A method and system for generating a directed lane centerline based on remote sensing images

Abstract

The application provides a kind of based on remote sensing image's directed lane center line generation method and system, method includes: the remote sensing image data is input into trained directed lane center line segmentation model, obtains the position prediction image and direction prediction image of lane center line;Line segment detection method is used to obtain line segment information from the position prediction image, the actual geographic coordinates corresponding to line segment are obtained using the coordinate information of remote sensing image data;The RGB data in the same place of the direction prediction image with the coordinate position of the line segment is extracted, and the coordinate difference (dx, dy) between the end and the beginning of the line segment is calculated according to the RGB data;According to the coordinate difference (dx, dy), the direction information of the current line segment is calculated;Iterate position prediction image and direction prediction image, according to the coordinate information and direction information of each line segment, adjust the head-tail connection relationship of each line segment, and obtain directed lane center line data.The application reduces the production cost of lane center line data.

Description

Technical Field

[0001] This invention relates to the field of high-precision mapping technology, and more specifically, to a method and system for generating directional lane centerlines based on remote sensing imagery. Background Technology

[0002] In the process of creating crowdsourced high-precision maps, lane centerline data is typically obtained by real-time scanning and image recognition using vehicle-mounted cameras, followed by a series of post-processing steps. This approach results in high production costs. Optical remote sensing data not only possesses advantages such as macroscopic scope, multi-source nature, real-time accuracy, and massive volume, but also enables detailed interpretation of target features with high spatial resolution. The extracted target information exhibits strict geographic coordinate transformation relationships. Therefore, how to apply optical remote sensing data to the generation process of lane centerline data to reduce the production cost of crowdsourced high-precision maps is a worthy research topic. Summary of the Invention

[0003] This invention addresses the technical problems existing in the prior art by providing a method and system for generating directional lane centerlines based on remote sensing imagery. It can automatically extract directional lane centerline data from remote sensing imagery data, thus solving the problem of high production costs for lane centerline data.

[0004] According to a first aspect of the present invention, a method for generating a directed lane centerline based on remote sensing imagery is provided, comprising:

[0005] Input remote sensing image data into a trained oriented lane centerline segmentation model to obtain the lane centerline position prediction image and direction prediction image.

[0006] Line segment information is obtained from the location prediction image using a line segment detection method, and the actual geographic coordinates corresponding to the line segment are obtained using the coordinate information of the remote sensing image data; RGB data of the same coordinate position as the line segment are extracted from the direction prediction image, and the coordinate difference (dx, dy) between the end and beginning of the line segment is calculated based on the RGB data; the direction information of the current line segment is calculated based on the coordinate difference (dx, dy).

[0007] Traverse the position prediction image and direction prediction image, adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment, and obtain the directed lane centerline data.

[0008] Based on the above technical solution, the present invention can also be improved as follows.

[0009] Optionally, the directed lane centerline segmentation model includes an encoder module, a dilated convolution module, and a decoder module connected in sequence. The output of the decoder module has two parallel output branches, one of which is used to output a position prediction image, and the other is used to output a direction prediction image.

[0010] Optionally, the training process of the directed lane centerline segmentation model includes:

[0011] Acquire and label the remote sensing image data dataset, use a sliding window of a preset size to crop the remote sensing image data and input it into the directed lane centerline segmentation model for training, and obtain the weight file of the directed lane centerline segmentation model.

[0012] By using backpropagation during the training process, the network weights are continuously changed to minimize the loss function, making the network output iteratively approach the true value, thus obtaining an optimized directed lane centerline segmentation model weight file. This weight file serves as the input to the directed lane centerline segmentation model.

[0013] Optionally, in the directed lane centerline segmentation model, a linear combination of cross-entropy loss and Dice loss is used as the loss function for segmenting position information, and L2 loss is used as the loss function for segmenting direction information.

[0014] Optionally, the remote sensing image dataset is labeled, including:

[0015] A buffer of a preset width is established for the true data of the lane centerline. The geographic coordinates of the buffer are converted into the pixel coordinates of the image data. The RGB values ​​of the area corresponding to the lane centerline are set to the first preset value, and the RGB values ​​of the other locations are set to the second preset value to obtain the location label image.

[0016] According to the size of the sliding window, each lane centerline is segmented, and the coordinate difference (dx, dy) between the end and the beginning of each segment and the total length l of the segment are calculated. Based on the coordinate difference (dx, dy) and the total length l of the segment, the RGB value of the lane centerline segment at the corresponding position is set to the third preset value, and the RGB value of the remaining positions is set to the fourth preset value, thus obtaining the direction label image.

[0017] Optionally, in the location label image, the first preset value is (a,a,a), and the second preset value is (b,b,b); in the direction label image, the third preset value is ((a+b) / 2+int(dx / l)*(a+b) / 2,(a+b) / 2+int(dy / l)*(a+b) / 2,(a+b) / 2), and the fourth preset value is ((a+b) / 2,(a+b) / 2,(a+b) / 2); where a and b are preset constants.

[0018] Optionally, the traversal of the position prediction image and direction prediction image, adjusting the start-end connection relationship of each line segment according to the coordinate and direction information of each line segment, to obtain directed lane centerline data, includes:

[0019] The location prediction image is traversed using a sliding window, and all line segment information is obtained from the location prediction image using a line segment detection method. The line segment information is then converted into corresponding actual geographic coordinates using the coordinate information of the remote sensing image data.

[0020] Based on the correspondence between the position prediction image and the direction prediction image, the RGB data corresponding to the position of the same line segment in the direction prediction image is extracted. The coordinate difference (dx, dy) between the end and beginning of the current line segment is calculated based on the RGB data of the same line segment. The direction information of the current line segment is restored based on the coordinate difference (dx, dy).

[0021] The connection between the beginning and end of each line segment is adjusted based on the actual geographical coordinates and direction information of each line segment to obtain the directional lane centerline data.

[0022] According to a second aspect of the present invention, a directional lane centerline generation system based on remote sensing imagery is provided, comprising:

[0023] The segmentation module is used to input remote sensing image data into a trained oriented lane centerline segmentation model to obtain the lane centerline position prediction image and direction prediction image.

[0024] The extraction module is used to obtain line segment information from the location prediction image using a line segment detection method, and to obtain the actual geographic coordinates corresponding to the line segment using the coordinate information of the remote sensing image data; it is also used to extract RGB data from the direction prediction image at the same coordinate position as the line segment, calculate the coordinate difference (dx, dy) between the end and beginning of the line segment based on the RGB data; and calculate the direction information of the current line segment based on the coordinate difference (dx, dy).

[0025] The connection module is used to traverse the position prediction image and the direction prediction image, and adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment to obtain the directed lane centerline data.

[0026] According to a third aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the processor is configured to execute a computer management program stored in the memory to implement the steps of the above-described method for generating a directional lane centerline based on remote sensing imagery.

[0027] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer management program is stored, wherein when executed by a processor, the computer management program implements the steps of the above-described method for generating a directed lane centerline based on remote sensing imagery.

[0028] This invention provides a method, system, electronic device, and storage medium for generating directional lane centerlines based on remote sensing imagery. It can directly generate directional lane centerline data using existing open-source remote sensing imagery data, effectively reducing the production cost of lane centerline data. Compared to existing solutions, its directional information extraction does not rely on trajectory GPS data, eliminates the need for actual vehicle data collection, reduces production complexity, and simplifies production steps. Furthermore, by using existing open-source remote sensing imagery data, data acquisition costs are low, and post-processing is minimal. Attached Figure Description

[0029] Figure 1 A flowchart of a method for generating a directed lane centerline based on remote sensing imagery provided by the present invention;

[0030] Figure 2 A flowchart of a method in a specific implementation scenario;

[0031] Figure 3 A schematic diagram of the composition of a directional lane centerline generation system based on remote sensing imagery provided by the present invention;

[0032] Figure 4 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0033] Figure 5 This is a schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. Detailed Implementation

[0034] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0035] In crowdsourced mapping, the physical meaning of the lane centerline is the geometric centerline between the lane edges. This spatial feature can be used to distinguish whether a pixel is on the lane centerline. At the same time, the directional arrows on the ground and the traffic rules implied in the road vehicle status can also be used to determine its directional information. Therefore, the above information can be obtained by image segmentation.

[0036] Figure 1 A flowchart of a method for generating directed lane centerlines based on remote sensing imagery provided by this invention is shown below. Figure 1 As shown, the method includes:

[0037] Input remote sensing image data into a trained oriented lane centerline segmentation model to obtain the lane centerline position prediction image and direction prediction image.

[0038] Line segment information is obtained from the location prediction image using a line segment detection method, and the actual geographic coordinates corresponding to the line segment are obtained using the coordinate information of the remote sensing image data; RGB data of the same coordinate position as the line segment are extracted from the direction prediction image, and the coordinate difference (dx, dy) between the end and beginning of the line segment is calculated based on the RGB data; the direction information of the current line segment is calculated based on the coordinate difference (dx, dy).

[0039] Traverse the position prediction image and direction prediction image, adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment, and obtain the directed lane centerline data.

[0040] Understandably, given the high production cost of lane centerline data in the background technology, and considering that optical remote sensing data possesses advantages such as macroscopicity, multi-source nature, realism, real-time performance, and massive volume, and can provide detailed interpretation of target features with high spatial resolution, with extracted target information exhibiting strict geographic coordinate transformation relationships, this invention proposes a directional lane centerline generation method based on remote sensing imagery. This method utilizes existing open-source remote sensing imagery data to directly generate directional lane centerline data, effectively reducing the production cost of lane centerline data. Compared to existing solutions, its directional information extraction does not rely on trajectory GPS data, eliminates the need for actual vehicle data collection, reduces production complexity, and simplifies production steps; using existing open-source remote sensing imagery data results in low data acquisition costs and fewer post-processing steps.

[0041] In one possible embodiment, the directed lane centerline segmentation model includes an encoder module, a dilated convolution module, and a decoder module connected in sequence. The output of the decoder module has two parallel output branches, one of which is used to output a position prediction image, and the other is used to output a direction prediction image.

[0042] Understandably, the model is based on a traditional UET structure, with the encoder performing downsampling and the decoder performing upsampling. For example... Figure 2 As shown, ResNet blocks 1-4 are used for encoding during each downsampling step to better extract feature information. At the same time, dilated convolutions (D-conv) are embedded between downsampling and upsampling to obtain a larger receptive field. In addition, two branches are used after the last decoder block 4 to perform position and orientation segmentation respectively, so that these two tasks can share most of the weights in the model.

[0043] In one possible embodiment, the training process of the directed lane centerline segmentation model includes:

[0044] Acquire and label the remote sensing image data dataset, use a sliding window of a preset size to crop the remote sensing image data and input it into the directed lane centerline segmentation model for training, and obtain the weight file of the directed lane centerline segmentation model.

[0045] By using backpropagation during the training process, the network weights are continuously changed to minimize the loss function, making the network output iteratively approach the true value, thus obtaining an optimized directed lane centerline segmentation model weight file. This weight file serves as the input to the directed lane centerline segmentation model.

[0046] Understandably, after the model is built, the training phase of the directed lane centerline segmentation model begins. The directed lane centerline segmentation model uses a sliding window of a preset size (e.g., 640*640) to capture image data as input, enabling it to better learn global information. After model training, the weight parameters at the convergence point of the loss value are saved, resulting in the final optimized directed lane centerline segmentation model, which can then be used in lane centerline data production.

[0047] In one possible embodiment, in the directed lane centerline segmentation model, the segmentation of position information uses a linear combination of cross-entropy loss and Dice loss as the loss function, and the segmentation of direction information uses L2 loss as the loss function.

[0048] In one possible embodiment, such as Figure 2 As shown, before training the model, training data needs to be prepared and preprocessed (e.g., labeled) to verify the model's accuracy in segmenting remote sensing image data. Labeling the remote sensing image dataset includes:

[0049] A buffer of preset width is established for the ground truth data of the lane centerline. The geographic coordinates of the buffer are converted into pixel coordinates of the image data. The RGB values ​​of the area corresponding to the lane centerline are set to the first preset value, and the RGB values ​​of the remaining locations are set to the second preset value to obtain the location label image. In this embodiment, the first preset value is set to (255,255,255), and the second preset value is set to (0,0,0). By binarizing the location label image, the positional features of the lane centerline in the remote sensing image data are segmented.

[0050] According to the size of the sliding window, each lane centerline is segmented. The coordinate difference (dx, dy) between the end and beginning of each segment and the total length l of the segment are calculated. Based on the coordinate difference (dx, dy) and the total length l of the segment, the RGB values ​​of the lane centerline segments at corresponding positions are set to a third preset value, and the RGB values ​​at other positions are set to a fourth preset value, thus obtaining the direction label image. In this embodiment, the RGB values ​​of the lane centerline segments in the direction label image at corresponding positions are set to (127 + int(dx / l)*127, 127 + int(dy / l)*127, 127), and the RGB values ​​at other positions are set to (127, 127, 127). Of course, the third and fourth preset values ​​can also be set to other suitable values ​​to make them clearly distinguishable from the RGB values ​​set in the position label image.

[0051] By establishing the positional correspondence between the lane centerlines in the direction label image and the lane centerlines in the location label image, and after obtaining the positional features of the lane centerlines in the location label image, the RGB values ​​of each segment of the lane centerlines in the direction label image can be extracted. This allows for the calculation of the coordinate difference (dx, dy) between the end and beginning of each segment. The coordinate difference (dx, dy) between the end and beginning of each segment can then be used to calculate the directional features of that segment.

[0052] In one possible embodiment, to clearly distinguish the RGB values ​​in the location label image and the direction label image, the location label image has a first preset value of (a,a,a) and a second preset value of (b,b,b); the direction label image has a third preset value of ((a+b) / 2+int(dx / l)*(a+b) / 2,(a+b) / 2+int(dy / l)*(a+b) / 2,(a+b) / 2), and a fourth preset value of ((a+b) / 2,(a+b) / 2,(a+b) / 2); where a and b are preset constants.

[0053] For example Figure 2As shown, if a is set to 255 and b is 0, then the RGB value of the area corresponding to the lane centerline in the location label image is (255, 255, 255), and the RGB value of the other locations is (0, 0, 0). In the direction label image, the RGB value of the lane centerline segments at the corresponding locations is (127 + int(dx / l) * 127, 127 + int(dy / l) * 127, 127), and the RGB value of the other locations is (127, 127, 127).

[0054] In one possible embodiment, the traversal of the position prediction image and direction prediction image, adjusting the start-end connection relationship of each line segment according to the coordinate and direction information of each line segment to obtain directed lane centerline data, includes:

[0055] The location prediction image is traversed using a sliding window. The line segment detection method in OpenCV is used to obtain all line segment information from the location prediction image. The line segment information is converted into corresponding actual geographic coordinates using the coordinate information of the remote sensing image data to obtain the location information of the line segments.

[0056] Based on the correspondence between the position prediction image and the direction prediction image, the RGB data corresponding to the position of the same line segment in the direction prediction image is extracted. The coordinate difference (dx, dy) between the end and beginning of the current line segment is calculated based on the RGB data of the same line segment. The direction information of the current line segment is restored based on the coordinate difference (dx, dy).

[0057] The connection between the beginning and end of each line segment is adjusted based on the actual geographical coordinates and direction information of each line segment to obtain continuous, directional lane centerline data.

[0058] Figure 3 A structural diagram of a directed lane centerline generation system based on remote sensing imagery provided in an embodiment of the present invention is shown below. Figure 3 As shown, a directed lane centerline generation system based on remote sensing imagery includes a segmentation module, an extraction module, and a connection module, wherein:

[0059] The segmentation module is used to input remote sensing image data into a trained oriented lane centerline segmentation model to obtain the lane centerline position prediction image and direction prediction image.

[0060] The extraction module is used to obtain line segment information from the location prediction image using a line segment detection method, and to obtain the actual geographic coordinates corresponding to the line segment using the coordinate information of the remote sensing image data; it is also used to extract RGB data from the direction prediction image at the same coordinate position as the line segment, calculate the coordinate difference (dx, dy) between the end and beginning of the line segment based on the RGB data; and calculate the direction information of the current line segment based on the coordinate difference (dx, dy).

[0061] The connection module is used to traverse the position prediction image and the direction prediction image, and adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment to obtain the directed lane centerline data.

[0062] It is understood that the directed lane centerline generation system based on remote sensing imagery provided by this invention corresponds to the directed lane centerline generation method based on remote sensing imagery provided in the foregoing embodiments. The relevant technical features of the directed lane centerline generation system based on remote sensing imagery can be referred to the relevant technical features of the directed lane centerline generation method based on remote sensing imagery, and will not be repeated here.

[0063] Please see Figure 4 , Figure 4 This is a schematic diagram illustrating an embodiment of the electronic device provided in this invention. For example... Figure 4 As shown, an embodiment of the present invention provides an electronic device 400, including a memory 410, a processor 420, and a computer program 411 stored in the memory 410 and executable on the processor 420. When the processor 420 executes the computer program 411, it performs the following steps:

[0064] Input remote sensing image data into a trained oriented lane centerline segmentation model to obtain the lane centerline position prediction image and direction prediction image.

[0065] Line segment information is obtained from the location prediction image using a line segment detection method, and the actual geographic coordinates corresponding to the line segment are obtained using the coordinate information of the remote sensing image data; RGB data of the same coordinate position as the line segment are extracted from the direction prediction image, and the coordinate difference (dx, dy) between the end and beginning of the line segment is calculated based on the RGB data; the direction information of the current line segment is calculated based on the coordinate difference (dx, dy).

[0066] Traverse the position prediction image and direction prediction image, adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment, and obtain the directed lane centerline data.

[0067] Please see Figure 5 , Figure 5 This is a schematic diagram illustrating an embodiment of a computer-readable storage medium provided by the present invention. (See diagram below.) Figure 5 As shown, this embodiment provides a computer-readable storage medium 500 on which a computer program 511 is stored. When the computer program 511 is executed by a processor, it performs the following steps:

[0068] Input remote sensing image data into a trained oriented lane centerline segmentation model to obtain the lane centerline position prediction image and direction prediction image.

[0069] Line segment information is obtained from the location prediction image using a line segment detection method, and the actual geographic coordinates corresponding to the line segment are obtained using the coordinate information of the remote sensing image data; RGB data of the same coordinate position as the line segment are extracted from the direction prediction image, and the coordinate difference (dx, dy) between the end and beginning of the line segment is calculated based on the RGB data; the direction information of the current line segment is calculated based on the coordinate difference (dx, dy).

[0070] Traverse the position prediction image and direction prediction image, adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment, and obtain the directed lane centerline data.

[0071] This invention provides a method, system, electronic device, and storage medium for generating directional lane centerlines based on remote sensing imagery. It can directly generate directional lane centerline data using existing open-source remote sensing imagery data, effectively reducing the production cost of lane centerline data. Compared to existing solutions, its directional information extraction does not rely on trajectory GPS data, eliminates the need for actual vehicle data collection, reduces production complexity, and simplifies production steps. Furthermore, by using existing open-source remote sensing imagery data, data acquisition costs are low, and post-processing is minimal.

[0072] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0073] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0078] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for generating directed lane centerlines based on remote sensing imagery, characterized in that, include: Acquire and label remote sensing image datasets, and then train a directed lane centerline segmentation model using the labeled remote sensing image datasets. The labeling of the remote sensing image datasets includes: A buffer of a preset width is established for the true data of the lane centerline. The geographic coordinates of the buffer are converted into the pixel coordinates of the image data. The RGB values ​​of the area corresponding to the lane centerline are set to the first preset value, and the RGB values ​​of the other locations are set to the second preset value to obtain the location label image. According to the size of the sliding window, each lane centerline is segmented, and the coordinate difference (dx, dy) between the end and the beginning of each segment and the total length l of the segment are calculated. Based on the coordinate difference (dx, dy) and the total length l of the segment, the RGB value of the lane centerline segment at the corresponding position is set to the third preset value, and the RGB value of the other positions is set to the fourth preset value, so as to obtain the direction label image. Input the remote sensing image data to be tested into the trained directional lane centerline segmentation model to obtain the position prediction image and direction prediction image of the lane centerline. Line segment information is obtained from the location prediction image using a line segment detection method, and the actual geographic coordinates corresponding to the line segment are obtained using the coordinate information of the remote sensing image data; RGB data of the same coordinate position as the line segment are extracted from the direction prediction image, and the coordinate difference (dx, dy) between the end and beginning of the line segment is calculated based on the RGB data; the direction information of the current line segment is calculated based on the coordinate difference (dx, dy). Traverse the position prediction image and direction prediction image, adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment, and obtain the directed lane centerline data.

2. The method for generating a directed lane centerline based on remote sensing imagery according to claim 1, characterized in that, The directed lane centerline segmentation model includes an encoder module, a dilated convolution module, and a decoder module connected in sequence. The output of the decoder module has two parallel output branches, one of which is used to output the position prediction image and the other is used to output the direction prediction image.

3. The method for generating a directed lane centerline based on remote sensing imagery according to claim 1, characterized in that, The training process of the directed lane centerline segmentation model includes: Based on the labeled remote sensing image data dataset, remote sensing image data is cropped with a sliding window of a preset size and input into the directed lane centerline segmentation model for training, thereby obtaining the weight file of the directed lane centerline segmentation model. By using backpropagation during the training process, the network weights are continuously changed to minimize the loss function, making the network output iteratively approach the true value, thus obtaining an optimized directed lane centerline segmentation model weight file. This weight file serves as the input to the directed lane centerline segmentation model.

4. The method for generating a directed lane centerline based on remote sensing imagery according to claim 3, characterized in that, In the directional lane centerline segmentation model, the segmentation of position information uses a linear combination of cross-entropy loss and Dice loss as the loss function, while the segmentation of direction information uses L2 loss as the loss function.

5. The method for generating a directed lane centerline based on remote sensing imagery according to claim 1, characterized in that, In the location label image, the first preset value is (a,a,a), and the second preset value is (b,b,b); in the direction label image, the third preset value is ((a+b) / 2+ int(dx / l)*(a+b) / 2,(a+b) / 2+ int(dy / l)*(a+b) / 2,(a+b) / 2), and the fourth preset value is ((a+b) / 2,(a+b) / 2,(a+b) / 2); where a and b are preset constants.

6. The method for generating a directed lane centerline based on remote sensing imagery according to claim 1, characterized in that, The traversal of the position prediction image and direction prediction image, based on the coordinate and direction information of each line segment, adjusts the connection relationship between the beginning and end of each line segment to obtain directed lane centerline data, including: The location prediction image is traversed using a sliding window, and all line segment information is obtained from the location prediction image using a line segment detection method. The line segment information is then converted into corresponding actual geographic coordinates using the coordinate information of the remote sensing image data. Based on the correspondence between the position prediction image and the direction prediction image, the RGB data corresponding to the position of the same line segment in the direction prediction image is extracted. The coordinate difference (dx, dy) between the end and beginning of the current line segment is calculated based on the RGB data of the same line segment. The direction information of the current line segment is restored based on the coordinate difference (dx, dy). The connection between the beginning and end of each line segment is adjusted based on the actual geographical coordinates and direction information of each line segment to obtain the directional lane centerline data.

7. A directional lane centerline generation system based on remote sensing imagery, characterized in that, include: The training module is used to acquire and annotate remote sensing image datasets, and then use these annotated datasets to train the directed lane centerline segmentation model. The annotation of the remote sensing image datasets includes: A buffer of a preset width is established for the true data of the lane centerline. The geographic coordinates of the buffer are converted into the pixel coordinates of the image data. The RGB values ​​of the area corresponding to the lane centerline are set to the first preset value, and the RGB values ​​of the other locations are set to the second preset value to obtain the location label image. According to the size of the sliding window, each lane centerline is segmented, and the coordinate difference (dx, dy) between the end and the beginning of each segment and the total length l of the segment are calculated. Based on the coordinate difference (dx, dy) and the total length l of the segment, the RGB value of the lane centerline segment at the corresponding position is set to the third preset value, and the RGB value of the other positions is set to the fourth preset value, so as to obtain the direction label image. The segmentation module is used to input remote sensing image data into a trained oriented lane centerline segmentation model to obtain the lane centerline position prediction image and direction prediction image. The extraction module is used to obtain line segment information from the location prediction image using a line segment detection method, and to obtain the actual geographic coordinates corresponding to the line segment using the coordinate information of the remote sensing image data; it is also used to extract RGB data from the direction prediction image at the same coordinate position as the line segment, calculate the coordinate difference (dx, dy) between the end and beginning of the line segment based on the RGB data; and calculate the direction information of the current line segment based on the coordinate difference (dx, dy). The connection module is used to traverse the position prediction image and the direction prediction image, and adjust the connection relationship between the beginning and end of each line segment according to the coordinate information and direction information of each line segment to obtain the directed lane centerline data.

8. An electronic device, characterized in that, The system includes a memory and a processor, wherein the processor is used to execute computer management programs stored in the memory to implement the steps of the method for generating a directed lane centerline based on remote sensing imagery as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, It stores a computer management program, which, when executed by a processor, implements the steps of a method for generating a directional lane centerline based on remote sensing imagery as described in any one of claims 1-6.

Images