Labeling method and device of impervious surface area, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a marking method and device for an impervious surface area, electronic equipment and a storage medium, and belongs to the technical field of image processing. The method comprises the following steps: inputting a high-resolution remote sensing image corresponding to a surface area to be processed into a pre-trained semantic segmentation model to obtain a surface coverage category corresponding to each pixel in the high-resolution remote sensing image; determining an impervious degree score corresponding to each pixel according to the land cover category; obtaining the pixel weight of each pixel, updating the impervious degree score of the corresponding pixel according to the pixel weight, and obtaining an updated impervious degree score corresponding to each pixel; determining a target area impervious degree score of the earth surface area to be processed according to the updated impervious degree score corresponding to each pixel; and when the impervious degree score of the target area exceeds a preset score threshold, determining that the earth surface area to be processed is an impervious surface area. According to the method and the device, the accuracy and the labeling efficiency of the impervious surface sample region labeling can be improved.

Description

Technical Field

[0001] This application relates to the field of image processing, specifically to a method, apparatus, electronic device, and storage medium for labeling impermeable surface areas. Background Technology

[0002] Impermeable surfaces refer to artificial or natural materials and structures that prevent water from seeping into the soil. Their presence disrupts the natural hydrological cycle, leading to increased surface runoff. Common types of impermeable surfaces include asphalt pavements, concrete pavements, buildings, sidewalks, compacted gravel, or soil. To assess the impact of surface areas on the hydrological cycle, predict surface runoff, and support urban planning and flood risk management, it is necessary to label the surface areas to be treated to determine whether they are impermeable surface sample areas. Impermeable surface sample areas include a certain area of ​​impermeable surface.

[0003] In related technologies, impermeable surface sample areas are determined by manual annotation. However, manual annotation is time-consuming, labor-intensive, and highly subjective, making it difficult to ensure consistency in annotation across different surface areas. This results in inaccurate annotation of impermeable surface sample areas and low annotation efficiency. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for labeling impermeable surface areas, aiming to improve the accuracy and efficiency of labeling impermeable surface sample areas.

[0005] To achieve the above objectives, one embodiment of this application provides a method for marking impermeable surface areas, including: Acquire high-resolution remote sensing images of the surface area to be processed; High-resolution remote sensing images are input into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image. The pre-trained semantic segmentation model is trained based on the sample high-resolution remote sensing images corresponding to the sample land surface areas and the labeled land cover categories corresponding to the sample high-resolution remote sensing images. The impermeability score for each pixel is determined based on the land cover category; Obtain the pixel weight of each pixel, update the waterproof rating of the corresponding pixel based on the pixel weight, and obtain the updated waterproof rating for each pixel; Based on the updated impermeability score corresponding to each pixel, the impermeability score of the target area of ​​the surface area to be processed is determined. When the impermeability score of the target area exceeds the preset score threshold, the surface area to be treated is determined to be an impermeable surface area.

[0006] In some embodiments, after obtaining the land cover category corresponding to each pixel in the high-resolution remote sensing image, the method further includes: Obtain the total area of ​​a high-resolution remote sensing image; The first pixel is defined as the pixel whose land cover category is water, and the total area of ​​the region corresponding to the first pixel is calculated to obtain the first area. Pixels with a land cover category of "no information" are identified as the second pixel, and the total area of ​​the region corresponding to the second pixel is calculated to obtain the second area. Divide the first area and / or the second area by the total area of ​​the image to obtain the water area ratio; When the proportion of water area exceeds a preset proportion threshold, the surface area to be processed is determined to be an undefined area.

[0007] In some embodiments, determining the impermeability score for each pixel based on the land cover category includes: Obtain multiple standard land surface categories, and the initial impermeability scores corresponding to each standard land surface category; For each pixel, a target standard land cover category matching the corresponding land cover category is determined from multiple standard land cover categories, and the impermeability score corresponding to the pixel is determined based on the initial impermeability score corresponding to the target standard land cover category.

[0008] In some embodiments, determining the impermeability score of a pixel based on the initial impermeability score corresponding to the target standard surface category includes: Obtain vegetation factors for the land surface area; When the land cover category corresponding to a pixel is a vegetation category, the initial impermeability score is updated based on the vegetation factor to obtain the updated initial impermeability score, and the updated initial impermeability score is determined as the impermeability score corresponding to the pixel.

[0009] In some embodiments, when the land cover category corresponding to a pixel is a vegetation category, the initial impermeability score is updated based on the vegetation factor to obtain the updated initial impermeability score, including: When the land cover category corresponding to the pixel is vegetation and the shooting time is spring or summer, the initial impermeability score is multiplied by the vegetation factor to obtain the updated initial impermeability score. When the land cover category corresponding to a pixel is vegetation and the shooting time is autumn or winter, the initial impermeability score is divided by the vegetation factor to obtain the updated initial impermeability score.

[0010] In some embodiments, obtaining the pixel weight of each pixel includes: Based on the land cover category corresponding to each pixel, multiple pixels are divided into regions to obtain at least one sub-land cover region; Obtain the normalization constant; For each pixel, determine the first coordinate of the center pixel of the sub-surface region where the pixel is located, and determine the weight exponential function based on the normalization constant and the first coordinate; Determine the second coordinates of each pixel in each sub-surface region, and determine the pixel weight of each pixel based on the second coordinates and the weight exponential function. The pixel weight of each pixel is negatively correlated with the distance to the center pixel of the corresponding sub-surface region.

[0011] In some embodiments, the impermeability score of the target area of ​​the surface region to be processed is determined based on the updated impermeability score corresponding to each pixel, including: For each sub-surface region, the updated impermeability score corresponding to each pixel in the current sub-surface region is calculated to obtain the sub-surface region score corresponding to the current sub-surface region; The sum of the pixel weights of each pixel in the current sub-surface region is calculated to obtain the sub-surface region weight corresponding to the current sub-surface region; Based on the sub-surface area score and sub-surface area weight, the impermeability score of the corresponding sub-surface area is determined. Based on the impermeability scores of all sub-surface areas, the impermeability score of the target area of ​​the surface area to be treated is determined.

[0012] In some embodiments, before inputting the high-resolution remote sensing image into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image, the method further includes: Obtain the high-resolution remote sensing image of the sample land surface area, and the labeled land cover category of the sample high-resolution remote sensing image; The high-resolution remote sensing image of the sample is input into the semantic segmentation model, and the predicted land cover category corresponding to each pixel in the high-resolution remote sensing image of the sample is output. Determine the difference between the predicted land cover category and the corresponding labeled land cover category for each pixel; The semantic segmentation model is iteratively trained based on the differences to obtain a pre-trained semantic segmentation model.

[0013] To achieve the above objectives, one embodiment of this application provides a marking device for impermeable surface areas, comprising: The acquisition module is used to acquire high-resolution remote sensing images of the surface area to be processed; The semantic segmentation module is used to input high-resolution remote sensing images into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image. The pre-trained semantic segmentation model is trained based on the sample high-resolution remote sensing images corresponding to the sample land surface areas and the labeled land cover categories corresponding to the sample high-resolution remote sensing images. The first scoring module is used to determine the impermeability score of each pixel based on the land cover category. The second scoring module is used to obtain the pixel weight of each pixel, update the waterproofness score of the corresponding pixel according to the pixel weight, and obtain the updated waterproofness score of each pixel. The target scoring module is used to determine the target area impermeability score of the surface area to be processed based on the updated impermeability score corresponding to each pixel. The impermeable surface area determination module is used to determine the surface area to be processed as an impermeable surface area when the impermeability score of the target area exceeds the preset score threshold.

[0014] To achieve the above objectives, one aspect of this application provides a computer-readable storage medium storing multiple instructions adapted for loading by a processor to execute the steps in the marking method for impermeable surface areas provided in this application.

[0015] To achieve the above objectives, one aspect of this application provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it implements the steps in the marking method for impermeable surface areas provided in this application.

[0016] To achieve the above objectives, one aspect of this application provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in the marking method for impermeable surface areas provided in this application.

[0017] This application provides a method, apparatus, electronic device, and storage medium for labeling impermeable surface areas. The method involves acquiring a high-resolution remote sensing image corresponding to the surface area to be processed; inputting the high-resolution remote sensing image into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image; training the pre-trained semantic segmentation model based on sample high-resolution remote sensing images corresponding to sample surface areas and the labeled land cover categories corresponding to the sample high-resolution remote sensing images; determining the impermeability score corresponding to each pixel based on the land cover category; obtaining the pixel weight of each pixel and updating the impermeability score of the corresponding pixel based on the pixel weight to obtain an updated impermeability score for each pixel; determining the target area impermeability score of the surface area to be processed based on the updated impermeability score for each pixel; and determining the surface area to be processed as an impermeable surface area when the target area impermeability score exceeds a preset score threshold.

[0018] This application embodiment automatically identifies the land cover category corresponding to each pixel in the land surface area by using a pre-trained semantic segmentation model. Based on the impermeability score corresponding to the pixel determined by the land cover category and the obtained pixel weight, the impermeability of the pixel area indicated by the pixel is quantitatively analyzed. Then, based on the updated impermeability scores of each pixel, it is determined whether the land surface area to be processed is an impermeable surface area. Since this application uses a unified, pre-trained semantic segmentation model for semantic segmentation and then uses a unified scoring rule for impermeability scoring, this application can accurately identify and divide the spatial distribution of various land cover types (such as buildings, roads, green spaces, water bodies, etc.) through pixel-level semantic segmentation, and accurately determine the corresponding impermeability score based on the pixel weight corresponding to each pixel in the cover. Therefore, this application embodiment can achieve accurate and efficient labeling of the land surface area to be processed.

[0019] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1This is a schematic diagram of the system framework corresponding to the marking method for impermeable surface areas provided in the embodiments of this application; Figure 2 This is a flowchart illustrating the method for marking impermeable areas provided in an embodiment of this application; Figure 3 This is a schematic diagram of a high-resolution remote sensing image provided in an embodiment of this application; Figure 4 This is a schematic diagram of the semantic segmentation results provided in an embodiment of this application; Figure 5 This is a schematic diagram of the land surface category scoring criteria provided in the embodiments of this application; Figure 6 This is a schematic diagram illustrating the scoring of the impermeability of the target area of ​​the surface region to be treated, provided in an embodiment of this application. Figure 7 This is a schematic diagram illustrating the effect of selecting the preset scoring threshold provided in the embodiments of this application; Figure 8 This is a schematic diagram of the module structure of the marking device for impermeable surface areas provided in an embodiment of this application; Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0022] To enable those skilled in the art to better understand the solutions of this application, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] It should be noted that in each specific embodiment of this application, when data processing requires obtaining relevant data about the surface area to be processed, permission or consent from the relevant personnel managing the surface area is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when this application embodiment requires obtaining sensitive personal information of relevant personnel, separate permission or consent from the relevant personnel is obtained through pop-ups or redirection to a confirmation page. Only after obtaining the separate permission or consent of the relevant personnel is the necessary surface area data to be processed for the normal operation of this application embodiment obtained.

[0024] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, programmable consumer computer devices, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0025] Before providing a detailed description of the embodiments of this application, the nouns and terms used in the embodiments of this application will be explained first. The nouns and terms used in the embodiments of this application shall be interpreted as follows: Impermeable surfaces are closed surfaces that prevent rainwater from naturally infiltrating. They are typically constructed from artificially paved materials (such as concrete, asphalt, brick, and roofs) found in cities, including buildings, roads, plazas, and parking lots. Because these surfaces are almost completely impermeable, most rainwater cannot infiltrate and instead rapidly forms surface runoff, easily causing urban flooding, exacerbating water pollution, and disrupting the natural water cycle. Therefore, impermeable surfaces are an important indicator for measuring the degree of urbanization and its impact on the ecological environment.

[0026] The above is an explanation of the terminology and concepts related to the labeling of impermeable areas in this application. Other concepts will be described later.

[0027] Next, the technical problems existing in the related technologies are described: To assess the impact of surface areas on the hydrological cycle, predict surface runoff, and support urban planning and flood risk management, it is necessary to label the surface areas to be treated to determine whether they are impermeable sample areas. Impermeable sample areas include a certain area of ​​impermeable surface.

[0028] In related technologies, impermeable surface sample areas are determined by manual annotation. However, manual annotation is time-consuming, labor-intensive, and highly subjective, making it difficult to ensure consistency in annotation across different surface areas. This results in inaccurate annotation of impermeable surface sample areas and low annotation efficiency.

[0029] For example, in remote sensing imagery of a certain urban area, there are large areas of asphalt roads, rooftops, and concrete plazas, all of which are typical impermeable surfaces, while lawns, green spaces, and bare soil are permeable surfaces. Related techniques rely on visual interpretation by professionals to manually delineate which areas belong to impermeable surface samples on the remote sensing imagery; however, manual annotation is time-consuming and labor-intensive, resulting in low efficiency for impermeable surface annotation in these techniques. Furthermore, due to differences in the understanding of classification rules among different annotators, the accuracy of impermeable surface annotation in these techniques is poor.

[0030] To address the aforementioned problems, this application provides a method, apparatus, electronic device, and storage medium for labeling impermeable surface areas. The method involves: acquiring a high-resolution remote sensing image corresponding to the surface area to be processed; inputting the high-resolution remote sensing image into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image; training the pre-trained semantic segmentation model based on sample high-resolution remote sensing images corresponding to sample surface areas and the labeled land cover categories corresponding to the sample high-resolution remote sensing images; determining the impermeability score corresponding to each pixel based on the land cover category; obtaining the pixel weight of each pixel and updating the impermeability score of the corresponding pixel based on the pixel weight to obtain an updated impermeability score for each pixel; determining the target area impermeability score of the surface area to be processed based on the updated impermeability score for each pixel; and determining the surface area to be processed as an impermeable surface area when the target area impermeability score exceeds a preset score threshold.

[0031] This application embodiment automatically identifies the land cover category corresponding to each pixel in the land surface area by using a pre-trained semantic segmentation model. Based on the impermeability score corresponding to the pixel determined by the land cover category and the obtained pixel weight, the impermeability of the pixel area indicated by the pixel is quantitatively analyzed. Then, based on the updated impermeability scores of each pixel, it is determined whether the land surface area to be processed is an impermeable surface area. Since this application uses a unified, pre-trained semantic segmentation model for semantic segmentation and then uses a unified scoring rule for impermeability scoring, this application can accurately identify and divide the spatial distribution of various land cover types (such as buildings, roads, green spaces, water bodies, etc.) through pixel-level semantic segmentation, and accurately determine the corresponding impermeability score based on the pixel weight corresponding to each pixel in the cover. Therefore, this application embodiment can achieve accurate and efficient labeling of the land surface area to be processed.

[0032] The specific details regarding the marking method, apparatus, electronic device, and storage medium for impermeable surface areas provided in the embodiments of this application will be described in detail below.

[0033] Please see Figure 1, Figure 1 This is a schematic diagram of the system framework corresponding to the method for marking impermeable surface areas provided in this application embodiment. The method for marking impermeable surface areas provided in this application embodiment can be applied to this system framework.

[0034] It includes terminal 140, Internet 130, gateway 120, server 110, etc.

[0035] Terminal 140 or server 110 can be a device that performs a method for marking impermeable surface areas.

[0036] Terminal 140 includes, but is not limited to, mobile phones, tablets, computers, and intelligent computing centers. Terminal 140 can be a single device or a collection of multiple devices. For example, multiple computers can be interconnected via a local area network, sharing a single monitor to work collaboratively, thus forming a terminal 140. Terminal 140 can communicate with the Internet 130 via wired or wireless means to exchange data.

[0037] Server 110 refers to a computer system that can provide certain services to terminal 140. Compared to ordinary terminal 140, server 110 has higher requirements in terms of stability, security, and performance. Server 110 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0038] Gateway 120, also known as an internetwork connector or protocol converter, is a computer system or device that acts as a translator, enabling network interconnection at the transport layer. It bridges the gap between two systems using different communication protocols, data formats, languages, or even completely different architectures. Gateways can also provide filtering and security functions. Messages sent from terminal 140 to server 110 are forwarded to the corresponding server 110 via gateway 120. Messages sent from server 110 to terminal 140 are also forwarded to the corresponding terminal 140 via gateway 120.

[0039] The embodiments of this application can be applied to various scenarios, such as urban flooding risk assessment, dynamic monitoring of land use changes, and urban planning. This application does not limit the scenarios in which the method for marking impermeable areas in this application can be used.

[0040] Next, we will describe the marking devices in the impermeable surface area from the perspective of... Figure 2 As shown, Figure 2 This is a flowchart illustrating the method for marking impermeable surfaces provided in this application embodiment. The method for marking impermeable surfaces is applied to a marking device for impermeable surfaces. Figure 2 The method may include, but is not limited to, the following steps 101 to 106. When the marking device for impermeable surfaces executes the marking method for impermeable surfaces, the specific process is as follows. It should be noted that this embodiment... Figure 3 The order of steps 101 to 106 is not specifically limited. The order of steps can be adjusted or some steps can be reduced or added according to actual needs.

[0041] Step 101: Obtain the high-resolution remote sensing image corresponding to the surface area to be processed; Step 102: Input the high-resolution remote sensing image into the pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image. The pre-trained semantic segmentation model is trained based on the sample high-resolution remote sensing image corresponding to the sample land surface area and the labeled land cover category corresponding to the sample high-resolution remote sensing image.

[0042] Step 103: Determine the impermeability score for each pixel based on the land cover category.

[0043] Step 104: Obtain the pixel weight of each pixel, update the impermeability score of the corresponding pixel according to the pixel weight, and obtain the updated impermeability score for each pixel.

[0044] Step 105: Determine the target area impermeability score of the surface area to be processed based on the updated impermeability score corresponding to each pixel.

[0045] Step 106: When the impermeability score of the target area exceeds the preset score threshold, the surface area to be treated is determined to be an impermeable surface area.

[0046] Steps 101 to 106 will be described in detail below.

[0047] In step 101, a high-resolution remote sensing image of the surface area to be processed is acquired.

[0048] The surface area to be processed refers to a geographical area on the Earth's surface that requires specific analysis, monitoring, or processing. This area is typically defined by geographic coordinates (such as latitude and longitude boundaries or polygonal areas), and it is a high-resolution imagery area requiring fine-tuning. For example, the surface area to be processed could be a 0.5-meter (m) × 0.5-meter (m) area. That is, by acquiring a high-resolution remote sensing image of the surface area to be processed, various different land features such as trees, watercourses, buildings, roads, and field ridges can be identified within the surface area.

[0049] In some embodiments, the surface area to be processed can also be determined by the following methods: Step 1: Define the number of sampling points and randomly sample the global land surface according to the number of sampling points. The random sampling method can be random sampling or stratified random sampling. Record the latitude and longitude coordinates of the sampling points and retain them to 6 decimal places. Control the spatial error of the sampling points to within 0.1 meters.

[0050] Step 2: Using the sample point as the center, determine a rectangular vector range area with the longest and widest possible equidistant dimensions. The length and width of the rectangle should be around 512 meters. The vector file should be in Keyhole Markup Language (KML) format for easy import and download later.

[0051] Step 3: Window the rectangular vector range area according to the target sample resolution to obtain the surface area to be processed.

[0052] For example, random sampling determines sampling points, and a rectangular vector range area with the sampling points as the center is determined with the length and width as equal as possible. The size of this area is 30m×30m, which means the area covered is as high as 900 square meters. In this case, a single pixel often mixes multiple land features (such as buildings, roads, and green spaces), so it is impossible to identify specific structures and boundaries, details are severely lacking, it is impossible to obtain the corresponding high-resolution image, and it cannot be used for fine-grained surface analysis. Therefore, 0.5m×0.5m can be selected as the target resolution, and the rectangular vector range area is windowed according to the target sample resolution to obtain the surface area to be processed, and subsequent data processing is performed on the 0.5m×0.5m surface area to be processed.

[0053] High-resolution remote sensing images, in particular, are surface image data of the area to be processed, acquired through long-range sensors such as satellites, aircraft, or drones without direct contact with the target ground features. High-resolution remote sensing images record the reflectance or radiation information of the area to be processed in different spectral bands (such as visible light, infrared, and microwave). For example... Figure 3 As shown, Figure 3 This is a schematic diagram of a high-resolution remote sensing image provided in an embodiment of this application. Figure 3 It includes high-resolution remote sensing images corresponding to four different surface areas to be processed, namely high-resolution remote sensing image A, high-resolution remote sensing image B, high-resolution remote sensing image C, and high-resolution remote sensing image D.

[0054] In step 102, the high-resolution remote sensing image is input into the pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image. The pre-trained semantic segmentation model is trained based on the sample high-resolution remote sensing image corresponding to the sample land surface area and the labeled land cover category corresponding to the sample high-resolution remote sensing image.

[0055] High-resolution remote sensing images are two-dimensional digital matrices composed of multiple pixels. Each pixel represents the average reflectance or radiance energy value of a unit area in the surface region to be processed. These pixels are organized in rows and columns and arranged in a regular pattern to form a complete image. The size of the pixels (i.e., spatial resolution) determines the degree to which the image can distinguish surface details. The entire high-resolution remote sensing image presents the structure and characteristics of the surface environment through the spatial distribution and numerical differences of these pixels.

[0056] Among them, the pre-trained semantic segmentation model is a deep learning model pre-trained with a large amount of labeled data, specifically designed for pixel-level image classification tasks. The pre-trained semantic segmentation model has the ability to identify the category of each pixel in a high-resolution remote sensing image, enabling it to distinguish vegetation, water bodies, buildings, etc., within the image.

[0057] The land cover category describes the actual land cover status of each pixel. Common land cover categories include forest, grassland, farmland, water bodies, bare land, roads, and buildings. In this embodiment, each pixel is assigned a land cover category label to represent the type of land cover in that pixel area. Figure 4 As shown, Figure 4 This is a schematic diagram of semantic segmentation results provided in an embodiment of this application. Semantic segmentation result A corresponds to the semantic segmentation result of high-resolution remote sensing image A, semantic segmentation result B corresponds to the semantic segmentation result of high-resolution remote sensing image B, semantic segmentation result C corresponds to the semantic segmentation result of high-resolution remote sensing image C, and semantic segmentation result D corresponds to the semantic segmentation result of high-resolution remote sensing image D. It should be noted that the pre-trained semantic segmentation model performs pixel-level semantic segmentation. The set of land cover categories corresponding to each pixel is the semantic segmentation result of the high-resolution remote sensing image corresponding to the high-resolution remote sensing image. The semantic segmentation result of the high-resolution remote sensing image includes the semantic segmentation result corresponding to at least one sub-land surface region.

[0058] In some embodiments, if the high-resolution remote sensing image features corresponding to the surface area to be processed are obvious and easily identifiable (such as large areas of water, bare land, urban built-up areas, etc.), a pre-trained semantic segmentation model can usually be used directly for rapid classification to effectively obtain its semantic labels. If the high-resolution remote sensing image features corresponding to the surface area to be processed are incomplete, such as being only a part of an object, the pre-trained semantic segmentation model may not be able to achieve semantic segmentation correctly. In this case, semantic segmentation can be performed on the high-resolution remote sensing image of the rectangular vector range of the area to be processed, so as to determine the land cover category corresponding to each pixel in the high-resolution remote sensing image based on rich and complete high-resolution remote sensing image features.

[0059] Furthermore, the pre-trained semantic segmentation model is obtained through the following steps: (1.1) Obtain the high-resolution remote sensing image of the sample land surface area and the label land cover category of the sample high-resolution remote sensing image; (1.2) Input the high-resolution remote sensing image of the sample into the semantic segmentation model and output the predicted land cover category corresponding to each pixel in the high-resolution remote sensing image of the sample; (1.3) Determine the difference between the predicted land cover category and the corresponding labeled land cover category for each pixel; (1.4) The semantic segmentation model is iteratively trained based on the differences to obtain a pre-trained semantic segmentation model.

[0060] The sample surface region refers to a representative geographical area selected from the Earth's surface, which serves as the dataset needed to train the semantic segmentation model. The definition of the sample surface region is similar to that of the surface region to be processed, and will not be repeated here. Unlike the surface region to be processed, the sample surface region typically covers a variety of terrains, climates, and land use types to ensure that the semantic segmentation model can learn the characteristics of land features in different environments.

[0061] High-resolution remote sensing images (HRRAS) refer to remote sensing imagery data acquired for sample land surface areas, used for training and validating semantic segmentation models. The definition of HRAS is similar to that of high-resolution remote sensing images and will not be repeated here. Land cover categories are the true land cover type of each pixel in the obtained HRAS image, determined through manual annotation or high-precision annotation. For example, land cover categories include, but are not limited to, water bodies, vegetation, buildings, and roads. Land cover categories constitute ground truth labels, used to supervise the training of the semantic segmentation model. Specifically, the difference between the predicted land cover category and the corresponding labeled land cover category for each pixel guides the parameter optimization of the semantic segmentation model, improving classification accuracy.

[0062] The difference between the predicted land cover category and the corresponding labeled land cover category for each pixel can be determined by a predefined loss function, such as cross-entropy loss or Dice loss. The loss function calculates the degree of mismatch between the semantic segmentation model's prediction and the true label. For example, if a pixel should be classified as "water" land cover, but the semantic segmentation model predicts it as "bare land," this error is recorded and converted into a numerical difference. This difference reflects the current classification accuracy of the semantic segmentation model; the larger the error, the more the model's prediction deviates from reality, thus providing a basis for parameter adjustments during subsequent iterative training of the semantic segmentation model.

[0063] Furthermore, based on the difference between the predicted result calculated in step (1.3) and the ground truth label, the backpropagation algorithm and optimizer (such as SGD or Adam) can be used to adjust the parameters inside the semantic segmentation model to reduce the error in the next prediction. The adjusted parameters include, but are not limited to, convolutional kernel weights, bias terms, batch normalization layer parameters, fully connected layer weights, and attention mechanism parameters. This process will be repeated multiple times on a large number of high-resolution remote sensing images (i.e., "iterative training"). Each iteration of training allows the semantic segmentation model to gradually learn a more accurate ability to represent land cover features. As training progresses, the semantic segmentation model's recognition accuracy for various land cover types continuously improves. When the loss value (difference) converges to a low level or reaches the preset number of training rounds, training stops, ultimately resulting in a stable, highly generalizable pre-trained semantic segmentation model. This pre-trained semantic segmentation model can be used for automatic and accurate land cover classification of subsequent high-resolution remote sensing images.

[0064] In some embodiments, after obtaining the land cover category corresponding to each pixel in the high-resolution remote sensing image, the method further includes: (2.1) Obtain the total area of ​​the high-resolution remote sensing image; (2.2) Determine the pixel whose land cover category is water body as the first pixel, and calculate the total area of ​​the region corresponding to the first pixel to obtain the first area; (2.3) Determine the pixels whose land cover category is no information category as the second pixel, and calculate the total area of ​​the region corresponding to the second pixel to obtain the second area; (2.4) Divide the first area and / or the second area by the total area of ​​the image to obtain the water area ratio; (2.5) When the proportion of water area exceeds the preset proportion threshold, the surface area to be treated is determined to be a permeable surface area.

[0065] The total image area can be the entire geographical area covered by the high-resolution remote sensing image in the real world, or it can be the proportional area of ​​the high-resolution remote sensing image. The area definitions of the first area and the second area are adaptively adjusted according to the definition of the total image area. After semantic segmentation, the land cover category corresponding to each pixel in the high-resolution remote sensing image can be obtained. For pixels whose land cover category is water, these pixels are uniformly designated as the first pixel, and the area of ​​the region of these first pixels is calculated to obtain the first area. The first area includes all pixels identified as water bodies (such as rivers, lakes, oceans, etc.) by the semantic segmentation model.

[0066] Furthermore, when high-resolution remote sensing images contain data quality defects, such as cloud cover, cloud shadows, haze, or sensor malfunctions causing obscuring or distorted surface information in the area to be processed, the semantic segmentation model may be unable to extract effective features and thus struggle to determine the land cover type. In this case, the land cover type of the corresponding pixel will be labeled as "no information category". For pixels with a land cover category of "no information category", these pixels are uniformly designated as second pixels, and the area of ​​these second pixels is calculated to obtain the second area, which includes all pixels identified as "no information category" by the semantic segmentation model.

[0067] In this embodiment, both the first pixel and the second pixel are identified as low-quality pixels. The water area ratio is determined based on the low-quality pixels. Once the water area ratio exceeds a preset ratio threshold, it is considered that the permeable surface ratio of the corresponding sample area to be processed is large. Thus, such surface areas can be identified as permeable surface areas, thereby quickly filtering out surface areas that do not meet the requirements of impermeable surfaces.

[0068] Furthermore, the water area ratio can be determined using the following formula: .in, Indicates the proportion of water area. Represents the total area of ​​the image. Indicates the first area. This represents the second area. In some embodiments, when... The surface area to be processed corresponding to the high-resolution remote sensing image can be considered as an undefined area.

[0069] Undefined areas refer to regions in high-resolution remote sensing image processing where water covers more than 95% of the image, making it difficult to effectively extract or identify surface information. These areas lack sufficient land cover features, making it impossible to perform routine surface classification, change detection, and other analytical tasks. In practical applications, undefined areas usually require special processing or are excluded from subsequent analysis.

[0070] Furthermore, if the surface area to be processed is an undefined area, it is directly filtered out from multiple surface areas to be processed, meaning that the surface area to be processed will no longer participate in the data processing steps 103 to 106. In this way, when there are a large number of surface areas to be processed, computational resources can be avoided on data with a high proportion of permeability, thereby greatly improving the overall annotation efficiency.

[0071] In step 103, the impermeability score corresponding to each pixel is determined based on the land cover category.

[0072] In other words, based on the land cover type classified for each pixel, a quantitative score representing its ability to impede rainwater infiltration is assigned. For example, artificial hard surfaces such as buildings and roads have high impermeability, so the impermeability score for built areas is 10 (the maximum impermeability score is set to 10); while water can directly enter water bodies, so the impermeability score for water bodies is 0 (the maximum impermeability score is set to 0). The impermeability scoring process transforms the semantic segmentation results into a continuous numerical layer required for urban hydrological analysis, so that the impermeability score of each pixel can be more accurately evaluated, thereby more accurately assessing the impermeability of the surface area to be processed.

[0073] The impermeability score for each pixel can be determined by referring to a predefined impermeable layer classification standard, or by pre-training a scoring model and having the model determine the score based on the land cover category corresponding to the pixel. In this embodiment, the impermeability score is obtained by referring to a predefined impermeable layer classification standard, which will be described in detail below.

[0074] In some embodiments, determining the impermeability score for each pixel based on the land cover category includes the following steps: (1.1) Obtain multiple standard land surface categories and the initial impermeability scores corresponding to each of the multiple standard land surface categories; (1.2) For each pixel, determine the target standard land cover category that matches the land cover category corresponding to the pixel from multiple standard land cover categories, and determine the impermeability score corresponding to the pixel based on the initial impermeability score corresponding to the target standard land cover category.

[0075] The standard land cover category refers to a predefined set of representative land cover type classification systems used to uniformly describe and identify land cover types in high-resolution remote sensing images, covering common natural and artificial land cover types. The initial impermeability score is a pre-set quantitative value for each standard land cover category, used to characterize the impermeability of the corresponding standard land cover category, and the value range can be set between 0 (completely permeable) and 10 (completely impermeable).

[0076] like Figure 5 As shown, Figure 5 This is a schematic diagram of the surface category scoring criteria provided in this application embodiment. In this application embodiment, the standard surface categories include eight categories: bare land, grassland, developed land, roads, trees, water bodies, farmland, and buildings. Furthermore, each standard surface category is assigned its corresponding initial impermeability. Specifically, the impermeability contribution score for each standard surface category is set to an integer from 0 to 10. The higher the score, the greater the contribution of the corresponding pixel to the impermeability. The standard surface categories are sorted from highest to lowest in terms of their ability to impede water penetration, and their corresponding contribution scores are as follows: Buildings, as completely hardened artificial structures, are impermeable to water and are the core source of impermeability contribution, corresponding to a contribution score of 10; Roads, mainly composed of asphalt, cement, and other hardened surfaces, are almost impermeable to water, and their contribution is second only to buildings, corresponding to a contribution score of 9; developed land is mostly plazas and pedestrian areas. Hardened areas such as roads have extremely low permeability and contribute slightly less than roads, corresponding to a contribution score of 8 points; bare land is naturally exposed soil, which may be compacted but still has a certain degree of permeability (some water can seep in when it rains), contributing less, corresponding to a contribution score of 2 points; the growth of vegetation such as trees, grasslands, and farmland depends on the soil, the root area is permeable, and there is no hardened surface, contributing very little, corresponding to a contribution score of 1 point; the surface of water bodies does not hinder water infiltration (water enters the water body directly), contributing nothing to impermeable surfaces, corresponding to a contribution score of 0 points.

[0077] It should be noted that, Figure 5 This is just an example; the standard surface category and the corresponding initial impermeability score can be set according to the actual situation. This application does not limit this.

[0078] Furthermore, for each pixel, a target standard land cover category matching the corresponding land cover category is determined from multiple standard land cover categories, and the impermeability score corresponding to the pixel is determined based on the initial impermeability score corresponding to the target standard land cover category.

[0079] In this step, for each pixel in the high-resolution remote sensing image, its corresponding land cover category is matched with a preset standard land cover category to find the most matching category (i.e., the "target standard land cover category"). Then, the initial impermeability score corresponding to the target standard land cover category is directly assigned to that pixel. For example, if the land cover category of a pixel is "road," and the initial impermeability score corresponding to "road" in the standard scoring system is 9, then the impermeability score of that pixel is 9. In this way, each pixel in the high-resolution remote sensing image is assigned a continuous score value, ultimately generating complete spatial distribution data reflecting the impermeable surface of the area to be processed.

[0080] In some embodiments, determining the impermeability score of a pixel based on the initial impermeability score corresponding to the target standard surface category includes: (1.2.1) Obtain vegetation factors of the land surface area; (1.2.2) When the land cover category corresponding to the pixel is a vegetation category, the initial impermeability score is updated according to the vegetation factor to obtain the updated initial impermeability score, and the updated initial impermeability score is determined as the impermeability score corresponding to the pixel.

[0081] Among them, vegetation factors are parameters used to quantitatively describe vegetation cover density, growth status, or vitality. They can comprehensively assess the long-term cover stability of vegetation and reflect the overall coverage and growth vitality of vegetation in the surface area to be treated. Vegetation factors are usually set within a preset range of values. For example, vegetation factors are usually set within the range of [0, 1]. The smaller the degree of vegetation difference, the closer the value of the vegetation factor is to 1; the greater the degree of vegetation difference, the closer the value of the vegetation factor is to 0.

[0082] Furthermore, vegetation factors can be determined based on the Normalized Difference Vegetation Index (NDVI); or, vegetation factors can be pre-set, such as pre-setting the value of the vegetation factor to 0.5; or, they can be obtained from local official data disclosure or calculation. This application does not limit the method for obtaining vegetation factors, and adjustments can be made according to actual circumstances.

[0083] Furthermore, the determined vegetation factors are used to refine the impermeability assessment of specific pixels. It should be noted that subsequent update operations are only triggered when a pixel is identified as a vegetation category in the initial semantic segmentation. In this embodiment, grassland, trees, and farmland all belong to the vegetation category. Once the conditions are met, the labeling device for the impermeable surface area updates the initial impermeability score of the pixel based on the vegetation factors. The principle of the update is that, unlike other categories, grassland, trees, and farmland areas have vegetation and / or crops growing there, and their growth density varies in different seasons, leading to subtle differences in the impermeability of the area. To better and more accurately assess the impermeability score of pixels with a vegetation cover category, the vegetation and / or crop density in the area needs to be averaged over the four seasons to obtain the vegetation factors. Then, the initial impermeability score is updated using these vegetation factors, and the updated initial impermeability score is determined as the impermeability score corresponding to the pixel. The impermeability score corresponding to the pixels whose land cover category is vegetation determined by steps (1.2.1) and (1.2.2) is more accurate, and the subsequent determination of the impermeability of the surface area to be treated is also more accurate based on the impermeability score.

[0084] In some embodiments, when the land cover category corresponding to a pixel is a vegetation category, the initial impermeability score is updated based on the vegetation factor to obtain the updated initial impermeability score, including: (A.1) Obtain the time of acquisition of the remote sensing image; (A.2) When the land cover category corresponding to the pixel is vegetation category and the shooting time is spring and summer, the initial impermeability score is multiplied by the vegetation factor to obtain the updated initial impermeability score. (A.3) When the land cover category corresponding to the pixel is vegetation and the shooting time is autumn or winter, the initial impermeability score is divided by the vegetation factor to obtain the updated initial impermeability score.

[0085] The capture time can be obtained by extracting precise timestamp information from the metadata of the high-resolution remote sensing image. The capture time is used to determine the seasonality of the high-resolution remote sensing image. It should be noted that determining the current season of the area to be processed also requires combining the latitude and longitude coordinates of the corresponding sample points in the high-resolution remote sensing image (the months of the season vary in different regions). For example, suppose the high-resolution remote sensing image corresponding to a certain surface area to be processed was captured on July 15th. The metadata shows that the center coordinates of this surface area are approximately 30 degrees North latitude and 120 degrees East longitude (Shanghai, China), located in the Northern Hemisphere. July is summer (June to August is usually defined as summer in Shanghai), therefore, this surface area is determined to be in summer when it was captured. Conversely, if another high-resolution remote sensing image of the surface area to be processed was captured at the same time and is located in the Southern Hemisphere, such as Sydney, Australia, then June is winter (June to August is usually defined as winter in Sydney), and it should be determined to be winter.

[0086] Furthermore, when the land cover category corresponding to a pixel is a vegetation category and the shooting time is in spring or summer, it means that the current vegetation and / or crop growth density of the corresponding pixel exceeds the annual average. Since the value of the vegetation factor is limited to the range of [0, 1], the initial impermeability score can be reduced by multiplying the initial impermeability score with the vegetation factor, and the updated initial impermeability score can be obtained.

[0087] Furthermore, when the land cover category corresponding to a pixel is a vegetation category and the shooting time is in autumn or winter, it means that the current vegetation and / or crop growth density of the corresponding pixel is lower than the annual average. Since the value of the vegetation factor is limited to the range of [0, 1], the initial impermeability score can be improved by dividing the initial impermeability score by the vegetation factor, and the updated initial impermeability score can be obtained.

[0088] Understandably, this application's embodiments introduce the acquisition time of high-resolution remote sensing images as a key judgment criterion and design a dynamic scoring adjustment mechanism coupled with the seasons. Instead of treating vegetation categories as static land cover with constant impermeability, it lowers their impermeability score during the vigorous growth seasons of spring and summer, and correspondingly increases it during the dormant or withered seasons of autumn and winter. Thus, this application's embodiments can more accurately and realistically assess land cover types with significant seasonal changes, such as farmland, grassland, and deciduous forests, thereby greatly improving the precision and realism of impermeability analysis. Ultimately, it can more accurately reflect the average impermeability of the surface area to be processed, achieving more accurate labeling of impermeable surface sample areas.

[0089] In step 104, the pixel weight of each pixel is obtained, and the impermeability score of the corresponding pixel is updated according to the pixel weight to obtain the updated impermeability score for each pixel.

[0090] In high-resolution remote sensing images, pixel weight is a numerical value assigned to each pixel to reflect its relative importance or reliability in subsequent analysis or scoring adjustments. Pixel weights typically range from 0 to 1. A high pixel weight indicates that the information from the corresponding pixel is more reliable or representative, and it should contribute more to the overall impermeability assessment; a low pixel weight indicates that the corresponding pixel has higher uncertainty or is more susceptible to interference, and should contribute less to the overall impermeability assessment.

[0091] Furthermore, the updated impermeability score is achieved by weighted fusion of the original impermeability score and pixel weights. Specifically, a conventional weighting method can be used, multiplying the original score of each pixel by its corresponding pixel weight, thereby reducing the impact of low-confidence pixels on the final impermeability assessment. Alternatively, a weighted average or function mapping method can be used, with the pixel weights input as adjustment factors into a nonlinear model for dynamic adjustment. After pixel weight correction, an updated impermeability score is obtained for each pixel. This updated impermeability score not only reflects the inherent impermeability characteristics of the land cover type but also incorporates spatial confidence, making the final determination of the impermeability of the surface area to be treated more accurate.

[0092] In some embodiments, obtaining the pixel weight of each pixel includes: (1.1) Divide multiple pixels into regions according to the land cover category corresponding to each pixel to obtain at least one sub-land cover region; (1.2) Obtain the normalization constant; (1.3) For each pixel, determine the first coordinate of the center pixel of the sub-surface region where the pixel is located, and determine the weight exponential function based on the normalization constant and the first coordinate; (1.4) Determine the second coordinates of each pixel in each sub-surface region, and determine the pixel weight of each pixel based on the second coordinates and the weight exponential function. The pixel weight of each pixel is negatively correlated with the distance of the center pixel of the corresponding sub-surface region.

[0093] First, the high-resolution remote sensing images after semantic segmentation undergo structured processing. Specifically, multiple pixels are divided into regions based on the land cover category corresponding to each pixel. The aim is to group spatially adjacent pixels with the same land cover category together to obtain at least one sub-land cover region. A sub-land cover region can be understood as a patch of homogeneous land cover. For example, a continuous area of ​​rooftops will be identified as a "building" sub-land cover region, and a complete road will constitute a "road" sub-land cover region, in order to transform discrete pixel-level classification results into land cover objects with clear spatial extent and geometric shape.

[0094] Furthermore, for each sub-surface region, the first coordinates of the center pixel of that sub-surface region are determined. These first coordinates represent the geometric center of the same type of material patch. Additionally, the labeling device for impermeable surfaces acquires a normalization constant. This normalization constant controls the shape of the subsequently determined weight exponential function; for example, it determines the rate at which pixel weights decay with distance. Then, the weight exponential function is determined based on the normalization constant and the first coordinates. : .in, This represents the second coordinate of any pixel within that sub-surface region. Represents the normalization constant. This indicates data index processing. This represents the first coordinate corresponding to the center pixel of the sub-surface region. It is the standard deviation of the Gaussian kernel (controls the weight decay rate, usually set to 30; the larger the value, the smoother the weight distribution). In addition, pixel weights are usually normalized so that the normalized values ​​are limited to the range [0, 1].

[0095] For example, when the first coordinate corresponding to the center pixel of a certain sub-surface region is When the weighting exponent function is: .

[0096] It is understandable that the pixel weight calculated using the weighted exponential function is negatively correlated with the distance from that pixel to the center pixel of its corresponding sub-surface region. This means that the closer a pixel is to the center of the sub-surface region, the higher its pixel weight value; conversely, the further away a pixel is from the center of the sub-surface region, the lower its pixel weight value. Thus, this embodiment of the application can reduce the influence of the edge parts of the surface region (usually mixed pixels or transition areas) on the overall score, highlighting the contribution of the core parts of the surface region, thereby making the final assessment result of the impermeability of the surface region to be treated more accurate.

[0097] In step 105, the target area impermeability score of the surface area to be processed is determined based on the updated impermeability score corresponding to each pixel.

[0098] In this embodiment of the application, the impermeability score of the target area of ​​the entire surface area to be treated is calculated based on the updated impermeability score of each pixel, so as to determine the overall impermeability layer coverage level of the surface area to be treated based on the impermeability score of the target area.

[0099] In some embodiments, the impermeability score of the target area of ​​the surface region to be processed is determined based on the updated impermeability score corresponding to each pixel, including: (1.1) For each sub-surface region, calculate the updated impermeability score corresponding to each pixel in the current sub-surface region to obtain the sub-surface region score corresponding to the current sub-surface region; (1.2) Calculate the sum of the pixel weights of each pixel in the current sub-surface region to obtain the sub-surface region weight corresponding to the current sub-surface region; (1.3) Determine the impermeability score of the sub-surface area corresponding to the sub-surface area based on the sub-surface area score and sub-surface area weight; (1.4) Based on the impermeability scores of all sub-surface areas, determine the impermeability score of the target area of ​​the surface area to be treated.

[0100] First, for each sub-surface region, the updated impermeability score for each pixel in the current sub-surface region is calculated using the following formula, thus obtaining the sub-surface region score for the current sub-surface region. : .in, This is represented as the initial impermeability score for each pixel in the current sub-surface region. This is represented as the pixel weight corresponding to that pixel. This represents the updated impermeability score obtained by updating the corresponding pixel weights. This indicates the number of pixels in the current sub-surface region. This indicates the weight of the sub-surface region corresponding to the current sub-surface region.

[0101] Furthermore, by overlaying the impermeability scores of all sub-surface areas within the surface area to be treated, the impermeability score of the target area corresponding to the surface area to be treated is obtained. .like Figure 6 As shown, Figure 6This is a schematic diagram illustrating the impermeability rating of the target area of ​​the surface region to be processed, provided in an embodiment of this application. Target area impermeability rating A corresponds to the impermeability rating of surface region A indicated by high-resolution remote sensing image A; target area impermeability rating B corresponds to the impermeability rating of surface region B indicated by high-resolution remote sensing image B; target area impermeability rating C corresponds to the impermeability rating of surface region C indicated by high-resolution remote sensing image C; and target area impermeability rating D corresponds to the impermeability rating of surface region D indicated by high-resolution remote sensing image D. The target area impermeability rating includes sub-surface region ratings corresponding to at least one sub-surface region, wherein, as... Figure 6 As shown, the sub-surface area scores corresponding to different sub-surface areas (in this example, the sub-surface area score ranges from [0,10]) are marked with different colors. The closer to red (the closer the sub-surface area score is to 10), the better the impermeability of the corresponding sub-surface area. The closer to blue (the closer the sub-surface area score is to 0), the worse the impermeability of the corresponding sub-surface area.

[0102] In step 106, when the impermeability score of the target area exceeds the preset score threshold, the surface area to be treated is determined to be an impermeable surface area.

[0103] The preset scoring threshold is a pre-defined scoring threshold s. The calculated impermeability score of the target area of ​​the surface to be treated is compared with this preset scoring threshold s to determine whether the surface to be treated is an impermeable area. If the impermeability score of the target area is not less than the preset scoring threshold... This indicates that the coverage ratio or overall intensity of impermeable surfaces (such as buildings, roads, etc.) within the surface area to be treated meets or exceeds the set standard. This surface area has a high degree of surface hardening and low permeability, and is therefore identified as an impermeable surface area. Furthermore, the impermeable surface area can be mapped to grid cells of a specified resolution using geographic coordinates, thereby generating training samples with images and labels for subsequent data processing.

[0104] In some embodiments, the preset scoring threshold is determined in the following manner: Before executing step 101, 2700 sample surface areas worldwide are first selected for training, and different integer initial scoring thresholds from 1 to 9 are selected to distinguish impermeable surface sample areas. For example... Figure 7 As shown, Figure 7This is a schematic diagram illustrating the effect of selecting the preset scoring threshold provided in the embodiments of this application. During the experiment, the sample discrimination accuracy also changes when different initial scoring thresholds are selected as the preset scoring thresholds. Specifically, when the total number of samples is 2700, the number of actual impermeable layer samples is 324, the number of actual permeable layer samples is 2376, and the number of eliminated samples is 481, the preset scoring threshold is set to 3 in turn when selecting the preset scoring threshold from different initial scoring thresholds (1 to 9). At this time, the overall accuracy (OA) of the overall sample labeling of the impermeable surface sample area is 98.07%, the precision of the labeling of a single sample in the impermeable surface sample area is 93.31%, the recall of the overall sample labeling of the impermeable surface sample area is 90.43%, and the F1 score of the overall sample in the impermeable surface sample area is 91.85%.

[0105] Furthermore, as can be seen from the figures, the accuracy rate of the impermeable surface sample area labeling in this application embodiment exceeds 93%. Therefore, this application embodiment can reliably and automatically determine the type of impermeable surface sample points, thereby minimizing the uncertainty caused by human intervention and providing a large number of impermeable surface samples quickly and accurately. In practical applications, this application embodiment can provide clear and quantitative decision-making basis for various aspects such as urban flooding risk zone identification, ecological control line delineation, and land use management through the rapid and accurate labeling of impermeable surface sample areas.

[0106] like Figure 8 As shown, Figure 8 This is a schematic diagram of the module structure of the marking device for impermeable surface areas provided in this application embodiment. The marking device 200 based on the impermeable surface area may include the following modules 201 to 206: The acquisition module 201 is used to acquire high-resolution remote sensing images corresponding to the surface area to be processed; The semantic segmentation module 202 is used to input high-resolution remote sensing images into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image. The pre-trained semantic segmentation model is trained based on the sample high-resolution remote sensing images corresponding to the sample land surface areas and the labeled land cover categories corresponding to the sample high-resolution remote sensing images. The first scoring module 203 is used to determine the impermeability score of each pixel according to the land cover category. The second scoring module 204 is used to obtain the pixel weight of each pixel, update the waterproofness score of the corresponding pixel according to the pixel weight, and obtain the updated waterproofness score of each pixel. The target scoring module 205 is used to determine the target area impermeability score of the surface area to be processed based on the updated impermeability score corresponding to each pixel. The impermeable surface area determination module 206 is used to determine the surface area to be processed as an impermeable surface area when the impermeability score of the target area exceeds the preset score threshold.

[0107] The specific implementation of the marking device for the impermeable surface area is basically the same as the specific implementation of the marking method for the impermeable surface area described above, and will not be repeated here.

[0108] In some embodiments, the semantic segmentation module 202 is used for: Obtain the total area of ​​a high-resolution remote sensing image; The first pixel is defined as the pixel whose land cover category is water, and the total area of ​​the region corresponding to the first pixel is calculated to obtain the first area. Pixels with a land cover category of "no information" are identified as the second pixel, and the total area of ​​the region corresponding to the second pixel is calculated to obtain the second area. Divide the first area and / or the second area by the total area of ​​the image to obtain the water area ratio; When the proportion of water area exceeds a preset proportion threshold, the surface area to be processed is determined to be an undefined area.

[0109] In some embodiments, the semantic segmentation module 202 is further configured to: Obtain the high-resolution remote sensing image of the sample land surface area, and the labeled land cover category of the sample high-resolution remote sensing image; The high-resolution remote sensing image of the sample is input into the semantic segmentation model, and the predicted land cover category corresponding to each pixel in the high-resolution remote sensing image of the sample is output. Determine the difference between the predicted land cover category and the corresponding labeled land cover category for each pixel; The semantic segmentation model is iteratively trained based on the differences to obtain a pre-trained semantic segmentation model.

[0110] In some embodiments, the first scoring module 203 is used for: Obtain multiple standard land surface categories, and the initial impermeability scores corresponding to each standard land surface category; For each pixel, a target standard land cover category matching the corresponding land cover category is determined from multiple standard land cover categories, and the impermeability score corresponding to the pixel is determined based on the initial impermeability score corresponding to the target standard land cover category.

[0111] In some embodiments, the first scoring module 203 is further configured to: Obtain vegetation factors for the land surface area; When the land cover category corresponding to a pixel is a vegetation category, the initial impermeability score is updated based on the vegetation factor to obtain the updated initial impermeability score, and the updated initial impermeability score is determined as the impermeability score corresponding to the pixel.

[0112] In some embodiments, the first scoring module 203 is further configured to: The time of capture of high-resolution remote sensing images; When the land cover category corresponding to the pixel is vegetation and the shooting time is spring or summer, the initial impermeability score is multiplied by the vegetation factor to obtain the updated initial impermeability score. When the land cover category corresponding to a pixel is vegetation and the shooting time is autumn or winter, the initial impermeability score is divided by the vegetation factor to obtain the updated initial impermeability score.

[0113] In some embodiments, the second scoring module 204 is used for: Based on the land cover category corresponding to each pixel, multiple pixels are divided into regions to obtain at least one sub-land cover region; Obtain the normalization constant; For each pixel, determine the first coordinate of the center pixel of the sub-surface region where the pixel is located, and determine the weight exponential function based on the normalization constant and the first coordinate; Determine the second coordinates of each pixel in each sub-surface region, and determine the pixel weight of each pixel based on the second coordinates and the weight exponential function. The pixel weight of each pixel is negatively correlated with the distance to the center pixel of the corresponding sub-surface region.

[0114] In some embodiments, the target scoring module 205 is used for: For each sub-surface region, the updated impermeability score corresponding to each pixel in the current sub-surface region is calculated to obtain the sub-surface region score corresponding to the current sub-surface region; The sum of the pixel weights of each pixel in the current sub-surface region is calculated to obtain the sub-surface region weight corresponding to the current sub-surface region; Based on the sub-surface area score and sub-surface area weight, the impermeability score of the corresponding sub-surface area is determined. Based on the impermeability scores of all sub-surface areas, the impermeability score of the target area of ​​the surface area to be treated is determined.

[0115] This application provides a method, apparatus, electronic device, and storage medium for labeling impermeable surface areas. The method involves acquiring a high-resolution remote sensing image corresponding to the surface area to be processed; inputting the high-resolution remote sensing image into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image; training the pre-trained semantic segmentation model based on sample high-resolution remote sensing images corresponding to sample surface areas and the labeled land cover categories corresponding to the sample high-resolution remote sensing images; determining the impermeability score corresponding to each pixel based on the land cover category; obtaining the pixel weight of each pixel and updating the impermeability score of the corresponding pixel based on the pixel weight to obtain an updated impermeability score for each pixel; determining the target area impermeability score of the surface area to be processed based on the updated impermeability score for each pixel; and determining the surface area to be processed as an impermeable surface area when the target area impermeability score exceeds a preset score threshold.

[0116] This application embodiment automatically identifies the land cover category corresponding to each pixel in the land surface area by using a pre-trained semantic segmentation model. Based on the impermeability score corresponding to the pixel determined by the land cover category and the obtained pixel weight, the impermeability of the pixel area indicated by the pixel is quantitatively analyzed. Then, based on the updated impermeability scores of each pixel, it is determined whether the land surface area to be processed is an impermeable surface area. Since this application uses a unified, pre-trained semantic segmentation model for semantic segmentation and then uses a unified scoring rule for impermeability scoring, this application can accurately identify and divide the spatial distribution of various land cover types (such as buildings, roads, green spaces, water bodies, etc.) through pixel-level semantic segmentation, and accurately determine the corresponding impermeability score based on the pixel weight corresponding to each pixel in the cover. Therefore, this application embodiment can achieve accurate and efficient labeling of the land surface area to be processed.

[0117] like Figure 9 As shown, Figure 9 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. The electronic device includes: The processor 301 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 302 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 302 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 302 and is called and executed by the processor 301 to execute the marking method for the impermeable surface area of ​​the embodiments of this application. Input / output interface 303 is used to implement information input and output; The communication interface 304 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 305 transmits information between various components of the device (e.g., processor 301, memory 302, input / output interface 303, and communication interface 304); The processor 301, memory 302, input / output interface 303, and communication interface 304 are connected to each other within the device via bus 305.

[0118] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for marking impermeable surface areas.

[0119] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0120] This application provides a method, apparatus, electronic device, and storage medium for labeling impermeable surface areas. The method involves acquiring a high-resolution remote sensing image corresponding to the surface area to be processed; inputting the high-resolution remote sensing image into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image; training the pre-trained semantic segmentation model based on sample high-resolution remote sensing images corresponding to sample surface areas and the labeled land cover categories corresponding to the sample high-resolution remote sensing images; determining the impermeability score corresponding to each pixel based on the land cover category; obtaining the pixel weight of each pixel and updating the impermeability score of the corresponding pixel based on the pixel weight to obtain an updated impermeability score for each pixel; determining the target area impermeability score of the surface area to be processed based on the updated impermeability score for each pixel; and determining the surface area to be processed as an impermeable surface area when the target area impermeability score exceeds a preset score threshold.

[0121] This application embodiment automatically identifies the land cover category corresponding to each pixel in the land surface area by using a pre-trained semantic segmentation model. Based on the impermeability score corresponding to the pixel determined by the land cover category and the obtained pixel weight, the impermeability of the pixel area indicated by the pixel is quantitatively analyzed. Then, based on the updated impermeability scores of each pixel, it is determined whether the land surface area to be processed is an impermeable surface area. Since this application uses a unified, pre-trained semantic segmentation model for semantic segmentation and then uses a unified scoring rule for impermeability scoring, this application can accurately identify and divide the spatial distribution of various land cover types (such as buildings, roads, green spaces, water bodies, etc.) through pixel-level semantic segmentation, and accurately determine the corresponding impermeability score based on the pixel weight corresponding to each pixel in the cover. Therefore, this application embodiment can achieve accurate and efficient labeling of the land surface area to be processed.

[0122] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0123] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0124] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0125] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0126] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0127] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0129] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0130] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0131] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0132] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for marking impermeable surface areas, characterized in that, include: Acquire high-resolution remote sensing images of the surface area to be processed; The high-resolution remote sensing image is input into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image. The pre-trained semantic segmentation model is trained based on the sample high-resolution remote sensing image corresponding to the sample land surface area and the labeled land cover category corresponding to the sample high-resolution remote sensing image. Determine the impermeability score for each pixel based on the land cover category; Obtain the pixel weight of each pixel, update the impermeability score of the corresponding pixel according to the pixel weight, and obtain the updated impermeability score for each pixel; Based on the updated impermeability score corresponding to each pixel, the target area impermeability score of the surface area to be processed is determined. When the impermeability score of the target area exceeds the preset score threshold, the surface area to be treated is determined to be an impermeable surface area.

2. The method for marking impermeable surface areas according to claim 1, characterized in that, After obtaining the land cover category corresponding to each pixel in the high-resolution remote sensing image, the method further includes: Obtain the total area of ​​the high-resolution remote sensing image; The pixel whose land cover category is water body is identified as the first pixel, and the total area of ​​the region corresponding to the first pixel is calculated to obtain the first area; The pixels whose land cover category is "no information" are identified as the second pixels, and the total area of ​​the regions corresponding to the second pixels is calculated to obtain the second area. Divide the first area and / or the second area by the total area of ​​the image to obtain the water area ratio; When the proportion of the water body area exceeds a preset proportion threshold, the surface area to be processed is determined to be an undefined area.

3. The method for marking impermeable surface areas according to claim 1, characterized in that, The step of determining the impermeability score corresponding to each pixel based on the land cover category includes: Obtain multiple standard land surface categories, and the initial impermeability scores corresponding to each of the multiple standard land surface categories; For each pixel, a target standard land cover category matching the land cover category corresponding to the pixel is determined from a plurality of standard land cover categories, and the impermeability score corresponding to the pixel is determined according to the initial impermeability score corresponding to the target standard land cover category.

4. The method for marking impermeable surface areas according to claim 3, characterized in that, The step of determining the impermeability score corresponding to the pixel based on the initial impermeability score corresponding to the target standard land surface category includes: Obtain vegetation factors for the aforementioned surface area; When the land cover category corresponding to the pixel is a vegetation category, the initial impermeability score is updated according to the vegetation factor to obtain the updated initial impermeability score, and the updated initial impermeability score is determined as the impermeability score corresponding to the pixel.

5. The method for marking impermeable surface areas according to claim 4, characterized in that, When the land cover category corresponding to the pixel is a vegetation category, the initial impermeability score is updated according to the vegetation factor to obtain the updated initial impermeability score, including: Obtain the capture time of the high-resolution remote sensing image; When the land cover category corresponding to the pixel is a vegetation category and the shooting time is spring or summer, the initial impermeability score is multiplied by the vegetation factor to obtain the updated initial impermeability score. When the land cover category corresponding to the pixel is a vegetation category and the shooting time is autumn or winter, the initial impermeability score is divided by the vegetation factor to obtain the updated initial impermeability score.

6. The method for marking impermeable surface areas according to claim 1, characterized in that, The step of obtaining the pixel weight of each pixel includes: Based on the land cover category corresponding to each pixel, multiple pixels are divided into regions to obtain at least one sub-land cover region; Obtain the normalization constant; For each pixel, determine the first coordinates of the center pixel of the sub-surface region where the pixel is located, and determine the weight exponential function based on the normalization constant and the first coordinates; A second coordinate is determined for each pixel in each sub-surface region, and a pixel weight is determined for each pixel based on the second coordinate and the weight exponential function, wherein the pixel weight of each pixel is negatively correlated with the distance to the center pixel of the corresponding sub-surface region.

7. The method for marking impermeable surface areas according to claim 6, characterized in that, The step of determining the impermeability score of the target area of ​​the surface region to be processed based on the updated impermeability score corresponding to each pixel includes: For each of the sub-surface regions, the updated impermeability score corresponding to each pixel in the current sub-surface region is calculated to obtain the sub-surface region score corresponding to the current sub-surface region; The sum of the pixel weights of each pixel in the current sub-surface region is calculated to obtain the sub-surface region weight corresponding to the current sub-surface region; Based on the sub-surface area score and the sub-surface area weight, the impermeability score of the sub-surface area corresponding to the sub-surface area is determined; Based on the impermeability scores of all the sub-surface regions, the target area impermeability score of the surface region to be treated is determined.

8. The method for marking impermeable surface areas according to claim 1, characterized in that, Before inputting the high-resolution remote sensing image into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image, the method further includes: Obtain high-resolution remote sensing images of sample land surface areas, and label land cover categories corresponding to the sample high-resolution remote sensing images; The high-resolution remote sensing image of the sample is input into the semantic segmentation model, and the predicted land cover category corresponding to each pixel in the high-resolution remote sensing image of the sample is output. Determine the difference between the predicted land cover category and the corresponding labeled land cover category for each pixel; The semantic segmentation model is iteratively trained based on the differences to obtain a pre-trained semantic segmentation model.

9. A marking device for impermeable surfaces, characterized in that, include: The acquisition module is used to acquire high-resolution remote sensing images of the surface area to be processed. The semantic segmentation module is used to input the high-resolution remote sensing image into a pre-trained semantic segmentation model to obtain the land cover category corresponding to each pixel in the high-resolution remote sensing image. The pre-trained semantic segmentation model is trained based on the sample high-resolution remote sensing image corresponding to the sample land surface area and the labeled land cover category corresponding to the sample high-resolution remote sensing image. The first scoring module is used to determine the impermeability score corresponding to each pixel based on the land cover category. The second scoring module is used to obtain the pixel weight of each pixel, update the impermeability score of the corresponding pixel according to the pixel weight, and obtain the updated impermeability score corresponding to each pixel. The target scoring module is used to determine the target area impermeability score of the surface area to be processed based on the updated impermeability score corresponding to each pixel. The impermeable surface area determination module is used to determine the surface area to be processed as an impermeable surface area when the impermeability score of the target area exceeds a preset score threshold.

10. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the marking method for impermeable surface areas as described in any one of claims 1 to 8.

11. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the marking method for impermeable surface areas as described in any one of claims 1 to 8.

Images