Land space planning remote sensing image segmentation method, device and system
Through multi-angle image matching and restoration technology, the problem of complex and low accuracy of traditional distortion repair methods is solved, and high-quality segmentation and analysis of remote sensing images in land space planning is realized.
Patent Information
- Application Number
- CN202510517948.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-24
- Publication Date
- 2025-05-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The distortion repair method of traditional land space planning remote sensing images is complex and has low accuracy, resulting in low image quality during subsequent image segmentation, affecting the accuracy of the analysis.
By obtaining the remote sensing images of land space planning at multiple shooting angles, determining the distorted block and its stable distribution coefficient, performing grayscale matching and distance angle matching, calculating the recovery joint coefficient, selecting the target block with the largest recovery joint coefficient for restoration, and finally segmenting the restored image according to the set segmentation strategy.
It improves the segmentation accuracy and reliability of remote sensing images in land space planning, ensures the improvement of image quality, and thus improves the accuracy of subsequent analysis.
Smart Images
Figure CN120047472A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and particularly relates to a method, device and system for segmenting remote sensing images of territorial spatial planning. Background Art
[0002] In the process of using an imaging device to capture remote sensing images of territorial spatial planning of a target area, since the imaging device often captures images at a relatively high position, when capturing a large range of ground areas, obvious bending will occur due to the curvature of the ground, resulting in perspective distortion. At the same time, since remote sensing images of territorial spatial planning mostly use wide-angle lenses for shooting, the perspective effect of the images will become extreme during shooting, straight lines may become curved, and the shapes and proportions of objects will also be distorted. In addition, when shooting at high altitudes, due to factors such as air flow, the working stability of the shooting device is affected, resulting in motion distortion in the imaging.
[0003] In related technologies, usually, an isometric model is used to calibrate the distorted remote sensing images of territorial spatial planning. However, this operation method has a large amount of data calculation, and there are non-calculable points at a 180° field of view angle, making it difficult to effectively repair the distorted images. In the subsequent image segmentation process, if the distorted images are segmented, it is easy to have the problem of low quality of the segmented images, thereby affecting the accuracy of the subsequent analysis of remote sensing images of territorial spatial planning. Summary of the Invention
[0004] In order to solve the technical problems that the traditional distortion repair scheme is complex in operation and low in accuracy, resulting in low quality of the subsequent segmented images, the purpose of the present invention is to provide a method, device and system for segmenting remote sensing images of territorial spatial planning. The specific technical solutions adopted are as follows: A method for segmenting remote sensing images of territorial spatial planning, the method comprising: Obtaining remote sensing images of territorial spatial planning at multiple shooting angles; Determining the distorted blocks in each remote sensing image of territorial spatial planning and the stable distribution coefficient of the distorted blocks; Performing gray-scale matching and angle-distance matching on each distorted block at each shooting angle with the distorted blocks corresponding to the spatial positions at the remaining shooting angles, to obtain the gray-scale matching coefficient and the angle-distance matching coefficient between each distorted block and the corresponding distorted blocks at the remaining shooting angles; Based on the stable distribution coefficient, the gray-scale matching coefficient, and the angle-distance matching coefficient, calculating the restoration joint coefficient between each distorted block and the corresponding distorted blocks at the remaining shooting angles; For each shooting angle, determine the target block with the largest restoration joint coefficient in the remaining shooting angles corresponding to each distortion block, and restore the corresponding distortion block based on the target block to generate a restored image for each shooting angle; Segment the restored image for each shooting angle according to the set segmentation strategy.
[0005] According to a remote sensing image segmentation method for territorial spatial planning provided by the present invention, determining the distortion blocks in each remote sensing image for territorial spatial planning includes: Segment each remote sensing image for territorial spatial planning according to a preset size to obtain a plurality of image blocks corresponding to each remote sensing image for territorial spatial planning; Extract the distortion blocks from the plurality of image blocks corresponding to each remote sensing image for territorial spatial planning.
[0006] According to a remote sensing image segmentation method for territorial spatial planning provided by the present invention, extracting the distortion blocks from the plurality of image blocks corresponding to each remote sensing image for territorial spatial planning includes: Determine the stable distribution coefficient corresponding to each image block; Use the image blocks in each remote sensing image for territorial spatial planning with the stable distribution coefficient higher than the preset stable distribution threshold as the distortion blocks.
[0007] According to a remote sensing image segmentation method for territorial spatial planning provided by the present invention, determining the stable distribution coefficient corresponding to each image block includes: Determine the pixel point distribution parameter corresponding to each pixel gray value in each image block, the number of types of pixel gray values in each image block, and the mean value of the distribution parameters; Subtract the pixel point distribution parameter corresponding to each pixel gray value from the mean value of the distribution parameters and square the result to obtain the pixel difference square value corresponding to each pixel gray value; Average the pixel difference square values corresponding to various pixel gray values in each image block to obtain the stable distribution coefficient corresponding to each image block.
[0008] According to a remote sensing image segmentation method for territorial spatial planning provided by the present invention, perform gray value matching on each distortion block at each shooting angle with the distortion blocks corresponding to the spatial positions in the remaining shooting angles, including: Extract two distortion blocks corresponding to the spatial positions from any two remote sensing images for territorial spatial planning with a local overlapping relationship at different shooting angles, and establish a plurality of block matching groups; Calculate the gray value matching coefficient between the two distortion blocks in each block matching group respectively.
[0009] A method for remote sensing image segmentation of territorial space planning provided by the present invention calculates the gray-scale matching coefficients between two distorted blocks in each block matching group respectively, including: Determine the gray-scale mean square error between two distorted blocks in each block matching group and the cosine similarity in the direction of the maximum gray-scale gradient; Divide the gray-scale mean square error by the cosine similarity to calculate the gray-scale matching coefficients between two distorted blocks in each block matching group.
[0010] A method for remote sensing image segmentation of territorial space planning provided by the present invention performs distance-angle matching on each distorted block at each shooting angle with the distorted blocks corresponding to the spatial positions at the remaining shooting angles respectively, including: Take each distorted block as the block to be matched, and take the distorted blocks corresponding to the spatial positions of the block to be matched at the remaining shooting angles as the auxiliary matching blocks; Obtain the corner points in the block to be matched and the auxiliary matching blocks respectively; Extract two first target corner points in the block to be matched, calculate the included angle between the line connecting the two first target corner points and the horizontal line, and obtain the connection included angle; Determine the target angle that is opposite to the connection included angle, and determine the angle parameter value based on the target angle; Determine two second target corner points corresponding to the two first target corner points in each auxiliary matching block respectively; Calculate the Euclidean distance between the two second target corner points in each auxiliary matching block respectively, and take the smallest Euclidean distance as the minimum Euclidean distance; Multiply the minimum Euclidean distance by the angle parameter value to obtain the distance-angle matching coefficients between each distorted block and the corresponding distorted blocks at the remaining shooting angles.
[0011] A method for remote sensing image segmentation of territorial space planning provided by the present invention calculates the restoration joint coefficients between each distorted block and the corresponding distorted blocks at the remaining shooting angles based on the stable distribution coefficient, the gray-scale matching coefficient, and the distance-angle matching coefficient, including: Multiply the gray-scale matching coefficient by the distance-angle matching coefficient to obtain an intermediate coefficient value; Divide the stable distribution coefficient by the intermediate coefficient value to obtain the restoration joint coefficient.
[0012] On the other hand, the present invention also provides a device for remote sensing image segmentation of territorial space planning. The device includes a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the program, it implements any one of the above-mentioned methods for remote sensing image segmentation of territorial space planning.
[0013] On the other hand, the present invention also provides a remote sensing image segmentation system for territorial spatial planning, and the system includes: an imaging device and the above-mentioned remote sensing image segmentation device for territorial spatial planning; The imaging device is connected to the remote sensing image segmentation device for territorial spatial planning. The imaging device is configured to capture remote sensing images of the territorial spatial planning of a target area at multiple shooting angles and send the remote sensing images of the territorial spatial planning at multiple shooting angles to the remote sensing image segmentation device for territorial spatial planning.
[0014] The present invention has the following beneficial effects: By obtaining remote sensing images of territorial spatial planning at multiple shooting angles, determining the distortion blocks in each remote sensing image of territorial spatial planning and the stable distribution coefficients of the distortion blocks, performing gray-level matching and distance-angle matching on each distortion block at each shooting angle with the distortion blocks corresponding to the spatial positions at the remaining shooting angles, obtaining the gray-level matching coefficients and distance-angle matching coefficients between each distortion block and the corresponding distortion blocks at the remaining shooting angles, and calculating the restoration joint coefficients between each distortion block and the corresponding distortion blocks at the remaining shooting angles based on the stable distribution coefficients, gray-level matching coefficients, and distance-angle matching coefficients. Subsequently, respectively determine the target blocks with the largest restoration joint coefficients at the remaining shooting angles corresponding to each distortion block at each shooting angle, restore the corresponding distortion blocks based on the target blocks, generate the restored images at each shooting angle, and finally segment the restored images at each shooting angle according to the set segmentation strategy. Since the restoration joint coefficients at the remaining shooting angles corresponding to each distortion block at each shooting angle are obtained by calculating the stable distribution coefficients, gray-level matching coefficients, and distance-angle matching coefficients before image segmentation, and then the corresponding distortion blocks are restored based on the target blocks with the largest restoration joint coefficients, each remote sensing image of territorial spatial planning can be gradually restored. Subsequently, segmenting the restored images can obtain a segmentation result with higher accuracy and better image quality, thereby improving the segmentation accuracy and reliability of the remote sensing images of territorial spatial planning. Description of the Drawings
[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.
[0016] Figure 1 It is a flowchart of a method for a remote sensing image segmentation method for territorial spatial planning provided by an embodiment of the present invention; Figure 2 A remote sensing image of territorial space planning at a shooting angle; Figure 3 Another remote sensing image of territorial space planning that has a partial overlapping relationship with the Figure 2 image shown at another shooting angle; Figure 4 The structural schematic diagram of a device for segmenting remote sensing images of territorial space planning provided by an embodiment of the present invention; Figure 5 The structural schematic diagram of a system for segmenting remote sensing images of territorial space planning provided by an embodiment of the present invention. Detailed implementation manners
[0017] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following combines the accompanying drawings and preferred embodiments to specifically describe a method, device, and system for segmenting remote sensing images of territorial space planning proposed according to the present invention, including their specific implementation manners, structures, features, and effects, as follows. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.
[0019] The following combines the attached Figures 1 to 5 Specifically describe the specific solutions of a method, device, and system for segmenting remote sensing images of territorial space planning provided by the present invention.
[0020] Please refer to Figure 1 , which shows the flowchart of a method for segmenting remote sensing images of territorial space planning provided by an embodiment of the present invention. As Figure 1 shown, the above-mentioned method for segmenting remote sensing images of territorial space planning specifically includes: Step 110: Obtain remote sensing images of territorial space planning at multiple shooting angles.
[0021] In practical applications, relevant image acquisition devices can be used to collect remote sensing images of territorial space planning. For example, the image acquisition device carried by a remote sensing platform can be used to collect remote sensing images of territorial space planning within the target range. In this embodiment, in order to reduce the comparison error, when obtaining remote sensing images of territorial space planning at any position, it is necessary to maintain shooting for a certain period of time as much as possible, and keep the pixel ratio and light input of the captured remote sensing images consistent.
[0022] In some embodiments, after obtaining the original remote sensing image of the territorial spatial planning, preprocessing operations such as denoising and grayscale conversion can be performed on the original remote sensing image of the territorial spatial planning to obtain the preprocessed remote sensing image of the territorial spatial planning, thereby further improving the image quality of the remote sensing image of the territorial spatial planning.
[0023] Step 120: Determine the distorted blocks in each remote sensing image of the territorial spatial planning and the stable distribution coefficient of the distorted blocks.
[0024] In this embodiment, the distorted block refers to the image block with different degrees of distortion in the remote sensing image of the territorial spatial planning. The stable distribution coefficient can characterize the gradual change degree of the image color distribution in the distorted block. The larger the stable distribution coefficient, the more gradual the image color distribution in the distorted block, and further the higher the possibility of perspective distortion in the distorted block.
[0025] Step 130: Perform grayscale matching and angle-distance matching on each distorted block at each shooting angle with the distorted blocks corresponding to the spatial positions at the remaining shooting angles respectively, to obtain the grayscale matching coefficient and the angle-distance matching coefficient between each distorted block and the corresponding distorted blocks at the remaining shooting angles.
[0026] It can be understood that the grayscale matching coefficient can characterize the similarity degree of the pixel grayscale value features between a certain distorted block and the corresponding distorted blocks at the remaining shooting angles. When the grayscale matching coefficient is smaller, it indicates that the similarity degree of the pixel grayscale value features between the distorted block and the corresponding distorted blocks at the remaining shooting angles is higher.
[0027] The angle-distance matching coefficient can characterize the possibility that the shooting angles are diagonal to each other and the internal stretching ratios are close for a certain distorted block, that is, the contribution degree to the restoration of a certain distorted block.
[0028] Step 140: Based on the stable distribution coefficient, the grayscale matching coefficient, and the angle-distance matching coefficient, calculate the restoration joint coefficient between each distorted block and the corresponding distorted blocks at the remaining shooting angles.
[0029] In this embodiment, the restoration joint coefficient can characterize the possibility of restoring the original image area based on the combination of the current distorted block and a certain corresponding distorted block at the remaining shooting angles. The larger the restoration joint coefficient, the greater the possibility of restoring the original image area.
[0030] Step 150: Respectively determine the target blocks with the largest restoration joint coefficients at the remaining shooting angles corresponding to each distorted block at each shooting angle, and restore the corresponding distorted blocks based on the target blocks to generate the restored images at each shooting angle.
[0031] It can be understood that in this embodiment, the distorted block with the largest restoration joint coefficient is taken as the target block, and the corresponding distorted block is restored, so that the corresponding image area before distortion can be restored to the greatest extent. After performing the restoration process on each distorted block in the remote sensing image of the territorial spatial planning according to the above similar operations, the restored image can be obtained.
[0032] In some embodiments, specifically, the distorted block can be restored by geometric inverse transformation, and preprocessing operations such as filtering and denoising can be performed again after the restoration operation to ensure better quality of the restored image.
[0033] Step 160: Segment the restored image at each shooting angle according to the set segmentation strategy.
[0034] It can be understood that the set segmentation strategy can represent the user's segmentation requirements for the restored image. For example, the set segmentation strategy can be a to-be-segmented area pre-input by the user or a to-be-segmented line pre-set.
[0035] The solution provided in this embodiment can accurately and reliably restore each distorted block according to the calculated restoration joint coefficient, and finally segment the restored image, which can make the quality of the segmented image better.
[0036] In one embodiment, determining the distorted blocks in each remote sensing image of the territorial spatial planning specifically includes: First, segment the remote sensing image of the territorial spatial planning at each shooting angle according to a preset size to obtain multiple image blocks corresponding to each remote sensing image of the territorial spatial planning.
[0037] In this embodiment, the preset size can be a preset segmentation area, such as the pixel size. Specifically, the remote sensing image of the territorial spatial planning can be segmented from left to right according to the pixel size, and at this time, image blocks of the same size can be obtained.
[0038] It should be noted that if the edge part cannot form an image block of pixel size during the image segmentation process, the zero-padding algorithm can be used to perform zero-padding operations along the edge to ensure that each image block is pixel size. In this way, the remote sensing image of the territorial spatial planning at the shooting angle can be expressed as: (1) Among them, represents the remote sensing image of the territorial spatial planning at the shooting angle , , represents the first image block, represents the th image block, .
[0039] Then, distortion blocks are extracted from multiple image blocks corresponding to each remote sensing image of the territorial spatial planning.
[0040] In a specific implementation, extracting distortion blocks from multiple image blocks corresponding to each remote sensing image of the territorial spatial planning specifically includes: First, determine the stable distribution coefficient corresponding to each image block.
[0041] In this embodiment, determining the stable distribution coefficient corresponding to each image block specifically includes: First, determine the pixel point distribution parameter corresponding to each pixel gray value in each image block, the number of types of pixel gray values in each image block, and the mean value of the distribution parameters corresponding to each image block.
[0042] In a specific implementation, determining the pixel point distribution parameter corresponding to each pixel gray value in each image block specifically includes: In the first step, obtain the pixel gray values of the 8 neighboring pixel points around each pixel point corresponding to each pixel gray value.
[0043] In the second step, calculate the absolute value of the difference between the pixel gray value of each pixel point and the pixel gray value of each neighboring pixel point among the corresponding 8 neighboring pixel points.
[0044] In the third step, calculate the mean value of the absolute values of the differences obtained for the 8 neighboring pixel points corresponding to each pixel point to obtain the mean value of the neighboring pixel differences of each pixel point.
[0045] In the fourth step, further calculate the mean value of the mean values of the neighboring pixel differences of all pixel points under each pixel gray value to obtain the pixel point distribution parameter corresponding to each pixel gray value.
[0046] In this embodiment, for the image block the th pixel gray value (2) In the formula, represents the pixel point distribution parameter corresponding to the pixel gray value in the image block , represents the th pixel point among all pixel points under the pixel gray value Represents the pixel gray value The total number of the following pixel points, Represents the gray value of the th neighborhood pixel point among the 8 neighborhood pixel points corresponding to the
[0047] In this embodiment, the distribution parameter mean value corresponding to the image block can be expressed as follows: (3) In the formula, represents the distribution parameter mean value corresponding to the image block , represents the image block in the th kind of pixel gray value corresponding pixel point distribution parameter, represents the number of types of pixel gray values in the image block .
[0048] Then, the difference between the pixel point distribution parameter corresponding to each kind of pixel gray value and the distribution parameter mean value is taken and squared to obtain the pixel difference square value corresponding to each kind of pixel gray value.
[0049] Finally, the mean value of the pixel difference square values corresponding to each pixel gray value in each image block is obtained to obtain the stable distribution coefficient corresponding to each image block.
[0050] In this embodiment, the stable distribution coefficient corresponding to the image block can be expressed as follows: (4) In the formula, represents the stable distribution coefficient corresponding to the image block , represents the image block in the th kind of pixel gray value corresponding pixel point distribution parameter, represents the distribution parameter mean value corresponding to the image block , represents the image block in the number of types of pixel gray values.
[0051] Then, the image blocks in each remote sensing image of territorial spatial planning with a stable distribution coefficient higher than the preset stable distribution threshold are used as distorted blocks.
[0052] In the image block , the pixel gray value The corresponding pixel distribution parameters The size determines whether it is a single color reference. When the stable distribution coefficient is larger, the color distribution of the image is more gradual, indicating that the image block is more prone to perspective distortion. Therefore, in this embodiment, the image block with a stable distribution coefficient higher than the preset stable distribution threshold is used as the distortion block.
[0053] In one embodiment, for each distortion block at each shooting angle, gray-scale matching is performed with the distortion blocks corresponding to the spatial positions at the remaining shooting angles, specifically including: First, two distortion blocks corresponding to spatial positions are extracted from any two national territorial space planning remote sensing images with a local overlapping relationship at different shooting angles, and multiple block matching groups are established.
[0054] In this embodiment, the two distortion blocks in the block matching group are respectively taken from two national territorial space planning remote sensing images with a local overlapping relationship at different shooting angles, and the spatial positions of the two distortion blocks in the block matching group correspond. Figure 2 and Figure 3 Exemplarily shows two national territorial space planning remote sensing images with a local overlapping relationship at two different shooting angles, where the area circled by the rectangular wireframe is the local overlapping area in the two images.
[0055] Then, the gray-scale matching coefficients between the two distortion blocks in each block matching group are calculated respectively.
[0056] In a specific implementation, calculating the gray-scale matching coefficients between the two distortion blocks in each block matching group specifically includes: First, determine the gray-scale mean square error and the cosine similarity in the direction of the maximum gray-scale gradient between the two distortion blocks in each block matching group.
[0057] Then, divide the gray-scale mean square error by the cosine similarity to calculate the gray-scale matching coefficients between the two distortion blocks in each block matching group.
[0058] In this embodiment, taking the distortion block in the national territorial space planning remote sensing image and the distortion block in the national territorial space planning remote sensing image as an example, the gray-scale matching coefficient between the two distortion blocks can be expressed as follows: (5) In the formula,[[]] represents the gray-scale matching coefficient between the distortion block and the distortion block , represents the distortion block with the distorted block of the mean squared error of grayscale, indicating the distorted block and the distorted block in the maximum grayscale gradient direction and the cosine similarity on.
[0059] It can be understood that in the distorted block and the distorted block if there are the same image elements, then the grayscale value features of the two distorted blocks are similar, so the distorted block and the distorted block of the mean squared error of grayscale the smaller, due to the continuous change of the shooting angle, even if the two distorted blocks describe the same real area, but their grayscale values are still different, but the gradient direction of the grayscale values between the two is probably the same as the gradient direction of the overall object grayscale value, so the cosine similarity the larger. When the grayscale matching coefficient is smaller, it indicates that the two distorted blocks are more similar, and the possibility that they describe the same regional element is higher.
[0060] In an ideal shooting situation, the remotely sensed image of the territorial space planning obtained should not have any problems such as distortion and color difference, and the ratio of each line in the remotely sensed image of the territorial space planning is also normal. When distortion occurs, in the distorted block, when any two points are selected and compared with the original image, there are differences in the length or angle of the line connecting the two points, and due to different shooting angles, the angle differences are different in different images, but although the upper limit of the stretched line segment length is uncontrollable, its lower limit is controlled by the original image. When the stretched line segment distance approaches the original length, it can be considered that the area where this line segment is located has not been stretched, and in the same image block that is stretched and the stretching directions are exactly opposite, these two image blocks can be used to restore the distorted block at the corresponding spatial position in the original image. Accordingly, in this embodiment, each distorted block at each shooting angle is respectively subjected to distance-angle matching with the distorted blocks corresponding to the spatial positions at the remaining shooting angles.
[0061] In one embodiment, each distorted block at each shooting angle is respectively subjected to distance-angle matching with the distorted blocks corresponding to the spatial positions at the remaining shooting angles, which specifically includes: The first step is to use each distorted block as a block to be matched, and use the distorted blocks corresponding to the spatial positions of the block to be matched at the remaining shooting angles as auxiliary matching blocks.
[0062] It can be understood that for a certain block to be matched, there are multiple auxiliary matching blocks corresponding to its spatial position.
[0063] In the second step, obtain the corner points in the block to be matched and the auxiliary matching block respectively.
[0064] In this embodiment, the corner points in all distorted blocks can be obtained by using a corner point detection algorithm. In practical applications, usually the points on the edge of the distorted block or other points with obvious features are selected as corner points to represent the main image information of the distorted block. Therefore, in this embodiment, multiple representative corner points can be selected in the same distorted block.
[0065] In some embodiments, the corner point detection algorithm can select relatively mature algorithms such as Harris corner point detection algorithm, FAST corner point detection algorithm, and ORB corner point detection algorithm.
[0066] In the third step, extract two first target corner points in the block to be matched, calculate the included angle between the line connecting the two first target corner points and the horizontal line, and obtain the connection angle.
[0067] In this embodiment, for the distorted block the connection angle formed by the line connecting any two first target corner points and the horizontal line can be calculated as follows: (6) In the formula, represents the connection angle formed by the line connecting any two first target corner points and the horizontal line in the distorted block , and both represent any two of the multiple corner points in the distorted block , , , and respectively represent the horizontal and vertical coordinates of the two first target corner points and .
[0068] In the fourth step, determine the target angle that is diagonal to the connection angle, and determine the angle parameter value based on the target angle.
[0069] It can be understood that adding 180 degrees to a certain angle, the angle difference becomes the opposite direction within the period, and the final result will be close to zero or a symmetric state, which is a manifestation of "cancellation". This is also the case in this embodiment. Therefore, it is necessary to find the angle parameter value that is closest to the target angle that is diagonal to the connection angle .
[0070] In this embodiment, based on the above connection angle , the angle parameter value can be expressed as follows: (7) In the formula, represents the angular parameter value, represents the target angle that is diagonal to the included angle of the connection line, ( ) represents the proximity judgment function.
[0071] It can be understood that the proximity judgment function is mainly used to judge whether the difference between two numerical values is within the allowable error range. First, two input numerical values to be compared need to be obtained, then the allowable error range is set, and then the difference between the two input numerical values is calculated. Finally, it is judged whether the calculated difference is within the allowable error range. If the judgment result is yes, it means that the two input numerical values are close; otherwise, it can be determined that the two input numerical values are not close. In this embodiment, the proximity judgment function is mainly used to determine the angular parameter value that is closest to the target angle that is diagonal to the included angle of the connection line.
[0072] Step 5: Determine the two second target corner points corresponding to the two first target corner points in each auxiliary matching block.
[0073] In this embodiment, the two second target corner points corresponding to the two first target corner points and in any one auxiliary matching module can be expressed as and .
[0074] Step 6: Calculate the corner point Euclidean distance between the two second target corner points in each auxiliary matching block respectively, and take the smallest corner point Euclidean distance as the minimum Euclidean distance.
[0075] In this embodiment, the minimum Euclidean distance can be calculated as follows: (8) In the formula, represents the minimum Euclidean distance, and represent the two second target corner points, represents the corner point Euclidean distance between the two second target corner points and , ( ) represents the minimum value function.
[0076] It can be understood that in the distortion blocks corresponding to the same spatial position under different shooting angles, as the shooting angle changes, the degree of distortion will also change accordingly. However, no matter how much perspective distortion occurs, the length will not be less than the original image length. Therefore, the smaller the corner point Euclidean distance between the two second target corner points, the closer it is to the original line segment before distortion.
[0077] In the seventh step, multiply the minimum Euclidean distance by the angle parameter value to obtain the distance-angle matching coefficient between each distorted block and the corresponding distorted block at the remaining shooting angles.
[0078] In this embodiment, the distance-angle matching coefficient can be calculated as follows: (9) In the formula, represents the distance-angle matching coefficient between the distorted block and the distorted block at any one of the remaining angles, represents the minimum Euclidean distance between two second target corner points and and represents the angle parameter value.
[0079] Through the above operations, the auxiliary matching block with the angle closest to the diagonal and the smallest internal stretching ratio with respect to the distorted block can be obtained using the distance-angle matching coefficient .
[0080] In one embodiment, based on the stable distribution coefficient, the gray-scale matching coefficient, and the distance-angle matching coefficient, the restoration joint coefficient between each distorted block and the corresponding distorted block at the remaining shooting angles is calculated, specifically including: First, multiply the gray-scale matching coefficient by the distance-angle matching coefficient to obtain an intermediate coefficient value; Then, divide the stable distribution coefficient by the intermediate coefficient value to obtain the restoration joint coefficient.
[0081] In this embodiment, the maximum restoration joint coefficient corresponding to the distorted block can be specifically calculated as follows: (10) In the formula, represents the maximum restoration joint coefficient corresponding to the distorted block , represents the stable distribution coefficient of the distorted block , represents the distance-angle matching coefficient between the distorted block and the distorted block at any one of the remaining angles, represents the gray-scale matching coefficient between the distorted block and the distorted block at any one of the remaining angles, ( ) represents the maximum value function.
[0082] In practical applications, in this embodiment, the distorted block corresponding to the maximum restoration combined coefficient is selected as the target block, and the distorted block at the corresponding spatial position under the current shooting angle is restored through the target block. Specifically, the distorted block at the corresponding spatial position under the current shooting angle can be restored by means of geometric inverse transformation.
[0083] After each distorted block is restored, preprocessing operations such as filtering and denoising can be performed again in the restored area to ensure the image quality of each image block after the restoration operation. After all the distorted blocks in the remote sensing image of the territorial spatial planning are restored and preprocessed, the restored image can be obtained, and then the restored image can be segmented as required.
[0084] Based on the same general inventive concept, the present invention also protects a device and system for segmenting remote sensing images of territorial spatial planning. The device and system for segmenting remote sensing images of territorial spatial planning provided by the present invention will be described below. The device and system for segmenting remote sensing images of territorial spatial planning described below can be correspondingly referred to the method for segmenting remote sensing images of territorial spatial planning described above.
[0085] Please refer to Figure 4 , which shows a schematic structural diagram of a device for segmenting remote sensing images of territorial spatial planning provided by an embodiment of the present invention. As Figure 4 shown, the above-mentioned device for segmenting remote sensing images of territorial spatial planning specifically includes: a processor 210, a communication interface 220, a memory 230, and a communication bus 240. Among them, the processor 210, the communication interface 220, and the memory 230 complete communication with each other through the communication bus 240. The processor 210 can call the logical instructions in the memory 230 to execute the method for segmenting remote sensing images of territorial spatial planning provided by each of the above embodiments.
[0086] In addition, when the logical instructions in the above-mentioned memory 230 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.
[0087] Please refer to Figure 5 , which shows a schematic structural diagram of a remote sensing image segmentation system for territorial spatial planning provided by an embodiment of the present invention. As Figure 5 shown, the above-mentioned remote sensing image segmentation system for territorial spatial planning specifically includes: an image acquisition device 310 and the remote sensing image segmentation device 320 for territorial spatial planning provided in each of the above embodiments.
[0088] The image acquisition device 310 is connected to the remote sensing image segmentation device 320 for territorial spatial planning. The image acquisition device 310 is used to capture remote sensing images of the target area for territorial spatial planning at multiple shooting angles and send the remote sensing images of the target area for territorial spatial planning at multiple shooting angles to the remote sensing image segmentation device 320 for territorial spatial planning.
[0089] In this embodiment, the image acquisition device can be an image acquisition device mounted on platforms such as satellites, airplanes, or drones, and can capture remote sensing images of the target area for territorial spatial planning at multiple shooting angles.
[0090] For the remote sensing image segmentation device and system for territorial spatial planning provided in this embodiment, since the stable distribution coefficient, gray-level matching coefficient, and distance-angle matching coefficient are calculated before image segmentation to obtain the restoration joint coefficient corresponding to each distortion block at each shooting angle under the remaining shooting angles, and then the corresponding distortion block is restored based on the target block with the largest restoration joint coefficient, each remote sensing image for territorial spatial planning can be gradually restored. Subsequently, the restored images are segmented, and a segmentation result with higher accuracy and better image quality can be obtained, thereby improving the segmentation accuracy and reliability of the remote sensing images for territorial spatial planning.
[0091] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0092] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.
Claims
1. A method for segmenting remote sensing images for land space planning, characterized in that: The method comprises: Acquire remote sensing images of national land space planning from multiple shooting angles; Determine the distorted blocks in each national land space planning remote sensing image and the stable distribution coefficient of the distorted blocks; Grayscale matching and angular matching are performed on each distorted block at each shooting angle and the distorted blocks corresponding to the spatial positions at other shooting angles, so as to obtain the grayscale matching coefficient and angular matching coefficient between each distorted block and the corresponding distorted blocks at other shooting angles; Based on the stable distribution coefficient, the grayscale matching coefficient and the distance angle matching coefficient, a restoration joint coefficient between each distorted block and corresponding distorted blocks at other shooting angles is calculated; The method for obtaining the stable distribution coefficient is as follows: determining the pixel distribution parameters corresponding to each pixel grayscale value in each image block, the number of types of pixel grayscale values in each image block, and the mean value of the distribution parameters; subtracting the pixel distribution parameters corresponding to each pixel grayscale value from the mean value of the distribution parameters and squaring them to obtain the pixel difference square value corresponding to each pixel grayscale value; averaging the pixel difference square values corresponding to various pixel grayscale values in each image block to obtain the stable distribution coefficient corresponding to each image block; Determine the target blocks with the largest restoration joint coefficients at other shooting angles corresponding to each distorted block at each shooting angle, restore the corresponding distorted block based on the target blocks, and generate a restored image at each shooting angle; According to the set segmentation strategy, the restored image at each shooting angle is segmented.
2. The method for segmenting remote sensing images for land space planning according to claim 1, characterized in that: Determine the distorted blocks in each national land space planning remote sensing image, including: The land space planning remote sensing image at each shooting angle is segmented according to a preset size to obtain a plurality of image blocks corresponding to each land space planning remote sensing image; The distorted blocks are extracted from multiple image blocks corresponding to each national land space planning remote sensing image.
3. The method for segmenting remote sensing images for land space planning according to claim 2, characterized in that: The distorted blocks are extracted from multiple image blocks corresponding to each national land space planning remote sensing image, including: Determine a stable distribution coefficient corresponding to each image block; The image blocks in each national land space planning remote sensing image whose stable distribution coefficient is higher than the preset stable distribution threshold are regarded as distorted blocks.
4. The method for segmenting remote sensing images for land space planning according to claim 1, characterized in that: Grayscale matching is performed on each distorted block at each shooting angle and the distorted blocks corresponding to the spatial positions at other shooting angles, including: Extract two distorted blocks corresponding to the spatial position from any two land space planning remote sensing images with local overlap under different shooting angles, and establish multiple block matching groups; The grayscale matching coefficients between the two distorted blocks in each block matching group are calculated respectively.
5. The method for segmenting remote sensing images for land space planning according to claim 4, characterized in that: The grayscale matching coefficients between the two distorted blocks in each block matching group are calculated respectively, including: Determine the grayscale mean square error and the cosine similarity in the direction of the maximum grayscale gradient between two distorted blocks in each block matching group; The grayscale mean square error is divided by the cosine similarity to calculate the grayscale matching coefficient between the two distorted blocks in each block matching group.
6. The method for segmenting remote sensing images for land space planning according to claim 1, characterized in that: The distorted blocks at each shooting angle are matched with the distorted blocks at the corresponding spatial positions at other shooting angles, including: Each distorted block is used as a block to be matched, and the distorted blocks corresponding to the spatial positions of the block to be matched under other shooting angles are used as auxiliary matching blocks; Respectively obtaining corner points in the to-be-matched block and the auxiliary matching block; Extracting two first target corner points in the to-be-matched block, calculating the angle between a corner point connection line between the two first target corner points and a horizontal line, and obtaining a connection angle; Determine a target angle that is diagonally opposite to the angle of the connecting line, and determine an angle parameter value based on the target angle; Determine two second target corner points in each auxiliary matching block that correspond to the two first target corner points respectively; Calculate the corner point Euclidean distance between two second target corner points in each auxiliary matching block respectively, and take the corner point Euclidean distance with the smallest value as the minimum Euclidean distance; The minimum Euclidean distance is multiplied by the angle parameter value to obtain an angle matching coefficient between each distorted block and corresponding distorted blocks at other shooting angles.
7. The method for segmenting remote sensing images for land space planning according to claim 1, characterized in that: Based on the stable distribution coefficient, the grayscale matching coefficient and the distance angle matching coefficient, a restoration joint coefficient between each distorted block and corresponding distorted blocks at other shooting angles is calculated, including: Multiplying the grayscale matching coefficient by the distance angle matching coefficient to obtain an intermediate coefficient value; The stable distribution coefficient is divided by the intermediate coefficient value to obtain the restored joint coefficient.
8. A remote sensing image segmentation device for land space planning, characterized in that: The device includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, it implements the land space planning remote sensing image segmentation method as described in any one of claims 1 to 7.
9. A remote sensing image segmentation system for national land space planning, characterized in that: The system comprises: an imaging device and a land space planning remote sensing image segmentation device as claimed in claim 8; The imaging device is connected to the land space planning remote sensing image segmentation device, and the imaging device is used to shoot the land space planning remote sensing images of the target area at multiple shooting angles, and send the land space planning remote sensing images at multiple shooting angles to the land space planning remote sensing image segmentation device.