A method for detecting road traffic markings based on vectorized aerial images

Abstract

The application provides a kind of road traffic marking detection method based on vectorization aerial photograph, it is related to road marking detection and unmanned aerial vehicle vision technical field.The method first obtains the road image of aerial photography;And the obtained road image is carried out edge enhancement preprocessing;Then the road marking line segment of the road image after edge enhancement preprocessing is extracted, and the road marking line segment candidate set C is obtained;Then the road marking line segment in the road marking line segment candidate set C is cropped using the cropping algorithm, and data enhancement is realized;Finally, the cropped road marking line segment image is used as the road marking sample set, the features of the original aerial photograph and the image in the road marking sample set are extracted, and the road marking is obtained by Hungarian matching algorithm.The method realizes the matching of aerial photograph and sample set image by Laplace spectrum and Hungarian matching algorithm, and then combined with the above feature information and other parameter information, the lane line type can be basically determined.

Description

Technical Field

[0001] This invention relates to the fields of road marking detection and UAV vision technology, and in particular to a method for detecting road traffic markings based on vectorized aerial images. Background Technology

[0002] With increasing traffic and pedestrian flow on roads, clear road markings can help drivers drive safely. During traffic enforcement patrols, road markings can also be used to monitor for vehicle violations. In rapidly developing technologies such as intelligent driving and autonomous driving, lane marking detection is also widely used to assist vehicles in maintaining their lanes.

[0003] Existing road marking detection technologies are generally based on the driver's perspective. However, these technologies have limited visibility within a single frame and can only identify a limited number of road marking types, making them unsuitable for processing aerial images. Road marking detection in UAV aerial images primarily relies on orthographic imaging and straight lane line detection. While manual methods can accurately detect road markings in low-altitude UAV images, they are time-consuming, labor-intensive, and inefficient.

[0004] Therefore, it is necessary to use aerial images to detect road markings, so as to avoid the inefficiency of manual methods and the errors caused by traditional methods of detecting road markings. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a road traffic marking detection method based on vectorized aerial images, thereby realizing the detection of road traffic markings.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a road traffic marking detection method based on vectorized aerial images, comprising the following steps:

[0007] Step 1: Obtain aerial images of the road;

[0008] By specifying the drone's flight altitude and deflection angle, and taking environmental factors into account, the drone can capture images of the road.

[0009] Step 2: Perform edge enhancement preprocessing on the acquired road image;

[0010] An edge enhancement algorithm based on the Sobel operator is used to preprocess road images acquired by aerial photography to reduce noise interference in the images and better extract the edges of road markings;

[0011] Step 3: Extract road marking segments from the road image after edge enhancement preprocessing to obtain a candidate set C of road marking segments, and output the three key points of each road marking: the start point, the middle point, and the end point.

[0012] The LSD algorithm is used to extract road marking segments directly from the road image after edge enhancement preprocessing, and the start point, midpoint and end point of each road marking are output through the segment set;

[0013] Let the image after edge enhancement be I. enhance After LSD extraction, the candidate set C of road marking segments is obtained, as shown in the following formula:

[0014] C = LSD(I enhance )

[0015] The i-th line segment c in the candidate set C of road markings i As shown in the formula below:

[0016]

[0017] Where i = 1, 2, ..., l, and l is the total number of line segments in set C. For line segment c i The starting coordinates, For line segment c i The endpoint coordinates, θ i ∈(-90°, 90°) is the line segment c i The angle of inclination, The x-coordinate of the midpoint of the line segment is given by the following formula:

[0018]

[0019]

[0020] Step 4: Use a pruning algorithm to prune the road marking segments in the candidate set C to achieve data augmentation;

[0021] The Cohen-Sutherland clipping algorithm is used to determine whether the line segment to be clipped is within the rectangle. This is then used to clip the line segments in the candidate set C. The specific steps include:

[0022] Step 4.1: Determine the endpoint coordinates P1(x1,y1) and P2(x2,y2) of the line segment to be cut, as well as the four boundary coordinates Xwl, Xwr, Ywb and Ywt of the sliding rectangular window;

[0023] Step 4.2: Perform binary region encoding on endpoints P1 and P2, and set the encodings of P1 and P2 to Code1 and Code2 respectively;

[0024] Step 4.3: If Code1 = Code2 = 0, it means that both endpoints P1 and P2 are inside the sliding rectangle window, so the clipped line segment is fully visible and should be retained, and step 4.7 should be executed; otherwise, step 4.4 should be executed.

[0025] Step 4.4: If Code1 & Code2 ≠ 0, it means that the two endpoints P1 and P2 of the line segment are both outside the sliding rectangle window. In this case, the clipped line segment is completely invisible and should be discarded. Then proceed to step 4.8. Otherwise, proceed to step 4.5.

[0026] Step 4.5: Determine the position of endpoint P1 within the sliding rectangular window. If P1 is within the rectangular sliding window, swap the codes and coordinates of endpoints P1 and P2 to ensure that endpoint P1 always remains outside the sliding rectangular window.

[0027] Step 4.6: Following the order from right to left, the binary region code of endpoint P1 is detected step by step. When the first "1" is detected, it is determined that the line segment intersects with a certain edge of the sliding rectangular window, and the intersection point P of the line segment and the edge is solved. Thus, the line segment is divided into two parts, P1P and P2P. The P1P segment is discarded, the coordinate values ​​of P and P1 are swapped, and step 4.2 is executed again to judge the newly obtained line segment P2P.

[0028] Step 4.7: Draw the fully visible part within the sliding rectangle window, which is the current road marking segment P1P2;

[0029] Step 4.8: The cropping algorithm ends;

[0030] Step 5: Use the cropped road marking line segment image obtained in Step 4 as the road marking sample set, extract features from the original aerial image I and the image J in the road marking sample set, and obtain the road markings through the Hungarian matching algorithm;

[0031] Step 5.1: Construct Laplace matrices L for the original aerial image I and the image J in the sample set. I L J Obtain the initial matching matrix M of feature points in aerial image I and image J in the sample set;

[0032] For matrix L respectively I and L J Singular value decomposition is performed, as shown in the following formula:

[0033] L I =Udiag(λ1…λ n )U T

[0034] L J =Sdiag(γ1…γ n )S T

[0035] Where λ1≥…≥λ n =0, γ1≥…≥γ n =0, U=(U1...U n S = (S1...S) n ) are all n×n orthogonal matrices, where n is the number of feature points, and diag(λ1…λ2) is an n×n orthogonal matrix. n ) is L I A diagonal matrix, whose diagonal elements λ1…λn are matrix L. I eigenvalues; diag(γ1…γ n ) is L J A diagonal matrix whose diagonal elements γ1…γ n For matrix L J The eigenvalues; matrices U and S are the eigenvalues ​​of matrix L. I L J The orthogonal matrix obtained by decomposition, where the eigenvectors (U1...U2) are... n (S1...S) n ) are column vectors of U and S respectively;

[0036] Perform sign correction on orthogonal matrix S to obtain The initial matching matrix of feature points in the cropped aerial image I and the aerial image J in the sample set is obtained. The i-th row vector in U and S represents the i-th point feature of image I and image J, respectively, denoted as U (i) S (i) ; through S (i) Perform sign correction to make it satisfy U (i) The distance between them is minimized; at this point, let the i-th feature point of image J be .

[0037] Step 5.2: Perform row and column reduction on the initial matching matrix M: Let the matrix M be subtracted from the smallest element of the current row from the first row to the last row, and then perform the same operation on each column so that each row and each column of the obtained matrix F contains a zero element.

[0038] Step 5.3: Replace the non-zero elements in matrix F with zero elements, and replace the zero elements with 1 elements to obtain the new matrix G;

[0039] Step 5.4: Select the row with the fewest 1 elements in the new matrix G, then filter the column with the fewest 1 elements in that row, select this element and set all elements in its row and column to zero;

[0040] Step 5.5: Repeat steps 5.3 and 5.4 until the new matrix F contains no 1 elements, thus obtaining an independent group of 1 elements;

[0041] Step 5.6: Determine whether the independent 1-element group obtained in Step 5.5 is the largest independent 1-element group of the initial matching matrix M. If not, find a way to expand the independent 1-element group obtained in Step 5.5 to obtain the largest independent 1-element group of the initial matching matrix M. If the number of 1-elements in the independent 1-element group is equal to the order of matrix M, then the optimal solution of the matching matrix M between the original aerial image and the aerial images in the sample set has been obtained, stop the calculation, and proceed to Step 5.8; otherwise, proceed to the next step 5.7.

[0042] Step 5.7: Process matrix F using the augmented path method to obtain the intermediate state matrix;

[0043] Make the fewest lines to cover all the 0 elements of matrix F, then find the smallest element in the part not covered by the lines; subtract the obtained smallest element from each row covered by the lines, and add the smallest element to each column covered by the lines to obtain the intermediate state matrix, and return to step 5.3.

[0044] Step 5.8: Let Z be the optimal solution matrix obtained in Step 5.6; the elements Z in matrix Z... ij When = 1, it is considered that the i-th feature point of aerial image I matches the j-th feature point of aerial image J in the sample set.

[0045] The beneficial effects of adopting the above technical solution are as follows: The road traffic marking detection method based on vectorized aerial images provided by this invention improves the algorithm's ability to detect and track lane lines through adaptive magnetization of coordinates, while generating some lane line-related feature information. By matching aerial images with sample set images using the Laplace spectrum and Hungarian matching algorithm, and then combining the aforementioned feature information and other parameters, the lane line type can be basically determined. This method simplifies the calculation process and reduces the computational load while minimizing the impact of changes in illumination intensity, weather, and shadows on road traffic marking detection, thus effectively achieving real-time airborne processing of road traffic marking detection and simultaneously detecting road traffic markings, thereby improving the positional accuracy of lane line detection. Attached Figure Description

[0046] Figure 1 A flowchart of a road traffic marking detection method based on vectorized aerial images provided in an embodiment of the present invention;

[0047] Figure 2 Aerial road images provided for embodiments of the present invention;

[0048] Figure 3 The road image provided in this embodiment of the invention is used to extract road marking line segments. Detailed Implementation

[0049] 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.

[0050] In this embodiment, a road traffic marking detection method based on vectorized aerial images is described, such as... Figure 1 As shown, it includes the following steps:

[0051] Step 1: Obtain aerial images of the road;

[0052] By specifying the drone's flight altitude and deflection angle, and taking into account environmental factors such as weather, shadows, and time, the drone completes its imagery of the road. Figure 2 As shown;

[0053] Step 2: Perform edge enhancement preprocessing on the acquired road image;

[0054] An edge enhancement algorithm based on the Sobel operator is used to preprocess road images obtained by aerial photography to reduce the interference of noise such as shadows in the image and to better extract the edges of road markings;

[0055] The Sobel operator's convolution kernels in the x and y directions are as follows:

[0056]

[0057] The image is convolved using the aforementioned convolution kernels, which maximizes the response of the vertical and horizontal edges. The Sobel operator calculates the magnitude and independent variable values ​​of the first or second-order horizontal and vertical gradients of the image, and finally calculates the maximum modulus along the angular direction to obtain the image edges.

[0058] Step 3: Extract road marking segments from the road image after edge enhancement preprocessing to obtain a candidate set C of road marking segments, and output the three key points of each road marking: the start point, the middle point, and the end point.

[0059] To extract the set containing road markings from the preprocessed road image, line segment extraction is required. This invention directly extracts road marking line segments from the edge-enhanced preprocessed road image using the LSD algorithm, such as... Figure 3 As shown, the starting point, intermediate point, and ending point of each road marking are output through a set of line segments;

[0060] The LSD algorithm is used to directly extract line segments from the road image after edge enhancement preprocessing. Let the edge-enhanced image be I.enhance After LSD extraction, the candidate set C of road marking segments is obtained, as shown in the following formula:

[0061] C = LSD(I enhance )

[0062] The i-th line segment c in the candidate set C of road markings i As shown in the formula below:

[0063]

[0064] Where i = 1, 2, ..., l, and l is the total number of line segments in set C. For line segment c i The starting coordinates, For line segment c i The endpoint coordinates, θ i ∈(-90°, 90°) is the line segment c i The angle of inclination, The x-coordinate of the midpoint of the line segment is given by the following formula:

[0065]

[0066]

[0067] Step 4: Use a pruning algorithm to prune the road marking segments in the candidate set C to achieve data augmentation;

[0068] The Cohen-Sutherland clipping algorithm is used to determine whether the line segment to be clipped is within the rectangle. This is then used to clip the line segments in the candidate set C. The specific steps include:

[0069] Step 4.1: Determine the endpoint coordinates P1(x1,y1) and P2(x2,y2) of the line segment to be cut, as well as the four boundary coordinates Xwl, Xwr, Ywb and Ywt of the sliding rectangular window;

[0070] Step 4.2: Perform binary region encoding on endpoints P1 and P2, and set the encodings of P1 and P2 to Code1 and Code2 respectively;

[0071] Step 4.3: If Code1 = Code2 = 0, it means that both endpoints P1 and P2 are inside the sliding rectangle window, so the clipped line segment is fully visible and should be retained, and step 4.7 should be executed; otherwise, step 4.4 should be executed.

[0072] Step 4.4: If Code1 & Code2 ≠ 0, it means that the two endpoints P1 and P2 of the line segment are both outside the sliding rectangle window. In this case, the clipped line segment is completely invisible and should be discarded. Then proceed to step 4.8. Otherwise, proceed to step 4.5.

[0073] Step 4.5: Determine the position of endpoint P1 within the sliding rectangular window. If P1 is within the rectangular sliding window, swap the codes and coordinates of endpoints P1 and P2 to ensure that endpoint P1 always remains outside the sliding rectangular window.

[0074] Step 4.6: Following the order from right (least significant bit) to left (most significant bit), the binary region code of endpoint P1 is detected step by step. When the first "1" is detected, it is determined that the line segment intersects with a certain edge of the sliding rectangular window, and the intersection point P of the line segment and the edge is solved. Thus, the line segment is divided into two parts, P1P and P2P. The P1P segment is discarded, the coordinate values ​​of P and P1 are swapped, and step 4.2 is executed again to judge the newly obtained line segment P2P.

[0075] Step 4.7: Draw the fully visible part within the sliding rectangle window, which is the current road marking segment P1P2;

[0076] Step 4.8: The cropping algorithm ends;

[0077] Step 5: Use the cropped road marking line segment image obtained in Step 4 as the road marking sample set, extract features from the original aerial image I and the image J in the road marking sample set, and obtain the road markings through the Hungarian matching algorithm;

[0078] Step 5.1: Construct Laplace matrices L for the original aerial image I and the image J in the sample set. I L J Obtain the initial matching matrix M of feature points in aerial image I and image J in the sample set;

[0079] For matrix L respectively I and L J Singular value decomposition is performed, as shown in the following formula:

[0080] L I =Udiag(λ1…λ n )U T

[0081] L J =Sdiag(γ1…γ n )S T

[0082] Where λ1≥…≥λ n =0, γ1≥…≥γn =0, U=(U1...U n S = (S1...S) n ) are all n×n orthogonal matrices, where n is the number of feature points, and diag(λ1…λ2) is an n×n orthogonal matrix. n ) is L I A diagonal matrix, whose diagonal elements λ1…λn are matrix L. I eigenvalues; diag(γ1…γ n ) is L J A diagonal matrix whose diagonal elements γ1…γ n For matrix L J The eigenvalues; matrices U and S are the eigenvalues ​​of matrix L. I L J The orthogonal matrix obtained by decomposition, where the eigenvectors (U1...U2) are... n (S1...S) n ) are column vectors of U and S respectively;

[0083] Perform sign correction on orthogonal matrix S to obtain The initial matching matrix of feature points in the cropped aerial image I and the aerial image J in the sample set is obtained. The i-th row vector in U and S represents the i-th point feature of image I and image J, respectively, denoted as U (i) S (i) ; through S (i) Perform sign correction to make it satisfy U (i) The distance between them is minimized; at this point, let the i-th feature point of image J be .

[0084] Step 5.2: Perform row and column reduction on the initial matching matrix M: Let the matrix M be subtracted from the smallest element of the current row from the first row to the last row, and then perform the same operation on each column so that each row and each column of the obtained matrix F contains a zero element.

[0085] Step 5.3: Replace the non-zero elements in matrix F with zero elements, and replace the zero elements with 1 elements to obtain the new matrix G;

[0086] Step 5.4: Select the row with the fewest 1 elements in the new matrix G, then filter the column with the fewest 1 elements in that row, select this element and set all elements in its row and column to zero;

[0087] Step 5.5: Repeat steps 5.3 and 5.4 until the new matrix F contains no 1 elements, thus obtaining an independent group of 1 elements;

[0088] Step 5.6: Determine whether the independent 1-element group obtained in Step 5.5 is the largest independent 1-element group of the initial matching matrix M. If not, find a way to expand the independent 1-element group obtained in Step 5.5. (Finding a way to expand the path means finding an element in the matrix that contains a 0 element but is not in the current independent 1-element group, and then using a search algorithm (such as depth-first search, breadth-first search, etc.) to find other 0 elements adjacent to that element and try to add them to the current independent 1-element group). This will yield the largest independent 1-element group of the initial matching matrix M. If the number of 1 elements in the independent 1-element group is equal to the order of matrix M, then the optimal solution of the matching matrix M between the original aerial image and the aerial images in the sample set has been obtained. Stop the calculation and proceed to Step 5.8; otherwise, proceed to the next step, 5.7.

[0089] Step 5.7: Process matrix F using the augmented path method to obtain the intermediate state matrix;

[0090] Make the fewest lines to cover all the 0 elements of matrix F, then find the smallest element in the part not covered by the lines; subtract the obtained smallest element from each row covered by the lines, and add the smallest element to each column covered by the lines to obtain the intermediate state matrix, and return to step 5.3.

[0091] Step 5.8: Let Z be the optimal solution matrix obtained in Step 5.6; the elements Z in matrix Z... ij When = 1, it is considered that the i-th feature point of aerial image I matches the j-th feature point of aerial image J in the sample set.

[0092] Matrix Z actually describes the best possible match between feature points in the original image and the sample set images. After obtaining the optimal matrix Z in step 5.8, we can examine the rows and columns of the elements that are 1. These rows and columns correspond to the points selected as matching feature points in the original aerial image and the selected points in the aerial images in the sample set, respectively. If there exists a matching scheme such that all 1 elements in the Z matrix form a maximal independent set of 1 elements, then an optimal matching result can be obtained, meaning that each feature point matches only one other feature point without any conflicts or repetitions. Therefore, it can be considered that the original aerial image matches the aerial images in the sample set. Simultaneously, maximizing the sum of the weights in matrix Z also means finding the optimal matching scheme. Specifically, the position of these 1 elements indicates that the i-th feature point in the original aerial image matches the j-th feature point in the aerial images in the sample set, meaning that the i-th feature point in the original aerial image is similar to or identical to the j-th feature point in the aerial images in the sample set.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A method for detecting road traffic markings based on vectorized aerial images, characterized in that: Includes the following steps: Step 1: Obtain aerial images of the road; Step 2: Perform edge enhancement preprocessing on the acquired road image; Step 3: Extract road marking segments from the road image after edge enhancement preprocessing to obtain a candidate set C of road marking segments, and output the three key points of each road marking: the start point, the middle point, and the end point. Step 4: Use a pruning algorithm to prune the road marking segments in the candidate set C to achieve data augmentation; Step 5: Use the cropped road marking line segment image as a road marking sample set, extract features from the original aerial image and the images in the road marking sample set, and obtain the road markings using the Hungarian matching algorithm; Step 5.1: Construct Laplace matrices L for the original aerial image I and the image J in the sample set. I L J Obtain the initial matching matrix M of feature points in aerial image I and image J in the sample set; For matrix L respectively I and L J Singular value decomposition is performed, as shown in the following formula: L I =Udiag(λ1…λ n )U T L J =Sdiag(γ1…γ n )S T Where λ1≥…≥λ n =0, γ1≥…≥γ n =0, U=(U1...U n S = (S1...S) n ) are all n×n orthogonal matrices, where n is the number of feature points, and diag(λ1…λ2) is an n×n orthogonal matrix. n ) is L I A diagonal matrix, whose diagonal elements λ1…λn are matrix L. I eigenvalues; diag(γ1…γ n ) is L J A diagonal matrix whose diagonal elements γ1…γ n For matrix L J The eigenvalues; matrices U and S are the eigenvalues ​​of matrix L. I L J The orthogonal matrix obtained by decomposition, where the eigenvectors (U1...U2) are... n (S1...S) n ) are column vectors of U and S respectively; Perform sign correction on orthogonal matrix S to obtain The initial matching matrix of feature points in the cropped aerial image I and the aerial image J in the sample set is obtained. The i-th row vector in U and S represents the i-th point feature of image I and image J, respectively, denoted as U. (i) S (i) ; through S (i) Perform sign correction to make it satisfy U (i) The distance between them is minimized; at this point, let the i-th feature point of image J be . Step 5.2: Perform row and column reduction on the initial matching matrix M: Let the matrix M be subtracted from the smallest element of the current row from the first row to the last row, and then perform the same operation on each column so that each row and each column of the obtained matrix F contains a zero element. Step 5.3: Replace the non-zero elements in matrix F with zero elements, and replace the zero elements with 1 elements to obtain the new matrix G; Step 5.4: Select the row with the fewest 1 elements in the new matrix G, then filter the column with the fewest 1 elements in that row, select this element and set all elements in its row and column to zero; Step 5.5: Repeat steps 5.3 and 5.4 until the new matrix F contains no 1 elements, thus obtaining an independent group of 1 elements; Step 5.6: Determine whether the independent 1-element group obtained in Step 5.5 is the largest independent 1-element group of the initial matching matrix M. If not, find a way to expand the independent 1-element group obtained in Step 5.5 to obtain the largest independent 1-element group of the initial matching matrix M. If the number of 1-elements in the independent 1-element group is equal to the order of matrix M, then the optimal solution of the matching matrix M between the original aerial image and the aerial images in the sample set has been obtained, stop the calculation, and proceed to Step 5.8; otherwise, proceed to the next step 5.

7. Step 5.7: Process matrix F using the augmented path method to obtain the intermediate state matrix; Make the fewest lines to cover all the 0 elements of matrix F, then find the smallest element in the part not covered by the lines; subtract the obtained smallest element from each row covered by the lines, and add the smallest element to each column covered by the lines to obtain the intermediate state matrix, and return to step 5.

3. Step 5.8: Let Z be the optimal solution matrix obtained in Step 5.6; the elements Z in matrix Z... ij When = 1, it is considered that the i-th feature point of aerial image I matches the j-th feature point of aerial image J in the sample set.

2. The method for detecting road traffic markings based on vectorized aerial images according to claim 1, characterized in that: Step 2 employs an edge enhancement algorithm based on the Sobel operator to preprocess the road image obtained from aerial photography, in order to reduce the interference of noise in the image and better extract the edges of road markings.

3. The method for detecting road traffic markings based on vectorized aerial images according to claim 2, characterized in that: Step 3 uses the LSD algorithm to directly extract road marking segments from the road image after edge enhancement preprocessing, and outputs the start point, middle point and end point of each road marking through the segment set; Let the image after edge enhancement be I. enhance After extracting line segments using LSD, the candidate set C of road marking line segments is obtained, as shown in the following formula: C=LSD(I enhance ) The i-th line segment c in the candidate set C of road markings i As shown in the formula below: Where i = 1, 2, ..., l, and l is the total number of line segments in set C. For line segment c i The starting coordinates, For line segment c i The endpoint coordinates, θ i ∈(-90°, 90°) is the line segment c i The angle of inclination, The x-coordinate of the midpoint of the line segment is given by the following formula:

4. The method for detecting road traffic markings based on vectorized aerial images according to claim 3, characterized in that: Step 4 uses the Cohen-Sutherland clipping algorithm to determine whether the line segment to be clipped is within the rectangle, thereby clipping the line segments in the candidate set C.

5. The method for detecting road traffic markings based on vectorized aerial images according to claim 4, characterized in that: The specific method for step 4 is as follows: Step 4.1: Determine the endpoint coordinates P1(x1,y1) and P2(x2,y2) of the line segment to be cut, as well as the four boundary coordinates Xwl, Xwr, Ywb and Ywt of the sliding rectangular window; Step 4.2: Perform binary region encoding on endpoints P1 and P2, and set the encodings of P1 and P2 to Code1 and Code2 respectively; Step 4.3: If Code1 = Code2 = 0, it means that both endpoints P1 and P2 are inside the sliding rectangle window, so the clipped line segment is fully visible and should be retained, and step 4.7 should be executed; otherwise, step 4.4 should be executed. Step 4.4: If Code1 & Code2 ≠ 0, it means that the two endpoints P1 and P2 of the line segment are both outside the sliding rectangle window. In this case, the clipped line segment is completely invisible and should be discarded. Then proceed to step 4.

8. Otherwise, proceed to step 4.

5. Step 4.5: Determine the position of endpoint P1 within the sliding rectangular window. If P1 is within the rectangular sliding window, swap the codes and coordinates of endpoints P1 and P2 to ensure that endpoint P1 always remains outside the sliding rectangular window. Step 4.6: Following the order from right to left, the binary region code of endpoint P1 is detected step by step. When the first "1" is detected, it is determined that the line segment intersects with a certain edge of the sliding rectangular window, and the intersection point P of the line segment and the edge is solved. Thus, the line segment is divided into two parts, P1P and P2P. The P1P segment is discarded, the coordinate values ​​of P and P1 are swapped, and step 4.2 is executed again to judge the newly obtained line segment P2P. Step 4.7: Draw the fully visible part within the sliding rectangle window, which is the current road marking segment P1P2; Step 4.8: The cropping algorithm ends.

Images