A road elevation detection method, medium and system
By using dual laser emitters and image verification technology, combined with optical cameras and 3D depth cameras, the problem of inaccurate full-width lane measurement was solved, achieving high-precision road surface elevation detection.
Patent Information
- Application Number
- CN202410520871.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-28
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-04-28
AI Technical Summary
Existing automatic measurement units for surface features of infrastructure are insufficient to meet the measurement requirements of the entire lane, resulting in inaccurate detection.
The method employs dual laser emitters to simultaneously emit two line structured beams, combined with an optical camera and a 3D depth camera. By verifying the images, it identifies normal and error signals in the reflected laser signals. Using images captured by a high-definition camera as a reference, it identifies and corrects discrete noise data in the reflected laser signals acquired by the 3D depth camera.
It enables full-width lane surface measurement, improves measurement accuracy, preserves actual ground feature data to the maximum extent, and effectively eliminates erroneous data.
Smart Images

Figure CN118274785B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of road surface detection, and in particular to a road surface elevation detection method, medium and system. Background Art
[0002] The full-width surface sensing unit is the core component for automatic measurement of surface features of transportation infrastructure. It can collect characteristic information such as 3D surface, longitudinal section, cross section, texture section, etc. of infrastructure such as highways and urban roads.
[0003] The existing sensor units for automatic measurement of surface features of infrastructure are affected by laser irradiation intensity and laser irradiation uniformity, and it is difficult for them to independently meet the requirements for automatic surface measurement of full-width lanes (width of more than 3.5m). At present, the surface measurement of full-width lanes at home and abroad all adopts the measurement method of combining two sets of sensor units. This design requires the data measured by the two sets of sensor units to be spliced and processed, which introduces errors and leads to inaccurate detection. Summary of the invention
[0004] The embodiments of the present invention provide a road elevation detection method, medium and system to solve the problem of inaccurate road elevation detection in the prior art.
[0005] In a first aspect, a road elevation detection method is provided, comprising:
[0006] When the inspection vehicle is driving on the road, the dual laser emitters carried by the inspection vehicle are controlled to simultaneously emit two lines of structured light toward the road surface;
[0007] Synchronously controlling the optical camera carried by the inspection vehicle to collect road surface images, and the 3D depth camera carried by the inspection vehicle to collect reflected laser signals of the line structured light reflected by the road surface;
[0008] Identify a normal signal and an error signal in the reflected laser signal through the road surface image and the corresponding reflected laser signal;
[0009] The normal signal in the reflected laser signal is used to calculate the road surface elevation.
[0010] In a second aspect, a computer-readable storage medium is provided, on which computer program instructions are stored; when the computer program instructions are executed by a processor, the road elevation detection method as described in the embodiment of the first aspect is implemented.
[0011] In a third aspect, a road elevation detection system is provided, comprising: a computer-readable storage medium as described in the embodiment of the second aspect.
[0012] In this way, the embodiment of the present invention, by providing a dual laser transmitter with the characteristics of intensiveness, wide field of view and multi-function, can be widely used on various small carriers, and can realize the surface measurement of the entire lane; the image taken by the high-definition camera of the optical camera is used as a reference to identify and correct the discrete noise data in the reflected laser signal collected by the 3D depth camera, thereby improving the accuracy of 3D measurement; through image verification, the actual ground feature data can be retained to the maximum extent, and the error data caused by dust, light and other factors can be effectively eliminated. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
[0014] Figure 1 is a flow chart of a road elevation detection method according to an embodiment of the present invention;
[0015] Figure 2 is a schematic diagram of a detection vehicle according to an embodiment of the present invention;
[0016] Figure 3 is a schematic diagram of a dual laser transmitter and 3D depth camera detection according to an embodiment of the present invention;
[0017] Figure 4 is a schematic diagram of the structure of a dual laser transmitter and 3D depth camera detection according to an embodiment of the present invention;
[0018] Figure 5 It is a schematic diagram of a dual laser transmitter emitting two line structured lights according to an embodiment of the present invention. DETAILED DESCRIPTION
[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0020] The embodiment of the present invention discloses a method for detecting road elevation. Figure 1 As shown, the method of the embodiment of the present invention includes the following steps:
[0021] Step S101: When the inspection vehicle is driving on the road, the dual laser emitters carried by the inspection vehicle are controlled to simultaneously emit two lines of structured light toward the road.
[0022] like Figures 2 to 5 As shown, in an embodiment of the present invention, a mounting bracket can be provided at the center of the upper part of the rear end of the inspection vehicle. A dual laser emitter can be installed at one end of the mounting bracket connected to the inspection vehicle. The dual laser emitter starts from the center point and simultaneously emits two line structured lights to both sides. In a specific embodiment, the illumination width of each structured light of the dual laser emitter is 0.5×(3.3~4.1)m, and the total illumination width is (3.3~4.1)m.
[0023] Step S102: synchronously controlling the optical camera carried by the inspection vehicle to collect road surface images, and the 3D depth camera carried by the inspection vehicle to collect reflected laser signals of the line structured light reflected by the road surface.
[0024] The optical camera and 3D depth camera of the embodiment of the present invention can be installed on the end of the mounting bracket away from the inspection vehicle. The optical camera adopts a high-definition camera. The 3D depth camera and the dual laser emitters together constitute a full-frame surface sensing device. The 3D depth camera usually adopts the time of flight (TOF) principle or the structured light principle to measure the depth information of the surface of the object to be measured. Taking the TOF principle as an example, the 3D depth camera measures the time difference from the emission of the emitted laser pulse to the reflection of the surface of the object to be measured to the 3D depth camera. Based on the known speed of light, the distance from the surface of the object to the camera can be calculated. The structured light principle reconstructs the three-dimensional shape of the object by projecting a special light coding pattern onto the surface of the object to be measured and analyzing the deformation of the coding pattern. Both principles can obtain three-dimensional data of the surface of the object.
[0025] The road image is denoted as I, with a size of M×N and a pixel value range of [0,255]. Suppose the reflected laser signal collected by the 3D depth camera is D, with a size of P×Q and a value range of [d min ,d max ], represents the distance from the 3D depth camera to the object being measured. Therefore, for the pixel point (x, y) in the road image, x∈[1,M], y∈[1,N], for the reflected laser signal at (p,q) collected by the 3D depth camera, p∈[1,P], q∈[1,Q].
[0026] Step S103: identifying a normal signal and an error signal in the reflected laser signal through the road surface image and the corresponding reflected laser signal.
[0027] Specifically, this step includes the following process:
[0028] 1. Convert each pixel of the road surface image into a first matrix, and convert the reflected laser signal into a second matrix.
[0029] Each first element in the first matrix is a pixel value of a pixel point at a corresponding position, and each second element in the second matrix is an intensity of a reflected laser signal at a corresponding position.
[0030] For road images, common image matrixing methods can be used. The image is divided into grids of equal size (equivalent to the pixel points described in the embodiment of the present invention), and each grid corresponds to an element in the matrix. The value of the element can be the average gray value, RGB value, etc. of the pixels in the grid area. Common image matrixing algorithms include bilinear interpolation, nearest neighbor interpolation, etc.
[0031] For the reflected laser signal, the original reflected laser signal needs to be preprocessed first to remove obvious outliers. Then, according to the resolution of the camera, the reflected laser signal is discretized into depth values corresponding to the camera pixels one by one. Each depth value corresponds to an element in the second matrix. It should be understood that due to the limitations of the field of view and resolution of the 3D depth camera, valid depth values may not be obtained in certain areas of the object surface, which will correspond to null values or default values in the second matrix.
[0032] Specifically, the first matrix is as follows:
[0033]
[0034] Specifically, the second matrix is as follows:
[0035]
[0036] Wherein, I(i,j) is the first element of the i-th row and j-th column in the first matrix, I(x,y) is the pixel value of the pixel point with coordinates (x,y) in the road surface image, M is the length of the road surface image, N is the width of the road surface image, [*] is a rounding operation, D(k,l) is the second element of the k-th row and l-th column in the second matrix, and D(p,q) is the intensity of the reflected laser signal at (p,q) collected by the 3D depth camera.
[0037] 2. Establish a mapping relationship between each second element in the second matrix and each first element in the first matrix.
[0038] This step requires camera calibration, that is, the three-dimensional space coordinates corresponding to each pixel are calculated based on the intrinsic parameters of the optical camera (such as focal length, principal point coordinates, etc.) and extrinsic parameters (position of the optical camera in the world coordinate system). Commonly used camera calibration algorithms include Zhang Zhengyou calibration method, OpenCV calibration method, etc. After calibration, a mapping function can be obtained to map the elements (pixel coordinates) in the first matrix to points in three-dimensional space. Then the intensity values in the second matrix need to be converted into three-dimensional space coordinates. This can be achieved by substituting the intensity values into the above mapping function. Finally, the three-dimensional space coordinates of each element in the second matrix are matched with the three-dimensional space coordinates of the corresponding element in the first matrix to establish a mapping relationship between elements.
[0039] Specifically, this step includes the following process:
[0040] (1) Calculate the first three-dimensional space point corresponding to each first element in the first matrix and the second three-dimensional space point corresponding to each second element in the second matrix.
[0041] Specifically, the calculation formula for the first three-dimensional space point is as follows:
[0042]
[0043] Specifically, the calculation formula for the second three-dimensional space point is as follows:
[0044]
[0045] in, K is the intrinsic parameter matrix of the 3D depth camera, f x 、f y are the focal lengths in the x and y directions, respectively, (c x ,c y ) is the coordinate of the principal point, R is the rotation matrix of the 3D depth camera in the world coordinate system, t is the translation matrix of the 3D depth camera in the world coordinate system, (u, v) is the two-dimensional position coordinate of (i, j) in the first matrix in the camera coordinate system, (x c ,y c ) is the two-dimensional position coordinate of (k,l) in the second matrix in the camera coordinate system.
[0046] (2) Traverse the first three-dimensional space points and obtain the first three-dimensional space point closest to each second three-dimensional space point.
[0047] The distance between two points can be calculated using the Euclidean distance formula.
[0048] (3) Establishing a mapping relationship between a second element in the second matrix corresponding to the second three-dimensional space point and a first element in the first matrix corresponding to the first three-dimensional space point closest to the second three-dimensional space point.
[0049] When two points are closest, the two points form a corresponding mapping relationship. Through this step, the corresponding relationship between the intensity data and the image data is established, which lays the foundation for the subsequent fusion processing.
[0050] 3. Based on the mapping relationship, determine whether the preset window centered on each second element in the second matrix has continuity, and preliminarily determine the second element at the center of the preset window with continuity as a noise element.
[0051] For each non-empty element in the second matrix, check whether the values of the elements around the corresponding position in the first matrix are continuous. Specifically, a window size can be set, and the difference of element values can be calculated within the window. If the difference is less than a certain threshold, it is considered that the element and its surrounding elements are continuous.
[0052] Specifically, determining whether the preset window centered on each second element in the second matrix has continuity includes the following process:
[0053] (1) For each second element in the second matrix, calculate the eigenvalue of the second element within a preset window centered on the second element.
[0054] The characteristic values include, but are not limited to, variance, absolute difference, relative difference, maximum value or minimum value, which can be selected according to actual conditions.
[0055] (2) If the feature value of the second element in the preset window centered on the second element is less than a first preset feature threshold, then it is determined that the preset window centered on the second element has continuity; otherwise, it is determined that the preset window centered on the second element has no continuity.
[0056] The preset window can be expressed as follows:
[0057]
[0058] Wherein, w is the width of the preset window, and h is the height of the preset window.
[0059] The first preset feature threshold may be adjusted according to the requirements of a specific application scenario. For example, for variance, the first preset feature threshold is usually between 0.1 and 0.5.
[0060] For example, when the eigenvalue is the variance, the eigenvalue of the second element in the preset window is calculated as follows:
[0061]
[0062] in, is the variance of the second element within the preset window centered at the second element (k, l).
[0063] This step can be used to preliminarily determine whether there are discrete points in the second element that may be noise.
[0064] 4. Cluster the second elements in the second matrix to determine whether each category has continuity, and preliminarily determine the second elements in the category with continuity as noise elements.
[0065] Clustering is performed on the second elements in the region surrounded by the second elements having continuity in the second matrix. Commonly used clustering algorithms include but are not limited to K-Means clustering, Mean-Shift clustering, DBSCAN clustering and other algorithms.
[0066] Taking the DBSCAN algorithm as an example, the advantage of this algorithm is that it can automatically discover clusters of any shape and effectively identify noise points.
[0067] The DBSCAN algorithm contains two important parameters: ∈ and minPts. ∈ is the neighborhood radius, that is, when the distance between two points is less than ∈, the two points are considered to be adjacent. minPts is the minimum number of neighboring points of the core point. The specific steps of the DBSCAN algorithm are as follows:
[0068] (1) Calculate the distance dist(p,q) between any two points p and q.
[0069] This distance can be calculated using the Euclidean distance formula between two points.
[0070] (2) Traverse each point p in D and calculate the number of points n in its ∈ neighborhood p .
[0071] (3) If n p ≥minPts, then p is called a core point.
[0072] (4) For each core point p, all points within its ∈ neighborhood are classified into the same category.
[0073] (5) For each non-core point q, if it is adjacent to a core point p, q is classified into the category of p.
[0074] (6) Points that are not adjacent to any core points are classified as noise points and marked as noise points.
[0075] For each clustered category, calculate the statistical characteristic values of the elements in the category, such as mean, variance, maximum value, minimum value, etc. If the characteristic value of the elements within the category is small, it means that the elements of the category have good continuity. Then the category is judged as an actual ground feature, and the data of all elements in the category are retained. On the contrary, if the characteristic value of the elements in a category is large, it means that the category is composed of some discrete and discontinuous points, and the category is preliminarily judged as noise data. It is worth noting that a reasonable threshold needs to be set in the judgment process to balance the relationship between noise removal and feature retention. If the threshold is set too large, the actual ground feature data may also be judged as noise and eliminated; if the threshold is set too small, the noise may not be effectively removed.
[0076] Specifically, this step includes the following process:
[0077] (1) Calculate the eigenvalue of the second element of each category.
[0078] The characteristic values include, but are not limited to, variance, absolute difference, relative difference, maximum value or minimum value, which can be selected according to actual conditions.
[0079] (2) If the characteristic value of the second element of the category is less than the second preset characteristic threshold, the category is determined not to be a noise category; otherwise, the category is determined to be a noise category.
[0080] Taking the variance as an example, after clustering, if the variance of the internal elements of a category is less than the second preset feature threshold τ c , then the category is judged as an actual ground feature, and all elements in the category are retained; conversely, if the internal variance of a category is greater than the second preset feature threshold τ c , then the category is determined to be a noise category. The setting of the second preset feature threshold needs to be adjusted in combination with the specific application scenario, and the variance threshold is usually between 0.1 and 1.
[0081] In addition, it should be understood that, taking the DBSCAN algorithm as an example, it will obtain a class of points that are not adjacent to any core points. These points are classified as noise categories, and this category does not have continuity.
[0082] This step preliminarily distinguishes noise data from actual ground feature data.
[0083] 6. For the second element preliminarily determined to be a noise element, the pixel information of the corresponding area of the road image is combined to determine whether there is a physical entity that matches the noise element in the corresponding area. If there is no physical entity that matches the noise element in the corresponding area, the noise element is marked as an error element.
[0084] For the second element initially determined to be a noise element, combined with the pixel information of the corresponding area of the road image, if there is a physical entity consistent with the noise element in the area, the noise element is retained, otherwise it is marked as an error element.
[0085] First, according to the established mapping relationship, the position of the second element preliminarily determined as a noise element in the second matrix is mapped to the corresponding position of the first matrix. Then, for each second element determined as a noise element, the neighborhood image block corresponding to its position on the road surface image is extracted. Through image processing algorithms, such as edge detection, texture analysis, etc., it is determined whether there is a physical entity in the image block that matches the noise element. If there is indeed a matching physical entity, it means that the second element is not noise, but an actual ground feature point, which should be retained. On the contrary, if there is no physical entity in the image block that matches the second element, it will be marked as an error point.
[0086] Specifically, this step includes the following process:
[0087] (1) According to the mapping relationship, determine the first element in the first matrix corresponding to the mapping relationship of the second element preliminarily determined to be a noise element.
[0088] Assume that the second element preliminarily determined to be a noise element is (k, l), and map it to the position (i, j) on the first matrix I according to the established mapping relationship.
[0089] (2) Acquire an image block of a preset size in the road surface image with a pixel point in the road surface image corresponding to the first element in the first matrix as the center.
[0090] Extract the image block B centered at (i, j) from the road image I ij , the process can be expressed as follows:
[0091] B ij ={I(i+m,j+n)|m∈[-r,r],n∈[-r,r]}.
[0092] Where r is the image patch radius.
[0093] (3) Calculate the eigenvalues of the image block.
[0094] The characteristic value of the image block includes, but is not limited to: the edge strength of the image block or the texture similarity of the image block.
[0095] Taking the edge strength as an example, edge detection is performed on the image block to obtain the edge strength, which is calculated as follows:
[0096]
[0097] Among them, E ij is the edge strength, For the gradient operator, operators such as Sobel, Prewitt or Canny can be used.
[0098] (4) If the feature value of the image block is greater than a third preset feature threshold, the second element in the second matrix corresponding to the center is determined as a physical entity; otherwise, the second element in the second matrix corresponding to the center is determined as an error element.
[0099] The third preset feature threshold can be reasonably set as needed. According to different feature values of the image block, the third preset feature threshold includes but is not limited to: edge detection threshold or texture similarity threshold. The setting of the third preset feature threshold needs to be adjusted in combination with the specific scene to achieve the best noise removal effect.
[0100] The third feature preset threshold is used as the edge detection threshold δ e Take the following as an example:
[0101] If E ij >δ e , then it is considered that there is an obvious physical entity edge at (i,j), and the corresponding depth value D(k,l) is retained, otherwise, (k,l) is marked as an error element. Therefore, by using the image edge information to further identify noise points, it is possible to avoid erroneously removing the actual ground feature points.
[0102] 7. Based on the marking of the error elements in the second matrix, a third matrix is established, and according to the third matrix, the corresponding reflected laser signal is marked as a normal signal or an error signal.
[0103] Specifically, a third matrix for marking can be used, which has the same dimension as the second matrix. Initially, the values of all elements of the third matrix are set to 0. Then, each second element in the second matrix is traversed. If the element is marked as an error element, the value of the third matrix element at the corresponding position is set to 1 or other marking value, and the third matrix corresponding to the second matrix with errors can be obtained.
[0104] According to the third matrix marked with errors, the original reflected laser signal is error marked for error marking of the final calculated ground features. First, it is necessary to establish a mapping relationship between the third matrix marked with errors and the original reflected laser signal. This mapping relationship is the corresponding relationship of the aforementioned reflected laser signal converted into the second matrix. Then traverse each element in the third matrix, and if the element is marked as an error element, the corresponding original reflected laser signal is also marked as an error signal. The marking method can be to add special marks, modify the data format, etc. Finally, when the elevation is finally calculated, the normal reflected laser signal is used for calculation, and the error signal is marked as an error point, thereby realizing the error marking of the final calculation result.
[0105] Specifically, this step includes the following process:
[0106] (1) Based on the determined error elements, a third matrix with the same dimension as the second matrix is constructed.
[0107] The value of the third element in the third matrix at the same position as the error element is 1, and the values of other third elements are 0.
[0108] Construct a third matrix M with the same dimensions as the second matrix D, with all initial values of the elements set to 0. For (k, l) marked as error elements, set M(k, l) = 1, and for the retained points, set M(k, l) = 0. That is:
[0109]
[0110] Finally, the third matrix M marked with errors is obtained.
[0111] This step separates the error points from the normal points in the depth data.
[0112] (2) The reflected laser signals corresponding to the third elements with values of 1 and 0 in the third matrix are marked as error signals and normal signals, respectively.
[0113] According to the mapping relationship between the second matrix D and the original reflected laser signal D, each element (k, l) in the third matrix M is traversed. If M(k, l) = 1, the corresponding original reflected laser signal D(p, q) is marked as an error signal. If M(k, l) = 0, the corresponding original reflected laser signal D(p, q) is marked as a normal signal.
[0114] This step feeds back the noise point information identified in the previous step to the original data, so that subsequent processing can effectively avoid erroneous data.
[0115] Step S104: Calculate the road surface elevation using the normal signal in the reflected laser signal.
[0116] Specifically, the depth of the corresponding position can be calculated by the intensity of the reflected laser signal, and the elevation of the road surface can be calculated by the depth. This calculation is a prior art and will not be described in detail here.
[0117] In addition, an embodiment of the present invention further provides a computer-readable storage medium, on which computer program instructions are stored; when the computer program instructions are executed by a processor, the road elevation detection method as described in the above embodiment is implemented.
[0118] In addition, an embodiment of the present invention further provides a road elevation detection system, comprising: a computer-readable storage medium as described in the above embodiment.
[0119] In summary, the embodiments of the present invention, by providing a dual laser transmitter with the characteristics of intensiveness, wide field of view and multi-function, can be widely used on various small carriers and can realize the surface measurement of the entire lane; the image captured by the high-definition camera of the optical camera is used as a reference to identify and correct the discrete noise data in the reflected laser signal collected by the 3D depth camera, thereby improving the accuracy of the 3D measurement; through image verification, the actual ground feature data can be retained to the maximum extent, while the error data caused by dust, light and other factors can be effectively eliminated.
[0120] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A road elevation detection method, It is characterized in that include: When the inspection vehicle is driving on the road, the dual laser emitters carried by the inspection vehicle are controlled to simultaneously emit two lines of structured light toward the road surface; Synchronously controlling the optical camera carried by the inspection vehicle to collect road surface images, and the 3D depth camera carried by the inspection vehicle to collect reflected laser signals of the line structured light reflected by the road surface; Identify a normal signal and an error signal in the reflected laser signal through the road surface image and the corresponding reflected laser signal; Calculating the road surface elevation using the normal signal in the reflected laser signal; The step of identifying a normal signal and an error signal in the reflected laser signal through the road surface image and the corresponding reflected laser signal comprises: Convert each pixel point of the road surface image into a first matrix, and convert the reflected laser signal into a second matrix, wherein each first element in the first matrix is a pixel value of a pixel point at a corresponding position, and each second element in the second matrix is an intensity of the reflected laser signal at a corresponding position; Establishing a mapping relationship between each second element in the second matrix and each first element in the first matrix; Based on the mapping relationship, determining whether a preset window centered on each second element in the second matrix has continuity, and preliminarily determining the second element at the center of the preset window with continuity as a noise element; Clustering the second elements in the second matrix, determining whether each category has continuity, and preliminarily determining the second elements in the category with continuity as noise elements; For the second element preliminarily determined to be a noise element, combining pixel information of a corresponding area of the road surface image, determining whether there is a physical entity that matches the noise element in the corresponding area; if there is no physical entity that matches the noise element in the corresponding area, marking the noise element as an error element; Based on the marking of the error elements in the second matrix, a third matrix is established, and according to the third matrix, the corresponding reflected laser signal is marked as a normal signal or an error signal; The step of determining whether there is a physical entity in the corresponding area that matches the noise element comprises: Determine, according to the mapping relationship, a first element in the first matrix corresponding to a mapping relationship of a second element preliminarily determined to be a noise element; Acquire, in the road surface image, an image block of a preset size centered at a pixel point in the road surface image corresponding to the first element in the first matrix; Calculating a characteristic value of the image block; If the eigenvalue of the image block is greater than a third preset eigenvalue threshold, determining the second element in the second matrix corresponding to the center as a physical entity; otherwise, determining the second element in the second matrix corresponding to the center as an error element; The step of marking the corresponding reflected laser signal as a normal signal or an error signal comprises: Based on the determined error element, construct a third matrix of the same dimension as the second matrix, wherein the value of the third element at the same position as the error element in the third matrix is 1, and the value of other third elements is 0; The reflected laser signals corresponding to the third elements with values of 1 and 0 in the third matrix are marked as error signals and normal signals, respectively.
2. The road elevation detection method according to claim 1, It is characterized in that The first matrix is: The second matrix is: Wherein, I(i,j) is the first element of the i-th row and j-th column in the first matrix, I(x,y) is the pixel value of the pixel point with coordinates (x,y) in the road surface image, M is the length of the road surface image, N is the width of the road surface image, [*] is a rounding operation, D(k,l) is the second element of the k-th row and l-th column in the second matrix, and D(p,q) is the intensity of the reflected laser signal at (p,q) collected by the 3D depth camera.
3. The road elevation detection method according to claim 1, It is characterized in that The step of establishing a mapping relationship between each second element in the second matrix and each first element in the first matrix comprises: Calculate a first three-dimensional space point corresponding to each first element in the first matrix, and a second three-dimensional space point corresponding to each second element in the second matrix; Traversing the first three-dimensional space points, and obtaining the first three-dimensional space point closest to each of the second three-dimensional space points; Establishing a mapping relationship between a second element in the second matrix corresponding to the second three-dimensional space point and a first element in the first matrix corresponding to the first three-dimensional space point closest to the second three-dimensional space point; Wherein, the calculation formula of the first three-dimensional space point includes: The calculation formula of the second three-dimensional space point includes: in, K is the intrinsic parameter matrix of the optical camera, f x 、f y are the focal lengths in the x and y directions, respectively, (c x ,c y ) is the coordinate of the principal point, R is the rotation matrix of the 3D depth camera in the world coordinate system, t is the translation matrix of the 3D depth camera in the world coordinate system, (u, v) is the two-dimensional position coordinate of (i, j) in the first matrix in the camera coordinate system, (x c ,y c ) is the two-dimensional position coordinate of (k,l) in the second matrix in the camera coordinate system.
4. The road elevation detection method according to claim 1, It is characterized in that The step of determining whether the preset window centered on each second element in the second matrix has continuity comprises: For each second element in the second matrix, calculating an eigenvalue of the second element within a preset window centered on the second element, wherein the eigenvalue includes: variance, absolute difference, relative difference, maximum value or minimum value; If the characteristic value of the second element in the preset window centered on the second element is less than the first preset characteristic threshold, the preset window centered on the second element is determined to be continuous; otherwise, the preset window centered on the second element is determined to be discontinuous.
5. The road elevation detection method according to claim 1, It is characterized in that The step of determining whether each category is continuous includes: Calculate the characteristic value of the second element of each category, wherein the characteristic value includes: variance, absolute difference, relative difference, maximum value or minimum value; If the characteristic value of the second element of the category is less than the second preset characteristic threshold, the category is determined to be continuous; otherwise, the category is determined to be discontinuous.
6. A computer-readable storage medium, Features: The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, the road elevation detection method according to any one of claims 1 to 5 is implemented.
7. A road elevation detection system, It is characterized in that include: The computer readable storage medium of claim 6.
Citation Information
Patent Citations
Real-time identification system and method of road surface textures for unmanned vehicle
CN108960060A
Road surface and automatic extraction method of marker lines thereof
CN111079611A
Pavement three-dimensional detection system and method based on bidirectional line structured light
CN114910046A
Road surface three-dimensional data measuring method and measuring system
CN116429023A
Pavement elevation three-dimensional mesh value detection system
CN203684048U
