CoGASWT-based disaster area road identification and detection method
By combining the CoGASWT method with MinRTV and an improved edge detection operator, the problems of incomplete road extraction and tree occlusion in disaster areas were solved, resulting in better road recognition and detection performance.
Patent Information
- Application Number
- CN202410999825.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-15
AI Technical Summary
Existing technologies for road extraction in disaster areas suffer from problems such as incomplete road extraction, unclear division between the background and road, and road obstruction by trees, which prevents road extraction. They are particularly ineffective in complex mountainous environments.
The CoGASWT-based approach is adopted, including HSV color space decomposition, MinRTV processing, edge detection, CoGASWT algorithm, connected component filtering, and missing road connection. By using minimum value filtering and improved edge detection operators, combined with interquartile range threshold filtering and eight-neighbor repair, the road extraction effect is improved.
It effectively removes tree interference, improves road edge detection accuracy, connects missing areas, enhances the integrity and accuracy of road extraction, and adapts to road recognition in different environments.
Smart Images

Figure BDA0004961249450000034 
Figure BDA0004961249450000035 
Figure BDA0004961249450000036
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of image processing, and in particular relates to a disaster area road recognition and detection method based on CoGASWT. Background Art
[0002] After a natural disaster occurs, using drones to take photos and analyze the roads in the disaster area is an important task for rapid detection and rescue assessment of the roads in the disaster area. Currently, advanced transmission technology can provide high-resolution image data, which provides an effective means for extracting roads in the disaster area. The next step is to analyze the obtained images.
[0003] Traditional methods for extracting roads in disaster areas are often affected by factors such as tree occlusion and uneven road color, resulting in incomplete road extraction. In order to overcome these problems, scholars have proposed a variety of algorithms to improve the effect of road extraction. Among them, methods based on clustering segmentation, Hough line detection, feature extraction and classification can obtain results in a short time, but in the case of complex and changeable mountain roads, they are often disturbed, resulting in inaccurate segmentation of roads and backgrounds. Another method is an algorithm based on deep neural networks, which can use a large amount of data for learning to achieve accurate road and background segmentation. However, due to the real-time requirements of disaster area road extraction, methods based on deep neural networks still have some challenges in practice. In addition, the traditional stroke width transformation algorithm is less used in disaster area road extraction, and it is easy to produce too many edges, incomplete road extraction, and more non-road areas detected on images with a lot of interference information and complex backgrounds. Therefore, this algorithm still needs further improvement.
[0004] In terms of connected domain analysis, in the past, when screening connected domains for stroke width transformation, the variance and coefficient of variation were mainly used for evaluation, but this method has poor robustness to mutation values. Therefore, a more effective connected domain analysis method needs to be proposed in disaster area road extraction to accurately screen out road areas. Another problem that needs to be solved is the difficulty in road extraction caused by tree occlusion. In traditional algorithms, the road with trees cannot be extracted because the impact of tree occlusion on the road cannot be fully considered. Therefore, for this situation, a solution needs to be proposed to better identify and extract the road area under tree occlusion.
[0005] In summary, the road extraction technology in mountainous disaster areas has made certain progress in recent years, but there are still some challenges and problems. Summary of the invention
[0006] The purpose of the present invention is to provide a disaster area road recognition and detection method based on CoGASWT to solve the problems of incomplete extraction of road parts and unclear division of background and road parts in the prior art. The present invention has better extraction effect in different environments, can improve the poor adaptability of the SWT algorithm when searching for corresponding edges, and connect the occluded and missing road parts.
[0007] In order to achieve the above object, the present invention adopts the following technical scheme:
[0008] The disaster area road recognition and detection method based on CoGASWT includes the following steps:
[0009] Step 1: Obtain aerial road images of the disaster area;
[0010] Step 2: Decompose the aerial road image of the disaster area obtained in step 1 into HSV color space to obtain the H channel image, which is recorded as image I. H ;
[0011] Step 3: For the image I obtained in step 2 H Perform MinRTV processing to obtain image I without tree interference distr ;
[0012] Step 4: For image I in step 3 distr Perform edge detection to obtain image I edge ;
[0013] Step 5: For the image I obtained in step 4 edge Processed by CoGASWT algorithm, image I is obtained swt ;
[0014] Step 6: For the image I obtained in step 5 swt Remove the small connected domains and perform corrosion operations on them in the horizontal and vertical directions to obtain image I con ;
[0015] Step 7: For the image I obtained in step 6 con For each connected domain in , arrange the values of each point from small to large, and calculate its lower quartile Q1, whose position is Calculate the middle quartile Q2, which is located at Calculate the upper quartile Q3, its location is Calculate the interquartile range Q3-Q1 and set the interquartile range threshold Q th , for values above the threshold Q th The connected domain is removed to obtain image I Q4 ;
[0016] Step 8: For the image I obtained in step 7Q4 Perform partial restoration based on eight neighborhoods to obtain the final image.
[0017] Furthermore, step 2 is specifically as follows: converting the aerial road image of the disaster area in the RGB color space into the HSV color space, and converting the image from R, G, and B color components into three components of H, S, and V, where H represents the hue of the image, S represents the saturation of the image, and V represents the brightness of the image.
[0018] Furthermore, in step 3, the image I obtained in step 2 H Perform MinRTV processing to obtain image I without tree interference distr , the expression is as follows:
[0019] G min (x,y)=min{I H (x,y),I H (x-1,y),I H (x+1,y),I H (x,y-1),I H (x,y+1),I H (x-1,y-1),I H (x-1,y+1),I H (x+1,y-1),I H (x+1,y+1)}
[0020] I distr =RTV(G min ,λ minRTV ,σ,ε)
[0021] In the formula, I H Represents the image matrix, I H (x, y) represents the value at (x, y) in the image matrix; λ minRTV represents the smoothing strength; σ represents the filter template size; ε is the coefficient to prevent the denominator from being zero;
[0022] The model of MinRTV is:
[0023]
[0024] in:
[0025]
[0026]
[0027]
[0028]
[0029] In the formula, G min is the minimum value image of the input, I distr is the output image, p is a local area, D x (p) and D y (p) is the total variation of the window in the x and y directions, respectively, x (p) is F y (p) are the window intrinsic variations in the x and y directions, ε is the regularization number, and represents the partial differential in two directions, g p is the Gaussian kernel function:
[0030]
[0031] In the formula, x p and p is the difference between the horizontal and vertical coordinates of the center pixel and the boundary pixel of the local area. σ is the filter template, which controls the spatial size of the window. min After filtering, we get the image I without tree interference. distr .
[0032] Furthermore, minRTV The value is 0.03, ε is 0.02, and the filter template σ is 3.
[0033] Furthermore, in step 4, the image I in step 3 is distr Perform edge detection to obtain image I edge , specifically including:
[0034] Step 4.1: For image I in step 3 distr Perform Gaussian filtering to obtain image I G , the formula is:
[0035]
[0036] Among them, I G is the filtered image, x distr and distr is the image I obtained in step 3 distr The horizontal and vertical coordinates of a point in the, δ is the standard deviation of the filter;
[0037] Step 4.2: Use the improved edge detection operator to calculate the image I obtained in step 4.1 G Gradient, the improved edge detection operator is as follows:
[0038]
[0039] Among them, Gride is the gradient value in the x or y direction, and * is the convolution operator;
[0040] Step 4.3: Get the magnitude and direction of the gradient obtained in step 4.2. The formula is as follows:
[0041]
[0042] Where L is the gradient amplitude at a certain point, and θ is the gradient angle at a certain point, which is between 0-180°;
[0043] Step 4.4: Perform non-maximum suppression on the gradient amplitude obtained in step 4.3;
[0044] Step 4.5: Perform double threshold processing on the image obtained in step 4.4 to classify edges and non-edges; calculate the length of the connected domain of each edge, set the edge with a connected domain length less than the threshold η to 0, and obtain the edge detection image I edge .
[0045] Furthermore, during edge detection, η is taken as 0.3 times the image width.
[0046] Furthermore, in step 5, the image I obtained in step 4 is edge The CoGASWT algorithm is used for processing, specifically:
[0047] Step 5.1: For image I edge Find the gradient direction of each edge point and store it;
[0048] Step 5.2: Search for the corresponding edge operation for each edge point, compare the gradient directions of the previous and next edge points of the edge point, and the difference between the gradient direction of the current point and the gradient directions of the two adjacent points is less than or equal to θ th In the case of , the search method with an angle of 0° and ±30° with the current point gradient direction is used; the difference between the gradient direction of the current point and the gradient directions of the two adjacent points is greater than or equal to θ th In the case of , the search method with an angle of 0°, ±30°, ±45° with the current point gradient direction is used, and the formula is as follows:
[0049]
[0050] Where dist represents the Euclidean distance between (x, y) and the point behind it and rounds it off, θ max Indicates the maximum value of the difference between the gradient direction of the current point and the gradient directions of the two adjacent points, θ th To determine the threshold, I is obtained by the above method. swt .
[0051] Furthermore, the threshold θ in step 5 th The value is Step 7 Qth The value is 5.
[0052] Furthermore, in step 8, the image I obtained in step 7 is Q4 Perform partial repair based on eight neighborhoods, specifically:
[0053] Step 8.1: Traverse image I Q4 Each point in and read the value of the point;
[0054] Step 8.2: For each point in step 8.1, follow the maximum search length L specified for that point. max Search as a condition;
[0055] Step 8.3: When the corresponding valued corresponding point is found under the conditions of step 8.2, this search path is connected; when the corresponding valued corresponding point is not found under the conditions of step 8.2, this path is abandoned and it is considered that this section does not belong to the road area.
[0056] Furthermore, L max The value is the value corresponding to each point at that location.
[0057] Compared with the prior art, the present invention has the following beneficial technical effects:
[0058] The present invention combines MinRTV with the CoGASWT algorithm, and further refines the detected roads through the connected domain screening rules and the algorithm for connecting defective roads. In view of the excessive interference information caused by the large number of trees in the image, the MinRTV algorithm combining minimum filtering and RTV is introduced, which greatly reduces the texture information of the tree area and better retains the edge information of the road. When detecting the edge of the road, the improved edge detection operator is used to better detect the edge. At the same time, the threshold screening technology is introduced to create good conditions for the next step of CoGASWT by removing the connected domain that is not considered to be the edge of the road. Using the interquartile range as the condition for connected domain screening can effectively avoid the influence of mutation values. Connecting and repairing the parts of the road that are missing and blocked by trees can solve the problems of defects and occlusion to a certain extent, which provides good image information for road extraction and subsequent information processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] The drawings in the specification are used to provide further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0060] Figure 1 It is a schematic diagram of the process of the present invention;
[0061] Figure 2 and Figure 3 The images are obtained by performing the main stages of the algorithm of the present invention on two images of disaster areas in mountainous areas, where (a) is the original image, (b) is the result of processing using only traditional image filtering and SWT algorithm, (c) is the result of morphological processing based on (b), (d) is the result of connected domain screening based on (c), and (e) is the final processing result obtained by using the repair algorithm based on (d). DETAILED DESCRIPTION
[0062] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0063] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0064] See also Figure 1 ,The disaster area road recognition and detection method based on CoGASWT includes the following steps:
[0065] Step 1: Obtain aerial road images of the disaster area.
[0066] Use image acquisition equipment to obtain aerial road images of the disaster area.
[0067] Step 2: In order to preliminarily distinguish the road from the background area, the image obtained in step 1 is decomposed in the HSV color space to obtain the H channel image denoted as I H ;
[0068] Step 3: For the image I obtained in step 2 H Perform MinRTV processing to obtain image I without tree interference distr , the expression is as follows:
[0069] G min (x,y)=min{I H (x,y),I H (x-1,y),I H (x+1,y),I H (x,y-1),I H (x,y+1),I H (x-1,y-1),I H (x-1,y+1),I H (x+1,y-1),I H (x+1,y+1)}
[0070] I distr =RTV(G min ,λ minRTV ,σ,ε)
[0071] In the formula, I H Represents the image matrix, I H (x, y) represents the value at (x, y) in the image matrix; λ minRTV represents the smoothing strength; σ represents the filter template size; ε is the coefficient to prevent the denominator from being zero.
[0072] The model of MinRTV is:
[0073]
[0074] in:
[0075]
[0076]
[0077]
[0078]
[0079] In the formula, G min is the minimum value image of the input, I distr is the output image, p is a local area, D x (p) and D y (p) is the total variation of the window in the x and y directions, respectively, x (p) is F y (p) are the window intrinsic variations in the x and y directions, ε is the regularization number, and represents the partial differential in two directions, g p is the Gaussian kernel function:
[0080]
[0081] In the formula, x p and p is the difference between the horizontal and vertical coordinates of the center pixel and the boundary pixel of the local area. σ is the filter template, which controls the spatial size of the window. min After filtering, we get the image I without tree interference. distr .
[0082] where λ minRTV The value is 0.03, ε is 0.02, and the filter template σ is 3. For the minimum image G min After filtering, the texture-free image I is obtained. disttr ;
[0083] Step 4: For image I in step 3 distr To perform edge detection, the specific steps are as follows:
[0084] Step 4.1: Perform Gaussian filtering on the input image. The formula is:
[0085]
[0086] Among them, I G is the filtered image, x distr and distr is the image I obtained in step 3 distr The horizontal and vertical coordinates of a point in the filter, δ is the standard deviation of the filter, where δ is 1.2;
[0087] Step 4.2: Use the improved edge detection operator to calculate the I of the image obtained in step 4.1 G Gradient. Its operator is as follows:
[0088]
[0089] Gride is the gradient value in the x or y direction, and * is the convolution operator;
[0090] Step 4.3: Get the value and direction of the gradient obtained in step 4.2. The formula is as follows:
[0091]
[0092] Where L is the gradient amplitude at a certain point, θ is the gradient angle at a certain point, between 0-180°;
[0093] Step 4.4: Perform non-maximum suppression on the gradient amplitude obtained in step 4.3;
[0094] Step 4.5: Perform double threshold processing on the image obtained in step 4.4 to classify edges and non-edges;
[0095] Furthermore, the length of the connected domain of each edge is calculated, and the edges with a connected domain length less than a threshold η are set to 0, and only larger edges are retained. Usually, the threshold η is 0.3 times the image width.
[0096] Step 5: For the image I obtained in step 4 edge The specific steps of CoGASWT algorithm processing are as follows:
[0097] Step 5.1: For image I edge Find the gradient direction of each edge point and store it;
[0098] Step 5.2: Search for the corresponding edge operation for each edge point, compare the gradient directions of the previous and next edge points of the edge point, and the difference between the gradient direction of the current point and the gradient directions of the two adjacent points is less than or equal to θ th In the case of , the search method with an angle of 0° and ±30° with the current point gradient direction is used. The difference between the gradient direction of the current point and the gradient direction of the two adjacent points is greater than or equal to θ th In the case of , the search method with an angle of 0°, ±30°, ±45° with the current point gradient direction is used. The formula is as follows:
[0099]
[0100] dist represents the Euclidean distance between (x, y) and the point behind it and is rounded. max Indicates the maximum value of the difference between the gradient direction of the current point and the gradient directions of the two adjacent points, θ th To determine the threshold, I is obtained by the above method. swt , where θ th Pick
[0101] Step 6: For the image I obtained in step 5 swt To remove the small connected domain, an erosion operation with a length of 2 is performed in the horizontal and vertical directions respectively. This operation can effectively remove the thin and long areas caused by the sudden change of stroke width caused by the missing of the corresponding edge in the image, and obtain I con ;
[0102] Step 7: For the image I obtained in step 6 con For each connected domain in , arrange the values of each point from small to large, and calculate its lower quartile Q1, whose position is Calculate the middle quartile Q2, which is located at Calculate the upper quartile Q3, its location is Calculate the interquartile range Q3-Q1 and set the interquartile range threshold Q th, remove the connected domains above the threshold and get I Q4 , where Q th The value is 5;
[0103] Step 8: For the image I obtained in step 7 Q4 Perform partial repair based on eight neighborhoods. The specific steps are as follows:
[0104] Step 8.1: Traverse image I Q4 Read the value of each point in
[0105] Step 8.2: For each point in step 8.1, follow the maximum search length L specified for that point. max As a condition to search, here L max Take the value of this point;
[0106] Step 8.3: When the corresponding valued corresponding point is found under the conditions of step 8.2, this search path is connected to repair the missing part; when the corresponding valued corresponding point is not found under the conditions of step 8.2, this path is abandoned and it is considered that this section does not belong to the road area.
[0107] Figure 2 (a) is an image of a mountain road, where part of the road edge is blocked by trees. Figure 2 (b) to (e) show the effects of different algorithms. Figure 2 (b) is the image obtained by using only the SWT algorithm for detection. It can be seen that there are many interference areas in the result due to excessive road interference information; Figure 2 (c) is the image obtained after MinRTV combined with the SWT algorithm. It can be seen that after the de-texturing operation, most of the tree interference in the image is removed; Figure 2 (d) is an image using MinRTV and CoGASWT and connected domain screening. It can be seen that the interference edge information in the image is largely removed, and the main part of the road is further clarified; Figure 2 (e) is the image after road defect repair, that is, the image obtained by this algorithm. This step can repair and connect the undetected defective road parts or the parts blocked by trees, thereby restoring the original state of the road.
[0108] Figure 3 (a) is also an image of a road in a mountainous area. It can be seen that some edges of the road are not obvious and some edges are covered by sand and dust. Figure 3 (b) to (e) also show the effects of different algorithms. Figure 3(b) is the image obtained by using only the SWT algorithm for detection. It can be seen that due to excessive road interference information, there are many interference areas in the result, and there are many small connected domain interferences; Figure 3 (c) is the image obtained after MinRTV combined with the SWT algorithm. It can be seen that after the de-texturing operation, some irregular connected domains exist; Figure 3 (d) is an image using MinRTV and CoGASWT and connected domain screening. The irregular connected domains in the image are removed by the set heuristic rules, and the main part of the road is further clarified; Figure 3 (e) is the image after road defect repair, that is, the image obtained by this algorithm. This step can repair and connect the undetected defective road parts or the parts blocked by trees, thereby restoring the original state of the road.
[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims of the invention.
Claims
1. The disaster area road recognition and detection method based on CoGASWT is characterized by: The following steps are involved: Step 1: Obtain aerial road images of the disaster area; Step 2: Decompose the aerial road image of the disaster area obtained in step 1 into HSV color space to obtain the H channel image, which is recorded as image I. H ; Step 3: For the image I obtained in step 2 H Perform MinRTV processing to obtain image I without tree interference distr ; Step 4: For image I in step 3 distr Perform edge detection to obtain image I edge ; Step 5: For the image I obtained in step 4 edge Processed by CoGASWT algorithm, image I is obtained swt ; Step 6: For the image I obtained in step 5 swt Remove the small connected domains and perform corrosion operations on them in the horizontal and vertical directions to obtain image I con ; Step 7: For the image I obtained in step 6 con For each connected domain in , arrange the values of each point from small to large, and calculate its lower quartile Q1, whose position is Calculate the middle quartile Q2, which is located at Calculate the upper quartile Q3, its location is Calculate the interquartile range Q3-Q1 and set the interquartile range threshold Q th , for values above the threshold Q th The connected domain is removed to obtain image I Q4 ; Step 8: For the image I obtained in step 7 Q4 Perform partial restoration based on eight neighborhoods to obtain the final image.
2. The disaster area road identification and detection method based on CoGASWT according to claim 1 is characterized in that: Step 2 is as follows: convert the aerial road image of the disaster area in the RGB color space to the HSV color space, and transform the image from R, G, and B color components to H, S, and V components, where H represents the hue of the image, S represents the saturation of the image, and V represents the brightness of the image.
3. The disaster area road identification and detection method based on CoGASWT according to claim 1 is characterized in that: In step 3, the image I obtained in step 2 H Perform MinRTV processing to obtain image I without tree interference distr , the expression is as follows: G min (x,y)=min{I H (x,y),I H (x-1,y),I H (x+1,y),I H (x,y-1),I H (x,y+1),I H (x-1,y-1),I H (x-1,y+1),I H (x+1,y-1),I H (x+1,y+1)} I distr =RTV(G min ,l minRTV ,in) In the formula, I H Represents the image matrix, I H (x, y) represents the value at (x, y) in the image matrix; λ minRTV represents the smoothing strength; σ represents the filter template size; ε is the coefficient to prevent the denominator from being zero; The model of MinRTV is: in: In the formula, G min is the minimum value image of the input, I distr is the output image, p is a local area, D x (p) and D y (p) is the total variation of the window in the x and y directions, respectively, x (p) is F y (p) are the window intrinsic variations in the x and y directions, ε is the regularization number, and represents the partial differential in two directions, g p is the Gaussian kernel function: In the formula, x p and p is the difference between the horizontal and vertical coordinates of the center pixel and the boundary pixel of the local area. σ is the filter template, which controls the spatial size of the window. min After filtering, we get the image I without tree interference. distr .
4. The method for identifying and detecting disaster area roads based on CoGASWT according to claim 3 is characterized in that: λ minRTV The value is 0.03, ε is 0.02, and the filter template σ is 3.
5. The method for identifying and detecting disaster area roads based on CoGASWT according to claim 1, characterized in that: In step 4, image I in step 3 distr Perform edge detection to obtain image I edge , specifically including: Step 4.1: For image I in step 3 distr Perform Gaussian filtering to obtain image I G , the formula is: Among them, I G is the filtered image, x distr and distr is the image I obtained in step 3 distr The horizontal and vertical coordinates of a point in the, δ is the standard deviation of the filter; Step 4.2: Use the improved edge detection operator to calculate the image I obtained in step 4.1 G Gradient, the improved edge detection operator is as follows: Among them, Gride is the gradient value in the x or y direction, and * is the convolution operator; Step 4.3: Get the magnitude and direction of the gradient obtained in step 4.
2. The formula is as follows: Where L is the gradient amplitude at a certain point, and θ is the gradient angle at a certain point, which is between 0-180°; Step 4.4: Perform non-maximum suppression on the gradient amplitude obtained in step 4.3; Step 4.5: Perform double threshold processing on the image obtained in step 4.4 to classify edges and non-edges; calculate the length of the connected domain of each edge, set the edge with a connected domain length less than the threshold η to 0, and obtain the edge detection image I edge .
6. The method for identifying and detecting disaster area roads based on CoGASWT according to claim 5, characterized in that: During edge detection, η is taken as 0.3 times the image width.
7. The method for identifying and detecting disaster area roads based on CoGASWT according to claim 1, characterized in that: In step 5, the image I obtained in step 4 is edge The CoGASWT algorithm is used for processing, specifically: Step 5.1: For image I edge Find the gradient direction of each edge point and store it; Step 5.2: Search for the corresponding edge operation for each edge point, compare the gradient directions of the previous and next edge points of the edge point, and the difference between the gradient direction of the current point and the gradient directions of the two adjacent points is less than or equal to θ th In the case of , the search method with an angle of 0° and ±30° with the current point gradient direction is used; the difference between the gradient direction of the current point and the gradient directions of the two adjacent points is greater than or equal to θ th In the case of , the search method with an angle of 0°, ±30°, ±45° with the current point gradient direction is used, and the formula is as follows: Where dist represents the Euclidean distance between (x, y) and the point behind it and rounds it off, θ max Indicates the maximum value of the difference between the gradient direction of the current point and the gradient directions of the two adjacent points, θ th To determine the threshold, I is obtained by the above method. swt .
8. The method for identifying and detecting disaster area roads based on CoGASWT according to claim 7 is characterized in that: The threshold θ in step 5 th The value is Step 7 Q th The value is 5.
9. The method for identifying and detecting disaster area roads based on CoGASWT according to claim 1, characterized in that: In step 8, the image I obtained in step 7 is Q4 Perform partial repair based on eight neighborhoods, specifically: Step 8.1: Traverse image I Q4 Each point in and read the value of the point; Step 8.2: For each point in step 8.1, follow the maximum search length L specified for that point. max Search as a condition; Step 8.3: When the corresponding points with values are found under the conditions of step 8.2, the search path is connected; If no corresponding point with value is found under the conditions of step 8.2, this section of the path is abandoned and it is considered that this section does not belong to the road area.
10. The method for identifying and detecting disaster area roads based on CoGASWT according to claim 1, characterized in that: L max The value is the value corresponding to each point at that location.
Citation Information
Cited By
Road settlement automatic detection method and system
CN120564155A
