Verification method, device and equipment for territorial change investigation proof data and medium

Abstract

The invention discloses a territorial change survey proof data verification method and device, equipment and a medium. The method comprises the steps of obtaining to-be-processed proof data and to-be-processed remote sensing data; filtering the to-be-processed proof data through a proof photo quality detection model to obtain filtered proof photo data; based on the filtered proof photo data, performing quality inspection on the to-be-processed remote sensing data to obtain target proof photo data, and extracting target remote sensing image data according to the proof shooting position; inputting the target proof photo data into a proof photo classification model to obtain a proof photo classification result, and inputting the target remote sensing image data into a remote sensing image classification model in parallel to obtain a remote sensing image classification result; and performing analysis processing on the proof photo classification result and the remote sensing image classification result to generate a verification result of the to-be-processed land change investigation proof. According to the scheme, the classification result can be accurately determined, and the accuracy of proof achievement verification is improved.

Description

Technical Field

[0001] The present invention generally relates to the field of data processing technology, and specifically relates to a method, device, equipment and medium for verifying land change investigation and evidence data. Background Art

[0002] With the continuous development and innovation of land survey work, land change survey is an important national condition and national strength survey. Through the investigation, monitoring, verification, summary, statistics and analysis of the current land use status, ownership changes and various land use management information in the natural year, it can fully grasp the dynamic changes of land resources, including changes in land use types, land cover, and land ownership. In order to protect land resources, promote sustainable land use, and promote land space rules, it is particularly important to verify the land change survey data.

[0003] At present, the relevant technology is to manually judge information such as evidence photo data and remote sensing image data. The actual judgment process needs to rely on a lot of manual experience, which is highly subjective and may result in a high error rate. At the same time, it requires a lot of human resources, resulting in a long review cycle for land change investigation and evidence data. Summary of the invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, it is desired to provide a method, device, equipment and medium for verifying land change investigation and evidence data.

[0005] In a first aspect, the present invention provides a method for verifying land change investigation and evidence data, the method comprising:

[0006] Acquire the land change survey data to be processed; the land change survey data to be processed includes the evidence data to be processed and the corresponding remote sensing data to be processed;

[0007] The evidence data to be processed is filtered through an evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the evidence data of land change surveys over the years and the corresponding evidence quality labeling results;

[0008] Based on the filtered evidence photo data, the remote sensing data to be processed is subjected to an evidence shooting position quality check to obtain target evidence photo data, and corresponding target remote sensing image data is extracted according to the evidence shooting position;

[0009] Inputting the target evidence photo data into the evidence photo classification model to perform image classification to obtain the evidence photo classification result, and in parallel inputting the target remote sensing image data into the remote sensing image classification model to perform image classification to obtain the remote sensing image classification result;

[0010] The classification results of the evidence photos and the classification results of the remote sensing images are analyzed and processed to generate verification results of the land change survey data to be processed.

[0011] In one embodiment, the evidence data to be processed is filtered through an evidence photo quality detection model to obtain filtered evidence photo data, including:

[0012] Inputting the evidence data to be processed into the evidence photo quality detection model for quality classification to obtain the evidence photo quality classification result;

[0013] The evidence data to be processed is filtered according to the quality classification result of the evidence photo to obtain the filtered evidence photo data.

[0014] In one embodiment, based on the filtered evidence photo data, the remote sensing data to be processed is subjected to an evidence shooting position quality check to obtain target evidence photo data, and corresponding target remote sensing image data is extracted according to the evidence shooting position, including:

[0015] According to the filtered evidence photo data, obtaining azimuth data during the shooting process and remote sensing image data corresponding to the evidence shooting location;

[0016] Determine whether the shooting position of the filtered evidence photo data covers the surroundings of the spot position of the remote sensing image data of the evidence shooting position;

[0017] When the four sides of the spot position are covered, the geometric center position of the spot of the remote sensing image data of the evidence shooting position is obtained, and based on the azimuth data, it is determined whether the filtered evidence photo data can surround the geometric center of the spot;

[0018] When surrounding the geometric center of the spot, the filtered evidence photo data surrounding the geometric center of the spot and covering the four sides of the spot position is used as the target evidence photo data;

[0019] The remote sensing data corresponding to the target evidence photo data is extracted according to the evidence shooting position as the target remote sensing image data.

[0020] In one embodiment, the target evidence photo data is input into an evidence photo classification model for image classification to obtain an evidence photo classification result, including:

[0021] Preprocessing the target evidence photo data to obtain processed target evidence photo data;

[0022] The processed target evidence photo data is input into the evidence photo classification model, and the processed target evidence photo data is classified by ground type and calculation to obtain the evidence photo classification result.

[0023] In one embodiment, the classification results of the evidence photos and the classification results of the remote sensing images are analyzed and processed to generate the verification results of the land change survey data to be processed, including:

[0024] Performing soft voting processing of integrated learning on the classification results of the evidence photos and the classification results of the remote sensing images to obtain target classification results;

[0025] A comprehensive analysis and judgment is performed on the target classification results to obtain the verification results of the land change survey data to be processed.

[0026] In one embodiment, the classification result of the evidence photo and the classification result of the remote sensing image are subjected to soft voting processing of integrated learning to obtain the target classification result, including:

[0027] Determine the land category of the evidence photo and the corresponding probability value of the evidence photo classification from the evidence photo classification result, and determine the remote sensing image category and the corresponding remote sensing image classification probability value from the remote sensing image classification result; the evidence photo classification probability value is output as the probability value corresponding to the land category of the evidence photo, and the remote sensing image classification probability value is output as the probability value corresponding to the remote sensing image category;

[0028] For each piece of land change survey data to be processed, the classification probability value of the evidence photo and the classification probability value of the remote sensing image are soft-voted with preset weights according to the preset weight distribution ratio to obtain the target classification result.

[0029] In one embodiment, the evidence photo classification model is constructed in the following manner:

[0030] Obtaining historical land change survey and evidence data; the historical land change survey and evidence data includes: evidence photo results data and time information obtained during the land survey and evidence photo shooting process, shooting azimuth information, location information, and pitch angle information; the historical land change survey and evidence data are annotated with land classification results;

[0031] Preprocessing the historical land change investigation and evidence data to obtain processed historical land change investigation and evidence data;

[0032] Dividing the processed historical land change survey and evidence data into a training set and a validation set according to a preset ratio;

[0033] Using the training set to train the initial evidence photo classification model, fine-tuning all layers of the model's neurons according to the number of data classifications and other parameters of the training set to obtain an evidence photo classification model to be verified, and inputting the verification set into the evidence photo classification model to be verified to obtain an output result;

[0034] A loss function is constructed according to the output result and the land category labeling result, and the parameters of the evidence photo classification model to be verified are trained using an iterative algorithm in accordance with the minimization of the loss function to obtain the evidence photo classification model.

[0035] In a second aspect, an embodiment of the present application provides a device for verifying land change investigation and evidence data, the device comprising:

[0036] An acquisition module is used to acquire the land survey change data to be processed; the land survey change data to be processed includes the evidence data to be processed and the corresponding remote sensing image data;

[0037] A filtering processing module, used for filtering the evidence data to be processed through an evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the evidence data of land change surveys in previous years and the corresponding evidence quality labeling results;

[0038] A quality inspection module, for performing a quality inspection of the evidence shooting position on the remote sensing data to be processed based on the filtered evidence photo data, obtaining target evidence photo data, and extracting corresponding target remote sensing image data according to the evidence shooting position;

[0039] An evidence photo classification module is used to input the target evidence photo data into the evidence photo classification model for image classification to obtain an evidence photo classification result, and in parallel input the target remote sensing image data into the remote sensing image classification model for image classification to obtain a remote sensing image classification result;

[0040] The analysis module is used to analyze and process the classification results of the evidence photos and the classification results of the remote sensing images to generate verification results of the land change survey data to be processed.

[0041] In a third aspect, an embodiment of the present application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for verifying land change survey and evidence data provided in the above embodiment.

[0042] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. The computer program is executed by a processor to implement the method for verifying land change investigation and evidence data provided in the above embodiment.

[0043] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0044] The verification method, device, equipment and medium for land change survey and evidence data provided by the embodiments of the present application include: obtaining land change survey data to be processed; the land change survey data to be processed include evidence data to be processed and corresponding remote sensing data to be processed; filtering the evidence data to be processed through an evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the land change survey and evidence data of previous years and the corresponding evidence quality annotation results; based on the filtered evidence photo data, performing an evidence shooting position quality check on the remote sensing data to be processed to obtain target evidence photo data, and extracting corresponding target remote sensing image data according to the evidence shooting position; inputting the target evidence photo data into the evidence photo classification model for image classification to obtain evidence photo classification results, and in parallel inputting the target remote sensing image data into the remote sensing image classification model for image classification to obtain remote sensing image classification results; analyzing and processing the evidence photo classification results and the remote sensing image classification results to generate verification results for the land change survey and evidence data to be processed. Compared with the prior art, this technical solution can detect the shooting conditions of evidence photos by inputting the data of evidence photos to be processed into the quality detection model of evidence photos, filter out evidence photos that do not meet the conditions, and obtain filtered evidence photo data. It can also perform a quality check on the shooting positions of evidence photos on the remote sensing image data to be processed, and accurately determine the target evidence photo data and target remote sensing image data that have passed the quality check, thereby ensuring that the data participating in the subsequent land classification of photos and images are valid data, and classify the target evidence photo data and target remote sensing image data by their respective corresponding classification models, and accurately determine the classification results of evidence photos and remote sensing images, reducing the workload of manual inspection of evidence photo results, and at the same time assisting manual verification of field survey work, and then performing integrated learning and combination processing on the classification results of evidence photos and remote sensing images, thereby obtaining the verification results of the land change survey evidence data to be processed, reducing the error probability of manual judgment, and effectively improving the accuracy of verification of corresponding evidence results. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1This is an application environment diagram of a method for verifying land change investigation and evidence data provided in an embodiment of the present application;

[0047] Figure 2 A flowchart of a method for verifying land change investigation and evidence data provided in an embodiment of the present application;

[0048] Figure 3 A flowchart of a method for generating verification results of pending land change investigation and evidence data provided in an embodiment of the present application;

[0049] Figure 4 A schematic diagram of a structure for constructing a classification model for evidence photos provided in an embodiment of the present application;

[0050] Figure 5 A schematic diagram of functional modules of a device for verifying land change investigation and evidence data provided by another embodiment of the present application;

[0051] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0054] As mentioned in the background technology, the relevant technology is to manually organize and judge information such as evidence photos and remote sensing images. However, this solution needs to rely on a lot of manual experience, which is highly subjective and may result in misjudgment. It also requires a lot of manpower, resulting in low efficiency in the review of land change investigation and evidence data.

[0055] Based on the above-mentioned defects, the present application provides a verification method for land change investigation and evidence data. Compared with the prior art, the technical solution can detect the shooting conditions of the evidence photos by inputting the evidence photo data to be processed into the evidence photo quality detection model, filter out the evidence photos that do not meet the conditions, and obtain the filtered evidence photo data. The shooting position quality inspection is performed on the evidence photo data around the remote sensing image position to be processed, and the target evidence photo data and target remote sensing image data that have passed the quality inspection are accurately determined, thereby ensuring that the data participating in the subsequent evidence photo classification are all valid data, and the target evidence photo data and the target remote sensing image data are respectively classified by their respective corresponding land classification models, which can accurately determine the land classification results of the evidence photos and the land classification results of the remote sensing images, reducing the workload of manual inspection of the evidence photo results, and can assist manual verification of field investigation and evidence work, and then integrate the evidence photo classification results and the remote sensing image classification results for learning and analysis, so as to obtain the verification results of the land change investigation and evidence data to be processed, reducing the error probability of manual judgment, and effectively improving the accuracy of the verification of the evidence results.

[0056] The verification method of land change investigation and evidence data provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately or integrated on the server 104. The terminal 102 can send the photographed evidence to the server 104. After receiving the pending land change investigation and evidence data, the server 104 filters the pending land change investigation and evidence data, checks the quality of the map location, and classifies the land class through the evidence photo classification model and the remote sensing image classification model, obtains the evidence photo land class classification result and the remote sensing image land class classification result, and then generates the corresponding verification result, so as to feedback to the terminal 102. In addition, in some embodiments, the verification method of the land change investigation data can also be implemented by the server 104 or the terminal 102 alone, such as the terminal 102 can directly perform the verification process on the pending land change investigation and evidence data, or the server 104 can obtain the pending land change investigation and evidence data from the data storage system, and perform the verification process on the pending land change investigation and evidence data.

[0057] Optionally, the above-mentioned evidence photo classification model can be constructed by the server 104 or the terminal 102 based on the evidence photo data of previous years through neural classification learning, and the above-mentioned remote sensing image classification model can be constructed and published by the server 104 or the terminal 102 based on the analysis of previous years' remote sensing images.

[0058] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0059] In an exemplary embodiment, Figure 2 As shown, a method for verifying land change investigation and evidence data is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include the following steps S201 to S205.

[0060] in:

[0061] Step S201, obtaining the land change survey data to be processed; the land change survey data to be processed includes the evidence data to be processed and the corresponding remote sensing data to be processed.

[0062] It should be noted that the above-mentioned land change survey data to be processed refers to the land change survey evidence data that needs to be reviewed and verified, and the evidence data to be processed refers to the data generated during the field survey and evidence monitoring process in the change survey, which may include field survey and monitoring results photo data, shooting time, shooting location, shooting azimuth and other related data, and the remote sensing data to be processed refers to satellite image information or aerial image data, and the aerial image information may include resolution information and acquisition time information, etc. Satellite image information may be the latest acquired satellite image data, and aerial images may be image data files with high ground resolution and short shooting time intervals. Among them, the above-mentioned evidence photo data to be processed and the remote sensing image data to be processed may be in the format of images, videos, or text. Among them, the shooting location of the evidence photo data to be processed and the monitoring map of the remote sensing image data to be processed should include evidence photos taken in the eight range directions related to the map that point to the center of the monitoring map.

[0063] Optionally, the above-mentioned evidence data to be processed can be collected by investigators from field investigations, determined by verifying changes and causes, or obtained through real-time positioning through geographic location.

[0064] The above-mentioned remote sensing data to be processed can be obtained through public data sources, or through real-time collection through drone or manned aircraft aerial photography, or through the purchase of satellite programming flight photography, or through the archived images of image service providers over the years. It can include the information of the spots to be monitored. The spots refer to an area with the same land use attributes on the ground through satellite image monitoring or field investigation. The spot information can include the range of the spots.

[0065] In this embodiment, since the land change survey data to be processed is obtained, good data guidance information can be provided for subsequent data filtering and quality detection.

[0066] Step S202, filtering the evidence data to be processed through the evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the evidence data of land change surveys in previous years and the corresponding evidence quality labeling results.

[0067] It should be noted that the above-mentioned evidence photo quality detection model is a pre-trained model for performing quality detection on the evidence photo data to be processed. The evidence photo quality detection model can be a machine learning model or other models for performing quality detection on the evidence photo data to be processed. The above-mentioned filtered evidence photo data is the evidence photo data to be processed whose quality has passed the inspection, that is, the filtered evidence photo data is the evidence photo data that meets the basic requirements for evidence shooting.

[0068] After obtaining the data of the evidence photos to be processed, the evidence photos to be processed are input into the evidence photo quality detection model for quality classification to obtain the evidence photo quality classification results, and the evidence photo data to be processed are classified and processed according to the evidence photo quality classification results to obtain the filtered evidence photo data. Among them, the evidence photo quality classification results are used to characterize the shooting information of the evidence data to be processed, for example, they may include shooting conditions, shooting exposure conditions, shooting angles, and shooting clarity conditions. The shooting conditions include close-up and long-range; the shooting exposure conditions include backlighting, non-backlighting, normal exposure, and abnormal exposure; the shooting angles include: missing shooting position, no missing shooting position; the shooting clarity conditions include: clear and unclear. Among them, general evidence shooting quality problems include incorrect shooting exposure, backlight shooting, too large a proportion of the sky in shooting, incorrect shooting of the subject, etc.

[0069] It can be understood that the above-mentioned evidence photo quality detection model can be pre-established based on the evidence photo data of previous years and the corresponding evidence quality annotation results. The evidence photo data of previous years can be obtained first and used as the image quality classification data set, and the shooting quality of the evidence photo data of previous years can be annotated. The evidence quality annotation results can be obtained by manual annotation, and then the evidence photo data of previous years are pre-processed to obtain the processed evidence photo data of previous years, and the processed evidence photo data of previous years are divided into a training set and a verification set according to a preset ratio. For example, 80% of the image quality classification data set can be used as the training set, and 20% of the image quality classification data set can be used as the verification set. The initial evidence photo quality detection model is then trained with the training set to obtain the evidence photo quality detection model to be verified, and the verification set is input into the evidence photo quality detection model to be verified to obtain the evidence photo detection results. A loss function is constructed according to the evidence photo detection results and the corresponding evidence photo quality annotation results. According to the minimization of the loss function, the parameters of the evidence photo quality detection model to be verified are trained using an iterative algorithm to obtain the evidence photo quality detection model. The evidence photo quality detection model can distinguish evidence photos of different qualities.

[0070] It is understandable that the above-mentioned computer device may be pre-configured with Pytorch and other necessary libraries, and the above-mentioned initial evidence photo quality detection model may be a ResNet model. By calling Pytorch and other necessary libraries to train the parameters in the ResNet model, the corresponding evidence image quality detection model is obtained. Among them, in the process of pre-processing the evidence data of land change surveys over the years, the photo size may be processed, normalized, pixel value processed, etc.

[0071] Specifically, taking the evidence data to be processed as photos as an example, after the evidence data to be processed is input into the evidence photo quality detection model, the evidence data to be processed can be preprocessed, and the scaling, cropping, resizing, image enhancement, conversion to Tersor format, data normalization, and pixel value processing are performed in sequence to obtain the processed evidence photo data to be processed, so that the processed evidence photo data to be processed is processed into data suitable for model input, and then passed through the evidence photo detection model. The evidence photo detection model may include: convolution layer, normalization layer, fully connected layer, activation function, etc., by inputting the processed evidence photo data to be processed into the convolution layer to extract image features, and through the normalization processing training process, the gradient propagation is stabilized and it helps to prevent overfitting, and then it is processed through the fully connected layer to obtain a fully connected vector, and then the fully connected vector is processed by the activation function to introduce nonlinearity, which helps the model learn complex patterns, thereby obtaining image quality classification results. The image quality classification result may include: various types of shooting information, such as exposure, texture, shooting content, and shooting conditions of near and far views, etc., specifically including: shooting near views and far views; shooting exposure conditions including backlighting, non-backlighting, normal exposure, and abnormal exposure; shooting angles including: shooting lack of position, shooting without lack of position; shooting clarity conditions including: clear and unclear.

[0072] After obtaining the quality classification results of the evidence photos, the shooting quality of the evidence photos to be processed can be comprehensively analyzed based on the shooting information, and the evidence photos to be processed with abnormal exposure, missing shooting position, and unclear shooting information in the shooting information are regarded as unqualified shooting data and removed, and the qualified shooting data are used as filtered evidence photo data, thereby eliminating evidence photos that do not meet the quality requirements and retaining only evidence photos with qualified quality for subsequent analysis. The data with the evidence photo quality classification result of unqualified shooting will not participate in subsequent data processing and will be submitted to manual interactive processing.

[0073] In this embodiment, the evidence data to be processed is input into the evidence photo quality detection model for quality classification, and the evidence photo quality classification results are accurately obtained. According to the evidence photo quality classification results, the evidence photos that do not meet the quality requirements are eliminated. Specifically, by checking the clarity of the photos, it is ensured that the photo quality is high enough to clearly show the details of the land use nature and land use type of the photographed plot, which is helpful for subsequent classification and analysis, and checking the exposure of the photos to avoid information loss due to overexposure or underexposure, thereby improving the readability of the image and the accuracy of the analysis, thereby improving the accuracy and reliability of the land change investigation evidence photos.

[0074] Step S203, based on the filtered evidence photo data, the remote sensing data to be processed is subjected to an evidence shooting position quality check to obtain target evidence photo data, and corresponding target remote sensing image data is extracted according to the evidence shooting position.

[0075] It should be noted that the above-mentioned filtered evidence photo data is the evidence photo data to be processed that meets the quality requirements, the target evidence photo data is the filtered evidence photo data that has passed the spot position quality inspection, and the target remote sensing image data is the filtered remote sensing image data that has passed the spot position quality inspection. The above-mentioned quality inspection content includes: checking whether the evidence shooting has missed any position, whether the evidence shooting is correctly aligned, etc.

[0076] In one embodiment, the present application also provides a specific implementation method for performing a quality check of the shooting position of the evidence photo on the remote sensing data to be processed to obtain the target evidence photo data and the target remote sensing image data, the method comprising:

[0077] According to the filtered evidence photo data, the azimuth data during the shooting process and the corresponding evidence shooting positioning data are obtained, and the filtered remote sensing image data of the corresponding position are obtained at the same time; it is determined whether the shooting position of the filtered evidence photo data covers the four sides of the spot position of the filtered remote sensing image data; when the four sides of the remote sensing image position are covered, the center position of the spot of the filtered remote sensing data can be obtained, and based on the azimuth data of the evidence photo, it is determined whether the filtered evidence photo data surrounds the center position of the spot of the remote sensing image data; when it surrounds the center position of the spot, the filtered evidence photo data surrounding the center position of the spot and covering the four sides of the spot position is used as the target evidence photo data; the filtered remote sensing image data corresponding to the target evidence photo data is used as the target remote sensing image data.

[0078] The azimuth angle data in the above shooting process refers to the angle information when the evidence photo data is shot, such as the angle of the camera relative to the true north direction, which can be obtained through the magnetic direction sensing device inside the camera, or through post-processing software, or through manual measurement. The above filtered remote sensing image data refers to the remote sensing image data to be processed corresponding to the filtered evidence photo data.

[0079] Specifically, after obtaining the filtered evidence photo data, determine the azimuth data during the shooting process and the corresponding filtered remote sensing image data. By obtaining the corresponding azimuth data, it can be ensured that the photo is taken from multiple angles. Then, the filtered evidence photo data is used in the spot position quality check of the remote sensing image data to confirm whether the evidence photo can cover the spot position. When the spot position is not covered, it needs to be returned to manual judgment processing; when the spot position is covered, calculate the spot geometric center position of the filtered remote sensing image data, and then determine whether the filtered evidence photo data surrounds the spot geometric center position based on the azimuth data, that is, determine whether the evidence photo is a valid photo related to the spot geometric center position. When it is determined to surround the spot geometric center position, the filtered evidence photo data surrounding the spot geometric center position and covering the spot position is used as the target evidence photo data, and then the filtered remote sensing image data corresponding to the target evidence photo data is used as the target remote sensing image data. In the case of a lack of evidence photos in a certain direction, mark it out for manual review.

[0080] In this embodiment, based on the filtered evidence photo data, the quality check of the shooting position of the evidence photo is performed on the remote sensing image data to be processed, ensuring that the target evidence photo data fully covers all key areas around the photographed picture spot, and checking whether the photos are taken from different angles and directions to ensure that all sides of the picture spot are recorded, which helps to fully understand the actual ground land use of the picture spot, and by obtaining the geometric center position of the picture spot, it can be ensured that the photo is taken around the center of the picture spot, rather than deviating from the center, thereby improving the accuracy and reliability of the evidence photo; and combined with the azimuth data of the photo, it can be verified whether the evidence photo is taken from multiple directions to ensure that all directions of the picture spot are covered, thereby ensuring the accuracy of the position, reducing manual review, and improving the accuracy of the evidence photo data.

[0081] Step S204, input the target evidence photo data into the evidence photo classification model for image classification to obtain the evidence photo classification result, and in parallel input the target remote sensing image data into the remote sensing image classification model for image classification to obtain the remote sensing image classification result.

[0082] It should be noted that the above-mentioned evidence photo classification model is a pre-trained model for land use classification of the content of evidence photos. The evidence photo classification model is a neural network model whose input is the target evidence photo data and output is the evidence photo classification result, and has the ability to classify the target evidence photo data and can predict the evidence photo classification result; the remote sensing image classification model is a neural network model whose input is the target remote sensing image data and output is the remote sensing image classification result, and has the ability to classify the target remote sensing image data and can predict the remote sensing image classification result. The above-mentioned evidence photo classification result is the processing result obtained after photo classification of the target evidence photo, which is used to identify the land use type information of the content shot in the target evidence photo data, so as to be able to quickly obtain the land type information and land type attribute characteristics of the target evidence photo data. The above remote sensing image classification results are the processing results obtained after image classification of the target remote sensing image data, which are used to identify the land use category information in the target remote sensing image, so as to be able to quickly obtain the land type information and land type attribute characteristics of the remote sensing image data, for example, it may include "paddy field", "dry land", "forest land", "grassland", "industrial land", "highway land", etc.

[0083] In the process of inputting the target evidence photo data into the evidence photo classification model to classify the photo content, the target evidence photo data may be preprocessed first to obtain the processed target evidence photo data; the processed target evidence photo data is input into the evidence photo classification model, and the processed target evidence photo data is classified into the photo location category to obtain the evidence photo classification result. Among them, the above-mentioned evidence photo classification model can use the classification model models such as Resnet18, Resnet34, Resnet50, Resnet101, etc. in the ImageNet dataset, a large visual database with an important position in the field of computer vision. Remote sensing image classification models can include Resnet18, Resnet34, Resnet50, Resnet101, etc.

[0084] Specifically, after obtaining the target evidence photo data, the target evidence photo data is preprocessed, which may be to adjust the image size of the target evidence photo data and normalize the pixel value processing, so that the target evidence photo data meets the input requirements of the data in the evidence photo classification model, thereby obtaining the processed target evidence photo data, and inputting the processed target evidence photo data into the evidence photo classification model, extracting the image features of the target evidence photo data through the convolution layer, and accelerating the training process through the normalization layer, stabilizing the gradient propagation, and helping to prevent overfitting, and then processing through the fully connected layer to obtain a fully connected vector, The fully connected vector is then processed through an activation function to introduce nonlinearity, which helps the model learn complex patterns and obtain the classification results of the evidence photos, which may include, for example, "agricultural land", "construction land", "unused land" and "others". "Agricultural land" may include "arable land", "garden", "grassland", "forest land", etc.; "construction land" may include "urban construction land", "rural residential areas", "public facilities land", etc.; "unused land" may include "wasteland", "saline-alkali land", "swamp", "sand land", etc.; "others" may include "water area", "tidal flat", "glacier", etc.

[0085] Similarly, after acquiring the target remote sensing image data, the target remote sensing image data can be classified into image land types in parallel. By preprocessing the target remote sensing image data, the image size of the target remote sensing image data can be adjusted and the pixel value can be normalized so that the target remote sensing image data meets the input requirements of the data in the remote sensing image classification model, thereby obtaining the processed target remote sensing image data, and inputting the processed target remote sensing image data into the remote sensing image classification model, and performing convolution layer, normalization layer, fully connected layer and activation function processing in sequence to obtain the remote sensing image classification result.

[0086] In the embodiment of the present application, by inputting the target evidence photo data into the evidence photo classification model for image classification, the land classification result of the evidence photo can be accurately determined, and the target remote sensing image data is input into the remote sensing image classification model for land classification, and the remote sensing image land classification result is accurately determined, thereby improving the accuracy of the classification result determination and facilitating improving the efficiency of the verification of the land change investigation and evidence results.

[0087] Step S205, analyzing and processing the classification results of the evidence photos and the classification results of the remote sensing images to generate verification results of the land change survey data to be processed.

[0088] It should be noted that the verification results of the above-mentioned land change survey and evidence data are used to characterize the verification status of the land change survey and evidence data, which may include: whether the pending land change survey and evidence data is verified qualified or unqualified. For the qualified verification, the corresponding predicted land classification results are included.

[0089] In one embodiment of this embodiment, a specific implementation method for generating the verification result of the pending land change investigation and evidence data is also provided. Figure 3 As shown, the method may include the following steps:

[0090] S301, performing soft voting processing of integrated learning on the classification results of the evidence photos and the classification results of the remote sensing images to obtain the target classification results.

[0091] It should be noted that the above-mentioned evidence photo classification results may include evidence photo categories and corresponding evidence photo classification prediction probability values. For example, if the evidence photo categories include "rural roads", "urban residential land" and "rural homestead", the evidence photo classification prediction probability value corresponding to "rural roads" is 10%, the evidence photo classification prediction probability value corresponding to "urban residential land" is 85%, and the evidence photo classification prediction probability value corresponding to "rural homestead" is 5%. The "urban residential land" with the largest final prediction probability value of the above-mentioned evidence photos is the final classification result.

[0092] Specifically, the category of evidence photos and the corresponding predicted probability value of evidence photos are determined from the classification results of evidence photos, and the category of remote sensing images and the corresponding predicted probability value of remote sensing images are determined from the classification results of remote sensing images; the photo prediction probability value is output as the probability value corresponding to the category of evidence photos, and the remote sensing image prediction probability value is output as the probability value corresponding to the category of remote sensing images; for each piece of land change investigation evidence data to be processed, the predicted probability value of the evidence photo and the predicted probability value of the remote sensing county are combined and learned by weighted soft voting according to the preset weight distribution ratio to obtain the target classification result.

[0093] It is understandable that when the computer device is configured with the PyTorch library and the Scikit-Learn library, the soft voting (Soft Voting) in the VotingClassifier module of the Scikit-Learn library can be called through the PyTorch interface to perform a combination strategy to perform a multi-model classification combination on the classification results of the evidence photos and the classification results of the remote sensing images. Thereby obtaining the final classification result of the target. Among them, in the multi-model classification combination, it is necessary to set the weight distribution ratio of the classification results of the evidence photos and the classification results of the remote sensing images in advance according to the importance of each classification result to the target classification result. For example, the weight value of the classification result of the evidence photos is 60%, and the weight value of the classification result of the remote sensing image is 40%. Then, for each pending land change investigation evidence data, the predicted probability value of the evidence photos and the predicted probability value of the remote sensing image are voted for by a combination strategy according to the preset weight distribution ratio, and the classification result with the highest probability voting value is used as the target classification result.

[0094] In this embodiment, all classification probabilities of various machine learning classification results are averaged using the software voting mechanism, averaging the prediction probabilities in various classification models, and also reflecting that when the probability of a single category is high, the classification probability will be correspondingly high after comprehensive averaging. At the same time, by using the software voting mode to integrate the image classification results of each classification model, the prediction information of each classification model can be fully utilized, and the corresponding prediction probability value is also taken into account, thereby reducing the image of the single classification model predicting errors and improving the accuracy of the target classification result determination.

[0095] S302: Comprehensively analyze and judge the target classification results to obtain the verification results of the land change survey data to be processed.

[0096] In this step, after obtaining the target classification results, the target classification results can be analyzed. Data with high classification probability values ​​can be directly judged as the corresponding class; data with low classification probability values ​​are assigned to manual review to determine the final land category; data that cannot meet the relevant requirements for evidence in the land change investigation are judged as unqualified; thereby obtaining the verification results of the final land change investigation evidence data to be processed.

[0097] The verification method of land change survey evidence data provided by the embodiment of the present application includes: obtaining land change survey data to be processed, the land change survey data to be processed includes evidence photo data to be processed and corresponding remote sensing data to be processed, and filtering the evidence data to be processed through an evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the land change survey evidence data of previous years and the corresponding evidence quality annotation results; based on the filtered evidence photo data, the remote sensing data to be processed is subjected to an evidence photo shooting position quality check to obtain target evidence photo data and target remote sensing image data; the target evidence photo data is input into the evidence photo classification model to classify the photo shooting content to obtain the evidence photo classification result, and the target remote sensing image data is input into the remote sensing image classification model in parallel to classify the image land type to obtain the remote sensing image classification result; the evidence photo classification result and the remote sensing image classification result are subjected to integrated learning soft voting combination processing to generate the verification result of the land change survey evidence data to be processed. Compared with the prior art, this technical solution can detect the photo shooting conditions by inputting the data of the evidence photos to be processed into the evidence photo quality detection model, filter out the evidence photos that do not meet the conditions, and obtain the filtered evidence photo data. It can also perform a quality check on the shooting positions of the evidence photos of the corresponding spots on the remote sensing image data to be processed, and accurately determine the target evidence photo data and target remote sensing image data that have passed the quality check, thereby ensuring that the data participating in the subsequent classification of evidence photos are all valid data, and classify the target evidence photo data and target remote sensing image data into land categories through their respective corresponding classification models, and can accurately determine the classification results of the evidence photos and the classification results of the remote sensing images, reducing the workload of manual inspection of the evidence results, and can assist manual verification of the field investigation and evidence photo work, and then analyze and process the classification results of the evidence photos and the remote sensing image, so as to obtain the verification results of the land change investigation and evidence data to be processed, reducing the error probability of manual judgment, and effectively improving the accuracy of the verification of the evidence results.

[0098] In another exemplary embodiment of the present application, in order to accurately classify and verify the land change survey and evidence data of this year, the present application embodiment provides an implementation method for continuously classifying and learning and optimizing relevant specific models using the land change survey and evidence data of previous years, such as Figure 4 As shown, the method may include the following steps:

[0099] Step S401, obtaining the data of evidence of land change surveys over the years; the data of evidence of land change surveys over the years include: the data of evidence photos and the time information obtained during the shooting process of the land survey evidence photos, the shooting direction information, the position information, the pitch angle information, and the data of evidence of land change surveys over the years are annotated with land classification results.

[0100] It should be noted that the above-mentioned data on evidence of land changes in previous years may be the total results data obtained after inspection and acceptance of field evidence of land changes in previous years in the detection area. The above-mentioned evidence photo results data may be represented in the format of photos, videos, etc., which may include the scope of the monitored map, basic information of the land type after the change is confirmed, etc. The time information for obtaining the survey and evidence data may be relevant information such as the shooting time and shooting angle of the survey and evidence photos.

[0101] Optionally, the above-mentioned historical land change survey and evidence data can be extracted from the historical land change survey results data of each year, and can also be extracted from other survey platforms. It only needs to meet the basic shooting requirements and land classification requirements. This embodiment does not impose any restrictions on the method of obtaining the historical evidence photo data.

[0102] After obtaining the data of land change investigation and evidence in previous years, data labeling can be performed through data cleaning and sorting to obtain the labeling results with corresponding land categories after inspection and verification. The land category labeling results are known land category results.

[0103] Step S402, pre-processing the data of investigation and evidence of land changes in previous years to obtain processed data of investigation and evidence of land changes in previous years.

[0104] After obtaining the historical land change survey and evidence data, the historical land change survey and evidence data can be pre-processed, and the processed historical land change survey and evidence data can be obtained by performing operations such as image size adjustment, scaling and cropping, image enhancement, Tersor conversion, and normalization. The processed historical land change survey and evidence data are reviewed and checked. The review and inspection content includes eliminating the unclear land category descriptions and obvious errors in the reverse shooting angles of the historical land change survey and evidence data, and separately annotating the evidence photo data that is obviously not a land category feature, such as the land category change of bulldozing, and the data of the evidence photo data is not an undetermined land category. Such data will affect the accuracy of deep learning land category determination and classification.

[0105] In this embodiment, the data of historical land change surveys and evidence are preprocessed to obtain processed data of historical land change surveys and evidence, which can make the processed data of historical land change surveys and evidence more regular, facilitate improving the model training effect, help reduce the complexity of the model, and improve the interpretability and generalization ability of the model.

[0106] Step S403, dividing the processed data of land change investigation and evidence collection in previous years into a training set and a validation set according to a preset ratio.

[0107] Step S404, using the training set to train the initial evidence photo classification model, fine-tuning the number of training set data classifications and other parameters of all layers of the model's neurons to obtain the evidence photo classification model to be verified, and inputting the verification set into the evidence photo classification model to be verified to obtain the output result.

[0108] Step S405, constructing a loss function according to the output results and the land category labeling results, and minimizing the loss function by using an iterative algorithm to train the parameters of the evidence photo classification model to be verified to obtain the evidence photo classification model.

[0109] After obtaining the processed data on land change surveys and evidence from previous years, the processed key features can be divided into a training set and a validation set. The training set is used for model training, and the model learns the mapping relationship between input and output; the validation set is used to adjust model parameters and prevent overfitting, while monitoring the specific fitting situation during model training.

[0110] In the process of data division, the processed data of land change survey and evidence in previous years can be divided into a training set and a validation set according to a certain ratio, for example, the training set is 80% of the processed data of land change survey and evidence in previous years, and the validation set is 20% of the processed data of land change survey and evidence in previous years. Specifically, the Pytorch library in the Python environment can be used to train the classification model of the land change survey and evidence photo data of previous years. In the process of building the evidence photo classification model, the training set can be input into the initial evidence photo classification model, the model parameters are trained to obtain the evidence photo classification model to be verified, and the validation set can be input into the evidence photo classification model to be verified to obtain the output result, and the loss function is constructed according to the output result and the land type labeling result corresponding to the validation set. According to the minimization of the loss function, the initial evidence photo classification model is iteratively trained to obtain the evidence photo classification model. Among them, the initial evidence photo classification model can be a model such as Resnet18, Resnet34, Resnet50, and Resnet101.

[0111] After obtaining the output result, a loss function can be constructed based on the output result and the labeled land class result, and the initial evidence photo classification model can be optimized according to the minimization of the loss function to obtain the evidence photo classification model. According to the difference between the output result and the land class labeling result, the parameters in the initial land class labeling result are updated to achieve the purpose of training the initial land class labeling result and obtain the evidence photo classification model. The updating of the parameters in the initial evidence photo classification model can be to update the matrix parameters such as the weight matrix and the bias matrix in the initial evidence photo classification model. The above-mentioned weight matrix and bias matrix include but are not limited to the matrix parameters in the convolution layer, feedforward network layer, and fully connected layer in the initial evidence photo classification model.

[0112] After obtaining the evidence photo classification model to be trained, the verification set can be input into the evidence photo classification model to be verified, and the hyperparameters in the model can be continuously adjusted through an iterative training algorithm according to the land category labeling results corresponding to the verification set, so that the evidence photo classification model to be verified corresponding to the optimal hyperparameters can be used as the evidence photo classification model.

[0113] Since the historical evidence photo data in this embodiment includes information on the time when the evidence photos were taken, it can be used as an additional parameter for the shooting of the evidence photos, thereby more comprehensively considering various information of the evidence photo data, thereby improving the accuracy of model training. By training the initial evidence photo classification model, the model parameters in the initial evidence photo classification model can be iteratively trained more accurately and comprehensively, making the constructed evidence photo classification model better, and thus making the determined evidence photo classification results more accurate, which is convenient for assisting manual verification of field investigation work, reduces manual workload, and thus improves verification efficiency and accuracy.

[0114] Similarly, the training process of the above remote sensing image classification model is similar to the training process of the evidence photo classification model, which is obtained by obtaining historical remote sensing image data, which are annotated with land classification results, and then preprocessing the historical remote sensing image data to obtain processed historical remote sensing image data, and dividing the processed historical remote sensing image data into training sets and verification sets according to a preset ratio, and then inputting the training set into the initial remote sensing image classification model to obtain the remote sensing image classification model to be verified, and inputting the verification set into the remote sensing image classification model to be verified to obtain the corresponding prediction results; according to the land classification results and the prediction results, the loss function is calculated, and according to the minimization of the loss function, the parameters of the remote sensing image classification model to be verified are trained using an iterative algorithm to obtain the remote sensing image classification model. Among them, the above historical remote sensing image data can be satellite images or high-definition aerial image information within the monitoring range of previous years in the monitoring area, including resolution information and acquisition time information.

[0115] In this embodiment, by training the evidence photo classification model and the remote sensing image classification model, different land coverage types can be identified and classified. According to the trained evidence photo classification model and remote sensing image classification model, the evidence photo results and remote sensing image data of this year can be automatically quality controlled and inspected, which effectively improves the accuracy of the evidence results and reduces the workload of manual inspection of the evidence results. At the same time, according to the image category and the preset probability, the target classification result can be determined more comprehensively, thereby effectively guiding the classification and identification of land types in this year's land change survey and evidence, reducing the probability of manual judgment errors, improving the judgment work efficiency, and facilitating the smooth implementation of the annual land change survey.

[0116] Based on the same inventive concept, the embodiment of the present application also provides a verification device for land change investigation and evidence data for implementing the verification method for land change investigation and evidence data involved in the above-mentioned. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of one or more verification devices for land change investigation and evidence data provided below can refer to the limitations of the verification method for land change investigation and evidence data above, and will not be repeated here.

[0117] In an exemplary embodiment, Figure 5 As shown, a verification device for land change investigation and evidence data is provided, and the verification device for land change investigation and evidence data includes:

[0118] The acquisition module 510 is used to acquire the land survey change data to be processed; the land survey change data to be processed includes the evidence data to be processed and the corresponding remote sensing data to be processed;

[0119] The filtering processing module 520 is used to filter the evidence data to be processed through the evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the evidence data of land change surveys in previous years and the corresponding evidence quality labeling results;

[0120] The quality inspection module 530 is used to perform a quality inspection of the evidence shooting position on the remote sensing data to be processed based on the filtered evidence photo data, obtain the target evidence photo data, and extract the corresponding target remote sensing image data according to the evidence shooting position;

[0121] The evidence photo classification module 540 is used to input the target evidence photo data into the evidence photo classification model to perform image classification to obtain the evidence photo classification result, and in parallel input the target remote sensing image data into the remote sensing image classification model to perform image classification to obtain the remote sensing image classification result;

[0122] The analysis module 550 is used to analyze and process the classification results of the evidence photos and the classification results of the remote sensing images to generate verification results of the land change survey data to be processed.

[0123] As an optional implementation, the filtering processing module 520 is specifically used to:

[0124] Input the evidence data to be processed into the evidence photo quality detection model for quality classification, and obtain the evidence photo quality classification result;

[0125] The evidence data to be processed is filtered according to the quality classification result of the evidence photo to obtain filtered evidence photo data.

[0126] As an optional implementation, the quality inspection module 530 is specifically used to:

[0127] According to the filtered evidence photo data, the azimuth data during the shooting process and the remote sensing image data corresponding to the evidence shooting position are obtained;

[0128] Determine whether the shooting location of the filtered evidence photo data covers the surrounding areas of the spot location of the remote sensing image data of the evidence shooting location;

[0129] When the area around the spot is covered, the geometric center of the spot of the remote sensing image data of the evidence shooting location is obtained, and based on the azimuth data, it is determined whether the filtered evidence photo data surrounds the geometric center of the spot;

[0130] When surrounding the geometric center of the spot, the filtered evidence photo data surrounding the geometric center of the spot and covering the four sides of the spot position is used as the target evidence photo data;

[0131] The remote sensing data corresponding to the target evidence photo data is extracted according to the evidence shooting position as the target remote sensing image data.

[0132] As an optional implementation, the image classification module 540 is specifically used for:

[0133] Preprocessing the target evidence photo data to obtain processed target evidence photo data;

[0134] The processed target evidence photo data is input into the evidence photo classification model, and the processed target evidence photo data is classified and calculated into land categories to obtain the evidence photo classification results.

[0135] As an optional implementation, the analysis module 550 is specifically used for:

[0136] Perform soft voting processing on the classification results of evidence photos and remote sensing images through ensemble learning to obtain the target classification results;

[0137] Comprehensively analyze and judge the target classification results to obtain the verification results of the land change survey data to be processed.

[0138] As an optional implementation, the analysis module 550 is further configured to:

[0139] Determine the land category of the evidence photo and the corresponding probability value of the evidence photo classification from the evidence photo classification results, and determine the remote sensing image category and the corresponding remote sensing image classification probability value from the remote sensing image classification results; the evidence photo classification probability value is output as the probability value corresponding to the land category of the evidence photo, and the remote sensing image classification probability value is output as the probability value corresponding to the remote sensing image category;

[0140] For each piece of land change survey data to be processed, the classification probability value of the evidence photo and the classification probability value of the remote sensing image are soft-voted with preset weights according to the preset weight distribution ratio to obtain the target classification result.

[0141] As an optional implementation, the evidence photo classification model is constructed in the following way:

[0142] Obtain the data of land change surveys and evidence collection over the years; the data of land change surveys and evidence collection over the years include: evidence photo results data and time information obtained during the shooting of land survey evidence photos, shooting azimuth information, location information, and pitch angle information. The data of land change surveys and evidence collection over the years are annotated with land classification results;

[0143] Pre-processing the data of investigation and evidence of land changes in previous years to obtain processed data of investigation and evidence of land changes in previous years;

[0144] Divide the processed historical land change survey and evidence data into a training set and a validation set according to a preset ratio;

[0145] The initial evidence photo classification model is trained using the training set, and the number of classifications and other parameters of the training set data are fine-tuned for all layers of the model's neurons to obtain the evidence photo classification model to be verified, and the verification set is input into the evidence photo classification model to be verified to obtain the output result;

[0146] A loss function is constructed according to the output results and the land type labeling results. According to the minimization of the loss function, an iterative algorithm is used to train the parameters of the evidence photo classification model to be verified to obtain the evidence photo classification model.

[0147] Among them, by implementing this classification model method, the technical solution can detect the photo shooting situation by inputting the data of the evidence photos to be processed into the evidence photo quality detection model, filter out the evidence photos that do not meet the conditions, and obtain the filtered evidence photo data, and perform a quality check on the location of the evidence photos taken on the remote sensing image data to be processed, and determine the target evidence photo data that passes the quality check, thereby ensuring that the data participating in the subsequent photo classification are all valid data, and ensuring the accuracy of the subsequent integrated learning, and classifying the target evidence photo data and the target remote sensing image data into target land use types through their respective corresponding classification models, which can accurately determine the classification results of the evidence photos and the classification results of the remote sensing images, reducing the workload of manual inspection and judgment of the results of the evidence photos, and can assist manual verification of the quality of field survey work, and then perform integrated learning and analysis on the classification results of the evidence photos and the remote sensing images, thereby obtaining the verification results of the pending land change survey evidence data, reducing the error probability of manual judgment, and effectively improving the accuracy of the evidence results.

[0148] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device (including a central processing unit and a graphics processing unit) is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store evidence photos and remote sensing image data and corresponding label processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for verifying land change survey evidence data is implemented.

[0149] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0150] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0151] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0152] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0153] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0154] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0155] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0156] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0157] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for verifying land change investigation and evidence data, characterized in that: The verification method of the land change investigation and evidence data includes: Acquire the land change survey data to be processed; the land change survey data to be processed includes the evidence data to be processed and the corresponding remote sensing data to be processed; The evidence data to be processed is filtered through an evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the evidence data of land change surveys over the years and the corresponding evidence quality labeling results; Based on the filtered evidence photo data, the remote sensing data to be processed is subjected to an evidence shooting position quality check to obtain target evidence photo data, and corresponding target remote sensing image data is extracted according to the evidence shooting position; Inputting the target evidence photo data into the evidence photo classification model for image classification to obtain the evidence photo classification result, and in parallel inputting the target remote sensing image data into the remote sensing image classification model for image classification to obtain the remote sensing image classification result; The classification results of the evidence photos and the classification results of the remote sensing images are analyzed and processed to generate verification results of the land change survey data to be processed.

2. The verification method of land change investigation and evidence data according to claim 1 is characterized in that: The evidence data to be processed is filtered through the evidence photo quality detection model to obtain filtered evidence photo data, including: Inputting the evidence data to be processed into the evidence photo quality detection model for quality classification to obtain the evidence photo quality classification result; The evidence data to be processed is filtered according to the quality classification result of the evidence photo to obtain the filtered evidence photo data.

3. The verification method of land change investigation and evidence data according to claim 1 is characterized in that: Based on the filtered evidence photo data, the remote sensing data to be processed is subjected to an evidence shooting position quality check to obtain target evidence photo data, and corresponding target remote sensing image data is extracted according to the evidence shooting position, including: According to the filtered evidence photo data, obtaining azimuth data during the shooting process and remote sensing image data corresponding to the evidence shooting location; Determine whether the shooting position of the filtered evidence photo data covers the surroundings of the spot position of the remote sensing image data of the evidence shooting position; When the surroundings of the spot position are covered, the geometric center position of the spot of the remote sensing image data of the evidence shooting position is obtained, and based on the azimuth data, it is determined whether the filtered evidence photo data can surround the geometric center of the spot; When surrounding the geometric center of the spot, the filtered evidence photo data surrounding the geometric center of the spot and covering the four sides of the spot position is used as the target evidence photo data; The remote sensing data corresponding to the target evidence photo data is extracted according to the evidence shooting position as the target remote sensing image data.

4. The verification method of land change investigation and evidence data according to claim 1 is characterized in that: The target evidence photo data is input into the evidence photo classification model for image classification to obtain the evidence photo classification results, including: Preprocessing the target evidence photo data to obtain processed target evidence photo data; The processed target evidence photo data is input into the evidence photo classification model, and the processed target evidence photo data is classified and calculated into ground type categories to obtain the evidence photo classification result.

5. The verification method of land change investigation and evidence data according to claim 1 is characterized in that: Analyze and process the classification results of the evidence photos and the classification results of the remote sensing images to generate verification results of the land change survey data to be processed, including: Performing soft voting processing of integrated learning on the classification results of the evidence photos and the classification results of the remote sensing images to obtain target classification results; A comprehensive analysis and judgment is performed on the target classification results to obtain the verification results of the land change survey data to be processed.

6. The verification method of land change investigation and evidence data according to claim 5 is characterized in that: The classification results of the evidence photos and the classification results of the remote sensing images are subjected to soft voting processing of integrated learning to obtain target classification results, including: Determine the land category of the evidence photo and the corresponding probability value of the evidence photo classification from the evidence photo classification result, and determine the remote sensing image category and the corresponding remote sensing image classification probability value from the remote sensing image classification result; the evidence photo classification probability value is output as the probability value corresponding to the land category of the evidence photo, and the remote sensing image classification probability value is output as the probability value corresponding to the remote sensing image category; For each piece of land change survey data to be processed, the classification probability value of the evidence photo and the classification probability value of the remote sensing image are soft-voted with preset weights according to the preset weight distribution ratio to obtain the target classification result.

7. The verification method of land change investigation and evidence data according to claim 1 is characterized in that: The evidence photo classification model is constructed in the following way: Obtaining historical land change survey and evidence data; the historical land change survey and evidence data includes: evidence photo results data and time information obtained during the land survey and evidence photo shooting process, shooting azimuth information, location information, and pitch angle information; the historical land change survey and evidence data are annotated with land classification results; Preprocessing the historical land change investigation and evidence data to obtain processed historical land change investigation and evidence data; Dividing the processed historical land change survey and evidence data into a training set and a validation set according to a preset ratio; Using the training set to train the initial evidence photo classification model, fine-tuning all layers of the model's neurons according to the number of data classifications and other parameters of the training set to obtain an evidence photo classification model to be verified, and inputting the verification set into the evidence photo classification model to be verified to obtain an output result; A loss function is constructed according to the output result and the land category labeling result, and the parameters of the evidence photo classification model to be verified are trained using an iterative algorithm in accordance with the minimization of the loss function to obtain the evidence photo classification model.

8. A verification device for land change investigation and evidence data, characterized in that: The verification device for the land change investigation and evidence data comprises: An acquisition module is used to acquire the land survey change data to be processed; the land survey change data to be processed includes the evidence data to be processed and the corresponding remote sensing data to be processed; A filtering processing module, used for filtering the evidence data to be processed through an evidence photo quality detection model to obtain filtered evidence photo data; the evidence photo quality detection model is constructed based on the evidence data of land change surveys in previous years and the corresponding evidence quality labeling results; A quality inspection module, for performing a quality inspection of the evidence shooting position on the remote sensing data to be processed based on the filtered evidence photo data, obtaining target evidence photo data, and extracting corresponding target remote sensing image data according to the evidence shooting position; An evidence photo classification module is used to input the target evidence photo data into the evidence photo classification model for image classification to obtain an evidence photo classification result, and in parallel input the target remote sensing image data into the remote sensing image classification model for image classification to obtain a remote sensing image classification result; The analysis module is used to analyze and process the classification results of the evidence photos and the classification results of the remote sensing images to generate verification results of the land change survey data to be processed.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for verifying land change survey data as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for verifying land change survey data described in any one of claims 1 to 7 is implemented.

Images