AI intelligent identification and classification method and system based on unmanned aerial vehicle image

Abstract

The invention discloses an AI intelligent identification and classification method and system based on an unmanned aerial vehicle image, and belongs to the field of image analysis. And importing the corrected image into an abnormal area searching strategy to search for the abnormal cultivated land area, importing the searched abnormal cultivated land area into an intelligent identification strategy to carry out occupation distinguishing of the abnormal cultivated land area, and carrying out occupation early warning according to the occupation distinguishing of the abnormal cultivated land area. The remote sensing image is subjected to effective noise reduction processing and accurate analysis, the intelligent recognition and classification effect is improved, and therefore the intelligent recognition and classification accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the field of image analysis, and specifically relates to an AI intelligent recognition and classification method and system based on drone images. Background Art

[0002] The "non-agriculturalization" of cultivated land refers to the use of cultivated land for production and business activities other than agricultural production, which mainly includes the following four situations. 1. The specific situation of illegal occupation of cultivated land for greening and afforestation; 2. The specific situation of illegal occupation of cultivated land for the construction of green corridors beyond the standard; 3. The specific situation of illegal occupation of cultivated land for digging fields, building lakes and landscaping; 4. Occupation of permanent basic farmland to expand nature reserves;

[0003] Remote sensing technology has the advantages of wide coverage, high efficiency, and real-time monitoring, and can provide efficient implementation information for farmland protection. At present, the domestic detection of farmland changes based on remote sensing images is still mainly based on manual visual judgment. Faced with the urgent and heavy task of detecting the "non-grainization and non-agriculturalization" of farmland, it is difficult to complete it efficiently, and it is necessary to carry out AI intelligent recognition. By comparing drone images from different periods, changes in the boundaries of cultivated land, the occupation of cultivated land, and signs of cultivated land degradation can be quickly identified. Existing technologies cannot effectively process remote sensing images during the process of intelligent recognition and classification, resulting in low intelligent recognition and classification effects, which in turn leads to poor intelligent recognition and classification accuracy. Most of the existing technologies have the above problems;

[0004] The above patents all have the problems raised by this background technology. In order to solve the problems raised by this background technology, this application designs an AI intelligent recognition and classification method and system based on drone images. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention proposes an AI intelligent recognition and classification method and system based on drone images.

[0006] To achieve the above purpose, the present invention provides the following technical solution: an AI intelligent recognition and classification method based on drone images, which includes the following specific steps:

[0007] Use drones to obtain image data within a specified range, and at the same time obtain weather data, and store them in a storage module;

[0008] Use drones to obtain remote sensing image data and weather data and import them into the image correction model for image correction processing;

[0009] The corrected image is imported into the abnormal area search strategy to search for abnormal areas of cultivated land;

[0010] The found abnormal cultivated land areas are introduced into the intelligent identification strategy to distinguish the occupation of the abnormal cultivated land areas;

[0011] Occupancy warning is issued based on the occupancy division of abnormal cultivated land areas.

[0012] It should be noted here that as a preferred technical solution for the AI ​​intelligent recognition and classification method based on drone images, the specific steps of using drones to obtain remote sensing image data, simultaneously obtaining weather condition data, and storing them in a storage module are as follows:

[0013] S11. The UAV flies along the designated flight path, obtains remote sensing images on the flight path through the camera module, and obtains weather humidity and visibility data during the flight through the weather acquisition module;

[0014] S12. Store the acquired weather data and remote sensing image data in a storage component, and simultaneously acquire pixel value data of each pixel point of the cultivated land image data.

[0015] It should be noted here that as a preferred technical solution for the AI ​​intelligent recognition and classification method based on drone images, the use of drones to obtain remote sensing image data and weather condition data and import them into the image correction model for image correction processing includes the following specific steps:

[0016] S21, real-time acquisition of pixel value data of each pixel point of remote sensing image data, weather humidity and visibility data during flight;

[0017] S22, the collected historical weather data and historical weather picture fuzzy data are divided into 80% parameter training set and 20% parameter test set; the 80% parameter training set is input into the correction neural network model for training to obtain the correction initial neural network model; the 20% parameter test set is used to test the correction initial neural network model, and the optimal initial neural network model that meets the set picture standard threshold is output as the correction neural network model, wherein the output strategy formula of the specific neuron in the correction neural network model is: in To correct the output of the w-th neuron in the p+1th layer of the neural network model, To correct the connection weights between the p-th layer neuron v and the p+1-th layer w-th neuron of the neural network model, To correct the output of neuron v in the pth layer of the neural network model, To correct the bias of the linear relationship between the p-th layer neuron v and the p+1-th layer m-term neuron of the neural network model, σ() is the Sigmoid activation function. ;

[0018] S23, obtaining the image corrected by the correction neural network model.

[0019] It should be noted here that as a preferred technical solution of the AI ​​intelligent recognition and classification method based on drone images, the method of importing the corrected image into the abnormal area search strategy to search for abnormal cultivated land areas includes the following specific steps:

[0020] S31, obtaining the pixel value of each pixel point of the corrected image of the set size area, and obtaining the average pixel value data of the pixels of the cultivated land area, and subtracting the pixel value of each pixel point of the corrected image of the area from the average pixel value data of the pixels of the cultivated land area to obtain the pixel difference;

[0021] S32, comparing the pixel difference of each pixel with the difference threshold, if the pixel difference is greater than or equal to the difference threshold, setting the pixel as an abnormal pixel, if the pixel difference is less than the difference threshold, not setting the pixel as an abnormal pixel;

[0022] S33, obtaining an area formed by abnormal pixels and setting it as the abnormal area, obtaining the regional contours of the abnormal area with different pixel gradients and the regional pixel averages with different pixel gradients. It should be noted here that the pixel gradient is 10.

[0023] It should be noted here that as a preferred technical solution of the AI ​​intelligent recognition and classification method based on drone images, the method of importing the found abnormal cultivated land area into the intelligent recognition strategy to distinguish the occupation of the abnormal cultivated land area includes the following specific contents:

[0024] S41, obtaining the regional contours of different pixel gradients and the regional pixel averages of different pixel gradients in the abnormal area, and obtaining the contours of construction components and the average pixel values ​​of construction components of various historical illegal constructions, wherein the constructions are: illegal occupation of cultivated land for afforestation, illegal occupation of cultivated land for building green corridors beyond the standard, illegal occupation of cultivated land for digging fields to build lakes and landscaping, and occupation of permanent basic farmland to expand nature reserves;

[0025] S42, importing the obtained regional contours of different pixel gradients of the abnormal area, the regional pixel averages of different pixel gradients, the construction component contours of the i-th historical illegal construction, and the average pixel values ​​of the construction components into the intelligent recognition analysis value calculation formula to calculate the intelligent recognition analysis value, wherein the calculation formula for the intelligent recognition analysis value of the abnormal area and the i-th historical illegal construction is: Wherein, a is the contour size ratio coefficient, Zi is the contour type of the construction component of the i-th historical illegal construction, C is the regional contour type of different pixel gradients in the abnormal area, exp() is the power of the natural constant e, Lz is the size of the contour of the z-th construction component of the i-th historical illegal construction, Lc is the size of the contour of the c-th construction component in the abnormal area, pz is the average pixel value of the z-th construction component of the i-th historical illegal construction, and pc is the average pixel value of the contour of the c-th construction component in the abnormal area;

[0026] S43. Obtain the historical illegal construction type corresponding to the minimum intelligent identification analysis value, and set it as the illegal construction type in the abnormal area.

[0027] It should be noted here that as a priority technical solution for the AI ​​intelligent recognition and classification method based on drone images, the occupancy warning based on the occupancy areas of abnormal cultivated land areas includes the following specific contents: uploading the types of illegal construction in the acquired abnormal areas to the early warning system of the management department, and the early warning system of the management department issues an occupancy warning based on the types of illegal construction in the abnormal areas.

[0028] The AI ​​intelligent recognition and classification system based on drone images is implemented based on the above-mentioned AI intelligent recognition and classification method based on drone images, and specifically includes a data acquisition module, an image processing module, an abnormal area search module, an occupancy distinction module, an occupancy warning module and a control module, wherein the data acquisition module uses a drone to acquire remote sensing image data and weather condition data at the same time, and stores them in a storage module;

[0029] The image processing module uses the UAV to obtain remote sensing image data and weather condition data and imports them into the image correction model to perform image correction processing;

[0030] The abnormal area search module is used to import the corrected image into the abnormal area search strategy to search for abnormal areas of cultivated land;

[0031] The occupancy differentiation module is used to import the found abnormal cultivated land area into the intelligent identification strategy to perform occupancy differentiation of the abnormal cultivated land area;

[0032] The occupancy warning module is used to provide occupancy warning according to the occupancy classification of abnormal areas of cultivated land;

[0033] The control module is used for controlling the operation of the data acquisition module, the image processing module, the abnormal area search module, the occupancy distinction module and the occupancy warning module.

[0034] An electronic device comprises: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0035] The processor executes the above-mentioned AI intelligent recognition and classification method based on drone images by calling the computer program stored in the memory.

[0036] A computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned AI intelligent recognition and classification method based on drone images.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention uses a drone to obtain remote sensing image data and weather condition data at the same time, and stores them in a storage module. The remote sensing image data and weather condition data obtained by the drone are imported into an image correction model to perform image correction processing, and the corrected image is imported into an abnormal area search strategy to search for abnormal cultivated land areas. The abnormal cultivated land areas found are imported into an intelligent recognition strategy to perform occupancy distinction of the abnormal cultivated land areas, and occupancy warning is performed according to the occupancy distinction of the abnormal cultivated land areas. Effective noise reduction processing and accurate analysis are performed on the remote sensing image, thereby improving the intelligent recognition and classification effect, thereby improving the intelligent recognition and classification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a schematic diagram of the overall process of the AI ​​intelligent recognition and classification method based on drone images of the present invention;

[0040] Figure 2 The figure is a schematic diagram of the overall framework of the AI ​​intelligent recognition and classification system based on drone images of the present invention; DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present application and its application or use.

[0042] Example 1

[0043] In order to solve the technical problems raised in the background technology, the present invention provides a preferred embodiment: Figure 1 As shown in the figure, the AI ​​intelligent recognition and classification method based on drone images includes the following specific steps:

[0044] Fly the drone to obtain remote sensing image data and weather condition data, and store them in the storage module;

[0045] Use drones to obtain remote sensing image data and weather data and import them into the image correction model for image correction processing;

[0046] The corrected image is imported into the abnormal area search strategy to search for abnormal areas of cultivated land;

[0047] The found abnormal cultivated land areas are introduced into the intelligent identification strategy to distinguish the occupation of the abnormal cultivated land areas;

[0048] Occupancy warning is issued based on the occupancy division of abnormal cultivated land areas.

[0049] It should be noted that in this embodiment, the specific steps of using a drone to obtain remote sensing image data and weather condition data and storing them in a storage module are as follows:

[0050] S11. The UAV flies along the designated flight path, obtains images of cultivated land within the designated range through the camera module, and obtains weather humidity and visibility data during the flight through the weather collection module;

[0051] To use drones to collect remote sensing images and weather data simultaneously, you need to follow the following steps:

[0052] 1. Preparation stage:

[0053] Survey area: Understand the situation of the operation area, determine the model of the aerial photography drone and the parameters of the aerial photography camera it carries according to the situation of the operation area; understand the local traffic conditions, select the take-off and landing site of the aerial photography drone and the backup take-off and landing site. At the same time, collect information on whether there are airports and military facilities within 30 kilometers of the operation area, obtain the support and cooperation of the local relevant departments, and conduct legal aerial photography.

[0054] Airspace application: Apply for a flight plan from the airspace management department and conduct legal aerial photography

[0055] Drone inspection: Make sure the drone has sufficient power and all components (such as cameras, data acquisition modules, etc.) are working properly.

[0056] Flight path planning: Plan the flight path of the drone according to the scope of the survey area. You can use special flight planning software or map services for planning.

[0057] Task allocation: Determine the specific requirements of the collection task, such as flight heading and lateral overlap, collection resolution, frequency, coverage, etc.

[0058] 2. Take-off preparation:

[0059] Collect weather conditions: ensure that the weather conditions during the flight are suitable, and that the wind, humidity and visibility meet the requirements for drone operation. Plan the time for aerial photography and the backup time for aerial photography, make a good flight plan, and make emergency plans for possible emergencies.

[0060] Data acquisition module calibration: Ensure that the weather acquisition module can accurately measure humidity and visibility.

[0061] 3. Flight process:

[0062] Takeoff: Release the drone and monitor its takeoff and ascent process.

[0063] Flight operation: The UAV flies according to the predetermined flight path and is strictly implemented in accordance with the Basic Requirements for Safe Operation of UAV Aerial Photography to ensure the personal and property safety of operators and the safety of production equipment and data.

[0064] Image acquisition: Use the camera module to take aerial photos of the specified range and obtain aerial data.

[0065] Synchronous data collection: Use the weather collection module to obtain real-time weather data (humidity and visibility).

[0066] 4. Data collection details:

[0067] Image acquisition:

[0068] The camera module should have sufficient resolution and appropriate focal length to be able to capture clear images of cultivated land.

[0069] Control the drone to take aerial photos at appropriate flight altitude, speed and overlap to ensure image quality.

[0070] Different shooting angles, shooting frequencies and flight overlaps can be used as needed.

[0071] Weather data collection:

[0072] The weather collection module should be synchronized with the drone control system to ensure the real-time nature of the data.

[0073] Weather data is collected regularly, in sync with image acquisition, so that the conditions under which the images were taken can be recorded and analyzed.

[0074] Emergency Plan: Develop and implement emergency measures in case of drone malfunction or operational error.

[0075] Through the above steps, it can be ensured that the drone flies safely on the specified flight path and effectively collects image data within the specified range and weather humidity and visibility data during the flight.

[0076] S12. Store the acquired weather data and image data in a storage component, and simultaneously acquire pixel value data of each pixel point of the cultivated land image data.

[0077] It should be noted that in this embodiment, obtaining the image data and weather condition data obtained by the drone and importing them into the image correction model for image correction processing includes the following specific steps:

[0078] S21, acquiring pixel value data of each pixel point of the image data, weather humidity and visibility data during the flight in real time;

[0079] S22, the collected historical weather data and historical weather picture fuzzy data are divided into 80% parameter training set and 20% parameter test set; the 80% parameter training set is input into the correction neural network model for training to obtain the correction initial neural network model; the 20% parameter test set is used to test the correction initial neural network model, and the optimal initial neural network model that meets the set picture standard threshold is output as the correction neural network model, wherein the output strategy formula of the specific neuron in the correction neural network model is: in To correct the output of the w-th neuron in the p+1th layer of the neural network model, To correct the connection weights between the p-th layer neuron v and the p+1-th layer w-th neuron of the neural network model, To correct the output of neuron v in the pth layer of the neural network model, To correct the bias of the linear relationship between the p-th layer neuron v and the p+1-th layer m-term neuron of the neural network model, σ() is the Sigmoid activation function;

[0080] S23, obtaining an image corrected by the correction neural network model;

[0081] The following Python code implements the corresponding steps:

[0082]

[0083]

[0084]

[0085] It should be noted that in this embodiment, importing the corrected image into the abnormal area search strategy to search for abnormal cultivated land areas includes the following specific steps:

[0086] S31, obtaining the pixel value of each pixel point of the corrected image of the set size area, and obtaining the average pixel value data of the pixels of the cultivated land area, and subtracting the pixel value of each pixel point of the corrected image of the use area from the average pixel value data of the pixels of the cultivated land area to obtain the pixel difference;

[0087] S32, comparing the pixel difference of each pixel with the difference threshold, if the pixel difference is greater than or equal to the difference threshold, setting the pixel as an abnormal pixel, if the pixel difference is less than the difference threshold, not setting the pixel as an abnormal pixel;

[0088] S33, obtaining an area formed by abnormal pixels and setting it as the abnormal area, obtaining the regional contours of the abnormal area with different pixel gradients and the regional pixel averages with different pixel gradients. It should be noted here that the pixel gradient is 10.

[0089] It should be noted that in this embodiment, the process of importing the found abnormal cultivated land area into the intelligent identification strategy to distinguish the occupation of the abnormal cultivated land area includes the following specific contents:

[0090] S41, obtaining the regional contours of different pixel gradients and the regional pixel averages of different pixel gradients in the abnormal area, and obtaining the contours of various historical illegal constructions and the average pixel values ​​of the construction components;

[0091] S42, importing the obtained regional contours of different pixel gradients of the abnormal area, the regional pixel averages of different pixel gradients, the construction component contours of the i-th historical illegal construction, and the average pixel values ​​of the construction components into the intelligent recognition analysis value calculation formula to calculate the intelligent recognition analysis value, wherein the calculation formula for the intelligent recognition analysis value of the abnormal area and the i-th historical illegal construction is: Wherein, a is the contour size ratio coefficient, Zi is the contour type of the construction component of the i-th historical illegal construction, C is the regional contour type of different pixel gradients in the abnormal area, exp() is the power of the natural constant e, Lz is the size of the contour of the z-th construction component of the i-th historical illegal construction, Lc is the size of the contour of the c-th construction component in the abnormal area, pz is the average pixel value of the z-th construction component of the i-th historical illegal construction, and pc is the average pixel value of the contour of the c-th construction component in the abnormal area;

[0092] S43. Obtain the historical illegal construction type corresponding to the minimum intelligent identification analysis value, and set it as the illegal construction type in the abnormal area.

[0093] It should be noted that in this embodiment, the occupation warning according to the occupation classification of abnormal cultivated land areas includes the following specific contents: uploading the acquired illegal construction types in the abnormal areas to the management department warning system, and the management department warning system issues occupation warning according to the illegal construction types in the abnormal areas;

[0094] It should be noted that in this embodiment, the beneficial effects are:

[0095] The present invention flies a drone to acquire remote sensing image data, acquires weather condition data at the same time, and stores the data in a storage module; the remote sensing image data and weather condition data acquired by the drone are imported into an image correction model to perform image correction processing; the corrected image is imported into an abnormal area search strategy to perform search for abnormal cultivated land areas; the abnormal cultivated land areas found are imported into an intelligent recognition strategy to perform occupation distinction of the abnormal cultivated land areas; occupation warning is performed according to the occupation distinction of the abnormal cultivated land areas; effective noise reduction processing and accurate analysis are performed on the remote sensing image, thereby improving the intelligent recognition and classification effect, thereby improving the intelligent recognition and classification accuracy.

[0096] Example 2

[0097] like Figure 2 As shown, an AI intelligent recognition and classification system based on drone images is implemented based on the above-mentioned AI intelligent recognition and classification method based on drone images, and specifically includes a data acquisition module, an image processing module, an abnormal area search module, an occupancy distinction module, an occupancy warning module and a control module, wherein the data acquisition module uses a drone to obtain remote sensing image data, and at the same time obtains weather condition data, and stores them in a storage module; the image processing module uses a drone to obtain remote sensing image data and weather condition data and imports them into an image correction model for image correction processing; the abnormal area search module is used to import the corrected image into the abnormal area search strategy to search for abnormal cultivated land areas; the occupancy distinction module is used to import the abnormal cultivated land areas found into the intelligent recognition strategy to perform occupancy distinction of the abnormal cultivated land areas; the occupancy warning module is used to perform occupancy warning according to the occupancy distinction of the abnormal cultivated land areas; the control module is used to control the operation of the data acquisition module, the image processing module, the abnormal area search module, the occupancy distinction module and the occupancy warning module.

[0098] Example 3

[0099] This embodiment provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0100] The processor executes the above-mentioned AI intelligent recognition and classification method based on drone images by calling the computer program stored in the memory.

[0101] The electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors and one or more memories, wherein the memory stores at least one computer program, and the computer program is loaded and executed by the processor to implement the AI ​​intelligent recognition and classification method based on drone images provided in the above method embodiment. The electronic device may also include other components for implementing the functions of the device. For example, the electronic device may also have components such as a wired or wireless network interface and an input and output interface to input and output data. This embodiment will not be described in detail here.

[0102] Example 4

[0103] This embodiment provides a computer-readable storage medium having a rewritable computer program stored thereon;

[0104] When the computer program runs on a computer device, the computer device executes the above-mentioned AI intelligent recognition and classification method based on drone images.

[0105] For example, the computer readable storage medium can be a read-only memory, a random access memory, a read-only CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0106] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, a process or function according to an embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. Computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or / and a wireless network. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

Claims

1. The AI ​​intelligent recognition and classification method based on drone images is characterized by: It includes the following specific steps: Use drones to obtain remote sensing image data within a specified range, and at the same time obtain weather data and store it in a storage module; Use drones to obtain remote sensing cultivated land image data and weather data within a specified range and import them into the image correction model for image correction processing; The corrected image is imported into the abnormal area search strategy to search for abnormal cultivated land areas; The found abnormal cultivated land areas are introduced into the intelligent identification strategy to distinguish the occupation of the abnormal cultivated land areas; Occupancy warning is issued based on the occupancy division of abnormal cultivated land areas.

2. The AI ​​intelligent recognition and classification method based on drone images as claimed in claim 1, characterized in that: The specific steps of the drone acquiring remote sensing image data within a specified range and weather condition data and storing them in a storage module are as follows: S11. The drone flies along the set flight path, obtains remote sensing images within the specified range through the camera module, and obtains weather humidity and visibility data during the flight through the weather acquisition module; S12, storing the acquired weather data and remote sensing image data in a storage component, and acquiring pixel value data of each pixel point of the cultivated land image data at the same time.

3. The AI ​​intelligent recognition and classification method based on drone images as claimed in claim 2 is characterized in that: The method of using a drone to obtain remote sensing image data and weather condition data and importing them into an image correction model for image correction processing includes the following specific steps: S21, real-time acquisition of pixel value data of each pixel point of remote sensing image data, weather humidity and visibility data during flight; S22, the acquired historical weather data and historical weather picture fuzzy data are divided into 80% parameter training set and 20% parameter test set; the 80% parameter training set is input into the correction neural network model for training to obtain the correction initial neural network model; the 20% parameter test set is used to test the correction initial neural network model, and the optimal initial neural network model that meets the set picture standard threshold is output as the correction neural network model, wherein the output strategy formula of the specific neuron in the correction neural network model is: in To correct the output of the w-th neuron in the p+1th layer of the neural network model, To correct the connection weights between the p-th layer neuron v and the p+1-th layer w-th neuron of the neural network model, To correct the output of neuron v in the pth layer of the neural network model, To correct the bias of the linear relationship between the p-th layer neuron v and the p+1-th layer m-term neuron of the neural network model, σ() is the Sigmoid activation function. ; S23, obtaining the image corrected by the correction neural network model.

4. The AI ​​intelligent recognition and classification method based on drone images as claimed in claim 3 is characterized in that: The step of importing the corrected image into the abnormal area search strategy to search for abnormal cultivated land areas includes the following specific steps: S31, obtaining the pixel value of each pixel point of the corrected image of the set size area, and obtaining the average pixel value data of the pixels of the cultivated land area, and subtracting the pixel value of each pixel point of the corrected image of the area from the average pixel value data of the pixels of the cultivated land area to obtain the pixel difference; S32, comparing the pixel difference of each pixel with the difference threshold, if the pixel difference is greater than or equal to the difference threshold, setting the pixel as an abnormal pixel, if the pixel difference is less than the difference threshold, not setting the pixel as an abnormal pixel; S33, obtaining an area formed by abnormal pixels and setting it as the abnormal area, obtaining the regional contours of the abnormal area with different pixel gradients and the regional pixel averages with different pixel gradients.

5. The AI ​​intelligent recognition and classification method based on drone images as claimed in claim 4 is characterized in that: The method of introducing the found abnormal cultivated land area into the intelligent identification strategy to distinguish the occupation of the abnormal cultivated land area includes the following specific contents: S41, obtaining the regional contours of different pixel gradients and the regional pixel averages of different pixel gradients in the abnormal area, and obtaining the contours of construction components and the average pixel values ​​of construction components of various historical illegal constructions, wherein the illegal constructions include: illegal occupation of cultivated land for afforestation, illegal occupation of cultivated land for building green corridors beyond the standard, illegal occupation of cultivated land for digging fields to build lakes and landscaping, and occupation of permanent basic farmland to expand nature reserves; S42, importing the obtained regional contours of different pixel gradients of the abnormal area, the regional pixel averages of different pixel gradients, the construction component contours of the i-th historical illegal construction, and the average pixel values ​​of the construction components into the intelligent recognition analysis value calculation formula to calculate the intelligent recognition analysis value, wherein the calculation formula for the intelligent recognition analysis value of the abnormal area and the i-th historical illegal construction is: Wherein, a is the contour size ratio coefficient, Zi is the contour type of the construction component of the i-th historical illegal construction, C is the regional contour type of different pixel gradients in the abnormal area, exp() is the power of the natural constant e, Lz is the size of the contour of the z-th construction component of the i-th historical illegal construction, Lc is the size of the contour of the c-th construction component in the abnormal area, pz is the average pixel value of the z-th construction component of the i-th historical illegal construction, and pc is the average pixel value of the contour of the c-th construction component in the abnormal area; S43. Obtain the historical illegal construction type corresponding to the minimum intelligent identification analysis value, and set it as the illegal construction type in the abnormal area.

6. The AI ​​intelligent recognition and classification method based on drone images as claimed in claim 5, characterized in that: The occupancy warning according to the occupation zone of abnormal cultivated land includes the following specific contents: uploading the obtained illegal construction types in the abnormal areas to the early warning system of the management department, and the early warning system of the management department issues an occupancy warning according to the illegal construction types in the abnormal areas.

7. An AI intelligent recognition and classification system based on drone images, which is implemented based on the AI ​​intelligent recognition and classification method based on drone images as claimed in any one of claims 1 to 6, characterized in that: It specifically includes a data acquisition module, an image processing module, an abnormal area search module, an occupancy distinction module, an occupancy warning module and a control module, wherein the data acquisition module is used for the drone to acquire remote sensing image data and weather condition data at the same time, and store them in the storage module; The image processing module is used to obtain remote sensing image data and weather condition data obtained by the drone and import them into the image correction model to perform image correction processing; The abnormal area search module is used to import the corrected image into the abnormal area search strategy to search for abnormal areas of cultivated land; The occupancy differentiation module is used to import the found abnormal cultivated land area into the intelligent identification strategy to perform occupancy differentiation of the abnormal cultivated land area; The occupancy warning module is used to provide occupancy warning according to the occupancy classification of abnormal areas of cultivated land; The control module is used for controlling the operation of the data acquisition module, the image processing module, the abnormal area search module, the occupancy distinction module and the occupancy warning module.

8. An electronic device comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; It is characterized in that the processor executes the AI ​​intelligent recognition and classification method based on drone images as described in any one of claims 1 to 6 by calling the computer program stored in the memory.

9. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer executes the AI ​​intelligent recognition and classification method based on drone images as described in any one of claims 1 to 6.

Images