Remote Sensing Map Region Recognition Method, Device, System and Readable Storage Medium

Abstract

The present application provides a method, apparatus, system, and readable storage medium for remote sensing map area recognition. The method for remote sensing map area recognition includes: obtaining a satellite remote sensing image map and marking data, where the image includes a restricted area of a wind farm layout, and the marking data is used to mark the area contour of the restricted area; generating a training sample set based on the image and the marking data; inputting the training sample set into a region recognition model to train the region recognition model; when the recognition accuracy of the region recognition model reaches a predetermined recognition accuracy, obtaining a trained target region recognition model and outputting the pixel data of the restricted area of the target region recognition model. In this way, the recognition accuracy is relatively high and the human input is small.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a method, device, system and readable storage medium for remote sensing map area recognition. Background Art

[0002] With the development of wind power resources, some wind farms are located near restricted areas such as residential areas, forest areas, protected areas, and red lines, thus triggering discussions on the relationship between the siting of wind farms and these restricted areas. In this process, it is necessary to first identify these restricted areas.

[0003] In the related art, engineers manually identify restricted areas based on remote sensing maps, which requires a large amount of manpower and has low accuracy. Summary of the Invention

[0004] The present application provides a method, device, system and readable storage medium for remote sensing map area recognition, which requires less manpower and has higher accuracy.

[0005] The present application provides a method for remote sensing map area recognition, including:

[0006] Obtaining a satellite remote sensing image map and marking data, where the image includes restricted areas of wind farm layouts, and the marking data is used to mark the regional contours of the restricted areas;

[0007] Generating a training sample set according to the image and the marking data;

[0008] Inputting the training sample set into a region recognition model to train the region recognition model;

[0009] When the recognition accuracy of the region recognition model reaches a predetermined recognition accuracy, obtaining a trained target region recognition model and outputting the pixel data of the restricted areas of the target region recognition model.

[0010] Further, the generating a training sample set according to the image and the marking data includes:

[0011] Performing image augmentation processing on the image to form an initial training sample set;

[0012] Normalizing the image format in the initial training sample set and marking the boundaries of the restricted areas in the initial training sample set according to the types of restricted areas to generate the training sample set.

[0013] Further, after obtaining the satellite remote sensing image map and the marking data, the method further includes:

[0014] Generating a data set according to the image and the marking data;

[0015] Divide the dataset according to the data ratios of the training set, the development set, and the test set to obtain the training sample set, the development sample set, and the test sample set;

[0016] Use the development sample set to adjust the hyperparameters of the region recognition model and update the region recognition model;

[0017] After inputting the training sample set into the region recognition model and training the region recognition model, the method further includes:

[0018] Input the test sample set into the trained region recognition model to test the trained region recognition model and obtain a test result;

[0019] Determine the recognition accuracy of the trained region recognition model according to the test result.

[0020] Further, determining the recognition accuracy of the trained region recognition model according to the test result includes:

[0021] Determine the recognition accuracy of the trained region recognition model according to the pixel data of the restricted region output by the target region recognition model and the error pixel data in the pixel data.

[0022] Further, after the recognition accuracy of the region recognition model reaches a predetermined recognition accuracy to obtain a trained target region recognition model and obtain the pixel data of the restricted region output by the target region recognition model, the method further includes:

[0023] Display the pixel data of the restricted region output by the target region recognition model on the front end to obtain the restricted region.

[0024] Further, the front end includes a satellite remote sensing image map image displayed in a Web GIS interface;

[0025] After displaying the pixel data of the restricted region output by the target region recognition model on the front end and determining the restricted region, the method further includes:

[0026] Obtain the region image selected by the user from the image;

[0027] Generate input data according to the region image;

[0028] Input the input data into the target region recognition model to output region pixel data;

[0029] Extract the boundary information of the restricted region according to the region pixel data;

[0030] Display the boundary information of the restricted area to the front end.

[0031] Furthermore, the target area recognition model is a plurality of target area recognition models. The step of displaying the boundary information of the restricted area to the front end includes:

[0032] Display the boundary information of the restricted area corresponding to the plurality of target area recognition models to the front end;

[0033] Determine the evaluation indexes of the plurality of target area recognition models;

[0034] Select the target area recognition model corresponding to the best evaluation index among the evaluation indexes according to the evaluation indexes;

[0035] Use the area image, the selected model, and the area pixel data output by the selected model as samples and store them in the database.

[0036] Furthermore, the step of obtaining the satellite remote sensing image map and the marking data includes:

[0037] Obtain the input information of the user, where the input information includes the satellite remote sensing image map and the marking data input by the user.

[0038] Furthermore, after obtaining the satellite remote sensing image map and the marking data input by the user, the method further includes:

[0039] Obtain the storage information in the database, where the storage information includes the satellite remote sensing image map and the marking data stored in the database;

[0040] If the data increment of the input information compared with the storage information is greater than the preset increment, generate a training sample set according to the input information, and input the training sample set into the area recognition model to train the area recognition model.

[0041] The present application provides a remote sensing map area recognition device, including:

[0042] A data acquisition module, configured to acquire a satellite remote sensing image map and marking data, where the image includes a restricted area of a wind farm layout, and the marking data is used to mark the area contour of the restricted area;

[0043] A sample set generation module, configured to generate a training sample set according to the image and the marking data;

[0044] A model training module, configured to input the training sample set into an area recognition model to train the area recognition model;

[0045] A model optimization module, configured to obtain a trained target region recognition model and output pixel data of a restricted region of the target region recognition model when the recognition accuracy of the region recognition model reaches a predetermined recognition accuracy.

[0046] The present application provides a remote sensing map region recognition method system, including one or more processors, configured to implement the method as described in any one of the above.

[0047] The present application provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, the method as described in any one of the above is implemented.

[0048] In some embodiments, the remote sensing map region recognition method of the present application generates a training sample set based on a satellite remote sensing image map and marker data for marking the region contour of a restricted region, and then inputs the training sample set into a region recognition model to train the region recognition model to obtain restricted region boundary information. When the recognition accuracy of the region recognition model reaches a predetermined recognition accuracy, a trained target region recognition model is obtained and pixel data of the restricted region of the target region recognition model is output. In this way, the region recognition model is automatically trained and automatically recognized, with little human input, and marker data is used to reduce interference from other various types of geographical information in the satellite remote sensing image map during recognition, resulting in relatively high recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The figure shows a schematic flow chart of the remote sensing map region recognition method provided by an embodiment of the present application;

[0050] Figure 2 As shown in Figure 1 Another schematic flow chart of the remote sensing map region recognition method shown;

[0051] Figure 3 As shown in Figure 1 An application schematic diagram of the remote sensing map region recognition method where the restricted region is farmland;

[0052] Figure 4 As shown in Figure 1 An application schematic diagram of the remote sensing map region recognition method where the restricted region is a residential area;

[0053] Figure 5 As shown in Figure 1 An identification result schematic diagram of the remote sensing map region recognition method where the restricted region is farmland;

[0054] Figure 6 As shown in Figure 1 A specific schematic flow chart of the remote sensing map region recognition method shown;

[0055] Figure 7 The figure shows a schematic diagram of modules of a remote sensing map area recognition device provided by an embodiment of the present application;

[0056] Figure 8 The figure shows a block diagram of modules of a remote sensing map area recognition method system provided by an embodiment of the present application. Detailed implementation manners

[0057] Here, exemplary embodiments will be described in detail, and examples thereof are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this specification. On the contrary, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0058] It should be noted that: In other embodiments, the steps of the corresponding methods are not necessarily executed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may also be combined into a single step for description in other embodiments.

[0059] To solve the technical problems of large human input and low accuracy, an embodiment of the present application provides a remote sensing map area recognition method. According to satellite remote sensing image map images and marking data for marking the area contours of restricted areas, a training sample set is generated. Then, the training sample set is input into a region recognition model to train the region recognition model, and boundary information of the restricted area is obtained. When the recognition accuracy of the region recognition model reaches a predetermined recognition accuracy, a trained target region recognition model is obtained and pixel data of the restricted area of the target region recognition model is output. In this way, the region recognition model is automatically trained and automatically recognized, with small human input and short time consumption. Moreover, by using the marking data, interference from other various geographical information in the satellite remote sensing image map images to the recognition is reduced, and the recognition accuracy is relatively high.

[0060] Figure 1 The figure shows a schematic flowchart of the remote sensing map area recognition method provided by an embodiment of the present application.

[0061] As Figure 1 shown, the remote sensing map area recognition method may include the following steps 110 to 140:

[0062] Step 110: Obtain a satellite remote sensing image map and marking data. The image includes restricted areas of the wind farm layout, and the marking data is used to mark the regional contours of the restricted areas.

[0063] The satellite remote sensing image map can serve as a high-definition image source. The satellite remote sensing image map can include an image of the area to be planned for the wind farm. The image of the area to be planned can include restricted areas to define the areas where the wind farm can be planned. The restricted areas refer to the areas that need to be identified and where wind farms cannot be established. The marking data is used to mark the longitude and latitude coordinates of multiple corner points of the image and the contours of different types of restricted areas. The above marking data is marked by the user and can be used as label information for training the model. The restricted area contour can include a restricted area contour line. Exemplarily, the restricted area contour line is an irregular red line.

[0064] In some embodiments, the above step 110 may include obtaining the user's input information, where the input information includes the satellite remote sensing image map and marking data input by the user. Among them, an interface for manual input is added to the system. In this way, the user's input information can facilitate the user's marking and drawing, provide scalability, and subsequently, the marking data and the satellite remote sensing image map can be used as samples and stored in the database. In other embodiments, a remote sensing image is obtained, and the satellite remote sensing image map is generated using the remote sensing image as the basis for subsequent training.

[0065] Among them, after the above step 110, the remote sensing map area recognition method further includes a first step of obtaining the stored information in the database, where the stored information includes the satellite remote sensing image map and marking data stored in the database. The second step is that if the data increment of the input information compared to the stored information is greater than a preset increment, and the data increment of the input information compared to the stored information being greater than the preset increment can reflect that the modified content is relatively large, then a training sample set is generated based on the input information, and the training sample set is input into the area recognition model to train the area recognition model. In this way, in order to avoid retraining the area recognition model from scratch due to little manual modification and little data update, when there are more data modifications, the area recognition model is updated for training to improve the effectiveness of training the area recognition model.

[0066] Among them, after the above step 110, the method may further include entering the pre-collected images and marking data into the database. Among them, a non-relational database is established, and its key values at least include the data entry index index, the image address, the longitude and latitude of the image corner points, and the boundary inflection point coordinates of different types of areas in the image; by storing the image in the non-relational database and marking the storage path of the image.

[0067] The above preset increment may be greater than or equal to 1%. When the preset increment is 1% of the data increment of the stored information, steps 120 to 140 are re-performed, so as to upgrade the region recognition model, and store the upgraded region recognition model in the database, so as to upgrade the database.

[0068] Step 120: Generate a training sample set according to the image and the marked data.

[0069] In some embodiments of the above step 120, the above step 120 may further include the first step of performing image augmentation processing on the image to form an initial training sample set. The second step is to normalize the image format in the initial training sample set. For example, all images in the data set are converted into data of R (red), G (green), and B (blue). And mark the boundaries of the restricted areas in the initial training sample set according to the types of restricted areas, and generate a training sample set. Among them, the types of restricted areas are used to reflect various restricted areas. The types of restricted areas may include, for example, but are not limited to, one or more of residential areas, forest areas, protected areas, farmlands, and red line areas. Exemplarily, mark the boundary of the residential area as 1, mark the boundary of the forest area as 2, mark the boundary of the farmland as 3, etc. In this way, by marking the boundaries of the restricted areas in the initial training sample set according to the types of restricted areas and classifying and marking, the training of the region recognition model is more targeted, and it is convenient to obtain the recognition results of the pixel data of the restricted areas subsequently. Moreover, normalizing the images facilitates the unified use of subsequent training.

[0070] Among them, the above first step may further include performing random flipping, cropping, and deformation processing on the image to form an initial training sample set. The number of the initial training sample sets exceeds N times the image set before the image augmentation processing. Exemplarily, N may include, but is not limited to, 100.

[0071] In some other embodiments of the above step 120, perform image augmentation processing on the image to form an initial training sample set, normalize the image format in the initial training sample set, and uniformly mark the boundaries of the restricted areas in the initial training sample set to generate a training sample set. In this way, a training sample set in a unified format can be used for subsequent training, which is convenient for unified use.

[0072] Step 130: Input the training sample set into the region recognition model to train the region recognition model.

[0073] The above-mentioned region recognition model is a model that needs to be trained and can be used to identify restricted regions in satellite remote sensing image maps through training. The region recognition model includes one or more of the Alexnet model, the VGG (Visual Geometry Group, simply referred to as the computer vision group) model, and the Resnet (deep residual network) model. Optionally, the VGG model includes the VGG-16 model. The Resnet model can include one or more of the Resnet-18 model, the Resnet-34 model, and the Resnet-50 model. In this way, these five classic convolutional neural network models can be used as alternative models, and users can choose which region recognition model to use according to their needs. Convert the fully connected layers of the five CNN (Convolutional Neural Networks) models into fully convolutional layers, use transposed convolutional layers with parameters and add skip connections, so as to adjust the CNN model into the corresponding FCN (Fully Conv Neural Networks) model. The embodiments of the present application can use any region recognition model for training or use, and the embodiments of the present application can also use multiple region recognition models for training or use each time. When using multiple region recognition models for training or use, it is convenient for users to select and use according to the evaluation indexes of each region recognition model. The detailed description is as follows.

[0074] Continue Figure 1 As shown, after the above step 110, the remote sensing map region recognition method further includes:

[0075] Step 1: Generate a dataset based on the image and marking data. The dataset is the data formed by the sample set. The dataset can be referred to as the sample set. Step 2: Divide the dataset according to the data ratios of the training set, development set, and test set to obtain a training sample set, a development sample set, and a test sample set. Among them, the data ratios of the training set, development set, and test set can be set according to user requirements. Generally, the sample size of the training set is greater than that of the development set and the test set respectively. The sample size of the training set can be about 10 times that of the development set and about 100 times that of the test set. Exemplarily, the data ratios of the training set, development set, and test set can be but are not limited to 90:9:1. Step 3: Use the development sample set to adjust the hyperparameters of the region recognition model and update the region recognition model. After performing Step 130 using the region recognition model subsequently, the remote sensing map region recognition method further includes Step 4: Input the test sample set into the trained region recognition model to test the trained region recognition model and obtain a test result. Step 5: Determine the recognition accuracy of the trained region recognition model based on the test result. In this way, adjusting the hyperparameters of the region recognition model with the development sample set and the test sample set can improve the accuracy of the region recognition model.

[0076] Further, use the test sample set to test one or more of the above five trained classical convolutional neural network models to verify the accuracy of the region recognition model.

[0077] Among them, the above Step 5 can further include determining the recognition accuracy of the trained region recognition model based on the pixel data of the restricted region output by the target region recognition model and the incorrect pixel data in the pixel data. Compare the actual pixel data of the restricted region in the map image with the pixel data of the restricted region output by the target region recognition model to determine the incorrect pixel data and the correct pixel data. This process can be assisted manually or automatically by the model. Exemplarily, the pixel data can include pixel points. There are 20,000 pixel points in the restricted region to be recognized, and 15,000 correct pixel points are output for the tested restricted region, with an accuracy rate of 75%.

[0078] According to the Satisficing and Optimizing Metrics, use the following recognition accuracy formula to determine the recognition accuracy of the trained region recognition model, where the Satisficing and Optimizing Metrics can include one or more of accuracy and recognition time. Here, the Satisficing and Optimizing Metrics is the recognition accuracy. The recognition accuracy formula is:

[0079]

[0080] Among them, Accuracy is the recognition accuracy, L is the total number of types of restricted areas. For example, when distinguishing residential areas, forest areas, and farmlands, L = 3, l is the number of types of restricted areas, m dev is the number of samples in the development sample set, i is the serial number, x is the number of pixels in the x direction of the image, y is the number of pixels in the y direction of the image, nx as a whole represents the total number of pixels in the x direction of the image corresponding to the sample, ny as a whole represents the total number of pixels in the y direction of the image corresponding to the sample, n, I symbols represent the number of samples for which the expression in {} is true. In this formula, it is the wrong pixel data, label is the marked value, and the subscript pred represents the model prediction value.

[0081] The above-mentioned satisfaction and optimization indicators may include accuracy and the average recognition time of the images in the development sample set, in seconds. For details, please refer to Table 1 below:

[0082] Table 1 Satisfaction and Optimization Indicators

[0083] Region recognition model Accuracy Recognition time (seconds) Model1 Accuracy 1 Recognition time 1 Model2 Accuracy 2 Recognition time 2 …… …… ……

[0084] Step 140, when the recognition accuracy of the region recognition model reaches the predetermined recognition accuracy, obtain the trained target region recognition model and output the pixel data of the restricted region of the target region recognition model.

[0085] In the embodiment of the present application, the region recognition model is automatically trained and automatically recognized, with less human input and shorter time consumption. And it uses marked data to reduce the interference of other various geographic information in the satellite remote sensing image map on the recognition, and improve the recognition accuracy.

[0086] After obtaining the target region recognition model, package it into the remote sensing map region recognition system, and integrate the remote sensing map region recognition system with GIS (Geographic Information System or Geo-Information system, the system is integrated with the geographic information system). GIS provides satellite remote sensing image map images, and the region recognition model in the remote sensing map region recognition system. The region recognition model obtains the satellite remote sensing image map images and executes the above steps 110 to 140 to determine which pixel data are the restricted regions to be recognized, and then gives the pixel data of the restricted regions to WebGIS for display.

[0087] Web GIS refers to GIS working on the Web network, which is the extension and development of traditional GIS on the network. With WebGIS technology as the front-end support, it can provide satellite remote sensing image map images, and at the same time share the database and the geographic information input by users. And the front-end can include a user interface to facilitate the display of satellite remote sensing image map images, etc. The detailed description is as follows.

[0088] Figure 2 As shown Figure 1 Another schematic flow diagram of the remote sensing map area recognition method shown above.

[0089] Such as Figure 2 As shown, after the above step 140, the method further includes displaying the pixel data of the restricted area output by the target area recognition model to the front end to obtain the restricted area. Such front-end display facilitates user use.

[0090] Among them, the front end includes the satellite remote sensing image map image displayed in the Web GIS (Geographic Information System or Geo-Information system, the system is the same as the geographic information system) interface. In this way, the system can provide the function of displaying existing data, and can provide the restricted areas and the types of restricted areas that have been entered into the database.

[0091] Combined with Figure 2 As shown, after displaying the pixel data of the restricted area output by the target area recognition model to the front end and determining the restricted area, or after step 140, the method further includes step 150 of obtaining the area image selected by the user from the image. The area image refers to the area selected by the user from the satellite remote sensing image map image. This area image can be greater than or equal to the restricted area, that is, it contains the restricted area, which is convenient for finding the restricted area in this area later. Step 160, generating input data according to the area image. The input data refers to the data obtained by converting the area image into a data format that meets the input requirements of the target area recognition model. Step 170, inputting the input data into the target area recognition model for recognition to output the area pixel data. The area pixel data refers to the pixel data of the area image. Step 180, extracting the boundary information of the restricted area according to the area pixel data. Step 190, displaying the boundary information of the restricted area to the front end. In this way, by cooperating with the satellite remote sensing image map image in the Web GIS, for the area image required by the user, the input data is automatically generated, and the input data is automatically input into the above-mentioned target area recognition model, the boundary information of the restricted area is automatically recognized and displayed to the user, which is convenient for the user to view and use, and does not require the user to manually identify, thus improving the recognition efficiency and accuracy.

[0092] Among them, the target area recognition model is multiple target area recognition models. The above step 170 can further include inputting the input data into the multiple target area recognition models respectively, and respectively obtaining the area pixel data output by the multiple target area recognition models. Subsequently, better results can be selected according to multiple results. The detailed description is as follows.

[0093] Step 180 may further include extracting boundary information of restricted areas corresponding to multiple target area recognition models based on regional pixel data.

[0094] Step 190 may further include a first step of presenting the boundary information of restricted areas corresponding to multiple target area recognition models to the front end, where the boundary information may be presented as an identifier or a label. A second step is to determine evaluation metrics for the multiple target area recognition models, where the evaluation metrics may be referred to as satisfaction and optimization metrics. The evaluation metrics are used to reflect the dimensions of the quality of the target area recognition models. The evaluation metrics may include one or more of accuracy and recognition time. A third step is to select the target area recognition model corresponding to the best evaluation metric among the evaluation metrics. The regional image, the selected model, and the regional pixel data output by the selected model are used as samples and stored in the database. In this way, multiple target area recognition models are available for system users to view. The target area recognition model corresponding to the best evaluation metric selected from multiple trained area recognition models, that is, the selected model, and then record the model selected by the user in the current selection. The regional image selected in this operation and the output of the selected model are used as samples and sent back to the background and stored in the database for backup to improve the generality of the selected model. In this way, by coupling Web GIS, the database, and image recognition, the selected model can be continuously trained by collecting and processing user input data to obtain good recognition results.

[0095] In the data table in the database, the stored data entry structure includes the unique identifier id of the mark for each record, the image storage location gismap_path corresponding to the record, the longitude and latitude coordinates boundary_points of the image corner points, and the restricted area flag special_area. Among them, id is the mark for each data record and is randomly generated during entry. The longitude and latitude coordinates are used to correspond to the position of the image on the Web GIS; in the restricted area flag, 1 and 2 are the numbers of restricted area types. For example, 1 represents a residential area and 2 represents a forest area, which can be extended according to actual needs. For each number, it corresponds to the inflection point coordinates of the image that belong to this type of restricted area. If it does not include this type, the list is empty.

[0096] In the embodiments of the present application, an effective and updatable remote sensing map area recognition system can be built by collecting images and marked data, training area recognition models, integrating target area recognition models, and sending back sample data. The remote sensing map area recognition system can serve the recognition of restricted areas in a wind farm and related design and survey work for the recognition of restricted areas in satellite remote sensing image maps. Using this remote sensing map area recognition system can greatly reduce the work input of engineers in image recognition, improve recognition accuracy, and improve work efficiency.

[0097] Figure 3 As shown Figure 1 in the application schematic diagram where the restricted area in the remote sensing map area recognition method shown is farmland. Figure 4 As shown Figure 1 in the application schematic diagram where the restricted area in the remote sensing map area recognition method shown is a residential area. Figure 5 As shown Figure 1 in the recognition result schematic diagram where the restricted area in the remote sensing map area recognition method shown is farmland. Figure 6 As shown Figure 1 in the specific process schematic diagram of the remote sensing map area recognition method shown.

[0098] Such as Figure 3 and Figure 5 the flat area shown contains many small pieces of farmland. Figure 4 The flat area shown contains many small pieces of residential areas, etc. Due to the simple terrain, on-site exploration has not been carried out, but during the design process of the wind farm, the location of the restricted area needs to be determined to avoid the impact of wind turbines.

[0099] Such as Figure 6 shown, based on the remote sensing map area recognition system of the embodiment of the present application, the Web GIS interface is positioned to the satellite remote sensing image map corresponding to the area to be planned in the wind farm, the input data is input into the target area recognition model for image recognition, and after the recognition is completed, the pixel data of the restricted area marked by different models is selected and viewed, such as Figures 3 to 5 shown. The user selects the correct pixel data of the restricted area and the selected target area recognition model, and transmits it back to the database, which can be used as training data together with the recognition result of the pixel data of the restricted area.

[0100] Figure 7 shown is the module schematic diagram of the remote sensing map area recognition device provided by the embodiment of the present application.

[0101] Such as Figure 7 shown, the remote sensing map area recognition device may include the following modules:

[0102] The data acquisition module 31 is used to acquire the satellite remote sensing image map and the marking data, the image includes the restricted area of the wind farm layout, and the marking data is used to mark the area contour of the restricted area.

[0103] The sample set generation module 32 is used to generate a training sample set according to the image and the marking data.

[0104] The model training module 33 is used to input the training sample set into the area recognition model to train the area recognition model.

[0105] The model optimization module 34 is configured to obtain a trained target area recognition model and output pixel data of a restricted area of the target area recognition model when the recognition accuracy of the area recognition model reaches a predetermined recognition accuracy.

[0106] For the implementation processes of the functions and roles of each module in the above device, please refer to the implementation processes of the corresponding steps in the above method for details, which will not be elaborated here.

[0107] Figure 8 The block diagram of the remote sensing map area recognition method system 40 provided by an embodiment of the present application is shown. The remote sensing map area recognition method system 40 includes one or more processors 41, which are configured to implement the remote sensing map area recognition method as described above.

[0108] In some embodiments, the remote sensing map area recognition method system 40 may include a computer-readable storage medium 49. The computer-readable storage medium 49 may store a program that can be called by the processor 41, and may include a non-volatile storage medium. In some embodiments, the remote sensing map area recognition method system 40 may include a memory 48 and an interface 47. In some embodiments, the remote sensing map area recognition method system 40 may further include other hardware according to actual applications.

[0109] The computer-readable storage medium 49 of the embodiment of the present application stores a program, which, when executed by the processor 41, is configured to implement the remote sensing map area recognition method as described above.

[0110] The present application may be in the form of a computer program product implemented on one or more computer-readable storage media 49 (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing program codes. The computer-readable storage medium 49 includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information may be computer-readable instructions, data structures, program modules, or other data. Examples of the computer-readable storage medium 49 include but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device.

[0111] The above remote sensing map area recognition method system may include an electronic device 70. Specifically, the electronic device may be: a server, etc. There is no limitation here. Any electronic device that can implement the embodiments of the present invention belongs to the protection scope of the present invention.

[0112] The foregoing is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this specification shall be included within the protection scope of this specification.

[0113] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, commodity or device. Without further limitation, the element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the said element.

Claims

1. A method for identifying remote sensing map regions, characterized in that, Including: Obtain the input information of the user. The input information includes the satellite remote sensing image map image input by the user and the marking data. The image includes the restricted area of the wind farm layout. The restricted area refers to the area that needs to be identified and where a wind farm cannot be established. The marking data is used to mark the area contour of the restricted area. If the data increment of the input information compared to the stored information is greater than the preset increment, generate a training sample set based on the input information. The stored information includes the satellite remote sensing image map image and the marking data stored in the database. Input the training sample set into the region recognition model to train the region recognition model. When the recognition accuracy of the region recognition model reaches the predetermined recognition accuracy, obtain the trained target region recognition model and output the pixel data of the restricted area of the target region recognition model.

2. The remote sensing map area recognition method according to claim 1, characterized in that The generating a training sample set based on the input information includes: Perform image augmentation processing on the image to form an initial training sample set. Normalize the image format in the initial training sample set and mark the boundaries of the restricted areas in the initial training sample set according to the types of restricted areas to generate the training sample set.

3. The remote sensing map area recognition method according to claim 1, characterized in that, After obtaining the satellite remote sensing image map image and the marking data input by the user, the method further includes: Generate a data set based on the image and the marking data. Divide the data set according to the data ratios of the training set, the development set, and the test set to obtain a training sample set, a development sample set, and a test sample set. Use the development sample set to adjust the hyperparameters of the region recognition model and update the region recognition model. After inputting the training sample set into the region recognition model to train the region recognition model, the method further includes: Input the test sample set into the trained region recognition model to test the trained region recognition model and obtain a test result. Determine the recognition accuracy of the trained region recognition model based on the test result.

4. The remote sensing map area recognition method according to claim 3, characterized in that, The determining the recognition accuracy of the trained region recognition model based on the test result includes: Determine the recognition accuracy of the trained region recognition model based on the pixel data of the restricted area output by the target region recognition model and the error pixel data in the pixel data.

5. The remote sensing map area recognition method according to claim 1, wherein After reaching the predetermined recognition accuracy of the region recognition model, obtaining the trained target region recognition model, and obtaining the pixel data of the restricted area output by the target region recognition model, the method further includes: Display the pixel data of the restricted area output by the target region recognition model to the front end to obtain the restricted area.

6. The remote sensing map area recognition method according to claim 5, characterized in that, The front end includes the satellite remote sensing image map image displayed in the WebGIS interface. After displaying the pixel data of the restricted area output by the target region recognition model to the front end to obtain the restricted area, the method further includes: Obtain the region image selected by the user from the image. Generate input data based on the region image. Input the input data into the target area recognition model to output area pixel data; Extract the boundary information of the restricted area based on the area pixel data; Display the boundary information of the restricted area to the front end.

7. The remote sensing map area recognition method according to claim 6, wherein, The target area recognition model is multiple target area recognition models. The displaying the boundary information of the restricted area to the front end includes: Display the boundary information of the restricted area corresponding to the multiple target area recognition models to the front end; Determine the evaluation indexes of the multiple target area recognition models; Select the target area recognition model corresponding to the best evaluation index in the evaluation indexes according to the evaluation indexes; Use the area image, the selected model and the area pixel data output by the selected model as samples and store them in the database.

8. The remote sensing map area recognition method according to claim 7, wherein, After obtaining the satellite remote sensing image map image and the marking data input by the user, the method further includes: Obtain the storage information in the database.

9. A remote sensing map area recognition device, characterized in that, Including: A data acquisition module, configured to acquire the input information of the user. The input information includes the satellite remote sensing image map image and the marking data input by the user. The image includes the restricted area of the wind farm layout. The restricted area refers to the area that needs to be recognized and where a wind farm cannot be established. The marking data is used to mark the area contour of the restricted area; A sample set generation module, configured to generate a training sample set according to the input information if the data increment of the input information compared with the storage information is greater than a preset increment. The storage information includes the satellite remote sensing image map image and the marking data stored in the database; A model training module, configured to input the training sample set into the area recognition model to train the area recognition model; A model optimization module, configured to obtain the trained target area recognition model and output the pixel data of the restricted area of the target area recognition model when the recognition accuracy of the area recognition model reaches a predetermined recognition accuracy.

10. A remote sensing map area recognition system, characterized in that, Includes one or more processors for implementing the remote sensing map area recognition method according to any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, Stored thereon is a program which, when executed by the processor, implements the method according to any one of claims 1-8.

Images