TMV early-stage remote sensing identification method and system based on RGB image

Abstract

According to the RGB image-based TMV (Tobacco Mosaic Virus) early-stage remote sensing identification method and system provided by the invention, a large-area tobacco planting area can be quickly covered through an unmanned aerial vehicle platform, full-automatic disease detection is realized, the requirement of manual intervention is greatly reduced, and the detection efficiency is improved; a common RGB camera is used for disease detection, the hardware cost is reduced, and the method can be suitable for planting different scales and types of tobaccos; by means of the image processing technology and the convolutional neural network, early symptoms of the TMV can be accurately recognized, early warning can be carried out in real time, prevention and treatment measures can be taken in time, economic losses are reduced, disease information can be easily obtained through corresponding interfaces, field management can be conveniently carried out, and different agricultural scenes can be conveniently used.

Description

Technical Field

[0001] The invention relates to the field of plant lesion identification, and in particular to a TMV early remote sensing identification method and system based on RGB images. Background Art

[0002] Tobacco mosaic virus (TMV) is a serious disease that widely affects the tobacco industry. TMV replicates and spreads in plant cells, causing tobacco leaves to become mottled and curled, which affects photosynthesis and ultimately leads to a decrease in tobacco yield and deterioration in quality. TMV spreads rapidly and is difficult to cure, making it a long-standing challenge in tobacco cultivation. In order to reduce economic losses, timely detection and effective management of TMV damage are essential.

[0003] At present, the detection of TMV mainly relies on traditional methods such as manual identification and laboratory testing. Manual identification is usually done by experienced farmers or agronomists by observing the lesion characteristics on tobacco leaves to judge the disease situation, but there are problems such as high misjudgment rate, low efficiency, and great influence by environmental factors. Although laboratory testing can provide more accurate results, it requires professional experimental equipment and technicians, is costly, and is not suitable for large-scale fields and rapid testing; and laboratory testing is usually verified for samples that have been found to be suspected of diseases, rather than an active global regional monitoring method, and there is a lag in detection.

[0004] In order to overcome the shortcomings of traditional methods, drones and remote sensing technology have been introduced for disease monitoring. Drones have efficient and flexible flight capabilities, can quickly cover large planting areas, collect crop image data, and combined with remote sensing technology, drones can obtain multi-dimensional data on plant health by carrying different types of sensors such as multispectral, thermal imaging, and RGB imaging. Hyperspectral imaging technology can capture the reflection information of plants in different spectral bands, thereby identifying the photosynthesis intensity and chlorophyll content of plants, and then detecting the health of plants. However, hyperspectral cameras are expensive and complex in data processing, making them unsuitable for widespread adoption by small and medium-sized farmers. In contrast, ordinary RGB cameras are inexpensive and easy to deploy, but the traditional view is that RGB cameras have limited capabilities in plant disease identification, mainly because RGB images intelligently obtain visible light information and are difficult to provide deeper spectral data for analysis. Summary of the invention

[0005] The purpose of the present invention is to provide a TMV early remote sensing identification method and system based on RGB images, plan the flight route of an unmanned aerial vehicle equipped with a high-resolution RGB camera, control the unmanned aerial vehicle along the flight route, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area, pre-process the RGB images to obtain tobacco leaf area sub-images; determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-images; input the color distribution characteristics, texture complexity, shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area; based on TMV infection The system can collect status data and generate disease distribution maps of ground areas; based on the disease distribution maps, it can carry out early warning and disease development trend prediction. Through the UAV platform, it can quickly cover large areas of tobacco planting areas to achieve fully automated disease detection, greatly reduce the need for manual intervention, and improve detection efficiency; it can use ordinary RGB cameras for disease detection to reduce hardware costs and can be applied to tobacco planting of different scales and types; with the help of image processing technology and convolutional neural networks, it can accurately identify the early symptoms of TMV, issue early warnings in real time, and take timely prevention and control measures to reduce economic losses. It can easily obtain disease information through the corresponding interface, facilitate field management, and use different agricultural scenarios.

[0006] The present invention is achieved through the following technical solutions:

[0007] The TMV early remote sensing recognition method based on RGB images includes:

[0008] Planning a flight path of a drone equipped with a high-resolution RGB camera, controlling the drone to fly along the flight path, and controlling the high-resolution RGB camera to collect RGB images of tobacco plants in a ground area; preprocessing the RGB images to obtain sub-images of tobacco leaf areas;

[0009] Determining the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-image; inputting the color distribution characteristics, texture complexity, shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area;

[0010] Based on the TMV infection status data, a disease distribution map of the ground area is generated; based on the disease distribution map, early warning and disease development trend prediction are performed.

[0011] Optionally, planning a flight path of a drone equipped with a high-resolution RGB camera, controlling the drone along the flight path, and controlling the high-resolution RGB camera to collect RGB images of tobacco plants in a ground area; and preprocessing the RGB images, including:

[0012] Acquire boundary information of a ground area through GIS, and divide the ground area into a plurality of parallel flight routes based on the boundary information; determine a flight altitude and a flight speed matching each flight route based on the terrain undulations and vegetation density of the ground area;

[0013] Determine the shooting field width of the high-resolution RGB camera based on the spacing between different flight routes; control the UAV to fly along the corresponding flight route and its matching flight altitude and flight speed, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area;

[0014] The RGB image is subjected to denoising, color correction and image segmentation, so as to segment the tobacco leaf region in the RGB image from the background and obtain a tobacco leaf region sub-image.

[0015] Optionally, determining the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image; inputting the color distribution characteristics, the texture complexity, the shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground region, including:

[0016] Extracting a color histogram of the tobacco leaf region sub-image, and extracting color distribution features of the tobacco leaf from the color histogram;

[0017] Performing gray-level co-occurrence matrix and local binary pattern processing on the tobacco leaf region sub-image to obtain the texture complexity of the tobacco leaf;

[0018] Processing the tobacco leaf region sub-image with a Canny edge detection algorithm to obtain shape and edge information of the tobacco leaf;

[0019] The color distribution characteristics, the texture complexity, and the shape and edge information are input into a convolutional neural network model to obtain the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area.

[0020] Optionally, based on the TMV infection status data, a disease distribution map of the ground area is generated; based on the disease distribution map, early warning and disease development trend prediction are performed, including:

[0021] Associating the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area with the map of the ground area through GIS to generate a disease distribution map of the ground area;

[0022] Based on the disease distribution map, sub-areas of the ground area where tobacco failure may occur are determined, so as to send early warning information to the target object; based on the disease distribution map and the environmental data corresponding to the ground area, the TMV damage diffusion trend information of the ground area is predicted.

[0023] The TMV early remote sensing recognition system based on RGB images includes:

[0024] A UAV flight and photography control module, used to plan a flight route of a UAV equipped with a high-resolution RGB camera, control the UAV along the flight route, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area;

[0025] An RGB image preprocessing module, used for preprocessing the RGB image to obtain a tobacco leaf region sub-image;

[0026] A tobacco leaf feature determination module, used to determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image;

[0027] A neural network analysis module, used for inputting the color distribution characteristics, the texture complexity, the shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area;

[0028] A disease distribution map generating module, used for generating a disease distribution map of the ground area based on the TMV infection status data;

[0029] The early warning and disease prediction module is used to provide early warning and disease development trend prediction based on the disease distribution map.

[0030] Optionally, the UAV flight and photography control module is used to plan a flight route of a UAV equipped with a high-resolution RGB camera, control the UAV along the flight route, and control the high-resolution RGB camera to collect RGB images of tobacco plants in a ground area, including:

[0031] Acquire boundary information of a ground area through GIS, and divide the ground area into a plurality of parallel flight routes based on the boundary information; determine a flight altitude and a flight speed matching each flight route based on the terrain undulations and vegetation density of the ground area;

[0032] Determine the shooting field width of the high-resolution RGB camera based on the spacing between different flight routes; control the UAV to fly along the corresponding flight route and its matching flight altitude and flight speed, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area;

[0033] The RGB image preprocessing module is used to preprocess the RGB image to obtain a tobacco leaf region sub-image, including:

[0034] The RGB image is subjected to denoising, color correction and image segmentation, so as to segment the tobacco leaf region in the RGB image from the background and obtain a tobacco leaf region sub-image.

[0035] Optionally, the tobacco leaf feature determination module is used to determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image, including:

[0036] Extracting a color histogram of the tobacco leaf region sub-image, and extracting color distribution features of the tobacco leaf from the color histogram;

[0037] Performing gray-level co-occurrence matrix and local binary pattern processing on the tobacco leaf region sub-image to obtain the texture complexity of the tobacco leaf;

[0038] Processing the tobacco leaf region sub-image with a Canny edge detection algorithm to obtain shape and edge information of the tobacco leaf;

[0039] The neural network analysis module is used to input the color distribution characteristics, the texture complexity, the shape and edge information into the neural network model to obtain the TMV infection status data of the tobacco plants in the ground area, including:

[0040] The color distribution characteristics, the texture complexity, and the shape and edge information are input into a convolutional neural network model to obtain the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area.

[0041] Optionally, the disease distribution map generating module is used to generate a disease distribution map of the ground area based on the TMV infection status data, including:

[0042] Associating the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area with the map of the ground area through GIS to generate a disease distribution map of the ground area;

[0043] The early warning and disease prediction module is used to perform early warning and disease development trend prediction based on the disease distribution map, including:

[0044] Based on the disease distribution map, sub-areas of the ground area where tobacco failure may occur are determined, so as to send early warning information to the target object; based on the disease distribution map and the environmental data corresponding to the ground area, the TMV damage diffusion trend information of the ground area is predicted.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present application provides a method and system for early remote sensing identification of TMV based on RGB images, plans a flight route of an unmanned aerial vehicle equipped with a high-resolution RGB camera, controls the unmanned aerial vehicle along the flight route, and controls the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area, pre-processes the RGB images to obtain tobacco leaf area sub-images; determines the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-images; inputs the color distribution characteristics, texture complexity, shape and edge information into a neural network model, obtains TMV infection status data of tobacco plants in the ground area; and based on the TMV infection status data, obtains TMV infection status data of tobacco plants in the ground area. Based on the disease distribution map, early warning and disease development trend prediction are carried out. The UAV platform can quickly cover a large area of ​​tobacco planting area to achieve fully automated disease detection, greatly reduce the need for manual intervention and improve detection efficiency. Ordinary RGB cameras are used for disease detection to reduce hardware costs and can be applied to tobacco planting of different scales and types. With the help of image processing technology and convolutional neural networks, the early symptoms of TMV can be accurately identified, early warnings can be issued in real time, and prevention and control measures can be taken in time to reduce economic losses. Disease information can be easily obtained through the corresponding interface, which is convenient for field management and the use of different agricultural scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0048] Figure 1 A schematic flow chart of the TMV early remote sensing recognition method based on RGB images provided by the present invention.

[0049] Figure 2 This is a structural diagram of the convolutional neural network model used in the TMV early remote sensing recognition method based on RGB images provided by the present invention.

[0050] Figure 3 This is a schematic diagram of the structure of the TMV early remote sensing recognition system based on RGB images provided by the present invention. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It is to be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some structures related to the present application are shown in the accompanying drawings, rather than all structures. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0052] The terms "including" and "having" and any variations thereof in this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0053] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0054] See also Figure 1 As shown, an embodiment of the present application provides a TMV early remote sensing recognition method based on RGB images. The TMV early remote sensing recognition method based on RGB images includes:

[0055] Planning a flight path of a UAV equipped with a high-resolution RGB camera, controlling the UAV along the flight path, and controlling the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area; preprocessing the RGB image to obtain a sub-image of the tobacco leaf area;

[0056] Determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-image; input the color distribution characteristics, the texture complexity, the shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area;

[0057] Based on the TMV infection status data, a disease distribution map of the ground area is generated; based on the disease distribution map, early warning and disease development trend prediction are carried out.

[0058] The beneficial effects of the above embodiments are as follows: the RGB image-based TMV early remote sensing identification method plans the flight route of a drone equipped with a high-resolution RGB camera, controls the drone along the flight route, and controls the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area, pre-processes the RGB images to obtain tobacco leaf area sub-images; determines the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-images; inputs the color distribution characteristics, texture complexity, shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area; and based on the TMV infection status Based on the disease distribution map, early warning and disease development trend prediction can be carried out. The UAV platform can quickly cover a large area of ​​tobacco planting area to achieve fully automated disease detection, greatly reduce the need for manual intervention and improve detection efficiency. Ordinary RGB cameras are used for disease detection to reduce hardware costs and can be applied to tobacco planting of different scales and types. With the help of image processing technology and convolutional neural networks, the early symptoms of TMV can be accurately identified, early warning can be carried out in real time, and prevention and control measures can be taken in time to reduce economic losses. Disease information can be easily obtained through the corresponding interface, which is convenient for field management and the use of different agricultural scenarios.

[0059] In another embodiment, a flight path of a drone equipped with a high-resolution RGB camera is planned, the drone is controlled to fly along the flight path, and the high-resolution RGB camera is controlled to collect an RGB image of tobacco plants in a ground area; and the RGB image is preprocessed, including:

[0060] The boundary information of the ground area is obtained through GIS, and based on the boundary information, the ground area is divided into a number of parallel flight routes; based on the terrain undulations and vegetation density of the ground area, the flight altitude and flight speed matching each flight route are determined;

[0061] Based on the spacing between different flight routes, the shooting field width of the high-resolution RGB camera is determined; the UAV is controlled to fly along the corresponding flight route and its matching flight altitude and flight speed, and the high-resolution RGB camera is controlled to collect RGB images of tobacco plants in the ground area;

[0062] The RGB image is subjected to denoising, color correction and image segmentation, so as to segment the tobacco leaf region in the RGB image from the background and obtain a tobacco leaf region sub-image.

[0063] In order to achieve flexible large-area RGB image capture of ground areas such as tobacco fields, this application can use drones equipped with RGB cameras as mobile data collection platforms, such as the DJI Matrice 350 quad-rotor drone equipped with a 20-megapixel RGB camera, which can provide high-resolution image acquisition capabilities. At the same time, the RGB camera also supports autofocus and dynamic light adjustment to adapt to different lighting conditions, ensuring that clear images can be captured in various environments (such as sunny or cloudy weather environments).

[0064] In addition, the flight path of the drone adopts an automatic optimization algorithm based on the geographic information system GIS. Specifically, the boundary information of the field is first obtained through GIS, and the air above the field is divided into several parallel flight paths. The spacing of these flight paths can be calculated and determined based on the shooting field width of the RGB camera to ensure seamless coverage of the entire field during the shooting process. The flight path planning also takes into account the terrain undulations and vegetation density, and automatically adjusts the flight altitude and flight speed of the drone to ensure that the collected RGB images have sufficient resolution, while avoiding obstacles and reducing image overlap. In addition, according to the battery life of the drone, the nodes for the drone's return and mid-way charging are dynamically planned to ensure that the mission efficiently completes the shooting operation of the entire field within a single flight.

[0065] The drone automatically flies along the corresponding flight path and collects RGB images of tobacco plants at regular intervals or in real time during the flight. The frequency of RGB image acquisition can be adjusted according to the growth cycle of tobacco plants and the frequency of disease occurrence. The collected RGB images will be transmitted to the ground station or cloud platform in real time via 4G or 5G network for subsequent processing.

[0066] In addition, during the RGB image acquisition process, due to the vibration of the drone during flight and the change of the ambient light, there may be noise in the RGB image. In order to improve the quality of the RGB image, the RGB image needs to be denoised. Preferably, the salt and pepper noise and Gaussian noise in the RGB image can be removed by median filtering and / or Gaussian filtering. The color information of the RGB image may change under different lighting conditions. In order to ensure the consistency of the image color and the accuracy of the subsequent analysis, the RGB image needs to be color corrected. Preferably, white balance adjustment and / or gamma correction can be used to adjust the ratio and overall brightness of the RGB channels of the image so that the color of the image is closer to the actual situation. Among them, gamma correction is mainly used to adjust the overall brightness and contrast of the image so that the color distribution is more uniform. After denoising and color correction, the tobacco leaf area in the RGB image needs to be segmented from the background. For this purpose, the GrabCut algorithm is used, which can effectively separate the tobacco leaf area from the background through iterative optimization based on the initial rectangular area provided by the user. The segmented tobacco leaf area will be used for subsequent color distribution characteristics, texture complexity, shape and edge information extraction.

[0067] In another embodiment, the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image are determined; the color distribution characteristics, the texture complexity, the shape and edge information are input into a neural network model to obtain TMV infection status data of tobacco plants in the ground area, including:

[0068] Extracting a color histogram of the tobacco leaf region sub-image, and extracting color distribution features of the tobacco leaf from the color histogram;

[0069] The tobacco leaf region sub-image is processed by gray-level co-occurrence matrix and local binary pattern to obtain the texture complexity of the tobacco leaf;

[0070] The tobacco leaf region sub-image is processed by the Canny edge detection algorithm to obtain the shape and edge information of the tobacco leaf;

[0071] The color distribution characteristics, the texture complexity, the shape and edge information are input into the convolutional neural network model to obtain the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area.

[0072] Color distribution characteristics are one of the important bases for identifying TMV. In order to ensure the accuracy of color distribution feature extraction, the color histogram method is used to perform color analysis on the tobacco leaf area in the RGB image. By calculating the color histogram of the RGB image, the main color distribution of the tobacco leaves is extracted. In the early stage of TMV infection, tobacco leaves usually show obvious mottled colors (such as alternating green and yellow), and these color changes can be quantified by analyzing the color histogram.

[0073] The occurrence of TMV is often accompanied by abnormal changes in the surface texture of tobacco leaves. In order to extract the surface texture features of tobacco leaves, two methods, gray-level co-occurrence matrix (GLCM) and local binary pattern (LBP), can be used. GLCM can calculate the spatial relationship between pixels in RGB images and describe the direction, contrast, roughness and other characteristics of the surface texture of tobacco leaves; while LBP is a texture description method based on pixel neighborhood. By converting the local area of ​​the image into a binary pattern, it can effectively capture the subtle texture changes on the surface of tobacco leaves. In addition, the curling and deformation of the edges of tobacco leaves are also one of the typical symptoms of TMV. For this reason, the Canny edge detection algorithm can be used to extract the edges of tobacco leaves. By calculating the gradient amplitude and direction of the RGB image, it can accurately capture the subtle changes in the edges of tobacco leaves. The extracted edge information can be used to analyze the shape changes of the leaves and further improve the recognition accuracy of TMV damage.

[0074] In actual neural network analysis, convolutional neural network CNN is used as the main recognition module. CNN has high feature extraction and pattern recognition capabilities in image classification tasks and is particularly suitable for processing complex image data. The CNN structure used in this application is as follows Figure 2 As shown, the details are as follows:

[0075] Step 1: The input layer of the RGB image part is a 254×254×3 RGB image, and the input layer of the texture and edge feature part is a 254×254×12 feature matrix. The first convolution layer consists of two convolution layers with 3*3 convolution kernels, and the step size and padding parameters are set to 1;

[0076] Step 2: The second, third, and fifth layers consist of a maximum pooling layer (size: 2, step size: 2) and two convolutional layers with 3*3 convolution kernels. The convolution kernel step size and padding parameters are both set to 1.

[0077] Step 3: The fourth layer consists of a maximum pooling layer (size: 2, stride: 2) and a convolution layer with a 3*3 convolution kernel. The convolution layer of the texture and edge feature part is then superimposed with the convolution layer of the RGB image part.

[0078] Step 4: The sixth layer consists of a maximum pooling layer (size: 2, step size: 2) followed by a fully connected layer (4096 nodes);

[0079] Step 5: The convolution part of the color feature consists of two convolution layers with 3*3 convolution kernels, and the step size and padding parameters are set to 1. The next three layers consist of a maximum pooling layer (size: 2, step size: 2) and a convolution layer with a 3*3 convolution kernel, and finally a fully connected layer (4096 nodes).

[0080] The fully connected layers of step 6, step 4 and step 5 are added together with the final fully connected layer (4096 nodes); the resulting output layer activation function is Softmax.

[0081] The training of this model uses an image dataset collected by drones in the experimental field and artificial standards. TMV was artificially inoculated on the "Yunyan 87" tobacco in the cluster stage in the experimental field. RGB image data was collected by drones on the second day (asymptomatic infection period) and the 11th day (bright vein period) after inoculation. The artificial images were then segmented and annotated to obtain 1,865 healthy sample images, 1,232 asymptomatic infection sample images, and 1,021 bright vein period sample images. After extracting the color, texture, and edge features of these images, they were input into the CNN model for training. The CNN model can automatically learn and identify features related to the disease. During the training process, we also used data augmentation techniques (such as image rotation, scaling, flipping, etc.) to increase the diversity of the data set and prevent the model from overfitting.

[0082] In order to improve the recognition accuracy of the model, the model was optimized through hyperparameter tuning, cross-validation and other methods. The parameter optimization used the AdamOptimizer optimizer with a learning rate of 0.01, and the training epoch and batch size were both set to 100; the accuracy of the optimized model on the test set reached more than 92%, which can effectively identify TMV.

[0083] In another embodiment, a disease distribution map of the ground area is generated based on the TMV infection status data; and early warning and disease development trend prediction are performed based on the disease distribution map, including:

[0084] The distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area are associated with the map of the ground area through GIS to generate a disease distribution map of the ground area;

[0085] Based on the disease distribution map, the sub-areas of the ground area where tobacco failure may occur are determined, so as to send early warning information to the target objects; based on the disease distribution map and the environmental data corresponding to the ground area, the TMV damage spread trend information of the ground area is predicted.

[0086] The beneficial effect of the above embodiment is that the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area are associated with the map of the ground area through GIS to generate a disease distribution map of the ground area. Users can intuitively view the distribution of TMV damage through the corresponding platform interface, so as to take corresponding prevention and control measures according to the severity of TMV damage. When a diseased area is detected, the platform will automatically send early warning information to target objects such as farmers or managers; the early warning information may include the specific location, severity and recommended treatment measures of the disease, and these early warning information can be pushed to the target object via SMS, email or mobile application to ensure that it can respond in time. In addition, the platform will also save historical monitoring data, and use these historical monitoring data to analyze and predict the development trend of the disease, and predict the TMV damage diffusion trend information of the field. Combined with external weather, soil and other environmental data, the system can provide more accurate disease management suggestions for the target object to help it take preventive measures before the disease occurs.

[0087] See also Figure 3 As shown, an embodiment of the present application provides a TMV early remote sensing recognition system based on RGB images. The TMV early remote sensing recognition system based on RGB images includes:

[0088] The UAV flight and photography control module is used to plan the flight path of the UAV equipped with a high-resolution RGB camera, control the UAV along the flight path, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area;

[0089] An RGB image preprocessing module is used to preprocess the RGB image to obtain a tobacco leaf region sub-image;

[0090] A tobacco leaf feature determination module is used to determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image;

[0091] A neural network analysis module is used to input the color distribution characteristics, the texture complexity, the shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area;

[0092] A disease distribution map generating module, used for generating a disease distribution map of the ground area based on the TMV infection status data;

[0093] The early warning and disease prediction module is used to provide early warning and disease development trend prediction based on the disease distribution map.

[0094] The beneficial effects of the above embodiments are as follows: the RGB image-based TMV early remote sensing identification system plans the flight route of a drone equipped with a high-resolution RGB camera, controls the drone along the flight route, and controls the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area, pre-processes the RGB images to obtain tobacco leaf area sub-images; determines the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-images; inputs the color distribution characteristics, texture complexity, shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area; and based on the TMV infection status Based on the disease distribution map, early warning and disease development trend prediction can be carried out. The UAV platform can quickly cover a large area of ​​tobacco planting area to achieve fully automated disease detection, greatly reduce the need for manual intervention and improve detection efficiency. Ordinary RGB cameras are used for disease detection to reduce hardware costs and can be applied to tobacco planting of different scales and types. With the help of image processing technology and convolutional neural networks, the early symptoms of TMV can be accurately identified, early warning can be carried out in real time, and prevention and control measures can be taken in time to reduce economic losses. Disease information can be easily obtained through the corresponding interface, which is convenient for field management and the use of different agricultural scenarios.

[0095] In another embodiment, the UAV flight and photography control module is used to plan a flight path of a UAV equipped with a high-resolution RGB camera, control the UAV along the flight path, and control the high-resolution RGB camera to collect RGB images of tobacco plants in a ground area, including:

[0096] The boundary information of the ground area is obtained through GIS, and based on the boundary information, the ground area is divided into a number of parallel flight routes; based on the terrain undulations and vegetation density of the ground area, the flight altitude and flight speed matching each flight route are determined;

[0097] Based on the spacing between different flight routes, the shooting field width of the high-resolution RGB camera is determined; the UAV is controlled to fly along the corresponding flight route and its matching flight altitude and flight speed, and the high-resolution RGB camera is controlled to collect RGB images of tobacco plants in the ground area;

[0098] The RGB image preprocessing module is used to preprocess the RGB image to obtain a tobacco leaf region sub-image, including:

[0099] The RGB image is subjected to denoising, color correction and image segmentation, so as to segment the tobacco leaf region in the RGB image from the background and obtain a tobacco leaf region sub-image.

[0100] In another embodiment, the tobacco leaf feature determination module is used to determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image, including:

[0101] Extracting a color histogram of the tobacco leaf region sub-image, and extracting color distribution features of the tobacco leaf from the color histogram;

[0102] The tobacco leaf region sub-image is processed by gray-level co-occurrence matrix and local binary pattern to obtain the texture complexity of the tobacco leaf;

[0103] The tobacco leaf region sub-image is processed by the Canny edge detection algorithm to obtain the shape and edge information of the tobacco leaf;

[0104] The neural network analysis module is used to input the color distribution characteristics, the texture complexity, the shape and edge information into the neural network model to obtain TMV infection status data of the tobacco plants in the ground area, including:

[0105] The color distribution characteristics, the texture complexity, the shape and edge information are input into the convolutional neural network model to obtain the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area.

[0106] In another embodiment, the disease distribution map generating module is used to generate a disease distribution map of the ground area based on the TMV infection status data, including:

[0107] The distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area are associated with the map of the ground area through GIS to generate a disease distribution map of the ground area;

[0108] The early warning and disease prediction module is used to provide early warning and disease development trend prediction based on the disease distribution map, including:

[0109] Based on the disease distribution map, the sub-areas of the ground area where tobacco failure may occur are determined, so as to send early warning information to the target objects; based on the disease distribution map and the environmental data corresponding to the ground area, the TMV damage spread trend information of the ground area is predicted.

[0110] The function and effect of the TMV early remote sensing identification system based on RGB images provided in this application are the same as those of the TMV early remote sensing identification method based on RGB images provided above, and will not be repeated here.

[0111] In general, the RGB image-based TMV early remote sensing identification method and system plans the flight route of a UAV equipped with a high-resolution RGB camera, controls the UAV along the flight route, and controls the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area, pre-processes the RGB image to obtain a tobacco leaf area sub-image; determines the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-image; inputs the color distribution characteristics, texture complexity, shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area; based on the TMV infection status data Based on the disease distribution map, early warning and disease development trend prediction are carried out. The UAV platform can quickly cover a large area of ​​tobacco planting area to achieve fully automated disease detection, greatly reduce the need for manual intervention and improve detection efficiency. Ordinary RGB cameras are used for disease detection to reduce hardware costs and can be applied to tobacco planting of different scales and types. With the help of image processing technology and convolutional neural networks, the early symptoms of TMV can be accurately identified, early warnings can be issued in real time, and prevention and control measures can be taken in time to reduce economic losses. Disease information can be easily obtained through the corresponding interface, which is convenient for field management and the use of different agricultural scenarios.

[0112] The above is only a specific implementation of the present invention, and any other improvements made based on the concept of the present invention are considered to be within the protection scope of the present invention.

Claims

1. The TMV early remote sensing recognition method based on RGB image is characterized by: include: Planning a flight path of a drone equipped with a high-resolution RGB camera, controlling the drone to fly along the flight path, and controlling the high-resolution RGB camera to collect RGB images of tobacco plants in a ground area; Preprocessing the RGB image to obtain a tobacco leaf region sub-image; Determining the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-image; inputting the color distribution characteristics, texture complexity, shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area; Based on the TMV infection status data, generating a disease distribution map of the ground area; Based on the disease distribution map, early warning and disease development trend prediction are carried out.

2. The TMV early remote sensing identification method based on RGB images as claimed in claim 1, characterized in that: Planning a flight path of a drone equipped with a high-resolution RGB camera, controlling the drone to fly along the flight path, and controlling the high-resolution RGB camera to collect RGB images of tobacco plants in a ground area; The RGB image is preprocessed, including: Acquire boundary information of a ground area through GIS, and divide the ground area into a plurality of parallel flight routes based on the boundary information; determine a flight altitude and a flight speed matching each flight route based on the terrain undulations and vegetation density of the ground area; Determine the shooting field width of the high-resolution RGB camera based on the spacing between different flight routes; control the UAV to fly along the corresponding flight route and its matching flight altitude and flight speed, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area; The RGB image is subjected to denoising, color correction and image segmentation, so as to segment the tobacco leaf region in the RGB image from the background and obtain a tobacco leaf region sub-image.

3. The TMV early remote sensing identification method based on RGB image as claimed in claim 1, characterized in that: Determining the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf area sub-image; inputting the color distribution characteristics, the texture complexity, the shape and edge information into a neural network model to obtain TMV infection status data of the tobacco plants in the ground area, including: Extracting a color histogram of the tobacco leaf region sub-image, and extracting color distribution features of the tobacco leaf from the color histogram; Performing gray-level co-occurrence matrix and local binary pattern processing on the tobacco leaf region sub-image to obtain the texture complexity of the tobacco leaf; Processing the tobacco leaf region sub-image with a Canny edge detection algorithm to obtain shape and edge information of the tobacco leaf; The color distribution characteristics, the texture complexity, and the shape and edge information are input into a convolutional neural network model to obtain the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area.

4. The TMV early remote sensing identification method based on RGB images as claimed in claim 1, characterized in that: Based on the TMV infection status data, generating a disease distribution map of the ground area; Based on the disease distribution map, early warning and disease development trend prediction are carried out, including: Associating the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area with the map of the ground area through GIS to generate a disease distribution map of the ground area; Based on the disease distribution map, sub-areas of the ground area where tobacco failure may occur are determined, so as to send early warning information to the target object; based on the disease distribution map and the environmental data corresponding to the ground area, the TMV damage diffusion trend information of the ground area is predicted.

5. The TMV early remote sensing recognition system based on RGB images is characterized by: include: A UAV flight and photography control module, used to plan a flight route of a UAV equipped with a high-resolution RGB camera, control the UAV along the flight route, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area; An RGB image preprocessing module, used for preprocessing the RGB image to obtain a tobacco leaf region sub-image; A tobacco leaf feature determination module, used to determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image; A neural network analysis module, used for inputting the color distribution characteristics, the texture complexity, the shape and edge information into a neural network model to obtain TMV infection status data of tobacco plants in the ground area; A disease distribution map generating module, used for generating a disease distribution map of the ground area based on the TMV infection status data; The early warning and disease prediction module is used to provide early warning and disease development trend prediction based on the disease distribution map.

6. The TMV early remote sensing recognition system based on RGB images as claimed in claim 5, characterized in that: The UAV flight and shooting control module is used to plan the flight route of the UAV equipped with a high-resolution RGB camera, control the UAV along the flight route, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area, including: Acquire boundary information of a ground area through GIS, and divide the ground area into a plurality of parallel flight routes based on the boundary information; determine a flight altitude and a flight speed matching each flight route based on the terrain undulations and vegetation density of the ground area; Determine the shooting field width of the high-resolution RGB camera based on the spacing between different flight routes; control the UAV to fly along the corresponding flight route and its matching flight altitude and flight speed, and control the high-resolution RGB camera to collect RGB images of tobacco plants in the ground area; The RGB image preprocessing module is used to preprocess the RGB image to obtain a tobacco leaf region sub-image, including: The RGB image is subjected to denoising, color correction and image segmentation, so as to segment the tobacco leaf region in the RGB image from the background and obtain a tobacco leaf region sub-image.

7. The TMV early remote sensing recognition system based on RGB images as claimed in claim 5, characterized in that: The tobacco leaf feature determination module is used to determine the color distribution characteristics, texture complexity, shape and edge information of the tobacco leaf region sub-image, including: Extracting a color histogram of the tobacco leaf region sub-image, and extracting color distribution features of the tobacco leaf from the color histogram; Performing gray-level co-occurrence matrix and local binary pattern processing on the tobacco leaf region sub-image to obtain the texture complexity of the tobacco leaf; Processing the tobacco leaf region sub-image with a Canny edge detection algorithm to obtain shape and edge information of the tobacco leaf; The neural network analysis module is used to input the color distribution characteristics, the texture complexity, the shape and edge information into the neural network model to obtain the TMV infection status data of the tobacco plants in the ground area, including: The color distribution characteristics, the texture complexity, and the shape and edge information are input into a convolutional neural network model to obtain the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area.

8. The TMV early remote sensing recognition system based on RGB images as claimed in claim 5, characterized in that: The disease distribution map generating module is used to generate a disease distribution map of the ground area based on the TMV infection status data, including: Associating the distribution data of tobacco plants infected with TMV and the number of leaf lesions in the ground area with the map of the ground area through GIS to generate a disease distribution map of the ground area; The early warning and disease prediction module is used to perform early warning and disease development trend prediction based on the disease distribution map, including: Based on the disease distribution map, sub-areas of the ground area where tobacco failure may occur are determined, so as to send early warning information to the target object; based on the disease distribution map and the environmental data corresponding to the ground area, the TMV damage diffusion trend information of the ground area is predicted.

Images