Intelligent checking method, device and equipment for actual planting condition of tobacco leaves and storage medium

Abstract

The invention relates to the technical field of tobacco leaf planting condition checking, and discloses an intelligent checking method, device and equipment for the actual tobacco leaf planting condition and a storage medium. A low-altitude unmanned aerial vehicle, big data geographic information service and artificial intelligence serve as technical main lines, land parcels serve as the core, multiple convenient and rapid route planning modes are designed, and the actual tobacco leaf planting condition is checked. An unmanned aerial vehicle automatically completes tobacco field ortho-photograph collection and tobacco field boundary and plant number recognition, boundary conversion and calculation are rapidly achieved through spatial information big data analysis and calculation, actual tobacco field planting geographic coordinates are obtained, spatial location calculation and analysis are conducted on existing plots bound with a contract and plots for planning tobacco planting, and the tobacco field planting quality is improved. And obtaining planting coverage information. The efficient operation mode can guarantee large-scale comprehensive checking, the collection process is basically automatic, a large amount of professional staff investment is not needed, the labor cost is reduced, the intelligent recognition calculation process is achieved, excessive human intervention is reduced, and the accuracy and authenticity of data results are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of verification of tobacco leaf planting conditions, and particularly to an intelligent verification method, device, equipment and storage medium for the actual tobacco leaf planting conditions. Background Art

[0002] Tobacco is an important cash crop and tax source in China. Efficiently and accurately verifying the actual tobacco leaf planting conditions to ensure sufficient and solid planting area is one of the basic tasks for the standardized management of flue-cured tobacco contracts, and is of great significance for controlling the tobacco leaf yield. In recent years, with the continuous development of the digital transformation of tobacco agriculture, the traditional verification methods for the actual tobacco leaf planting conditions can no longer meet the modern tobacco leaf production work.

[0003] Most of the traditional verifications of the actual tobacco leaf planting conditions are carried out manually on-site after randomly selecting some tobacco fields. By measuring methods such as a seedling measuring instrument and a tape measure, the planting area is measured plot by plot, and the plant spacing density is counted, etc. This method for verifying the actual tobacco leaf planting conditions has the following defects: (1) The manual on-site verification has low work efficiency and takes a long time, and a large amount of manpower and material resources need to be invested; (2) Most of the traditional verifications randomly select some fields for verification, and it is difficult to ensure the integrity of the data by the sampling method; (3) The existing UAV verification methods have relatively high professional requirements, requiring the ability to operate the UAV for data collection, and the collection flight routes are single; (4) The method of manually measuring the plot area and counting the number of planted plants on-site is prone to errors during the process, resulting in incorrect data results, and it is impossible to ensure the accuracy and authenticity of the data.

[0004] Therefore, how to solve the problems of low work efficiency, long time consumption, large personnel investment and low verification quality in the verification of the actual tobacco leaf planting conditions is a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The present invention provides an intelligent verification method, device, equipment and storage medium for the actual tobacco leaf planting conditions, aiming to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides an intelligent verification method for the actual tobacco leaf planting conditions, including:

[0007] Collecting tobacco field training image data and training a tobacco field boundary recognition model and a pond cleaning and plant counting model;

[0008] Deploying the tobacco field boundary recognition model and the pond cleaning and plant counting model on a tobacco field recognition server;

[0009] Generating a UAV flight collection route for the tobacco fields in the target area according to the tobacco field plot range of the target area, and performing UAV flight and orthophoto collection to obtain a set of tobacco field images to be verified;

[0010] Import the tobacco field image set to be verified into the tobacco field recognition server, and use the multi-threaded concurrent recognition method to drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images of the tobacco field in the tobacco field image set to be verified, so as to obtain the tobacco field boundary pixel result and the tobacco plant pixel position;

[0011] Convert the tobacco field boundary pixel result and the tobacco plant pixel position into the spatial location of the tobacco field and the tobacco plant, and perform spatial location calculation and analysis based on the existing contract-bound plots and the planned tobacco-growing plot patches to obtain the verification result of the actual tobacco planting situation.

[0012] Optionally, collect tobacco field training image data and train the tobacco field boundary recognition model and the clearing pond point plant model, specifically including:

[0013] Collect tobacco field training image data; wherein, the tobacco field training image data includes the first tobacco field training image data collected by the unmanned aerial vehicle at the first height in the training tobacco field area and the second tobacco field training image data collected at the second height in the training tobacco field area, and the first height is greater than the second height;

[0014] Use the first tobacco field training image data to train and obtain the tobacco field boundary recognition model, and based on the tobacco field boundary recognition model and the second tobacco field training image data, construct the clearing pond point plant model.

[0015] Optionally, use the first tobacco field training image data to train and obtain the tobacco field boundary recognition model, and based on the tobacco field boundary recognition model and the second tobacco field training image data, construct the clearing pond point plant model, specifically including:

[0016] Screen the collected first tobacco field training image data at a preset photo-taking time interval, and label the pixel boundaries of the tobacco field range in the screened tobacco field training images to form single or multiple tobacco field plot boundary data;

[0017] Divide the data set according to a ratio, and based on the DETR target instance segmentation model and the backbone network of ConvNext tiny, perform tobacco field boundary recognition training to obtain the tobacco field boundary recognition model;

[0018] Input the second tobacco field training image data into the tobacco field boundary recognition model, segment the tobacco field in each second tobacco field training image, and label the tobacco plants in each tobacco field;

[0019] Divide the labeled second tobacco field training images according to a ratio, establish a yolov5 model based on the convolutional attention mechanism, and perform tobacco field tobacco plant recognition training to obtain the clearing pond point plant model.

[0020] Optionally, according to the range of tobacco fields in the target area, generate a UAV flight acquisition route for the tobacco fields in the target area and execute UAV flight and orthophoto acquisition to obtain a set of tobacco field images to be verified, specifically including:

[0021] Obtain the range of tobacco fields and DEM image data in the target area, and plan the UAV flight acquisition route and route flight parameters;

[0022] Determine the DEM elevation value through each route node in the UAV flight acquisition route and set the UAV route flight parameters using the route flight parameters. Based on the principle of a constant height difference between the UAV and surface features, calculate the flight height of each route node;

[0023] Drive the UAV to perform the acquisition of tobacco field images to be verified at each route node in the flight acquisition route according to the route flight parameters and flight height.

[0024] Optionally, plan the UAV flight acquisition route, specifically including:

[0025] For tobacco fields in the target area with target plot vector boundary data: Batch-select the polygon vector boundary data of the plots to be verified. According to the polygon vector boundary data of the plots, calculate the coordinates of each vertex of the plot polygon, and then calculate the arithmetic mean of all vertices to obtain the center point coordinates of the plot. Determine each route node of the UAV flight acquisition route with the center point coordinates;

[0026] For tobacco fields in the target area without target plot vector boundaries and with a scattered planting distribution: According to the actual on-site planting situation and combined with the existing remote sensing image map, click on the target points of the tobacco-growing plots on the map, and automatically extract the coordinates of these points as route nodes;

[0027] For tobacco fields without target plot vector boundaries and with a contiguous planting distribution: Outline several planting areas, calculate the minimum bounding matrix of each area, calculate the center point coordinates of each minimum bounding matrix, and determine each route node of the UAV flight acquisition route with the center point coordinates.

[0028] Optionally, import the set of tobacco field images to be verified into the tobacco field recognition server, and use the multi-threaded concurrent recognition method to drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images in the set of tobacco field images to be verified, and obtain the tobacco field boundary pixel results and tobacco plant pixel positions, specifically:

[0029] Import the set of tobacco field images to be verified into the tobacco field recognition server, and drive the tobacco field recognition server to distribute several tobacco field images to be verified in the set of tobacco field images to be verified to each processing queue according to the current queuing situation of each processing queue, so that the completion time of each processing queue is the same;

[0030] Based on Redis and the lock mechanism, multi-threaded concurrent recognition control is performed on each processing queue to obtain the tobacco field boundary pixel results and the pixel positions of tobacco plants. Specifically, it includes the following steps:

[0031] Lock mechanism: Obtain the lock through lock.lock to ensure that the image processing of the same processing queue pipeline will not be carried out simultaneously, avoiding resource contention. If the lock acquisition fails, return directly to avoid blocking other tasks;

[0032] Obtain image information from Redis: Use redisUtil.getHashAll to obtain the images to be recognized from the Redis cache. If there is no image data, return directly;

[0033] Call image recognition: Send the image to the AI recognition service by calling the recognizeImage method; if the recognition result is empty, indicating failure, set the server status serverStatus to false;

[0034] Process the recognition result: If there is a recognition result, parse the JSON data returned by the tobacco field recognition server, and extract the edge data in the image and the pixel points of the tobacco plants. If the data is valid, call processRecognitionResult for further processing and save it to the database;

[0035] Release the lock: Ensure that the lock is released in finally regardless of whether the process ends normally.

[0036] Optionally, convert the tobacco field boundary pixel results and the pixel positions of tobacco plants into the spatial location of the tobacco field and tobacco plants, and perform spatial location calculation and analysis based on the existing contract-bound plots and the planned tobacco-growing plot patches to obtain the verification results of the actual tobacco-growing situation. Specifically, it includes:

[0037] Obtain the image size, sensor size, focal length, and flight altitude from the photo and sensor, and calculate the ground sampling distance gsd, which is used to represent the actual ground distance corresponding to each pixel;

[0038] For each pixel point, convert its offset relative to the center into the actual ground distance, and calculate the longitude and latitude coordinates corresponding to the pixel according to the azimuth angle theta. Construct the set of longitude and latitude coordinates after conversion of all polygon pixel points into a polygon object, and use it as the tobacco field plot polygon;

[0039] Based on the existing contract-bound plots and the planned tobacco plot patches, search for matching plot data according to the center point of the tobacco field plot polygon. If there is, update the plot information according to the existing recognition result. If not, generate a new plot and insert it into the database together with the plot information.

[0040] In addition, to achieve the above object, the present invention further provides an intelligent verification device for the actual tobacco planting situation, including:

[0041] A training module, configured to collect tobacco field training image data and train a tobacco field boundary recognition model and a pond cleaning point plant model;

[0042] A deployment module, configured to deploy the tobacco field boundary recognition model and the pond cleaning point plant model on a tobacco field recognition server;

[0043] A collection module, configured to generate a UAV flight collection route for the tobacco field in the target area according to the tobacco field plot range of the target area, execute UAV flight and orthophoto collection, and obtain a set of tobacco field images to be verified;

[0044] An identification module, configured to import the set of tobacco field images to be verified into the tobacco field recognition server, and drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images of the tobacco field in the set of tobacco field images to be verified in a multi-threaded concurrent identification manner, so as to obtain a tobacco field boundary pixel result and a tobacco plant pixel position;

[0045] A verification module, configured to convert the tobacco field boundary pixel result and the tobacco plant pixel position into the spatial location positions of the tobacco field and the tobacco plants, perform spatial location calculation and analysis based on the existing contract-bound plots and the planned tobacco-growing plot patches, and obtain a verification result of the actual tobacco planting situation.

[0046] In addition, to achieve the above object, the present invention further provides an intelligent verification device for the actual tobacco planting situation. The intelligent verification device for the actual tobacco planting situation includes: a memory, a processor, and an intelligent verification program for the actual tobacco planting situation stored on the memory and executable on the processor. When the intelligent verification program for the actual tobacco planting situation is executed by the processor, the steps of the intelligent verification method for the actual tobacco planting situation described in any one of the above are implemented.

[0047] In addition, to achieve the above object, the present invention further provides a storage medium. An intelligent verification program for the actual tobacco planting situation is stored on the storage medium. When the intelligent verification program for the actual tobacco planting situation is executed by a processor, the steps of the intelligent verification method for the actual tobacco planting situation described in any one of the above are implemented.

[0048] The beneficial effects of the present invention are as follows: A method, device, equipment and storage medium for intelligent verification of the actual tobacco planting situation are proposed. By taking low-altitude drones, big data geographic information services and artificial intelligence as the main technical lines and the plot as the core, a variety of convenient and fast route planning methods are designed to enable the drone to automatically complete the acquisition of orthophoto images of tobacco fields, the identification of tobacco field boundaries and the number of planted plants. Through spatial information big data analysis and calculation, the boundary conversion and calculation are quickly realized, and the actual planting geographic coordinates of the tobacco field are obtained. At the same time, spatial location calculation and analysis are carried out with the existing contract-bound plots and the planned tobacco-growing plot patches to obtain the corresponding planting coverage information. The efficient operation method can ensure comprehensive verification on a large scale. The acquisition process is basically automated, without the need for a large number of professional staff, reducing labor costs. The intelligent identification and calculation process reduces excessive human intervention and can ensure the accuracy and authenticity of the data results. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment solution of the present invention;

[0050] Figure 2 It is a schematic flow chart of the embodiment of the intelligent verification method for the actual tobacco planting situation of the present invention;

[0051] Figure 3 It is an application example diagram for identifying the tobacco field boundary of the present invention;

[0052] Figure 4 It is an application example diagram for identifying the clear pond and plant count of the present invention;

[0053] Figure 5 It is one of the application example diagrams for spatial location analysis of the contract-bound plots of the present invention;

[0054] Figure 6 It is the second application example diagram for spatial location analysis of the contract-bound plots of the present invention;

[0055] Figure 7 It is a structural block diagram of the embodiment of the intelligent verification device for the actual tobacco planting situation of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0058] As Figure 1 shown, Figure 1 is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment solution of the present invention.

[0059] As Figure 1 shown, the device may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0060] Those skilled in the art can understand that Figure 1 the structure of the device shown in

[0061] does not constitute a limitation on the device, and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Figure 1 shown, the memory 1005, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a smart verification program for the actual tobacco planting situation.

[0062] In Figure 1 the terminal shown, the network interface 1004 is mainly used to connect to the background server and perform data communication with the background server; the user interface 1003 is mainly used to connect to the client (user side) and perform data communication with the client; and the processor 1001 may be used to call the smart verification program for the actual tobacco planting situation stored in the memory 1005 and perform the following operations:

[0063] Collect tobacco field training image data and train a tobacco field boundary recognition model and a pond cleaning point plant model;

[0064] Deploy the tobacco field boundary recognition model and the pond cleaning point plant model on the tobacco field recognition server;

[0065] Generate a UAV flight acquisition route for the tobacco fields in the target area according to the range of the tobacco field plots in the target area, and execute UAV flight and orthophoto acquisition to obtain a set of tobacco field images to be verified;

[0066] Import the set of tobacco field images to be verified into the tobacco field recognition server, and use the multi-threaded concurrent recognition method to drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images of tobacco fields in the set of tobacco field images to be verified, so as to obtain the tobacco field boundary pixel result and the tobacco plant pixel position;

[0067] Convert the tobacco field boundary pixel result and the tobacco plant pixel position into the spatial location of the tobacco field and the tobacco plant, and perform spatial location calculation and analysis based on the existing contract-bound plots and the planned tobacco-growing plot patches to obtain the verification result of the actual tobacco leaf planting situation.

[0068] The specific embodiments of the present invention applied to the device are basically the same as those of the following embodiments of the intelligent verification method for the actual tobacco leaf planting situation, and will not be elaborated here.

[0069] An embodiment of the present invention provides an intelligent verification method for the actual tobacco leaf planting situation, referring to Figure 2 , Figure 2 is a schematic flowchart of the intelligent verification method for the actual tobacco leaf planting situation in the embodiment of the present invention. The intelligent verification method for the actual tobacco leaf planting situation includes the following steps:

[0070] S100: Collect tobacco field training image data, and train a tobacco field boundary recognition model and a pond cleaning point plant model;

[0071] S200: Deploy the tobacco field boundary recognition model and the pond cleaning point plant model on the tobacco field recognition server;

[0072] S300: Generate a UAV flight acquisition route for the tobacco fields in the target area according to the range of the tobacco field plots in the target area, and execute UAV flight and orthophoto acquisition to obtain a set of tobacco field images to be verified;

[0073] S400: Import the set of tobacco field images to be verified into the tobacco field recognition server, and use the multi-threaded concurrent recognition method to drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images of tobacco fields in the set of tobacco field images to be verified, so as to obtain the tobacco field boundary pixel result and the tobacco plant pixel position;

[0074] S500: Convert the tobacco field boundary pixel result and the tobacco plant pixel position into the spatial location of the tobacco field and the tobacco plant, and perform spatial location calculation and analysis based on the existing contract-bound plots and the planned tobacco-growing plot patches to obtain the verification result of the actual tobacco leaf planting situation.

[0075] It should be noted that the current verification method for the actual tobacco leaf planting situation has the following defects: (1) The manual on-site verification has low work efficiency and takes a long time, requiring a large amount of manpower and material resources; (2) Most traditional verifications randomly select some plots for verification, and it is difficult to ensure the integrity of data by sampling; (3) The existing UAV verification method has relatively high professional requirements, requiring the ability to operate UAVs for data collection and having a single collection route; (4) The method of manually measuring the plot area and counting the number of planted plants on-site is prone to errors during the process, resulting in incorrect data results and unable to ensure the accuracy and authenticity of the data.

[0076] In this embodiment, with low-altitude UAVs, big data geographic information services, and artificial intelligence as the technical main lines and plots as the core, a variety of convenient and fast route planning methods are designed to enable the UAV to automatically complete the collection of orthophoto images of tobacco fields, the identification of tobacco field boundaries and the number of planted plants. Through spatial information big data analysis and calculation, the boundary conversion and calculation are quickly realized, and the actual planting geographic coordinates of the tobacco fields are obtained. At the same time, spatial location calculation and analysis are carried out with the existing contract-bound plots and the mapped patches of planned tobacco-growing plots to obtain the corresponding planting coverage information. The efficient operation method can ensure comprehensive verification on a large scale. The collection process is basically automated, without the need for a large number of professional staff to be invested, reducing labor costs. The intelligent identification and calculation process reduces excessive human intervention and can ensure the accuracy and authenticity of the data results.

[0077] To explain the present invention more clearly, the following provides specific examples of the intelligent verification method for the actual tobacco leaf planting situation of the present invention, which includes the following specific implementation steps:

[0078] Step 1.1, collect low-altitude real-scene photos of the UAV, covering all stages of the entire growth cycle of tobacco leaves and the situation of plastic films on the tobacco ridges, etc. (Advantage: Construct a dataset completely based on the production environment, and the model trained using this dataset will also achieve good results in the production environment);

[0079] Step 1.2, screen the collected UAV photos at a certain interval (4 - 5 photos);

[0080] Step 1.3, mark the pixel boundaries of the tobacco field range to form single or multiple tobacco field boundary data;

[0081] Step 1.4, divide the dataset in a ratio of 8:2, with 80% of the pictures as the training set and 20% as the validation set. Based on the DETR-based object instance segmentation model, select ConvNext tiny as the backbone network, train 200 times, and train the tobacco field boundary recognition model to obtain the tobacco field boundary recognition model;

[0082] Step 2.1, collect photos with the UAV flight altitude below 60m, input the pictures into the tobacco field boundary recognition model generated in Step 1, and segment the tobacco fields;

[0083] Step 2.2, intercept the largest rectangle of each tobacco field, and convert all non-tobacco field data to 0;

[0084] Step 2.3, label the tobacco plants in each tobacco field (Advantage: use the tobacco field recognition model to segment the tobacco fields, and then conduct pond cleaning and plant counting for each tobacco field to achieve full detection from the surface to the points);

[0085] Step 2.4, improve the yolov5 model by adding a convolutional attention mechanism to enhance the model's feature extraction ability; divide the dataset in the ratio of 8:2, with 80% of the pictures as the training set and 20% as the validation set. Based on the improved yolov5 model, train the pond cleaning and plant counting model to obtain the pond cleaning and plant counting model;

[0086] Step 3.1, deploy the test model and write the calling API interface in combination with the business;

[0087] Step 3.2, after the docking test is completed, export the deployed model to the application server;

[0088] Step 4.1, obtain the data collection area and the target plot (in actual verification, there are 3 situations. One is that there is vector boundary data of the target plot, one is that there is no vector boundary of the target plot and the planting distribution is scattered, and one is that there is no vector boundary of the target plot and the planting distribution is contiguous; Advantage: specifically plan the flight route according to the actual planting situation, which can save verification time and cost and avoid unnecessary redundant collection);

[0089] Step 4.2, obtain the DEM image data of the target area;

[0090] Step 4.3, plan the UAV flight collection route according to the area and the plot (plan the route for the 3 situations in Step 3.1 respectively. For the first situation, batch select the vector boundary data of the plots that need to be verified, and automatically generate the center point of each plot as the collection flight point (Method: first calculate the coordinates of each vertex of the plot polygon vector boundary data, and then calculate the arithmetic mean of all vertices to obtain the center point coordinates of the plot); for the second situation, according to the actual planting situation on site, combine the existing remote sensing image map, click on the target point of the tobacco-growing plot on the map, and automatically extract the coordinates of this point as the collection flight point; for the third situation, circle the planting area, calculate the minimum circumscribed rectangle of the target area, and automatically generate the orthoimage flight route);

[0091] Step 4.4, set the flight parameters (flight altitude, speed, overlap rate, etc.);

[0092] Step 4.5, calculate and obtain the DEM elevation value through the route nodes (photo-taking position points), calculate the flight altitude of each flight point, so that the aircraft always maintains a unified height with the surface features, and achieve "flying along the surface".

[0093] Step 4.6, select the current UAV model for collection (related to UAV imaging parameters);

[0094] Step 4.7, complete takeoff and collect orthophotos;

[0095] Step 4.8, pull and upload the photos to the server;

[0096] Step 5.1, import the photos collected in Step 4 into the model deployed in Step 3, automatically complete the identification of the tobacco field boundary and the positions of tobacco plants, and obtain the pixel results of the tobacco field boundary and the pixel positions of tobacco plants; (Advantage: Process the pictures in the form of a pipeline, and send them to the AI recognition service to obtain the recognition results. After processing the recognition results, store the results in the database, and use Redis and lock mechanisms to manage concurrency to ensure that multiple pipelines do not conflict.)

[0097] The detailed technology is as follows:

[0098] ① Lock mechanism: Obtain the lock through lock.lock to ensure that the picture processing of the same pipeline will not be carried out simultaneously, and avoid resource contention. If the lock acquisition fails, return directly to avoid blocking other tasks.

[0099] ② Obtain picture information from Redis: Use redisUtil.getHashAll to obtain the pictures to be recognized from the Redis cache. If there is no picture data, return directly.

[0100] ③ Call picture recognition: Send the picture to the AI recognition service by calling the recognizeImage method. If the recognition result is empty, it means failure, and set the server status serverStatus to false.

[0101] ④ Process the recognition result: If there is a recognition result, parse the JSON data returned by the AI and extract the edge data and pixel points of tobacco plants in the picture. If the data is valid, call processRecognitionResult for further processing (such as saving to the database).

[0102] ⑤ Release the lock: Ensure that the lock is released in finally regardless of whether the process ends normally or not.

[0103] Such as Figure 3 and Figure 4As shown, they are respectively the application example diagrams of tobacco field boundary recognition (the upper two diagrams are input images, and the lower two diagrams are the corresponding images output by the tobacco field boundary recognition model) and the application example diagrams of clearing pond plant recognition (the upper two diagrams are input images, and the lower two diagrams are the corresponding images output by the clearing pond plant recognition model).

[0104] Step 5.2: Use the pixel-to-coordinate algorithm to convert the tobacco field boundary pixels and tobacco plant pixels to obtain the spatial location of the tobacco field and tobacco plants;

[0105] The detailed calculation steps are as follows:

[0106] ① First, obtain necessary parameters such as image size, sensor size, focal length, and flight altitude from the photo and sensor;

[0107] ② Calculate the ground sampling distance gsd, which represents the actual ground distance corresponding to each pixel;

[0108] ③ For each pixel point, convert its offset (x and y) relative to the center to the actual ground distance d, and calculate the longitude and latitude coordinates corresponding to the pixel according to its azimuth angle theta;

[0109] ④ Construct a polygon object, that is, the tobacco field plot polygon, from the set of longitude and latitude coordinates obtained by converting all polygon pixel points;

[0110] ⑤ If the generated polygon is invalid or empty, try to repair the topology and regenerate a valid polygon;

[0111] Step 6.1: Conduct spatial location calculation and analysis based on the existing contract-bound plots and planned tobacco-growing plot patches

[0112] The detailed calculation steps are as follows:

[0113] ① Based on the existing contract-bound plots and planned tobacco-growing plot patches, search for matching plot data according to the polygon center point. If there is a match, update the relevant plot information according to the existing recognition results. If no matching plot is found, generate a new plot and insert it into the database together with relevant information (such as area, tobacco-growing status, etc.);

[0114] ② Through spatial location calculation and analysis, statistically calculate in real time the tobacco-growing area of the contract-bound plots, the non-tobacco-growing area of the contract-bound plots, the tobacco-growing area within the planned tobacco-growing plots outside the contract-bound plots, and the tobacco-growing area outside the planned tobacco-growing plots outside the contract-bound plots, and present them in the form of "one map" to assist in the adjustment and assessment management of tobacco leaf production contracts and provide analysis and decision-making.

[0115] As Figure 5 and Figure 6 shown, Figure 5It is an application example diagram for the spatial location analysis of contract-bound plots (the three diagrams are respectively the photos collected by the drone, the output image of model recognition, and the analysis result is the area where tobacco is not planted within the contract binding. As shown in the figure, the area marked in red is the area where tobacco is not planted within the contract binding). Figure 6 It is another application example diagram for the spatial location analysis of contract-bound plots (the three diagrams are respectively the photos collected by the drone, the output image of model recognition, and the analysis result is the area where tobacco is planted outside the contract binding. As shown in the figure, the area marked in pink is the area where tobacco is planted outside the contract binding).

[0116] In this embodiment, considering that the traditional verification method for the actual tobacco planting situation can no longer meet the modern tobacco production work, it not only has low work efficiency, takes a long time, but also requires a large amount of manpower and material resources, and the verification result is greatly affected by humans. Based on the application of technologies such as low-altitude drones, artificial intelligence, machine learning, cloud computing, big data, global positioning technology, and geographic information system, using the JAVA development language and building with the PostgreSQL database, a method, device, equipment, and storage medium for intelligent verification of the actual tobacco planting situation are provided, which realizes taking the plot as the core, designing a variety of convenient and fast flight route planning methods, enabling the drone to automatically complete the patrol inspection and AI recognition calculation, and providing real-time, efficient, and accurate verification services for the actual tobacco planting situation in the industry. The intelligent verification method can not only greatly improve the work efficiency, reduce the personnel cost, but also reduce excessive human intervention and improve the accuracy and authenticity of the verification data.

[0117] Refer to Figure 7 , Figure 7 It is the structural block diagram of the embodiment of the intelligent verification device for the actual tobacco planting situation of the present invention.

[0118] As Figure 7 shown, the intelligent verification device for the actual tobacco planting situation proposed in the embodiment of the present invention includes:

[0119] The training module 10 is used to collect the training image data of the tobacco field and train the tobacco field boundary recognition model and the pond cleaning and single-plant models;

[0120] The deployment module 20 is used to deploy the tobacco field boundary recognition model and the pond cleaning and single-plant models on the tobacco field recognition server;

[0121] The acquisition module 30 is used to generate a drone flight acquisition route for the tobacco field in the target area according to the tobacco field plot range of the target area and execute the drone flight and orthophoto acquisition to obtain the image set of the tobacco field to be verified;

[0122] An identification module 40, configured to import a tobacco field image set to be verified into a tobacco field identification server, and drive the tobacco field identification server to perform tobacco field boundary identification and tobacco plant position identification on several orthophoto images of the tobacco field in the tobacco field image set to be verified in a multi-threaded concurrent identification manner, so as to obtain a tobacco field boundary pixel result and a tobacco plant pixel position;

[0123] A verification module 50, configured to convert the tobacco field boundary pixel result and the tobacco plant pixel position into spatial location positions of the tobacco field and the tobacco plants, perform spatial location calculation and analysis based on existing contract-bound plots and planned tobacco-growing plot patches, and obtain a verification result of the actual tobacco leaf planting situation.

[0124] Other embodiments or specific implementation manners of the intelligent verification device for the actual tobacco leaf planting situation of the present invention may refer to the above method embodiments, and will not be elaborated here.

[0125] In addition, the present invention also provides an intelligent verification device for the actual tobacco leaf planting situation, where the intelligent verification device for the actual tobacco leaf planting situation includes: a memory, a processor, and an intelligent verification program for the actual tobacco leaf planting situation stored on the memory and executable on the processor. When the intelligent verification program for the actual tobacco leaf planting situation is executed by the processor, the steps of the intelligent verification method for the actual tobacco leaf planting situation as described above are implemented.

[0126] The specific implementation manner of the intelligent verification device for the actual tobacco leaf planting situation of the present application is basically the same as that of the above embodiments of the intelligent verification method for the actual tobacco leaf planting situation, and will not be elaborated here.

[0127] In addition, the present invention also provides a readable storage medium, where the readable storage medium includes a computer-readable storage medium, on which an intelligent verification program for the actual tobacco leaf planting situation is stored. The readable storage medium may be Figure 1 the memory 1005 in a terminal, or at least one of ROM (Read-Only Memory), RAM (Random Access Memory), magnetic disk, and optical disk. The readable storage medium includes several instructions for causing an intelligent verification device for the actual tobacco leaf planting situation with a processor to execute the intelligent verification method for the actual tobacco leaf planting situation described in each embodiment of the present invention.

[0128] The specific implementation manner in the readable storage medium of the present application is basically the same as that of the above embodiments of the intelligent verification method for the actual tobacco leaf planting situation, and will not be elaborated here.

[0129] It should be understood that in the description of this specification, the descriptions with reference to terms such as "one embodiment", "another embodiment", "other embodiments", or "the first embodiment to the Nth embodiment" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0130] It should be noted that in this text, the term "comprising", "including", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or system comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article, or system. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article, or system comprising that element.

[0131] The serial numbers of the above embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments.

[0132] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc) as described above and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0133] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An intelligent verification method for the actual tobacco leaf planting situation, characterized in that, Including: Collecting tobacco field training image data, and training a tobacco field boundary recognition model and a pond cleaning and plant point model; Deploying the tobacco field boundary recognition model and the pond cleaning and plant point model on a tobacco field recognition server; Generating a UAV flight collection route for the tobacco field in the target area, performing UAV flight and orthophoto acquisition, and obtaining a tobacco field image set to be verified; Importing the tobacco field image set to be verified into the tobacco field recognition server, and using the multi-threaded concurrent recognition method to drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images in the tobacco field image set to be verified, so as to obtain the tobacco field boundary pixel result and the tobacco plant pixel position; Converting the tobacco field boundary pixel result and the tobacco plant pixel position into the spatial location of the tobacco field and the tobacco plant, and performing spatial location calculation and analysis based on the existing contract-bound plots and the planned tobacco-growing plot patches, so as to obtain the verification result of the actual tobacco planting situation.

2. The intelligent verification method for the actual tobacco leaf planting situation according to claim 1, wherein, Collecting tobacco field training image data, and training a tobacco field boundary recognition model and a pond cleaning and plant point model, specifically including: Collecting tobacco field training image data; wherein, the tobacco field training image data includes the first tobacco field training image data collected by the UAV at the first height in the training tobacco field area and the second tobacco field training image data collected at the second height in the training tobacco field area, and the first height is greater than the second height; Training a tobacco field boundary recognition model using the first tobacco field training image data, and constructing a pond cleaning and plant point model based on the tobacco field boundary recognition model and the second tobacco field training image data.

3. The intelligent verification method for the actual tobacco leaf planting situation according to claim 2, wherein Training a tobacco field boundary recognition model using the first tobacco field training image data, and constructing a pond cleaning and plant point model based on the tobacco field boundary recognition model and the second tobacco field training image data, specifically including: Screening the collected first tobacco field training image data at a preset photo-taking time interval, and annotating the pixel boundary of the tobacco field range for the screened tobacco field training images to form single or multiple tobacco field plot boundary data; Dividing the data set according to a ratio, and performing tobacco field boundary recognition training based on the object instance segmentation model of DETR and the backbone network of ConvNext tiny to obtain a tobacco field boundary recognition model; Inputting the second tobacco field training image data into the tobacco field boundary recognition model, segmenting the tobacco field in each second tobacco field training image, and annotating the tobacco plants in each tobacco field; Dividing the annotated second tobacco field training images according to a ratio, establishing a yolov5 model based on the convolutional attention mechanism, and performing tobacco plant recognition training for the tobacco field to obtain a pond cleaning and plant point model.

4. The intelligent verification method for the actual tobacco leaf planting situation according to claim 1, characterized in that, Generating a UAV flight collection route for the tobacco field in the target area, performing UAV flight and orthophoto acquisition, and obtaining a tobacco field image set to be verified, specifically including: Obtaining the tobacco field plot range and DEM image data of the target area, and planning the UAV flight collection route and the route flight parameters; Determining the DEM elevation value through each route node in the UAV flight collection route and setting the UAV route flight parameters using the route flight parameters, and calculating the flight height of each route node based on the principle that the height difference between the UAV and the surface features is constant. Drive the drone to fly according to the flight route parameters and flight altitude, and perform the acquisition of the to-be-verified tobacco field images at each route node in the flight acquisition route of the drone.

5. The intelligent verification method for the actual tobacco leaf planting situation according to claim 4, wherein, Plan the flight acquisition route of the drone, specifically including: For the tobacco field plots with target plot vector boundary data in the target area: Batch-select the polygon vector boundary data of the plots to be verified, calculate the coordinates of each vertex of the plot polygon according to the polygon vector boundary data of the plot, and then calculate the arithmetic mean of all vertices to obtain the center point coordinates of the plot, and determine each route node of the drone flight acquisition route with the center point coordinates; For the tobacco field plots without target plot vector boundaries and with scattered planting distributions in the target area: According to the actual planting situation on site and combined with the existing remote sensing image map, click on the target points of the tobacco-growing plots on the map, and automatically extract the coordinates of these points as route nodes; For the tobacco field plots without target plot vector boundaries and with contiguous planting distributions: Outline several planting areas, calculate the minimum circumscribed matrix of each area, calculate the center point coordinates of each minimum circumscribed matrix, and determine each route node of the drone flight acquisition route with the center point coordinates.

6. The intelligent verification method for the actual tobacco leaf planting situation according to claim 1, wherein Import the to-be-verified tobacco field image set into the tobacco field recognition server, and drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images in the to-be-verified tobacco field image set in a multi-threaded concurrent recognition manner to obtain the tobacco field boundary pixel results and tobacco plant pixel positions, specifically: Import the to-be-verified tobacco field image set into the tobacco field recognition server, and drive the tobacco field recognition server to distribute several to-be-verified tobacco field images in the to-be-verified tobacco field image set to each processing queue according to the current queuing situation of each processing queue, so that the completion time of each processing queue is the same; Perform multi-threaded concurrent recognition control on each processing queue based on Redis and the lock mechanism to obtain the tobacco field boundary pixel results and tobacco plant pixel positions, specifically including the steps: Lock mechanism: Obtain the lock through lock.lock to ensure that the image processing of the same processing queue pipeline will not be carried out simultaneously, avoid resource contention, and if the lock acquisition fails, return directly to avoid blocking other tasks; Obtain image information from Redis: Use redisUtil.getHashAll to obtain the images to be recognized from the Redis cache. If there is no image data, return directly; Call image recognition: Send the image to the AI recognition service by calling the recognizeImage method; If the recognition result is empty, indicating failure, set the server status serverStatus to false; Process the recognition result: If there is a recognition result, parse the JSON data returned by the tobacco field recognition server, and extract the edge data and pixel points of the tobacco plants in the image. If the data is valid, call processRecognitionResult for further processing and save it to the database; Release the lock: Ensure that the lock is released in finally regardless of whether the process ends normally.

7. The intelligent verification method for the actual tobacco leaf planting situation according to claim 1, characterized in that, Convert the tobacco field boundary pixel results and tobacco plant pixel positions into the spatial location of the tobacco field and tobacco plants, and perform spatial location calculation and analysis based on the existing contract-bound plots and planned tobacco-growing plot patches to obtain the verification results of the actual tobacco leaf planting situation, specifically including: Obtain the image size, sensor size, focal length, and flight altitude from the photos and sensors, and calculate the ground sampling distance (gsd) to represent the actual ground distance corresponding to each pixel; For each pixel point, convert its offset relative to the center into the actual ground distance, calculate the longitude and latitude coordinates corresponding to the pixel according to the azimuth angle theta, and construct the set of longitude and latitude coordinates after conversion of all polygon pixel points into a polygon object, which is used as the tobacco field plot polygon; Based on the existing contract-bound plots and planned tobacco plot patches, search for matching plot data according to the center point of the tobacco field plot polygon. If there is, update the plot information according to the existing recognition results. If not, generate a new plot and insert it into the database together with the plot information.

8. An intelligent verification device for the actual tobacco leaf planting situation, characterized in that, Including: A training module for collecting tobacco field training image data and training a tobacco field boundary recognition model and a pond cleaning point and plant model; A deployment module for deploying the tobacco field boundary recognition model and the pond cleaning point and plant model on the tobacco field recognition server; A collection module for generating a drone flight collection route for the tobacco field in the target area according to the tobacco field plot range of the target area, executing drone flight and orthophoto collection to obtain a set of tobacco field images to be verified; An identification module for importing the set of tobacco field images to be verified into the tobacco field recognition server, and using the multi-threaded concurrent identification method to drive the tobacco field recognition server to perform tobacco field boundary recognition and tobacco plant position recognition on several orthophoto images of the tobacco field in the set of tobacco field images to be verified, and obtaining the tobacco field boundary pixel results and tobacco plant pixel positions; A verification module for converting the tobacco field boundary pixel results and tobacco plant pixel positions into the spatial location of the tobacco field and tobacco plants, and performing spatial location calculation and analysis based on the existing contract-bound plots and planned tobacco-growing plot patches to obtain the verification results of the actual tobacco leaf planting situation.

9. An intelligent verification device for the actual tobacco leaf planting situation, characterized in that, The intelligent verification device for the actual tobacco leaf planting situation includes: a memory, a processor, and an intelligent verification program for the actual tobacco leaf planting situation stored on the memory and operable on the processor. When the intelligent verification program for the actual tobacco leaf planting situation is executed by the processor, the steps of the intelligent verification method for the actual tobacco leaf planting situation according to any one of claims 1 to 7 are implemented.

10. A storage medium, characterized in that, An intelligent verification program for the actual tobacco leaf planting situation is stored on the storage medium. When the intelligent verification program for the actual tobacco leaf planting situation is executed by the processor, the steps of the intelligent verification method for the actual tobacco leaf planting situation according to any one of claims 1 to 7 are implemented.

Images