional point positions according to the future extension vector feature, and determine the future data feature set corresponding to the future regional point positions according to the future point position feature set;

[0047] Decompose the future data feature set to obtain the concentration attribute feature, the crop abnormal feature and the time trigger feature;

[0048] Determine whether the future area point is the concentrated abnormal area according to the concentrated attribute characteristics. If so, collect the area image information corresponding to the concentrated abnormal area, determine the occlusion points of crop occlusion existing in the area image information, determine the available sampling points according to the area image information and the crop abnormal characteristics, filter the available sampling points using the occlusion points as filtering conditions, and perform time setting on the filtered available sampling points according to the time trigger characteristics to obtain refined sampling points;

[0049] If not, collect the non-area image information corresponding to the non-concentrated abnormal area, determine the collection sampling points according to the non-area image information and the crop abnormal characteristics, and perform time setting on the collection sampling points according to the time trigger characteristics to obtain refined sampling points.

[0050] In a preferred example of the present application, it can be further configured that: after inputting the crop image characteristics into the trained abnormal recognition model for recognition to obtain an image recognition result, it further includes:

[0051] Determine whether there is crop abnormal information in the image recognition result. If so, determine the target abnormal position where the crop abnormal information exists based on the position coordinate information, and send the target abnormal position to the target terminal for control display.

[0052] In a second aspect, the present application provides a crop abnormal recognition system based on deep learning, adopting the following technical solution:

[0053] A crop abnormal recognition system based on deep learning, including:

[0054] An information acquisition module, used to acquire area image information, where the area image information is the crop images and growth condition information of different types of crops in the target area at different growth stages within the historical period;

[0055] An image processing module, used to preprocess the area image information to obtain precision image information:

[0056] An information determination module, used to determine the image characteristics of crop abnormality and the corresponding crop abnormal information based on the precision image information;

[0057] An information binding module, used to bind the crop abnormal information to the crop image corresponding to the image characteristics to obtain a crop abnormal image;

[0058] A sampling analysis module, used to perform sampling point analysis on the crop abnormal image to obtain abnormal sampling points;

[0059] An instruction generation module, configured to generate a crop collection instruction according to the abnormal sampling point, control an image acquisition device to perform crop image acquisition on the abnormal sampling point, and obtain a current crop image;

[0060] A model training module, configured to create an abnormal recognition model, and input the image features and crop abnormal information as training samples into the abnormal recognition model for training to obtain a trained abnormal recognition model;

[0061] An image recognition module, configured to extract crop image features and corresponding position coordinate information in the current crop image, and input the crop image features into the trained abnormal recognition model for recognition to obtain an image recognition result.

[0062] In a possible implementation manner, when the image processing module preprocesses the regional image information to obtain precision image information, it is specifically configured to:

[0063] Perform denoising and image enhancement processing on the crop images in the regional image information to obtain a primary processed image;

[0064] Determine target crops with crop abnormalities according to the growth condition information in the regional image information, and determine target images with target crops in the primary processed image;

[0065] Adjust the size of the target image according to the proportional distribution of the target crops in the target image to obtain a secondary processed image;

[0066] Update and replace the primary processed image according to the secondary processed image to obtain an updated primary processed image;

[0067] Update and replace the crop image according to the updated primary processed image to obtain an updated crop image;

[0068] Associate and summarize the updated crop image and the growth condition information to obtain precision image information.

[0069] In another possible implementation manner, when the sampling and analysis module performs sampling point analysis on the crop abnormal image to obtain abnormal sampling points, it is specifically configured to:

[0070] Construct a regional three-dimensional model in the target area according to the crop abnormal image;

[0071] Determine a concentrated abnormal area that meets the preset conditions and a scattered abnormal area that does not meet the preset conditions according to the three-dimensional model of the area, where the preset condition is that the number of abnormal crops that meet a specific distance exceeds 3 plants;

[0072] Respectively take the concentrated abnormal area and the scattered abnormal area as the abnormal area points, and connect the abnormal area points in the order of the abnormal occurrence time nodes to obtain an abnormal trend curve;

[0073] Conduct a regular extension analysis on the abnormal trend curve to obtain a future abnormal trend curve with different abnormal occurrence probabilities in the future cycle;

[0074] Determine the future area points corresponding to the future abnormal trend curve, and generate refined sampling points according to the concentration attribute of the future area points;

[0075] Summarize the refined sampling points to obtain abnormal sampling points.

[0076] In another possible implementation manner, when the sampling analysis module conducts a regular extension analysis on the abnormal trend curve to obtain a future abnormal trend curve with different abnormal occurrence probabilities in the future cycle, it specifically is used for:

[0077] Determine the point extension vectors for different abnormal area points to extend to the next abnormal area point and the data feature set corresponding to each abnormal area point according to the abnormal trend curve;

[0078] Based on the point extension vectors, determine the extension vector features, and add the extension vector features to the data feature set to obtain a point feature set;

[0079] Arrange each point feature set in a matrix according to time sequence to obtain a point feature matrix corresponding to each point feature set;

[0080] Deduce the point feature sets according to the time period of the point feature matrix to obtain multiple predicted point matrices in the future cycle;

[0081] Determine the future point feature set according to multiple predicted point matrices, and import the future point feature set into the three-dimensional model of the area to obtain a future abnormal trend curve with different abnormal occurrence probabilities in the future cycle.

[0082] In another possible implementation manner, when the sampling analysis module constructs the three-dimensional model of the area in the target area according to the abnormal crop image, it specifically is used for:

[0083] Extract the crop distribution information from the abnormal crop images, and convert the crop distribution information into crop feature points;

[0084] Convert the crop feature points into a group of crop raster objects, and use the first raster function to obtain the source of the maximum elevation value at all planar spatial positions in the group of crop raster objects;

[0085] Record the source of the maximum elevation value at all planar spatial positions in the group of crop raster objects as a position raster in the form of integer raster data;

[0086] Based on the position raster, select the raster pixels at the corresponding positions in the group of crop raster objects to obtain a crop layer raster data composed of single or multiple bottom elevation rasters, and convert the crop layer raster data into crop occupied area data;

[0087] Based on different depth feature types in the group of crop raster objects, using the crop occupied area data as the basis, and taking the distribution range of the crop layer corresponding to the depth feature type in the updated vector surface data as the condition, generate the spatial multi-dimensional data corresponding to the crop layer raster data of different depth feature types, and summarize the spatial multi-dimensional data corresponding to each crop layer raster data to obtain a regional three-dimensional model.

[0088] In another possible implementation manner, when the sampling analysis module determines the future regional point positions corresponding to the future abnormal trend curve and generates refined sampling points according to the concentration attributes of the future regional point positions, it is specifically used for:

[0089] Determine a future point feature set according to the future abnormal trend curve;

[0090] Retrieve the future extension vector feature and the future point feature set in the future point feature set;

[0091] Determine the future regional point positions according to the future extension vector feature, and determine the future data feature set corresponding to the future regional point positions according to the future point feature set;

[0092] Decompose the future data feature set to obtain a concentration attribute feature, a crop abnormality feature, and a time trigger feature;

[0093] Determine whether the future regional point is the concentrated abnormal area according to the concentrated attribute characteristics. If so, collect the regional image information corresponding to the concentrated abnormal area, determine the occlusion points of crop occlusion existing in the regional image information, determine the available sampling points according to the regional image information and the crop abnormal characteristics, filter the available sampling points with the occlusion points as the filtering conditions, and perform time setting on the filtered available sampling points according to the time trigger characteristics to obtain refined sampling points;

[0094] If not, collect the non-regional image information corresponding to the non-concentrated abnormal area, determine the collection sampling points according to the non-regional image information and the crop abnormal characteristics, and perform time setting on the collection sampling points according to the time trigger characteristics to obtain refined sampling points.

[0095] In another possible implementation manner, the system further includes: a control display module, wherein,

[0096] The control display module is used to determine whether there is crop abnormal information in the image recognition result. If so, determine the target abnormal position where the crop abnormal information exists based on the position coordinate information, and send the target abnormal position to the target terminal for control display.

[0097] In a third aspect, the present application provides an electronic device, adopting the following technical solution:

[0098] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned method for crop abnormal recognition based on deep learning are implemented.

[0099] In a fourth aspect, the present application provides a computer storage medium, with the following technical solution:

[0100] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for crop abnormal recognition based on deep learning are implemented.

[0101] In summary, the present application has the following beneficial technical effects:

[0102] Images and growth status information of different types of crops in the target area during different growth stages within the historical period are obtained, providing a comprehensive data basis for subsequent anomaly detection. Then, the regional image information is preprocessed to improve the image accuracy, making the anomaly features clearer and more distinguishable. Based on the accurate image information, it is possible to accurately determine the image features with anomalies and their corresponding crop anomaly information, providing accurate target positioning for subsequent operations. By corresponding and binding the crop anomaly information to the crop images, crop anomaly images are obtained, making the anomaly information more intuitive and facilitating subsequent analysis and processing. Analyzing the sampling points of the crop anomaly images can accurately find the abnormal sampling points, thereby guiding the image acquisition device to conduct targeted acquisition of the abnormal area, improving the acquisition efficiency and accuracy. By creating an anomaly recognition model and inputting the image features and crop anomaly information as training samples, the model can learn the anomaly features, improving the accuracy of anomaly recognition. In the subsequent recognition process, inputting the current crop image features into the trained anomaly recognition model can quickly obtain the image recognition results, achieving rapid response and accurate recognition of crop anomalies, thus improving the recognition efficiency of crop anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 is a flowchart of a method for crop anomaly recognition based on deep learning in one embodiment of the present application.

[0104] Figure 2 is a schematic structural diagram of a crop anomaly recognition system based on deep learning in one embodiment of the present application.

[0105] Figure 3 is a schematic block diagram of a principle of an electronic device in one embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0106] The following is a further detailed description of the present application in conjunction with the attached Figure 1 to the attached Figure 3 drawings.

[0107] This specific embodiment is only an interpretation of the present application and does not limit the present application. After reading this specification, those skilled in the art can make modifications to this embodiment without creative contributions as needed, but as long as they are within the scope of the claims of the present application, they are protected by the patent law.

[0108] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of this application without creative efforts shall fall within the scope of protection of this application.

[0109] In addition, the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, unless otherwise specified.

[0110] The embodiments of this application will be further described in detail below with reference to the accompanying drawings of the specification.

[0111] The embodiments of this application provide a method for identifying crop anomalies based on deep learning, which is executed by an electronic device. The electronic device can be a server or a terminal device. Among them, the server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The terminal device and the server can be directly or indirectly connected through wired or wireless communication methods, which are not limited in the embodiments of this application. As Figure 1 shown, the method includes:

[0112] Step S10: Obtain regional image information.

[0113] Among them, the regional image information is the crop images and growth status information of different types of crops in different growth stages in the target area during the historical period.

[0114] For the embodiments of this application, the regional image information represents the image data and its related information collected for a specific geographical area. The historical period refers to a past time range for review and analysis. The target area refers to a specific geographical area of concern where image information needs to be obtained. Different types of crops cover various crop species planted in this area, such as wheat, corn, rice, etc. Different growth stages refer to each growth period of crops from sowing to harvesting, such as the seedling stage, flowering stage, maturity stage, etc. The crop images refer to the crop photos or video frames taken at the above different growth stages. The growth status information is a detailed description of the growth situation of crops, which may include height, leaf area index, pest and disease conditions, etc.

[0115] In the embodiment of the present application, a database containing historical image data is established or accessed. The database should cover images of all relevant crops during the historical period of the target area. Secondly, the images are classified and labeled according to the crop type and growth stage to ensure that the required information can be retrieved quickly. Finally, image recognition technology and data analysis tools are used to extract the growth status information in the images, such as measuring the leaf area index of the crops through image analysis software.

[0116] Step S11: Preprocess the regional image information to obtain precision image information.

[0117] For the embodiment of the present application, the crop images in the regional image information are denoised and image-enhanced to obtain a primary processed image. The target crops with abnormal growth are determined according to the growth status information in the regional image information, and the target images containing the target crops in the primary processed image are determined. The target images are resized according to the proportion distribution of the target crops in the target images to obtain secondary processed images. The primary processed image is updated and replaced according to the secondary processed images to obtain an updated primary processed image. The crop images are updated and replaced according to the updated primary processed image to obtain updated crop images. The updated crop images and the growth status information are associated and summarized to obtain precision image information.

[0118] In the embodiment of the present application, denoising processing refers to removing the noise points or interference factors in the image to improve the clarity of the image. Image enhancement processing refers to enhancing the visual features such as contrast and brightness of the image through a series of technical means to make the details in the image more prominent. The primary processed image refers to the crop image after denoising and image enhancement processing. The target crops refer to the crop types with abnormalities (such as pests and diseases, malnutrition, etc.) determined according to the growth status information. The target images refer to the parts containing the target crops in the primary processed image. Resizing refers to scaling the target images according to the proportion of the target crops in the target images to ensure that the sizes of the crops in the images are consistent. The secondary processed images refer to the target images after resizing. Update and replacement refer to replacing the corresponding parts of the original image with the processed images. The precision image information refers to the image dataset with higher precision and practical value obtained by associating and summarizing the updated crop images and the growth status information.

[0119] For example, in an area image information, it contains an image of a wheat field and corresponding growth condition information (such as the leaves being yellowish, which may indicate a lack of nutrients). By performing denoising and image enhancement processing on the wheat field image, a clearer wheat field image with more prominent details (i.e., the first-level processed image) can be obtained. Then, based on the growth condition information, it is determined that there is a problem of nutrient deficiency in this wheat field (i.e., the target crop), and the part of the first-level processed image that contains this wheat field (i.e., the target image) is found. Next, according to the proportion of the wheat field in the target image, size adjustment is performed to obtain a wheat field image with consistent size (i.e., the second-level processed image). Finally, this wheat field image with consistent size is used to replace the corresponding part in the original image to obtain the updated first-level processed image. Then, the wheat field image in the updated first-level processed image is associated and summarized with the growth condition information to obtain a wheat field image containing accurate position and growth condition information (i.e., the precision image information).

[0120] For this application, algorithms such as median filtering and Gaussian filtering are used to perform denoising processing on the crop image to reduce noise interference in the image. Then, methods such as histogram equalization and contrast stretching are used for image enhancement processing to improve the clarity and contrast of the image. Next, based on the abnormal indicators (such as leaf color, growth rate, etc.) in the growth condition information, the target crop is determined, and the target image containing the target crop is marked in the first-level processed image. After that, the proportion of the target crop in the target image is calculated, and the target image is scaled according to the proportion to obtain a second-level processed image with consistent size. Then, the second-level processed image is compared with the first-level processed image to find the corresponding position and perform replacement to obtain the updated first-level processed image. Then, the crop image in the updated first-level processed image is compared with the original image to find the corresponding position and perform replacement to obtain the updated crop image. Finally, the updated crop image is associated and stored with the growth condition information to form the precision image information.

[0121] Step S12: Based on the precision image information, determine the image features with crop anomalies and the corresponding crop anomaly information corresponding to the image features.

[0122] For the embodiments of this application, image processing technologies such as color space conversion, edge detection, and texture analysis are used to preprocess and extract features from the precision image information. Then, machine learning algorithms such as support vector machine (SVM) and decision tree are used to classify and identify the extracted features to determine the abnormal image features. Finally, based on the classification results and knowledge in the agricultural field, the abnormal features are explained and diagnosed to obtain the crop anomaly information.

[0123] Step S13: According to the image features, bind the crop anomaly information to the crop image correspondingly to obtain the crop anomaly image.

[0124] For the embodiments of the present application, according to the knowledge and experience in the agricultural field, a correspondence table between features and abnormal information is established. Then, the features in the image are traversed, and the corresponding abnormal information is bound to the features in the form of marks or annotations according to the correspondence table. Finally, a crop abnormal image containing abnormal information is generated and saved. For example, in a crop image showing a cornfield, if it is recognized that the corn leaves in a certain area show abnormal yellow color, then the image features (such as color) of this area can be correspondingly bound to the crop abnormal information of "nutrient deficiency". The processed crop abnormal image will display marks or annotations in this yellow area, indicating the problem of nutrient deficiency in this area.

[0125] Step S14: Analyze the sampling points of the crop abnormal image to obtain abnormal sampling points.

[0126] For the embodiments of the present application, a regional three-dimensional model within the target area is constructed based on the crop abnormal image. According to the regional three-dimensional model, a concentrated abnormal area that meets the preset conditions and a scattered abnormal area that does not meet the preset conditions are determined. The preset condition is that the number of abnormal crops satisfying a specific distance exceeds 3 plants. The concentrated abnormal area and the scattered abnormal area are respectively used as abnormal area points, and the abnormal area points are connected in the order of the abnormal occurrence time nodes to obtain an abnormal trend curve. The abnormal trend curve is analyzed for regular extension to obtain a future abnormal trend curve with different abnormal occurrence probabilities in the future period. The future area points corresponding to the future abnormal trend curve are determined, and refined sampling points are generated according to the concentration attribute of the future area points. The refined sampling points are summarized to obtain abnormal sampling points.

[0127] Specifically, the concentrated abnormal area and the scattered abnormal area are the results of dividing the abnormal crops in the regional three-dimensional model according to the preset conditions. Among them, the concentrated abnormal area refers to an area where the number of abnormal crops exceeds a preset threshold (such as 3 plants) within a certain spatial range, and the scattered abnormal area refers to an area where the abnormal crops are distributed relatively scattered and do not meet the preset conditions. The abnormal area points refer to the specific position points representing the concentrated abnormal area and the scattered abnormal area, which are used for marking and analyzing on the map. The abnormal trend curve is a curve connected according to the abnormal area points and the order of the abnormal occurrence time nodes, which is used to show the change trend of crop abnormalities over time. The regular extension analysis refers to analyzing the abnormal trend curve to predict the probabilities of different abnormalities occurring in the future period and generating a future abnormal trend curve. The future area points refer to the possible abnormal area positions determined according to the future abnormal trend curve. The refined sampling points are more specific sampling positions generated based on the future area points according to the concentration attribute (such as abnormal density, abnormal type, etc.), which are used for further on-site investigation or sampling analysis.

[0128] In the embodiments of the present application, image processing algorithms and machine learning techniques are used to identify and classify abnormal crops in a three-dimensional model, and concentrated abnormal areas and scattered abnormal areas are divided according to preset conditions. Then, the geographic information system is used to mark the abnormal area points on the map, and an abnormal trend curve is generated according to the order of the time nodes when the abnormality occurs. After that, time series analysis or a machine learning model is used to perform regular extension analysis on the abnormal trend curve, predict the probabilities of different abnormalities occurring in the future cycle, and generate a future abnormal trend curve. Finally, the future area points are determined according to the future trend curve, and refined sampling points are generated at these points according to concentrated attributes such as abnormal density and type. For example, in an abnormal crop image, if the corn leaves in a certain area are generally yellow and the number exceeds 3 plants, then this area can be divided into a concentrated abnormal area and marked as an abnormal area point on the map. By connecting multiple abnormal area points and the order of the time nodes when the abnormality occurs, an abnormal trend curve showing the yellowing trend of corn leaves can be generated. Further, by performing regular extension analysis on this curve, the probability of the occurrence of the yellowing abnormality of corn leaves in the future cycle can be predicted, and the area positions (future area points) where abnormalities may exist in the future can be determined. Finally, refined sampling points are generated at these positions for on-site investigation.

[0129] Specifically, when constructing a regional three-dimensional model, the crop distribution information in the abnormal crop image is extracted and converted into crop feature points. The crop feature points are converted into a crop raster object group, and the first raster function is used to obtain the source of the maximum elevation value of all plane spatial positions in the crop raster object group. The source of the maximum elevation value of all plane spatial positions in the crop raster object group is recorded as a position raster in the form of integer raster data. Based on the position raster, the raster cells corresponding to the positions in the crop raster object group are selected to obtain a crop layer raster data composed of single or multiple bottom elevation rasters, and the crop layer raster data is converted into crop occupancy data. Based on different depth feature types in the crop raster object group, taking the crop occupancy data as the basis and the distribution range of the crop layer corresponding to the depth feature type in the updated vector surface data as the condition, spatial multi-dimensional data corresponding to the crop layer raster data of different depth feature types are generated, and the spatial multi-dimensional data corresponding to each crop layer raster data are summarized to obtain a regional three-dimensional model.

[0130] Specifically, determine the point extension vectors for the points in different abnormal regions extending to the points in the next abnormal region and the data feature set corresponding to each point in the abnormal region based on the abnormal trend curve. Determine the extension vector features based on the point extension vectors, and add the extension vector features to the data feature set to obtain the point feature set. Arrange each point feature set in a matrix according to the time sequence to obtain the point feature matrix corresponding to each point feature set. Deduce the point feature set in the future period based on the point feature matrix according to the time period to obtain multiple predicted point matrices in the future period. Determine the future point feature set according to the multiple predicted point matrices, and import the future point feature set into the regional three-dimensional model to obtain the future abnormal trend curve with different abnormal occurrence probabilities in the future period.

[0131] In the embodiments of the present application, spatial analysis techniques are used to calculate the point extension vectors. For example, methods such as Euclidean distance or Manhattan distance are used to determine the spatial relationship between points. Then, data such as the abnormal type, occurrence time, and environmental conditions of each point are collected to form a data feature set. After that, features such as the direction and speed of the point extension vectors are added to the data feature set to obtain the point feature set. Subsequently, these sets are arranged into a point feature matrix according to the time series. Finally, time series analysis, machine learning, or deep learning algorithms are used to deduce the point feature matrix to obtain the predicted point matrix in the future period, and the future point feature set is extracted from it.

[0132] Specifically, determine the future point feature set according to the future abnormal trend curve, and retrieve the future extension vector features and the future point feature set in the future point feature set. Determine the future regional points according to the future extension vector features, determine the future data feature set corresponding to the future regional points according to the future point feature set, disassemble the future data feature set to obtain the concentrated attribute features, crop abnormal features, and time trigger features. Determine whether the future regional points are concentrated abnormal regions according to the concentrated attribute features. If so, collect the regional image information corresponding to the concentrated abnormal region, and determine the occlusion points of the crop occlusion existing in the regional image information. Determine the available sampling points according to the regional image information and the crop abnormal features, filter the available sampling points using the occlusion points as filtering conditions, and set the time for the filtered available sampling points according to the time trigger features to obtain the refined sampling points. If not, collect the non-regional image information corresponding to the non-concentrated abnormal region, determine the collection sampling points according to the non-regional image information and the crop abnormal features, and set the time for the collection sampling points according to the time trigger features to obtain the refined sampling points.

[0133] Step S15: Generate a crop collection instruction according to the abnormal sampling point, control the image collection device to collect crop images at the abnormal sampling point, and obtain the current crop image.

[0134] Specifically, the system first detects the abnormal sampling point through the farmland monitoring system and automatically generates a crop collection instruction containing information such as point coordinates, collection time, and collection angle. Then, the system sends this instruction to the drone image collection device deployed in the farmland. After receiving the instruction, the drone will automatically fly to the designated point, adjust to the preset shooting height and angle, and start the camera for image collection. After the collection is completed, the drone sends the current crop image back to the system server through wireless transmission.

[0135] Step S16: Create an abnormal recognition model, and use the image features and crop abnormal information as training samples to input into the abnormal recognition model for training, and obtain the trained abnormal recognition model.

[0136] In the embodiment of the present application, the abnormal recognition model includes but is not limited to a neural network model.

[0137] Step S17: Extract the crop image features and the corresponding position coordinate information in the current crop image, and input the crop image features into the trained abnormal recognition model for recognition to obtain an image recognition result.

[0138] Specifically, determine whether there is crop abnormal information in the image recognition result. If so, based on the position coordinate information, determine the target abnormal position where the crop abnormal information exists, and send the target abnormal position to the target terminal for control display.

[0139] In the embodiments of the present application, images and growth status information of different types of crops in different growth stages in the target area within the historical period are obtained, providing a comprehensive data basis for subsequent anomaly detection. Then, the regional image information is preprocessed to improve the image accuracy, making the anomaly features more clearly distinguishable. Based on the accurate image information, the image features with anomalies and the corresponding crop anomaly information can be accurately determined, providing an accurate target location for subsequent operations. By corresponding and binding the crop anomaly information to the crop image, a crop anomaly image is obtained, making the anomaly information more intuitive and facilitating subsequent analysis and processing. By performing sampling point analysis on the crop anomaly image, the anomaly sampling points can be accurately found, thereby guiding the image acquisition device to perform targeted acquisition on the anomaly area, improving the acquisition efficiency and accuracy. By creating an anomaly recognition model and inputting the image features and crop anomaly information as training samples, the model can learn the anomaly features, improving the accuracy of anomaly recognition. In the subsequent recognition process, by inputting the current crop image features into the trained anomaly recognition model, the image recognition result can be quickly obtained, realizing the rapid response and accurate recognition of crop anomalies, thereby improving the recognition efficiency of crop anomalies.

[0140] The above embodiments introduce a deep learning-based crop anomaly recognition method from the perspective of the method flow. The following embodiments introduce a deep learning-based crop anomaly recognition system from the perspective of virtual modules or virtual units. For details, see the following embodiments.

[0141] Embodiments of the present application provide a deep learning-based crop anomaly recognition system 20, as Figure 2 shown Figure 2 is a schematic structural diagram of a deep learning-based crop anomaly recognition system provided by embodiments of the present application. The system 20 may specifically include:

[0142] An information acquisition module 21, configured to acquire regional image information, where the regional image information is crop images and growth status information of different types of crops in different growth stages in the target area within the historical period;

[0143] An image processing module 22, configured to preprocess the regional image information to obtain accurate image information;

[0144] An information determination module 23, configured to determine image features with crop anomalies and corresponding crop anomaly information based on the accurate image information;

[0145] An information binding module 24, configured to correspondingly bind the crop anomaly information to the crop image according to the image features to obtain a crop anomaly image;

[0146] The sampling and analysis module 25 is used to perform sampling point analysis on the crop anomaly image to obtain the abnormal sampling points;

[0147] The instruction generation module 26 is used to generate a crop collection instruction according to the abnormal sampling points, control the image acquisition device to collect crop images at the abnormal sampling points, and obtain the current crop image;

[0148] The model training module 27 is used to create an anomaly recognition model, and input the image features and crop anomaly information as training samples into the anomaly recognition model for training to obtain the trained anomaly recognition model;

[0149] The image recognition module 28 is used to extract the crop image features and the corresponding position coordinate information in the current crop image, and input the crop image features into the trained anomaly recognition model for recognition to obtain the image recognition result.

[0150] In a possible implementation manner of the embodiment of the present application, when the image processing module 22 preprocesses the regional image information to obtain the precision image information, it is specifically used for:

[0151] Perform denoising and image enhancement processing on the crop images in the regional image information to obtain a primary processed image;

[0152] Determine the target crops with crop anomalies according to the growth condition information in the regional image information, and determine the target images with target crops in the primary processed image;

[0153] Adjust the size of the target image according to the proportional distribution of the target crops in the target image to obtain a secondary processed image;

[0154] Update and replace the primary processed image according to the secondary processed image to obtain the updated primary processed image;

[0155] Update and replace the crop images according to the updated primary processed image to obtain the updated crop images;

[0156] Associate and summarize the updated crop images and the growth condition information to obtain the precision image information.

[0157] In another possible implementation manner of the embodiment of the present application, when the sampling and analysis module 25 performs sampling point analysis on the crop anomaly image to obtain the abnormal sampling points, it is specifically used for:

[0158] Construct a regional three-dimensional model within the target area according to the crop anomaly image;

[0159] Determine the concentrated anomaly area that meets the preset conditions and the scattered anomaly area that does not meet the preset conditions based on the regional three-dimensional model. The preset condition is that the number of abnormal crops satisfying a specific distance exceeds 3 plants;

[0160] Respectively take the concentrated anomaly area and the scattered anomaly area as the anomaly area points, and connect the anomaly area points in the order of the anomaly occurrence time nodes to obtain an anomaly trend curve;

[0161] Conduct a regular extension analysis on the anomaly trend curve to obtain a future anomaly trend curve with different anomaly occurrence probabilities in the future period;

[0162] Determine the future area points corresponding to the future anomaly trend curve, and generate refined sampling points according to the concentration attribute of the future area points;

[0163] Summarize the refined sampling points to obtain the anomaly sampling points.

[0164] Another possible implementation manner of the embodiment of the present application. When the sampling analysis module 25 conducts a regular extension analysis on the anomaly trend curve to obtain a future anomaly trend curve with different anomaly occurrence probabilities in the future period, it specifically is used for:

[0165] Determine the point extension vector for the extension of different anomaly area points to the next anomaly area point and the data feature set corresponding to each anomaly area point according to the anomaly trend curve;

[0166] Determine the extension vector feature based on the point extension vector, and add the extension vector feature to the data feature set to obtain a point feature set;

[0167] Arrange each point feature set in a matrix according to the time sequence to obtain a point feature matrix corresponding to each point feature set;

[0168] Deduce the point feature set according to the time period for the point feature matrix to obtain multiple predicted point matrices in the future period;

[0169] Determine the future point feature set according to the multiple predicted point matrices, and import the future point feature set into the regional three-dimensional model to obtain a future anomaly trend curve with different anomaly occurrence probabilities in the future period.

[0170] Another possible implementation manner of the embodiment of the present application. When the sampling analysis module 25 constructs a regional three-dimensional model within the target area based on the abnormal crop image, it specifically is used for:

[0171] Extract the crop distribution information in the abnormal crop image, and convert the crop distribution information into crop feature points;

[0172] Convert the crop feature points into a group of crop raster objects, and use the first raster function to obtain the source of the maximum elevation value at all planar spatial positions in the group of crop raster objects;

[0173] Record the source of the maximum elevation value at all planar spatial positions in the group of crop raster objects as a position raster in the form of integer raster data;

[0174] Based on the position raster, select the raster pixels at the corresponding positions in the group of crop raster objects to obtain a crop layer raster data composed of single or multiple bottom elevation rasters, and convert the crop layer raster data into crop occupied area data;

[0175] Based on different depth feature types in the group of crop raster objects, taking the crop occupied area data as the basis and the distribution range of the crop layer corresponding to the depth feature type in the updated vector surface data as the condition, generate spatial multi-dimensional data corresponding to the crop layer raster data of different depth feature types, and summarize the spatial multi-dimensional data corresponding to each crop layer raster data to obtain a regional three-dimensional model.

[0176] Another possible implementation manner of the embodiment of the present application. When the sampling analysis module 25 determines the future regional points corresponding to the future abnormal trend curve and generates refined sampling points according to the concentration attribute of the future regional points, it is specifically used for:

[0177] Determine the future point feature set according to the future abnormal trend curve;

[0178] Retrieve the future extension vector feature and the future point feature set in the future point feature set;

[0179] Determine the future regional points according to the future extension vector feature, and determine the future data feature set corresponding to the future regional points according to the future point feature set;

[0180] Decompose the future data feature set to obtain the concentration attribute feature, the crop anomaly feature, and the time trigger feature;

[0181] Determine whether the future regional points are concentrated abnormal areas according to the concentration attribute feature. If so, collect the regional image information corresponding to the concentrated abnormal areas, determine the occlusion points of the crop occlusion existing in the regional image information, determine the available sampling points according to the regional image information and the crop anomaly feature, filter the available sampling points with the occlusion points as the filtering condition, and perform time setting on the filtered available sampling points according to the time trigger feature to obtain refined sampling points;

[0182] Otherwise, non-region image information corresponding to the non-central abnormal region is collected, sampling points are determined based on the non-region image information and the crop abnormal features, and time settings are made for the sampling points according to the time trigger features to obtain refined sampling points.

[0183] Another possible implementation manner of the embodiment of the present application, the system 20 further includes: a control display module, wherein,

[0184] The control display module is configured to determine whether there is crop abnormal information in the image recognition result. If so, a target abnormal position with crop abnormal information is determined based on the position coordinate information, and the target abnormal position is sent to the target terminal for control display.

[0185] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described crop abnormal recognition system 20 based on deep learning can refer to the corresponding process in the foregoing method embodiment, and will not be described herein again.

[0186] An electronic device is provided in an embodiment of the present application, such as Figure 3 shown. Figure 3 FIG. is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Figure 3 The electronic device 300 shown includes: a processor 301 and a memory 303. Among them, the processor 301 and the memory 303 are connected, such as connected through a bus 302. Optionally, the electronic device 300 may further include a transceiver 304. It should be noted that in practical applications, the transceiver 304 is not limited to one, and the structure of the electronic device 300 does not constitute a limitation to the embodiment of the present application.

[0187] The processor 301 may be a CPU (Central Processing Unit, central processor), a general-purpose processor, a DSP (Digital Signal Processor, data signal processor), an ASIC (Application Specific Integrated Circuit, application-specific integrated circuit), an FPGA (Field Programmable Gate Array, field programmable gate array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can implement or execute various exemplary logical blocks, modules and circuits described in combination with the disclosure of the present application. The processor 301 may also be a combination for implementing a computing function, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0188] The bus 302 may include a path for transmitting information between the above components. The bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 3 it is only represented by a thick line in Figure 3 , but it does not mean that there is only one bus or one type of bus.

[0189] The memory 303 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or it may also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0190] The memory 303 is used to store the application program code for executing the solution of this application, and is controlled by the processor 301 to execute. The processor 301 is used to execute the application program code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0191] Among them, the electronic device includes but is not limited to: mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. It may also be a server, etc. Figure 3 The shown electronic device is only an example and should not bring any limitations to the functions and usage scopes of the embodiments of this application.

[0192] The embodiments of this application provide a computer-readable storage medium, on which a computer program is stored. When it runs on a computer, it enables the computer to execute the corresponding content in the foregoing method embodiments.

[0193] It should be understood that although the steps in the flowchart of the accompanying drawings are shown sequentially according to the indication of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless there is a clear indication in this document, there is no strict order restriction for the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0194] The above are only partial embodiments of the present application. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for identifying crop anomalies based on deep learning, characterized in that: include: Acquire regional image information, wherein the regional image information is crop images and growth status information of different types of crops in the target area at different growth stages within a historical period; Preprocessing the image information of the region to obtain high-precision image information; Determine, based on the precision image information, image features with abnormal farming degree and crop abnormality information corresponding to the image features; Binding the crop abnormality information to the crop image according to the image features to obtain the crop abnormality image; Performing sampling point analysis on the abnormal crop image to obtain abnormal sampling points; Generate a crop acquisition instruction according to the abnormal sampling point, control the image acquisition device to acquire the crop image at the abnormal sampling point, and obtain the current crop image; Creating an abnormality recognition model, and inputting the image features and crop abnormality information as training samples into the abnormality recognition model for training, thereby obtaining a trained abnormality recognition model; The crop image features in the current crop image and the position coordinate information corresponding to the crop image features are extracted, and the crop image features are input into the trained abnormality recognition model for recognition to obtain an image recognition result.

2. The method for identifying crop anomalies based on deep learning according to claim 1, characterized in that: The preprocessing of the regional image information to obtain precision image information includes: Performing denoising and image enhancement processing on the crop image in the regional image information to obtain a primary processed image; Determine the target crop with crop abnormality according to the growth status information in the regional image information, and determine the target image with the target crop in the primary processed image; Resizing the target image according to the distribution ratio of the target crop to the target image to obtain a secondary processed image; Updating and replacing the primary processed image according to the secondary processed image to obtain an updated primary processed image; updating and replacing the crop image according to the updated primary processed image to obtain an updated crop image; The updated crop image and the growth status information are associated and summarized to obtain precision image information.

3. The method for identifying crop anomalies based on deep learning according to claim 1, characterized in that: The performing sampling point analysis on the abnormal crop image to obtain abnormal sampling points includes: constructing a regional three-dimensional model within the target area according to the crop abnormality image; Determine, according to the regional three-dimensional model, concentrated abnormal areas that meet preset conditions and dispersed abnormal areas that do not meet preset conditions, wherein the preset condition is that the number of abnormal crops that meet a specific distance exceeds 3 plants; The concentrated abnormal area and the dispersed abnormal area are respectively used as abnormal area points, and the abnormal area points are connected according to the order of abnormal occurrence time nodes to obtain an abnormal trend curve; Perform regular extension analysis on the abnormal trend curve to obtain future abnormal trend curves with different abnormal occurrence probabilities in future periods; Determine the future regional points corresponding to the future abnormal trend curve, and generate refined sampling points according to the concentrated attributes of the future regional points; The refined sampling points are summarized to obtain abnormal sampling points.

4. The method for identifying crop anomalies based on deep learning according to claim 3, characterized in that: The regular extension analysis of the abnormal trend curve is performed to obtain a future abnormal trend curve with different abnormal occurrence probabilities in future periods, including: Determine, according to the abnormal trend curve, a point extension vector extending from different abnormal area points to the next abnormal area point and a data feature set corresponding to each abnormal area point; Determine an extended vector feature based on the point extended vector, and add the extended vector feature to the data feature set to obtain a point feature set; Arrange each of the point feature sets in a matrix in time sequence to obtain a point feature matrix corresponding to each point feature set; Deducing the point feature matrix from the point feature set according to the time period to obtain multiple predicted point matrices in the future period; A future point feature set is determined according to a plurality of prediction point matrices, and the future point feature set is imported into the regional three-dimensional model to obtain a future anomaly trend curve with different anomaly occurrence probabilities in future periods.

5. The method for identifying crop anomalies based on deep learning according to claim 3, characterized in that: The constructing of the regional three-dimensional model within the target area according to the crop abnormality image comprises: Extracting crop distribution information from the crop abnormality image, and converting the crop distribution information into crop feature points; Converting the crop feature points into a crop grid object group, and using a first grid function to obtain the source of the maximum elevation value of all plane spatial positions in the crop grid object group; The maximum elevation value source of all plane spatial positions in the crop grid object group is recorded as a position grid in the form of integer grid data; Selecting grid pixels at corresponding positions in the crop grid object group based on the position grid, obtaining crop layer grid data composed of a single or multiple bottom elevation grids, and converting the crop layer grid data into crop land area data; Based on the different depth feature types in the crop raster object group, taking the crop area data as a basis and taking the distribution range of the crop layer corresponding to the depth feature type in the updated vector surface data as a condition, the spatial multidimensional data corresponding to the crop layer raster data of the different depth feature types are generated, and the spatial multidimensional data corresponding to each crop layer raster data are summarized to obtain the regional three-dimensional model.

6. The method for identifying crop anomalies based on deep learning according to claim 3, characterized in that: The determining of the future regional points corresponding to the future abnormal trend curve and generating refined sampling points according to the concentrated attributes of the future regional points includes: Determine a future point feature set according to the future abnormal trend curve; Retrieving the future extended vector feature and the future point feature set in the future point feature set; Determine a future regional point according to the future extended vector feature, and determine a future data feature set corresponding to the future regional point according to the future point feature set; Decomposing the future data feature set to obtain concentrated attribute features, crop abnormality features, and time trigger features; Determine whether the future regional point is the concentrated abnormal area according to the concentrated attribute characteristics, if so, collect the regional image information corresponding to the concentrated abnormal area, and determine the occlusion points of crops in the regional image information, determine the available sampling points according to the regional image information and the crop abnormality characteristics, use the occlusion points as filtering conditions to filter the available sampling points, and set the time of the filtered available sampling points according to the time trigger characteristics to obtain refined sampling points; If not, the non-regional image information corresponding to the non-concentrated abnormal area is collected, and the sampling points are determined according to the non-regional image information and the abnormal characteristics of the crops, and the time of the sampling points is set according to the time trigger characteristics to obtain refined sampling points.

7. The method for identifying crop anomalies based on deep learning according to claim 1, characterized in that: The step of inputting the crop image features into the trained abnormality recognition model for recognition to obtain an image recognition result further includes: Determine whether there is crop abnormality information in the image recognition result. If so, determine the target abnormal position where the crop abnormality information exists based on the position coordinate information, and send the target abnormal position to the target terminal for control display.

8. A crop anomaly recognition system based on deep learning, characterized in that: include: An information acquisition module, used to acquire regional image information, wherein the regional image information is crop images and growth status information of different types of crops in a target area at different growth stages within a historical period; An image processing module, used for preprocessing the regional image information to obtain high-precision image information; An information determination module, used to determine image features with abnormal crop degree and abnormal crop information corresponding to the image features based on the precision image information; An information binding module, used for binding the crop abnormality information to the crop image according to the image features to obtain the crop abnormality image; A sampling analysis module, used for performing sampling point analysis on the abnormal crop image to obtain abnormal sampling points; An instruction generation module is used to generate a crop acquisition instruction according to the abnormal sampling point, control the image acquisition device to acquire the crop image at the abnormal sampling point, and obtain the current crop image; A model training module is used to create an abnormality recognition model, and input the image features and crop abnormality information as training samples into the abnormality recognition model for training, so as to obtain a trained abnormality recognition model; The image recognition module is used to extract the crop image features in the current crop image and the position coordinate information corresponding to the crop image features, and input the crop image features into the trained abnormality recognition model for recognition to obtain an image recognition result.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executes a method for identifying abnormality of crops based on deep learning as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the crop abnormality identification method based on deep learning as claimed in any one of claims 1 to 7.

Images