Bridge detection method and device based on remote sensing image water body
By obtaining water masks from time series remote sensing images and extracting river skeleton information, determining bridge distribution and training detection models, the problem of low bridge detection accuracy in the prior art is solved, and higher detection accuracy and reliability are achieved.
Patent Information
- Application Number
- CN202411637372.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-15
- Publication Date
- 2025-04-11
AI Technical Summary
In the prior art, bridge detection methods rely on prior knowledge or known road information, and there are few samples in remote sensing images, resulting in low bridge detection accuracy.
By obtaining the water mask from the time series remote sensing image, extracting the skeleton information and breakpoint information of the river, determining the bridge distribution information, and labeling and extracting the bridge based on this information, bridge sample data is obtained, and used to train the bridge detection model.
The detection accuracy of bridges in remote sensing images is improved, the dependence on prior knowledge and road information is reduced, and the accuracy and reliability of bridge detection is enhanced.
Smart Images
Figure CN119810634A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of remote sensing image processing, and in particular to a bridge detection method and device based on remote sensing image water body. Background Art
[0002] In remote sensing images, bridges are an important ground building target; target detection of bridges in remote sensing images is widely used in urban planning and management, transportation safety, natural disaster detection and response, environmental monitoring and protection, etc. Therefore, improving the target detection accuracy of bridges is of great significance.
[0003] In the related technology, knowledge-based extraction methods and threshold segmentation-based methods are usually used to detect bridges. The former usually relies on prior knowledge, resulting in insufficient bridge extraction accuracy and main direction deviation; the latter has poor detection effect on bridges under complex backgrounds and changing lighting conditions; in addition, the existing technology also uses a road-based bridge detection method, which relies on known road information and limits the bridge detection range. When the road information is incomplete or inaccurate, it will lead to inaccurate bridge detection results; in addition, when the above method is applied to remote sensing image bridge detection, few samples are collected, which leads to low bridge detection accuracy and poor effect in multi-target detection results. Summary of the invention
[0004] The present invention provides a bridge detection method and device based on remote sensing image water body, which are used to solve the defects of the bridge detection method in the prior art that the bridge detection method relies on prior knowledge or known road information, and when applied to remote sensing image bridge detection, the number of samples collected is small, resulting in low bridge detection accuracy, thereby improving the bridge detection accuracy in remote sensing images.
[0005] The present invention provides a bridge detection method based on remote sensing image water body, comprising: Obtaining a water body mask of a target area from a time series remote sensing image, wherein the target area includes a river and a bridge; Extracting skeleton information and breakpoint information of the river from the water body mask, and determining bridge distribution information in the target area according to the skeleton information and the breakpoint information; Annotating and extracting bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data; The bridge detection model is trained by using the bridge sample data to achieve bridge detection.
[0006] According to a bridge detection method based on water bodies in remote sensing images provided by the present invention, the method of obtaining a water body mask of a target area from a time series remote sensing image comprises: Acquire target time series remote sensing image data, wherein the target time series remote sensing image data includes one of medium-resolution remote sensing images, long-time series remote sensing images, and medium-resolution long-time series remote sensing images; Dividing the target area into a plurality of sub-areas at different levels based on water body factors in the target time series remote sensing image data; The water body mask is obtained from the sub-areas of the multiple different levels according to a water body extraction rule preset in a priori knowledge base of the water body, wherein the water body extraction rule includes operations of preprocessing, water body index calculation and threshold setting.
[0007] According to a bridge detection method based on remote sensing image water bodies provided by the present invention, the target area includes a plurality of water body masks; After obtaining the water body mask of the target area from the time series remote sensing image, the method includes: The contour area of each water body mask is calculated, and at least one of the fragments and holes in the water body mask is repaired according to the contour area to obtain a new water body mask.
[0008] According to a bridge detection method based on remote sensing image water body provided by the present invention, the skeleton information and breakpoint information of the river are extracted from the water body mask, including: extracting the skeleton information from the water body mask using a morphological image processing algorithm; A skeleton feature is extracted according to the skeleton information, and the breakpoint information is extracted from the water body mask according to the skeleton feature; wherein the skeleton feature is used to indicate that the skeleton in the skeleton information has a width of 1 pixel.
[0009] According to a bridge detection method based on remote sensing image water body provided by the present invention, the bridge detection model is trained by the bridge sample data, comprising: The target model is iteratively trained based on the bridge sample data to obtain the bridge detection model; wherein the target model includes at least one of a traditional machine learning model or a neural network model.
[0010] According to a bridge detection method based on remote sensing image water body provided by the present invention, after obtaining the bridge detection model, the method further includes: Based on the bridge detection model, bridge detection is performed on a single-view high-resolution image to obtain a bridge detection result.
[0011] The present invention also provides a bridge detection device based on remote sensing image water body, comprising: A first extraction module is used to obtain a water body mask of a target area from a time series remote sensing image, wherein the target area includes a river and a bridge; A second extraction module is used to extract skeleton information and breakpoint information of the river from the water body mask, and determine the bridge distribution information in the target area according to the skeleton information and the breakpoint information; A third extraction module is used to mark and extract bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data to train a bridge detection model; The training and detection module is used to train the bridge detection model through the bridge sample data to realize bridge detection.
[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the remote sensing image-based water body bridge detection method as described above is implemented.
[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the bridge detection method based on remote sensing image water body as described in any one of the above is implemented.
[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned bridge detection methods based on remote sensing image water bodies.
[0015] The present invention provides a bridge detection method and device based on remote sensing image water bodies. The water body mask of the target area is obtained from the time series remote sensing image, and the bridge distribution information in the target area is determined according to the skeleton information and the breakpoint information. The bridges in the time series remote sensing image are then labeled and extracted according to the bridge distribution information to obtain bridge sample data. The bridge detection model is trained based on the data to realize bridge detection, thereby improving the detection accuracy of bridges in remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0017] Figure 1 It is one of the flow charts of the bridge detection method based on remote sensing image water body provided by the present invention.
[0018] Figure 2 This is the second flow chart of the bridge detection method based on remote sensing image water body provided by the present invention.
[0019] Figure 3 It is a structural schematic diagram of a bridge detection device based on remote sensing image water body provided by the present invention.
[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0022] Combine the following Figure 1-Figure 3 The present invention describes a bridge detection method and device based on remote sensing image water body.
[0023] Figure 1 is one of the flow charts of the bridge detection method based on remote sensing image water body provided by the present invention, such as Figure 1 As shown, the method comprises the following steps: Step 110: Acquire a water body mask of a target area from time series remote sensing images, where the target area includes rivers and bridges.
[0024] In this step, time series remote sensing images are collected by remote sensing satellites such as Sentinel-2 and Landsat. Such images have the continuity and consistency of time series.
[0025] In this embodiment, the time series remote sensing images are remote sensing images with a long time series and a medium resolution.
[0026] In this embodiment, water masks of different regions may be automatically extracted from the time series remote sensing images by a preset water mask extraction program or algorithm as sample data for bridge detection.
[0027] Specifically, obtaining a water body mask of a target area from a time series remote sensing image includes: obtaining target time series remote sensing image data, the target time series remote sensing image data includes one of medium-resolution remote sensing images, long-time series remote sensing images, and medium-resolution long-time series remote sensing images; dividing the target area into a plurality of sub-areas at different levels based on water body factors in the target time series remote sensing image data; obtaining a water body mask from the plurality of sub-areas at different levels according to a water body extraction rule preset in a priori knowledge base of the water body, the water body extraction rule including preprocessing, water body index calculation, and threshold setting operations.
[0028] In this embodiment, a threshold-based water mask extraction algorithm can be used to obtain the water mask of the target area. For example, the spectral characteristics of the water body in the remote sensing image are used to select a suitable band to build a model; by analyzing the water body spectral characteristic curve, a suitable threshold is determined to separate the water body and non-water body areas; and then a multi-band method (such as the spectral relationship method and the water body index method) is used to calculate the water body index, and the corresponding water body mask is extracted from the medium-resolution long-time series remote sensing image according to the preset segmentation threshold and water body index.
[0029] In this embodiment, various band information of each month of a target area in a year is collected based on long-term series remote sensing image data; then, the maximum NDVI value (NDVI_MAX) of the target area in a year, the median reflectance of the near-infrared band in summer, and the night light intensity distribution are calculated based on the band information, which are used for the subsequent calculation of multiple water body indices; finally, for each sub-area, the corresponding water body mask is determined based on the water body index and the corresponding segmentation threshold.
[0030] In this embodiment, the NDVI is calculated using the following formula: NDVI = (NIR - R) / (NIR + R); Among them, NIR is the reflectance of the near-infrared band, and R is the reflectance of the red band. Then the corresponding NDVI is calculated for all remote sensing images of each year, and the maximum NDVI value of each pixel in the year is obtained.
[0031] In this embodiment, the near-infrared band reflectivity is extracted from the screened remote sensing images, and the extracted near-infrared band reflectivity is statistically analyzed and the median is calculated; night light brightness data covering the target area, such as NPP-VIIRS data, is collected, and the highlighted areas in the night light brightness data are used as candidate bridge areas. Finally, the extracted building areas are vectorized to generate building masks.
[0032] Step 120: extract the skeleton information and breakpoint information of the river from the water body mask, and determine the bridge distribution information in the target area based on the skeleton information and breakpoint information.
[0033] In this step, the skeleton information of the river includes the skeleton line of the river; on the skeleton line of the river, the bridge usually appears as a breakpoint or an obvious width change area; by analyzing the geometric characteristics and topological relationship of the skeleton line, potential bridge breakpoints can be identified.
[0034] Specifically, extracting the skeleton information and breakpoint information of the river from the water body mask includes: extracting the skeleton information from the water body mask using a morphological image processing algorithm; extracting skeleton features based on the skeleton information, and extracting breakpoint information from the water body mask based on the skeleton features; wherein the skeleton features are used to indicate that the skeleton in the skeleton information has a width of 1 pixel.
[0035] In this embodiment, the morphological image processing algorithm includes a Skeletonize algorithm; the algorithm first performs binarization, corrosion, thinning and skeletonization operations on the image to finally obtain the skeleton structure of the image; the morphological image processing algorithm also includes a Zhang-Suen thinning algorithm, which gradually thins the image by iteratively deleting pixels that meet specific conditions until the skeleton of the image is obtained; the morphological image processing algorithm also includes a MedialAxis Transform algorithm, which forms a skeleton by calculating the maximum disk of each point in the image and finding the center lines of these disks.
[0036] Specifically, the skeleton information and breakpoint information of the river are extracted through the following steps: (1) Refine the water body mask to obtain the initial skeleton line of the river.
[0037] In this embodiment, during the thinning process, it is necessary to ensure the continuity and integrity of the skeleton line while removing burrs and redundant structures; for example, topological constraints and morphological operations such as expansion and erosion are combined to further optimize and correct the skeleton line.
[0038] It should be noted that due to reasons such as bridge spans and river breaks, there may be breakpoints in the skeleton line.
[0039] In this embodiment, the breakpoints may be connected using expansion and thinning operations in mathematical morphology and topological relationship analysis.
[0040] For example, set appropriate expansion coefficients and refinement rules to ensure the accuracy and rationality of breakpoint connections.
[0041] (2) Propose bridge breakpoints from the river skeleton.
[0042] In this embodiment, potential bridge breakpoints can be identified by analyzing the geometric features and topological relationships (i.e., skeleton features) of the skeleton lines. Machine learning or deep learning algorithms can also be used to extract and classify features of the skeleton lines, thereby improving the accuracy and efficiency of bridge breakpoint identification.
[0043] (3) Determine the bridge distribution.
[0044] In this embodiment, the specific location of the bridge is determined based on the identified bridge breakpoints in combination with remote sensing images and GIS data.
[0045] Specifically, a coordinate converter of image row and column coordinates and longitude and latitude coordinates is constructed according to the geographic information and projection coordinates carried by the water body mask; the longitude and latitude of the skeleton and the breakpoints are calculated using the coordinate converter; The skeleton and breakpoints are saved as line and point types as shapefile files, and imported into geographic information software to obtain the distribution of bridges (points) in the time series remote sensing images.
[0046] Step 130: annotate and extract bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data.
[0047] In this step, the skeleton is extracted from the water body mask, and the skeleton endpoint detection method is used to obtain the river breakpoints and thereby obtain the approximate spatial information of the bridge.
[0048] Specifically, we first detect the breakpoints or mutation points of the river, and then infer the distribution of bridges based on the location and characteristics of the river breakpoints and the prior knowledge of bridges (such as bridges are usually located above rivers, the width and length of bridges, etc.).
[0049] In this embodiment, a machine learning algorithm (such as a support vector machine, a random forest, etc.) may be used to classify river breakpoints to distinguish bridge breakpoints from other types of breakpoints.
[0050] In this embodiment, in the time series remote sensing images, the bridges in the time series remote sensing images are identified and labeled by extracting the features of the bridges (such as shape, size, texture, brightness, etc.).
[0051] In this embodiment, vector data or image data of the bridge is extracted from the remote sensing image according to the marked bridge position and range; a pixel-based extraction method (such as threshold segmentation, edge detection, etc.) or an object-based extraction method (such as object-oriented classification, etc.) can be adopted.
[0052] In this embodiment, bridge samples may also be manually or automatically annotated according to the bridge positions detected by breakpoints in the high-resolution single-moment remote sensing image to construct bridge sample data.
[0053] Step 140: Train a bridge detection model using bridge sample data to implement bridge detection.
[0054] In this step, the bridge sample data is calibrated and extracted from the entire time series remote sensing image by the method of the above steps 110 to 130 as training samples, which can effectively increase the number and quality of bridge samples.
[0055] In this embodiment, the bridge sample data can be preprocessed as follows: the collected bridge sample data (image data) is cleaned to remove duplicate, invalid or erroneous data; the bridge sample data is preprocessed, such as cropping, scaling, denoising, etc., to further improve the image quality; the bridge sample data is standardized to ensure that data from different sources are consistent and comparable.
[0056] In this embodiment, features of the bridge are extracted from the preprocessed data, such as structural features (such as bridge type, span, height, etc.), material features (such as concrete strength, steel type, etc.), and environmental features (such as climate, geological conditions, etc.). Image processing techniques (such as edge detection, texture analysis, shape analysis, etc.) can also be used to extract features from image data, and the most useful features for model training are selected based on the needs and objectives of the bridge detection model.
[0057] Finally, the extracted features are used to train the target model. After the target model converges, a usable bridge detection model is obtained, which is used to accurately detect bridges in remote sensing images.
[0058] In this embodiment, the detection objects of the water body based on the remote sensing image include but are not limited to bridges, ships, buildings or other structures, that is, the target area includes at least one of the bridges, ships, buildings or other structures and the river.
[0059] In this embodiment, when the detection object is a ship, a building or other structures, the target detection method based on the remote sensing image water body is the same as the above steps 110 to 140, and will not be repeated in this embodiment.
[0060] The bridge detection method based on remote sensing image water body provided by the embodiment of the present invention obtains the water body mask of the target area from the time series remote sensing image, determines the bridge distribution information in the target area according to the skeleton information and the breakpoint information, and then labels and extracts the bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data, and uses this to train the bridge detection model to realize bridge detection, thereby improving the detection accuracy of the bridges in the remote sensing image.
[0061] In some embodiments, the target area includes multiple water body masks; after obtaining the water body masks of the target area from the time series remote sensing images, the method includes: calculating the contour area of each water body mask, and repairing at least one of the fragments and holes in the water body mask according to the contour area to obtain a new water body mask.
[0062] In this embodiment, the water body mask may be a binary image, in which the water body area is white or 1, and other areas are black or 0.
[0063] Specifically, first read the binary image containing the water body mask, and then use the findContours function in the image processing library (such as OpenCV) to find the contours of all water body areas. For each found contour, use the contourArea function to calculate its area.
[0064] In this embodiment, a large water body area that needs to be preserved and fragments or holes that need to be repaired are distinguished by setting an area threshold, and the area threshold can be set according to user needs.
[0065] In this embodiment, the repair area may be a fragment or a hole, or may be a fragment and a hole.
[0066] Specifically, all contours are traversed, and contours with an area smaller than a threshold are marked as fragments that need to be repaired. Morphological operations (such as dilation operations) are used to fill holes (i.e., small black holes), or image repair algorithms (such as the inpaint function in OpenCV) are used to repair these areas to obtain a new water mask.
[0067] In this embodiment, an image segmentation and region merging algorithm (such as a watershed algorithm, a region growing algorithm or a graph theory method) may also be used to merge small-area fragments into a nearby large water body area.
[0068] The bridge detection method based on remote sensing image water bodies provided by an embodiment of the present invention calculates the contour area of each water body mask and repairs at least one of the fragments and holes in the water body mask according to the contour area, thereby reducing the loss in the water body mask and improving the quality of the extracted bridge sample data.
[0069] In some embodiments, training a bridge detection model using bridge sample data includes: iteratively training a target model based on the bridge sample data to obtain a bridge detection model; wherein the target model includes at least one of a traditional machine learning model or a neural network model.
[0070] In this embodiment, the target model includes a support vector machine (SVM), a random forest (RF), or a convolutional neural network (CNN), etc.
[0071] In this embodiment, one of the traditional machine learning models and the neural network model can be used to implement feature extraction and regression analysis to obtain a bridge detection model; or the bridge features can be extracted through the traditional machine learning model, and the bridge features can be learned through the neural network model to obtain a bridge detection model.
[0072] In this embodiment, the bridge features extracted from the bridge sample data are used as the input of the target model, and the model is iteratively trained with the type or maintenance status of the bridge as a label; during the training process, the parameters and structure of the model need to be continuously adjusted to improve the accuracy and generalization ability of the model; during the model training process, cross-validation, early stopping and other strategies can be used to prevent the model from overfitting.
[0073] In this embodiment, a test data set of bridge samples is used to evaluate the trained model. The evaluation indicators include accuracy, recall rate, F1 score, etc., and the performance of the model is quantitatively analyzed based on the evaluation results, and then the model is optimized, such as adjusting model parameters, increasing the number of features, or improving feature extraction methods.
[0074] In some embodiments, after obtaining the bridge detection model, the method further includes: performing bridge detection on the single-view high-resolution image based on the bridge detection model to obtain a bridge detection result.
[0075] In this embodiment, a single-view high-resolution image can capture minute details on the bridge surface, including cracks, rust, spalling and other damage characteristics, providing more detailed observation data for bridge detection, which can further improve the accuracy of bridge detection.
[0076] The bridge detection method based on remote sensing image water bodies provided in an embodiment of the present invention obtains a bridge detection model by iteratively training at least one of the traditional machine learning models or neural network models based on bridge sample data, thereby improving the detection efficiency and accuracy of bridges in remote sensing images.
[0077] Figure 2 This is the second flow chart of the bridge detection method based on remote sensing image water body provided by the present invention. Figure 2 In the illustrated embodiment, the bridge detection method based on remote sensing image water body can also be implemented through the following steps: S1. Collect long-term series remote sensing image and other data; S2. Mark the target identification area and confirm the sub-areas at different levels; S3. For the target area, set the water body extraction rules according to the prior knowledge base of the water body, and extract the water body mask of the target area; S4. Use the water body mask as the initial sample and pre-process the sample; S5. Use the processed sample to extract the river skeleton and breakpoints; S6. Construct a coordinate converter, process the breakpoint information, and obtain the final data set; S7. According to the preliminary results, mark the bridge samples to obtain the final data set; S8. Use the final data set as the training sample, and use the deep learning target detection model to train the bridge detection results of the target area.
[0078] The following is a description of the bridge detection device based on remote sensing image water body provided by the present invention. The bridge detection device based on remote sensing image water body described below and the bridge detection method based on remote sensing image water body described above can be referenced to each other.
[0079] Figure 3 Schematic diagram of the structure of the bridge detection device based on remote sensing image water body provided by the present invention. Figure 3 As shown, the bridge detection device based on remote sensing image water body includes: a first extraction module 310, a second extraction module 320, a third extraction module 330 and a training and detection module 340.
[0080] A first extraction module 310 is used to obtain a water body mask of a target area from a time series remote sensing image, where the target area includes rivers and bridges; The second extraction module 320 is used to extract the skeleton information and breakpoint information of the river from the water body mask, and determine the bridge distribution information in the target area according to the skeleton information and breakpoint information; The third extraction module 330 is used to mark and extract bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data for training the bridge detection model; The training and detection module 340 is used to train the bridge detection model through bridge sample data to achieve bridge detection.
[0081] The remote sensing image water body based bridge detection device provided by the embodiment of the present invention obtains the water body mask of the target area from the time series remote sensing image, determines the bridge distribution information in the target area according to the skeleton information and the breakpoint information, and then labels and extracts the bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data, and uses this to train the bridge detection model to realize bridge detection, thereby improving the detection accuracy of the bridges in the remote sensing image.
[0082] Figure 4 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 4As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute a bridge detection method based on a remote sensing image water body, the method comprising: obtaining a water body mask of a target area from a time series remote sensing image, the target area including a river and a bridge; extracting skeleton information and breakpoint information of the river from the water body mask, and determining the bridge distribution information in the target area according to the skeleton information and the breakpoint information; annotating and extracting the bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data; and training a bridge detection model through the bridge sample data to achieve bridge detection.
[0083] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0084] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the bridge detection method based on remote sensing image water bodies provided by the above methods, and the method includes: obtaining a water body mask of a target area from a time series remote sensing image, wherein the target area includes rivers and bridges; extracting skeleton information and breakpoint information of the river from the water body mask, and determining the bridge distribution information in the target area based on the skeleton information and breakpoint information; marking and extracting bridges in the time series remote sensing image based on the bridge distribution information to obtain bridge sample data; and training a bridge detection model through the bridge sample data to realize bridge detection.
[0085] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the bridge detection method based on remote sensing image water bodies provided by the above-mentioned methods, the method comprising: obtaining a water body mask of a target area from a time series remote sensing image, wherein the target area includes rivers and bridges; extracting skeleton information and breakpoint information of the river from the water body mask, and determining the bridge distribution information in the target area based on the skeleton information and breakpoint information; marking and extracting bridges in the time series remote sensing image based on the bridge distribution information to obtain bridge sample data; and training a bridge detection model using the bridge sample data to achieve bridge detection.
[0086] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0087] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A bridge detection method based on remote sensing image water body, characterized in that: include: Obtaining a water body mask of a target area from a time series remote sensing image, wherein the target area includes a river and a bridge; Extracting skeleton information and breakpoint information of the river from the water body mask, and determining bridge distribution information in the target area according to the skeleton information and the breakpoint information; Annotating and extracting bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data; The bridge detection model is trained by using the bridge sample data to achieve bridge detection.
2. The bridge detection method based on remote sensing image water body according to claim 1 is characterized in that: The method of obtaining a water body mask of a target area from a time series remote sensing image comprises: Acquire target time series remote sensing image data, wherein the target time series remote sensing image data includes one of medium-resolution remote sensing images, long-time series remote sensing images, and medium-resolution long-time series remote sensing images; Dividing the target area into a plurality of sub-areas at different levels based on water body factors in the target time series remote sensing image data; The water body mask is obtained from the sub-areas of the multiple different levels according to a water body extraction rule preset in a priori knowledge base of the water body, wherein the water body extraction rule includes operations of preprocessing, water body index calculation and threshold setting.
3. The bridge detection method based on remote sensing image water body according to claim 1 is characterized in that: The target area includes a plurality of water body masks; After obtaining the water body mask of the target area from the time series remote sensing image, the method includes: The contour area of each water body mask is calculated, and at least one of the fragments and holes in the water body mask is repaired according to the contour area to obtain a new water body mask.
4. The bridge detection method based on remote sensing image water body according to claim 1 is characterized in that: The step of extracting the skeleton information and breakpoint information of the river from the water body mask comprises: extracting the skeleton information from the water body mask using a morphological image processing algorithm; A skeleton feature is extracted according to the skeleton information, and the breakpoint information is extracted from the water body mask according to the skeleton feature; wherein the skeleton feature is used to indicate that the skeleton in the skeleton information has a width of 1 pixel.
5. The bridge detection method based on remote sensing image water body according to claim 1 is characterized in that: The training of the bridge detection model by using the bridge sample data comprises: The target model is iteratively trained based on the bridge sample data to obtain the bridge detection model; wherein the target model includes at least one of a traditional machine learning model or a neural network model.
6. The bridge detection method based on remote sensing image water body according to claim 5 is characterized in that: After obtaining the bridge detection model, the method further includes: Based on the bridge detection model, bridge detection is performed on a single-view high-resolution image to obtain a bridge detection result.
7. A bridge detection device based on remote sensing image water body, characterized in that: include: A first extraction module is used to obtain a water body mask of a target area from a time series remote sensing image, wherein the target area includes a river and a bridge; A second extraction module is used to extract skeleton information and breakpoint information of the river from the water body mask, and determine the bridge distribution information in the target area according to the skeleton information and the breakpoint information; A third extraction module is used to mark and extract bridges in the time series remote sensing image according to the bridge distribution information to obtain bridge sample data to train a bridge detection model; The training and detection module is used to train the bridge detection model through the bridge sample data to realize bridge detection.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the bridge detection method based on remote sensing image water body as described in any one of claims 1 to 6 is implemented.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the bridge detection method based on remote sensing image water body as claimed in any one of claims 1 to 6 is implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the bridge detection method based on remote sensing image water body as claimed in any one of claims 1 to 6 is implemented.