An obstacle monitoring method, device, apparatus and storage medium

Abstract

The application discloses an obstacle monitoring method, device and equipment and a storage medium. The method comprises the following steps: acquiring point cloud data collected by a laser radar, including at least two screening boxes; generating at least one near-view boundary curve based on the at least two screening boxes; generating at least one far-view fitting curve in a continuously iterative manner based on the at least two screening boxes; obtaining an orbit boundary curve based on the near-view boundary curve and the far-view fitting curve; screening obstacle point cloud from the point cloud data according to at least one orbit boundary curve; and determining obstacle information according to the obstacle point cloud. According to the technical scheme of the application, the calculation amount of the system can be greatly reduced, and the accuracy of obstacle monitoring can be improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle technology, and in particular to an obstacle monitoring method, device, equipment and storage medium. Background Technology

[0002] Railway transportation is an important carrier of people's movement and is widely recognized as a comfortable, fast and safe mode of transportation.

[0003] The complex environment of railway construction areas poses a significant challenge to train safety. Existing methods for detecting obstacles on tracks can be broadly categorized into two types: pure vision algorithms based on image data and pure point cloud detection algorithms based on lidar.

[0004] Pure vision algorithms based on image data are prone to false positives or false negatives because cameras are very sensitive to light and shadow. During the day, shadows caused by objects blocking the light, and poor lighting at night, result in low image quality.

[0005] Currently, most pure point cloud detection algorithms based on LiDAR detect obstacles by filtering out backgrounds in consecutive frames or matching them with offline maps. This method requires calculations using all the collected point clouds, resulting in high computational complexity and difficulty. Furthermore, the accuracy of obstacle detection is relatively low because the collected point clouds may contain interfering points. Summary of the Invention

[0006] This invention provides an obstacle monitoring method, device, equipment, and storage medium. By fitting point cloud data collected by lidar to obtain a track boundary curve, obstacle point clouds are obtained from the point cloud data based on the track boundary curve, thereby determining obstacle information. This method can greatly reduce the computational load of the system and improve the accuracy of obstacle monitoring.

[0007] According to one aspect of the present invention, an obstacle detection method is provided, comprising:

[0008] Acquire point cloud data collected by LiDAR, including at least two filter boxes;

[0009] Generate at least one near-field boundary curve based on the at least two filter boxes;

[0010] Based on the at least two filter boxes, at least one vision fitting curve is generated through continuous iteration;

[0011] The track boundary curve is obtained based on the near-view boundary curve and the far-view fitted curve;

[0012] Obstacle point clouds are filtered from the point cloud data based on at least one track boundary curve;

[0013] Obstacle information is determined based on the obstacle point cloud.

[0014] According to another aspect of the present invention, an obstacle detection device is provided, the obstacle detection device comprising:

[0015] The point cloud data acquisition module is used to acquire point cloud data collected by LiDAR, including at least two filter boxes;

[0016] A near-field boundary curve generation module is used to generate at least one near-field boundary curve based on the at least two filter boxes;

[0017] The vision fitting curve generation module is used to generate at least one vision fitting curve based on the at least two filter boxes through continuous iteration.

[0018] The track boundary curve generation module is used to obtain the track boundary curve based on the near-view boundary curve and the far-view fitting curve.

[0019] An obstacle point cloud filtering module is used to filter obstacle point clouds from the point cloud data based on at least one track boundary curve;

[0020] An obstacle information determination module is used to determine obstacle information based on the obstacle point cloud.

[0021] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0022] At least one processor; and

[0023] A memory communicatively connected to the at least one processor; wherein,

[0024] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the obstacle detection method according to any embodiment of the present invention.

[0025] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the obstacle detection method according to any embodiment of the present invention.

[0026] This invention, through the acquisition of point cloud data collected by a LiDAR, includes at least two filtering boxes; at least one near-field boundary curve is generated based on the at least two filtering boxes; at least one far-field fitting curve is generated iteratively based on the at least two filtering boxes; a track boundary curve is obtained based on the near-field boundary curve and the far-field fitting curve; obstacle point clouds are filtered from the point cloud data according to the at least one track boundary curve; and obstacle information is determined based on the obstacle point clouds. This solves the problem of false detection or missed detection caused by the camera's high sensitivity to light and shadow, shadows caused by objects blocking the image during the day, and poor image quality due to poor lighting at night. By fitting the track boundary curve from the point cloud data collected by the LiDAR, and then filtering the obstacle point cloud from the point cloud data according to the track boundary curve to determine obstacle information, the computational load of the system can be greatly reduced and the accuracy of obstacle detection can be improved.

[0027] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart of an obstacle detection method according to an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of the structure of an obstacle monitoring device according to an embodiment of the present invention;

[0031] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] Example 1

[0035] Figure 1 This is a flowchart illustrating an obstacle detection method provided in an embodiment of the present invention. This embodiment is applicable to obstacle detection situations. The method can be executed by the obstacle detection device in this embodiment, which can be implemented in software and / or hardware, such as... Figure 1 As shown, the method specifically includes the following steps:

[0036] S110: Acquire point cloud data collected by LiDAR, including at least two filter boxes.

[0037] To improve accuracy and obtain as much ground point cloud data as possible, the lidar can be installed in the middle of the train's roof.

[0038] The filter box includes a certain number of point clouds. The filter box can be a cube or a box of other shapes; this embodiment of the invention does not impose any limitations on this.

[0039] Specifically, the method for acquiring point cloud data collected by LiDAR can be as follows: Initial point cloud data is acquired using LiDAR; at least two filter boxes are determined based on the initial point cloud data; and point cloud data is then determined based on the at least two filter boxes. Alternatively, the method can be as follows: The center point coordinates of the first filter box are set; the centroids of all point clouds within the corresponding filter box are obtained based on the first filter box; the center point coordinates of the second filter box are determined based on the centroids of all point clouds within the first filter box; and this process is repeated until the center point coordinates of the Nth filter box are determined based on the centroids of all point clouds within the (N-1)th filter box. Another method for acquiring point cloud data from LiDAR can be as follows: The center point coordinates of the first left-hand filter box are set; the centroids of all point clouds within the corresponding filter box are obtained based on the first left-hand filter box; the center point coordinates of the second left-hand filter box are determined based on the centroids of all point clouds within the first left-hand filter box; and this process is repeated until the center point coordinates of the Nth left-hand filter box are determined based on the centroids of all point clouds within the (N-1)th left-hand filter box. Set the center point coordinates of the first filter box on the right; obtain the centroid of all point clouds in the corresponding filter box based on the first filter box on the right; determine the center point coordinates of the second filter box on the right based on the centroid of all point clouds in the first filter box on the right; repeatedly execute the process of determining the center point coordinates of the Nth filter box on the right based on the centroid of all point clouds in the (N-1)th filter box on the right. To improve speed, the left and right filter boxes can be determined in parallel, but this embodiment of the invention does not impose any restrictions on this.

[0040] S120, generate at least one near-field boundary curve based on the at least two filter boxes.

[0041] Specifically, the method for generating at least one close-up boundary curve based on the at least two filter boxes can be as follows: generate at least one close-up boundary curve based on the center point coordinates of the at least two filter boxes. Alternatively, the method can be as follows: generate the close-up boundary curve of the left track based on the center point coordinates of the N filter boxes corresponding to the left track, and generate the close-up boundary curve of the right track based on the center point coordinates of the N filter boxes corresponding to the right track. Another possible method is: generate a first target set based on the center point coordinates of the N filter boxes corresponding to the left track; generate a second target set based on the center point coordinates of the N filter boxes corresponding to the right track; generate the close-up boundary curve of the left track based on the first target set; and generate the close-up boundary curve of the right track based on the second target set.

[0042] S130, Based on the at least two filter boxes, at least one vision fitting curve is generated through continuous iteration.

[0043] Specifically, based on the at least two filter boxes, the method of generating at least one distant view fitting curve through continuous iteration can be as follows: When calculating the first distant view point, it is obtained by calculating all the preceding near view points; when calculating the second distant view point, it is obtained by calculating all the preceding near view points plus the first distant view point; starting from the third distant view point, it is calculated according to the following rules: if M is greater than the threshold for the number of filter boxes, then a fitting curve is generated by the center point coordinates of the first M-1 filter boxes, and the center point coordinates of the Mth filter box are obtained in the extension direction of the fitting curve, where M is a positive integer not less than 3; the above operation is repeated until the constraint condition is met, and a distant view fitting curve is generated according to the center point coordinates of all filter boxes obtained in the extension direction of the fitting curve; the above operation is repeated to obtain the distant view fitting curves of the left track and the right track. Based on the at least two filter boxes, the method of generating at least one distant view fitting curve through continuous iteration can also be as follows: When calculating the first distant view point, it is obtained by calculating all the preceding near view points; when calculating the second distant view point, it is obtained by calculating all the preceding near view points plus the first distant view point; starting from the third distant view point, it is calculated based on the following rules: if the number of point clouds in the Mth filter box is less than the number of point clouds threshold, then a fitting curve is generated by the center point coordinates of the preceding M-1 filter boxes, and the center point coordinates of the Mth filter box are obtained in the extension direction of the fitting curve, where M is a positive integer not less than 3; the above operation is repeated until the constraint condition is met, and a distant view fitting curve is generated based on the center point coordinates of all filter boxes obtained in the extension direction of the fitting curve; the above operation is repeated to obtain the distant view fitting curve of the left track and the distant view fitting curve of the right track. Based on the at least two filter boxes, the method of generating at least one distant view fitting curve through continuous iteration can also be as follows: When calculating the first distant view point, it is obtained by calculating all the preceding near view points; when calculating the second distant view point, it is obtained by calculating all the preceding near view points plus the previous distant view point; starting from the third distant view point, it is calculated based on the following rules: if there is no point cloud data in the Mth filter box, then a fitting curve is generated by using the center point coordinates of the preceding M-1 filter boxes, and the center point coordinates of the Mth filter box are obtained in the extension direction of the fitting curve, where M is a positive integer not less than 3; the above operation is repeated until the constraint conditions are met, and a distant view fitting curve is generated based on the center point coordinates of all filter boxes obtained in the extension direction of the fitting curve; the above operation is repeated to obtain the distant view fitting curves of the left track and the right track.

[0044] S140, the track boundary curve is obtained based on the near-field boundary curve and the far-field fitting curve.

[0045] The near-field boundary curve may include a left-side near-field boundary curve and a right-side near-field boundary curve.

[0046] The distant view fitting curve may include: a distant view fitting curve on the left and a distant view fitting curve on the right.

[0047] It should be noted that the near-field curve refers to the curve within the detection range of the lidar, while the far-field curve is the curve outside the detection range of the lidar.

[0048] The track boundary curves include: a left track boundary curve and a right track boundary curve. The track boundary curves may also include: an upper track boundary curve. This embodiment of the invention does not impose any limitations on this.

[0049] Specifically, the method for obtaining the track boundary curve based on the near-view boundary curve and the far-view fitting curve can be as follows: A left-side boundary fitting curve is obtained based on the left-side near-view boundary curve and the far-view fitting curve; a right-side boundary fitting curve is obtained based on the right-side near-view boundary curve and the far-view fitting curve; the track boundary curve is obtained based on the left-side boundary fitting curve and / or the right-side boundary fitting curve. For example, if the user selects the left-side boundary fitting curve, the left-side boundary fitting curve is translated, and the track boundary curve is determined based on the translated curve and the left-side boundary fitting curve. If the user selects both the left-side and right-side boundary fitting curves, the track boundary curve is determined based on both.

[0050] The method for obtaining the track boundary curve based on the near-view boundary curve and the far-view fitting curve can also be as follows: obtain the left boundary fitting curve based on the left near-view boundary curve and the far-view fitting curve; obtain the right boundary fitting curve based on the right near-view boundary curve and the far-view fitting curve; calculate the confidence level of the left boundary fitting curve and the confidence level of the right boundary fitting curve, and compare their magnitudes; define the boundary fitting curve with higher confidence as the track boundary curve, and obtain the track boundary curve on the other side by translation.

[0051] The method for obtaining the track boundary curve based on the near-view boundary curve and the far-view fitting curve can also be as follows: Obtain the left boundary fitting curve based on the left near-view boundary curve and the far-view fitting curve; obtain the right boundary fitting curve based on the right near-view boundary curve and the far-view fitting curve; calculate the confidence scores of the left and right boundary fitting curves and compare their magnitudes; define the boundary fitting curve with the higher confidence score as the track boundary curve, and obtain the track boundary curve on the other side by translation; if the confidence score of the left boundary fitting curve is higher than that of the right boundary fitting curve, then generate the track upper boundary curve based on the point cloud with Z-coordinate values ​​greater than the coordinate value threshold in the N filter boxes corresponding to the left track; if the confidence score of the right boundary fitting curve is higher than that of the right boundary fitting curve, then generate the track upper boundary curve based on the point cloud with Z-coordinate values ​​greater than the coordinate value threshold in the N filter boxes corresponding to the right track.

[0052] S150, filter obstacle point clouds from the point cloud data based on the at least one track boundary curve.

[0053] The obstacle point cloud refers to the point cloud within the track.

[0054] Specifically, the method for filtering obstacle point clouds from the point cloud data based on at least one track boundary curve can be as follows: based on the left track boundary curve and the right track boundary curve, point clouds located within the track are filtered from the point cloud data, and the point clouds within the track are identified as obstacle point clouds.

[0055] Another method for filtering obstacle point clouds from the point cloud data based on at least one track boundary curve is as follows: Based on the left track boundary curve, the right track boundary curve, and the upper track boundary curve, point clouds located within the track are selected from the point cloud data, and these point clouds are identified as obstacle point clouds. For example, since the left, right, and upper track boundary curves have been identified, the (x, y, z) coordinates of all point clouds are compared with these curves to filter out point clouds located within the track. It should be noted that when the number of filtered point clouds is greater than 0, it is determined that there are obstacles on the track. The number of obstacles can be determined using a clustering algorithm; for example, the number of clusters corresponds to the number of obstacles.

[0056] S160, Determine obstacle information based on the obstacle point cloud.

[0057] Specifically, the method for determining obstacle information based on obstacle point clouds can be as follows: clustering the obstacle point clouds to obtain at least one point cloud set; determining obstacle information based on the position coordinates and parameter information of the bounding boxes corresponding to each point cloud set. Alternatively, the method can be as follows: clustering the obstacle point clouds to obtain at least one point cloud set; acquiring the target position information and feature information of obstacles in the image to be identified captured by a camera; projecting the position coordinates of each bounding box onto the image to be identified to obtain first position information of each bounding box relative to the image to be identified; fusing the feature information of the obstacle and the parameter information of the bounding boxes based on the first position information and the target position information to obtain obstacle information.

[0058] Optionally, the acquisition of point cloud data collected by the lidar includes at least two filter boxes, including:

[0059] Set the coordinates of the center point of the first filter box;

[0060] Based on the first filter box, obtain the centroids of all point clouds in the corresponding filter box;

[0061] The center point coordinates of the second filter box are determined based on the centroids of all point clouds in the first filter box;

[0062] The process is repeated until the center point coordinates of the Nth filter box are obtained by determining the center point coordinates of the Nth filter box based on the centroids of all point clouds in the (N-1)th filter box, and so on. Here, N is a positive integer not less than 2.

[0063] The method for setting the center point coordinates of the first filter box can be as follows: manually select the center point coordinates of the first filter box. For example, it can be that a section of track point cloud near the front of the train is manually selected through pass-through filtering to form the first filter box.

[0064] The filter box can be a 3D box, which can compensate for the shortcomings of manually selecting the new center of gravity, and the size of the 3D box can also adapt to tracks with different curvatures.

[0065] It should be noted that within a short distance, the center coordinates of the two filter boxes are not much different. The center coordinates of the next filter box and the center coordinates of the current filter box differ only in the direction of movement. For example, if the center coordinates of the current filter box are (x, y, z), the center coordinates of the next filter box are (x, y+1, z).

[0066] Specifically, one way to determine the center point coordinates of the second filter box based on the centroids of all point clouds in the first filter box is to use the average of the centroid coordinates of all point clouds in the first filter box as the center point coordinates of the second filter box. Another way is to calculate the center point coordinates of the second filter box using the following formula:

[0067]

[0068]

[0069]

[0070] Where x2 is the x-coordinate of the center point of the second filter box, y2 is the y-coordinate of the center point of the second filter box, z2 is the z-coordinate of the center point of the second filter box, and scale x The scaling factor corresponding to the x-coordinate, scale y The scaling factor corresponding to the y-coordinate, scale z The scaling factor corresponding to the z-coordinate. This is the average x-coordinate of the centroids of all point clouds in the first filter box. This is the average y-coordinate of the centroids of all point clouds in the first filter box. Let be the average z-coordinate of the centroids of all point clouds in the first filter box, and n be the total number of point clouds in the first filter box. (x1) i Let (y1) be the x-coordinate of the centroid of the i-th point cloud in the first filter box. i Let (z1) be the y-coordinate of the centroid of the i-th point cloud in the first filter box. i Let z be the z-coordinate of the centroid of the i-th point cloud in the first filter box.

[0071] Specifically, the method of repeatedly determining the center point coordinates of the Nth filter box based on the centroids of all point clouds in the (N-1)th filter box, until the center point coordinates of all filter boxes are obtained, can be based on the following formula:

[0072]

[0073]

[0074]

[0075] Where, x center Let x and y be the center coordinates of the Nth filter box. center Let z be the y-coordinate of the center point of the Nth filter box. centerLet z be the z-coordinate of the center point of the Nth filter box, and scale. x The scaling factor corresponding to the x-coordinate, scale y The scaling factor corresponding to the y-coordinate, scale z The scaling factor corresponding to the z-coordinate. This represents the average x-coordinate of the centroids of all point clouds in the (N-1)th filter box. This represents the average y-coordinate of the centroids of all point clouds in the (N-1)th filter box. Let x be the average z-coordinate of the centroids of all point clouds in the (N-1)th filter box, where n is the total number of point clouds in the (N-1)th filter box. i Let x and y be the centroid of the i-th point cloud in the (N-1)-th filter box. i Let z be the y-coordinate of the centroid of the i-th point cloud in the (N-1)-th filter box. i Let z be the z-coordinate of the centroid of the i-th point cloud in the (N-1)-th filter box.

[0076] Optionally, generating at least one near-field boundary curve based on the at least two filter boxes includes:

[0077] Generate the first target set based on the center point coordinates of the N filter boxes corresponding to the left track;

[0078] Generate a second target set based on the center point coordinates of the N filter boxes corresponding to the right track;

[0079] Generate the close-up boundary curve of the left track based on the first target set;

[0080] Generate the close-up boundary curve of the right track based on the second target set.

[0081] The first target set includes the center point coordinates of the N filter boxes corresponding to the left track, and the second target set includes the center point coordinates of the N filter boxes corresponding to the right track.

[0082] Specifically, the method for generating the near-field boundary curve of the left track based on the first target set can be as follows: the near-field boundary curve of the left track is obtained by fitting the center point coordinates of the N filter boxes corresponding to the left track in the first target set.

[0083] Specifically, the method for generating the near-field boundary curve of the right track based on the second target set can be as follows: the near-field boundary curve of the right track is obtained by fitting the center point coordinates of the N filter boxes corresponding to the right track in the second target set.

[0084] Optionally, based on the at least two filter boxes, at least one prospect fitting curve is generated through iterative processes, including:

[0085] If there is no point cloud data in the Mth filter box, a fitting curve is generated using the center point coordinates of the first M-1 filter boxes, and the center point coordinates of the Mth filter box are obtained in the extension direction of the fitting curve, where M is a positive integer not less than 3.

[0086] Repeat the above operation until the constraints are met, and generate a vision fitting curve based on the coordinates of the center points of all the filter boxes obtained in the extension direction of the fitting curve.

[0087] Repeat the above steps to obtain the prospect fitting curves for the left track and the right track.

[0088] The constraint can be that M is greater than the preset number of filter boxes, the constraint can be that there is no point cloud data in Q consecutive filter boxes, or the constraint can be that the number of point clouds in Q consecutive filter boxes is less than the number of point clouds threshold.

[0089] Specifically, the further the distance, the more dispersed the point cloud distribution becomes. At a certain distance, if there is no point cloud data within the filter box, or the number of point clouds in the filter box is less than a threshold, determining the center point coordinates of the next filter box using a near-field approach may result in either being unable to obtain the center point coordinates of the next filter box or obtaining inaccurate coordinates. Therefore, it is necessary to fit the center point coordinates of the previous filter boxes to obtain a fitted curve, and then determine the center point coordinates of the next filter box based on its extension direction. If there is no point cloud data in the next three consecutive filter boxes, the process ends. Connecting the center point coordinates of the fitted filter boxes together yields the far-field fitted curve.

[0090] Specifically, by generating a fitting curve using the center point coordinates of the first M-1 filter boxes, and obtaining the center point coordinates of the Mth filter box in the extension direction of the fitting curve, the interruptions caused by faults and sparsity in the orbital point cloud can be avoided.

[0091] Optionally, the constraint is that there is no point cloud data in Q consecutive filter boxes, where Q is a positive integer not less than 2.

[0092] Optionally, the orbital boundary curve is obtained based on the near-field boundary curve and the far-field fitted curve, including:

[0093] The boundary fitting curve on the left is obtained based on the near-view boundary curve and the far-view fitting curve on the left.

[0094] The boundary fitting curve on the right is obtained based on the near-view boundary curve and the far-view fitting curve on the right.

[0095] Calculate the confidence scores of the boundary fitting curve on the left and the boundary fitting curve on the right, and compare their magnitudes.

[0096] The boundary fitting curve with high confidence is defined as the orbital boundary curve, and the orbital boundary curve on the other side is obtained by translation.

[0097] Specifically, the confidence level of the left-side boundary fitting curve can be calculated as follows: The confidence level of the left-side boundary fitting curve can be determined by obtaining the total number of filter boxes corresponding to the left track, the sum of the total number of filter boxes corresponding to the left track and the total number of filter boxes corresponding to the right track, the residual terms corresponding to the left track, and the sum of the residual terms corresponding to the left track and the right track. Alternatively, the confidence level of the left-side boundary fitting curve can be calculated based on the following formula: Among them, confidence left N represents the confidence level of the fitted curve on the left boundary. left N represents the total number of filter boxes corresponding to the left track. total E is the sum of the total number of filter boxes corresponding to the left track and the total number of filter boxes corresponding to the right track. left E represents the residual term corresponding to the left track. total This is the sum of the residual terms corresponding to the left track and the residual terms corresponding to the right track.

[0098] Specifically, the confidence level of the right-side boundary fitting curve can be calculated as follows: The confidence level of the right-side boundary fitting curve can be determined by summing the total number of filter boxes corresponding to the right track, the total number of filter boxes corresponding to the left track, and the total number of filter boxes corresponding to the right track, as well as the residual terms corresponding to the right track and the residual terms corresponding to the left track. Alternatively, the confidence level of the right-side boundary fitting curve can be calculated based on the following formula: Among them, confidence right N represents the confidence level of the boundary fitting curve on the right. right N represents the total number of filter boxes corresponding to the right track. total E is the sum of the total number of filter boxes corresponding to the left track and the total number of filter boxes corresponding to the right track. right E represents the residual term corresponding to the right track. total This is the sum of the residual terms corresponding to the left track and the residual terms corresponding to the right track.

[0099] Specifically, the boundary fitting curve with high confidence is defined as the orbital boundary curve. The method for obtaining the orbital boundary curve on the other side by translation can be as follows: If the confidence level of the left boundary fitting curve is greater than that of the right boundary fitting curve, then the orbital boundary curve on the other side is obtained by translating the left boundary fitting curve. If the confidence level of the left boundary fitting curve is less than that of the right boundary fitting curve, then the orbital boundary curve on the other side is obtained by translating the right boundary fitting curve. If the confidence level of the left boundary fitting curve is equal to that of the right boundary fitting curve, then the orbital boundary curve on the other side is obtained by translating either the right or left boundary fitting curve. This embodiment of the invention does not impose limitations on this method.

[0100] It should be noted that because the point cloud inside the curve is sparse or severely missing, there is no valid point cloud for the track over a long range, so the detection must be terminated. However, the point cloud outside the curve is very rich, and the point cloud inside the lane can be detected by using the point cloud outside alone. The corresponding point cloud inside can be obtained by simply translating the point cloud outside.

[0101] Specifically, the method of obtaining the track boundary curve on the other side by translation can be as follows: the track width is translated by the boundary fitting curve with high confidence to obtain the track boundary curve on the other side; another method of obtaining the track boundary curve on the other side by translation can be as follows: the track width and the curvature of the track curve are translated by the boundary fitting curve with high confidence to obtain the track boundary curve on the other side.

[0102] Optionally, the center point coordinates of the Nth filter box are determined based on the centroids of all point clouds in the (N-1)th filter box, including:

[0103] The coordinates of the center point of the Nth filter box are determined using the following formula:

[0104]

[0105]

[0106]

[0107] Where, x center Let x and y be the center coordinates of the Nth filter box. center Let z be the y-coordinate of the center point of the Nth filter box. center Let z be the z-coordinate of the center point of the Nth filter box, and scale. x The scaling factor corresponding to the x-coordinate, scale y The scaling factor corresponding to the y-coordinate, scale zThe scaling factor is the z-coordinate, n is the total number of point clouds in the (N-1)th filter box, and x is the z-coordinate. i Let x and y be the centroid of the i-th point cloud in the (N-1)-th filter box. i Let z be the y-coordinate of the centroid of the i-th point cloud in the (N-1)-th filter box. i Let z be the z-coordinate of the centroid of the i-th point cloud in the (N-1)-th filter box.

[0108] in, This represents the average x-coordinate of the centroids of all point clouds in the (N-1)th filter box. This represents the average y-coordinate of the centroids of all point clouds in the (N-1)th filter box. It is the average z-coordinate of the centroid of all point clouds in the (N-1)th filter box.

[0109] Optionally, the confidence level of the boundary fitting curve on the left can be calculated based on the following formula:

[0110]

[0111] Among them, confidence left N represents the confidence level of the fitted curve on the left boundary. left N represents the total number of filter boxes corresponding to the left track. total E is the sum of the total number of filter boxes corresponding to the left track and the total number of filter boxes corresponding to the right track. left E represents the residual term corresponding to the left track. total This is the sum of the residual terms corresponding to the left track and the residual terms corresponding to the right track.

[0112] in, f(k j ) is the curve fitting equation function, k j Let h be the x-coordinate of the j-th fitted point. j Let y be the y-coordinate of the j-th fitted point, and m be the total number of fitted points.

[0113] Optionally, determining obstacle information based on the obstacle point cloud includes:

[0114] Cluster the obstacle point cloud to obtain at least one point cloud set;

[0115] Obstacle information is determined based on the position coordinates and parameter information of the bounding box corresponding to each point cloud set.

[0116] The obstacle information may include the number of obstacles. The obstacle information may also include the location of the obstacles, but this embodiment of the invention does not impose any limitations on this.

[0117] The bounding box parameter information may include the displacement of the bounding box and the velocity of the bounding box, which are not limited in this embodiment of the invention.

[0118] Specifically, the method for determining obstacle information based on the position coordinates and parameter information of the bounding boxes corresponding to each point cloud set can be as follows: if the number of point clouds in the bounding box is greater than a first quantity threshold, then an obstacle is determined to exist, and the number of bounding boxes with a point cloud quantity greater than the first quantity threshold is determined as the number of obstacles. Another method for determining obstacle information based on the position coordinates and parameter information of the bounding boxes corresponding to each point cloud set can be as follows: acquire the target position information and feature information of the obstacles in the image to be identified captured by the camera; project the position coordinates of each bounding box onto the image to be identified to obtain the first position information of each bounding box relative to the image to be identified; fuse the feature information of the obstacle and the parameter information of the bounding box based on the first position information and the target position information to obtain the obstacle information. A further method for determining obstacle information based on the position coordinates and parameter information of the bounding boxes corresponding to each point cloud set can be as follows: acquire the position coordinates of the bounding boxes corresponding to each point cloud set; determine the displacement of the bounding boxes based on their position coordinates in adjacent frame point cloud images. Finally, obstacle information is determined based on the position coordinates and displacement of the bounding boxes corresponding to each point cloud set.

[0119] In a specific example, due to the limitations of the LiDAR's sensing capabilities, track detection can only be performed within a certain range. Track and obstacle detection at longer ranges can only be addressed with coarse algorithms, offering some accuracy for larger obstacles. For straight tracks, the actual location can be used to determine if it's a long straight section. If it is a long straight section, with the near-field track known, extending the track line to select obstacles within the lane can significantly increase the warning distance, providing timely and effective warnings for large obstacles. If a point cloud warning is received, a telephoto camera is used to locate the corresponding image area for 2D obstacle detection, providing a visual interface for train safety personnel to assess. The latter half of the near-field track detection, due to the availability of only one side of the track point cloud, is also used only for warning processing. A global visual inspection is performed on the image area corresponding to the track point cloud to determine the presence of obstacles.

[0120] Optionally, obstacle information can be determined based on the position coordinates and parameter information of the bounding box corresponding to each point cloud set, including:

[0121] Acquire the target location and feature information of obstacles in the image to be identified captured by the camera;

[0122] The position coordinates of each bounding box are projected onto the image to be identified to obtain the first position information of each bounding box relative to the image to be identified;

[0123] Based on the first location information and the target location information, the feature information of the obstacle and the parameter information of the bounding box are fused to obtain obstacle information.

[0124] Specifically, all obstacle point clouds are clustered, and obstacles are identified based on the number of points in each cluster. There's no need to determine the obstacle type; a sufficient number of points in the cloud reliably identifies an obstacle. However, when the point cloud consists of sparse, scattered points, an auxiliary camera is needed for accurate obstacle identification.

[0125] The camera may be a telephoto camera.

[0126] The target location information of the obstacle can be the coordinates of the center point of the target detection box, the length of the target detection box, and the width of the target detection box.

[0127] The feature information can be the type of obstacle, such as a person, an animal, or a motor vehicle.

[0128] Specifically, the method of projecting the position coordinates of each bounding box onto the image to be identified to obtain the first position information of each bounding box relative to the image to be identified can be as follows: based on the projection density and camera extrinsic parameters, the centroid coordinates of each bounding box are projected onto the image to be identified to obtain the first position information of each bounding box relative to the image to be identified.

[0129] Specifically, the method for obtaining the target location information and feature information of obstacles in the image to be identified captured by the camera can be as follows: input the image to be identified into the recognition model to obtain the target location information and feature information of obstacles in the image to be identified.

[0130] Specifically, the obstacle information can be obtained by fusing the feature information of the obstacle and the parameter information of the bounding box based on the first location information and the target location information. If the distance between the first location information and the target location information is less than a distance threshold, the feature information of the obstacle corresponding to the target location information and the parameter information of the bounding box corresponding to the first location information are fused.

[0131] In a specific example, for long-distance detection, a telephoto lens is used to capture the image to be identified. This captured image is then input into a target detection algorithm. This algorithm can detect common target types with distinct features, such as people, animals, and motor vehicles. The output of the target detection algorithm is the coordinates of the center point and the width and height of the target bounding box. Similarly, a deep neural network-based track detection algorithm can output the pixel coordinates of the track boundary in the image at a distance. Then, a curve fitting algorithm is used to obtain the equation representing the track in the image coordinates. Based on this equation and the target bounding box information, safety warnings can be issued for obstacles at extremely long distances.

[0132] The technical solution of this embodiment acquires point cloud data collected by lidar, including at least two filtering boxes; generates at least one near-field boundary curve based on the at least two filtering boxes; generates at least one far-field fitting curve based on the at least two filtering boxes through continuous iteration; obtains at least one track boundary curve based on the near-field boundary curve and the far-field fitting curve; filters obstacle point clouds from the point cloud data according to the at least one track boundary curve; and determines obstacle information based on the obstacle point clouds. This solves the problem of false detection or missed detection caused by the camera's high sensitivity to light and shadow, shadows caused by objects blocking the light during the day, and poor image quality due to poor lighting at night. By deleting point clouds that are not within the range of the left and right tracks, obstacle point clouds are obtained, and obstacle information is obtained based on these obstacle point clouds. Since point clouds that are not within the range of the left and right tracks do not constitute railway obstacles, deleting point clouds that are not within the range of the left and right tracks can greatly reduce the computational load of the system and improve the accuracy of obstacle detection.

[0133] Example 2

[0134] Figure 2 This is a schematic diagram of an obstacle monitoring device provided in an embodiment of the present invention. This embodiment is applicable to obstacle monitoring applications. The device can be implemented using software and / or hardware, and can be integrated into any device that provides obstacle monitoring functionality, such as… Figure 2 As shown, the obstacle monitoring device specifically includes: a point cloud data acquisition module 210, a near-view boundary curve generation module 220, a far-view fitting curve generation module 230, a track boundary curve generation module 240, an obstacle point cloud screening module 250, and an obstacle information determination module 260.

[0135] The point cloud data acquisition module 210 is used to acquire point cloud data collected by the lidar, including at least two filter boxes.

[0136] The foreground boundary curve generation module 220 is used to generate at least one foreground boundary curve based on the at least two filter boxes;

[0137] The vision fitting curve generation module 230 is used to generate at least one vision fitting curve based on the at least two filter boxes through continuous iteration.

[0138] The track boundary curve generation module 240 is used to obtain the track boundary curve based on the near-view boundary curve and the far-view fitting curve;

[0139] The obstacle point cloud filtering module 250 is used to filter obstacle point clouds from the point cloud data based on at least one track boundary curve;

[0140] The obstacle information determination module 260 is used to determine obstacle information based on the obstacle point cloud.

[0141] Optionally, the point cloud data acquisition module is specifically used for:

[0142] Set the coordinates of the center point of the first filter box;

[0143] Based on the first filter box, obtain the centroids of all point clouds in the corresponding filter box;

[0144] The center point coordinates of the second filter box are determined based on the centroids of all point clouds in the first filter box;

[0145] The process is repeated until the center point coordinates of the Nth filter box are obtained by determining the center point coordinates of the Nth filter box based on the centroids of all point clouds in the (N-1)th filter box, and so on. Here, N is a positive integer not less than 2.

[0146] Optionally, the near-field boundary curve generation module is specifically used for:

[0147] Generate the first target set based on the center point coordinates of the N filter boxes corresponding to the left track;

[0148] Generate a second target set based on the center point coordinates of the N filter boxes corresponding to the right track;

[0149] Generate the close-up boundary curve of the left track based on the first target set;

[0150] Generate the close-up boundary curve of the right track based on the second target set.

[0151] Optionally, the prospect fitting curve generation module is specifically used for:

[0152] If there is no point cloud data in the Mth filter box, a fitting curve is generated using the center point coordinates of the first M-1 filter boxes, and the center point coordinates of the Mth filter box are obtained in the extension direction of the fitting curve, where M is a positive integer not less than 3.

[0153] Repeat the above operation until the constraints are met, and generate a vision fitting curve based on the coordinates of the center points of all the filter boxes obtained in the extension direction of the fitting curve.

[0154] Repeat the above steps to obtain the prospect fitting curves for the left track and the right track.

[0155] Optionally, the constraint is that there is no point cloud data in Q consecutive filter boxes, where Q is a positive integer not less than 2.

[0156] Optionally, the track boundary curve generation module is specifically used for:

[0157] The boundary fitting curve on the left is obtained based on the near-view boundary curve and the far-view fitting curve on the left.

[0158] The boundary fitting curve on the right is obtained based on the near-view boundary curve and the far-view fitting curve on the right.

[0159] Calculate the confidence scores of the boundary fitting curve on the left and the boundary fitting curve on the right, and compare their magnitudes.

[0160] The boundary fitting curve with high confidence is defined as the orbital boundary curve, and the orbital boundary curve on the other side is obtained by translation.

[0161] Optionally, the point cloud data acquisition module is specifically used for:

[0162] The coordinates of the center point of the Nth filter box are determined using the following formula:

[0163]

[0164]

[0165]

[0166] Where, x center Let x and y be the center coordinates of the Nth filter box. center Let z be the y-coordinate of the center point of the Nth filter box. center Let z be the z-coordinate of the center point of the Nth filter box, and scale. x The scaling factor corresponding to the x-coordinate, scale y The scaling factor corresponding to the y-coordinate, scale z The scaling factor is the z-coordinate, n is the total number of point clouds in the (N-1)th filter box, and x is the z-coordinate. i Let x and y be the centroid of the i-th point cloud in the (N-1)-th filter box. i Let z be the y-coordinate of the centroid of the i-th point cloud in the (N-1)-th filter box. iLet z be the z-coordinate of the centroid of the i-th point cloud in the (N-1)-th filter box.

[0167] Optionally, the track boundary curve generation module is specifically used for:

[0168]

[0169] Among them, confidence left N represents the confidence level of the fitted curve on the left boundary. left N represents the total number of filter boxes corresponding to the left track. total E is the sum of the total number of filter boxes corresponding to the left track and the total number of filter boxes corresponding to the right track. left E represents the residual term corresponding to the left track. total This is the sum of the residual terms corresponding to the left track and the residual terms corresponding to the right track.

[0170] Optionally, the obstacle information determination module is specifically used for:

[0171] Cluster the obstacle point cloud to obtain at least one point cloud set;

[0172] Obstacle information is determined based on the position coordinates and parameter information of the bounding box corresponding to each point cloud set.

[0173] Optionally, the obstacle information determination module is specifically used for:

[0174] Acquire the target location and feature information of obstacles in the image to be identified captured by the camera;

[0175] The position coordinates of each bounding box are projected onto the image to be identified to obtain the first position information of each bounding box relative to the image to be identified;

[0176] Based on the first location information and the target location information, the feature information of the obstacle and the parameter information of the bounding box are fused to obtain obstacle information.

[0177] The above-mentioned products can perform the obstacle monitoring method provided in any embodiment of the present invention, and have the corresponding functional modules and beneficial effects of performing the method.

[0178] The technical solution of this embodiment acquires point cloud data collected by LiDAR, including at least two filtering boxes; generates at least one near-field boundary curve based on the at least two filtering boxes; generates at least one far-field fitting curve based on the at least two filtering boxes through continuous iteration; obtains a track boundary curve based on the near-field boundary curve and the far-field fitting curve; filters obstacle point clouds from the point cloud data according to the at least one track boundary curve; and determines obstacle information based on the obstacle point clouds. This solves the problem of false detection or missed detection caused by the camera's high sensitivity to light and shadow, shadows caused by objects blocking the light during the day, and poor image quality due to poor lighting at night. It can improve the accuracy of obstacle detection while reducing the amount of computation.

[0179] Example 3

[0180] Figure 3 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0181] like Figure 3 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0182] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0183] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as obstacle detection methods:

[0184] Acquire point cloud data collected by LiDAR, including at least two filter boxes;

[0185] Generate at least one near-field boundary curve based on the at least two filter boxes;

[0186] Based on the at least two filter boxes, at least one vision fitting curve is generated through continuous iteration;

[0187] The track boundary curve is obtained based on the near-view boundary curve and the far-view fitted curve;

[0188] Obstacle point clouds are filtered from the point cloud data based on at least one track boundary curve;

[0189] Obstacle information is determined based on the obstacle point cloud.

[0190] In some embodiments, the obstacle detection method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the obstacle detection method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the obstacle detection method by any other suitable means (e.g., by means of firmware).

[0191] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0192] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0193] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0194] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0195] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0196] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0197] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and no limitation is imposed herein.

[0198] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An obstacle detection method, characterized in that, include: Acquire point cloud data collected by LiDAR, including at least two filter boxes; Generate at least one near-field boundary curve based on the at least two filter boxes; Based on the at least two filter boxes, at least one vision fitting curve is generated through continuous iteration; The track boundary curve is obtained based on the near-view boundary curve and the far-view fitted curve; Obstacle point clouds are filtered from the point cloud data based on at least one track boundary curve; Obstacle information is determined based on the obstacle point cloud; Wherein, the near-field boundary curve is the curve within the detection range of the lidar, and the far-field fitting curve is the curve outside the detection range of the lidar; The acquisition of point cloud data collected by the lidar includes at least two filter boxes, including: Set the coordinates of the center point of the first filter box; Based on the first filter box, obtain the centroids of all point clouds in the corresponding filter box; The center point coordinates of the second filter box are determined based on the centroids of all point clouds in the first filter box; The process is repeated until the center point coordinates of the Nth filter box are obtained by determining the center point coordinates of the Nth filter box based on the centroids of all point clouds in the (N-1)th filter box, and so on. Here, N is a positive integer not less than 2.

2. The method according to claim 1, characterized in that, The center point coordinates of the Nth filter box are determined based on the centroids of all point clouds in the (N-1)th filter box, including: The coordinates of the center point of the Nth filter box are determined using the following formula: ； ； ； in, The center point of the Nth filter box coordinate, The center point of the Nth filter box coordinate, The center point of the Nth filter box coordinate, for The scaling factor corresponding to the coordinates for The scaling factor corresponding to the coordinates for The scaling factor corresponding to the coordinates The total number of point clouds in the (N-1)th filter box. For the (N-1)th filter box The center of gravity of a point cloud coordinate, For the (N-1)th filter box The center of gravity of a point cloud coordinate, For the (N-1)th filter box The center of gravity of a point cloud coordinate.

3. The obstacle monitoring method according to any one of claims 1 to 2, characterized in that, Generate at least one near-field boundary curve based on the at least two filter boxes, including: Generate the first target set based on the center point coordinates of the N filter boxes corresponding to the left track; Generate a second target set based on the center point coordinates of the N filter boxes corresponding to the right track; Generate the close-up boundary curve of the left track based on the first target set; Generate the close-up boundary curve of the right track based on the second target set.

4. The obstacle detection method according to claim 3, characterized in that, Based on the aforementioned at least two filter boxes, at least one prospect fitting curve is generated through iterative processes, including: If there is no point cloud data in the Mth filter box, a fitting curve is generated using the center point coordinates of the first M-1 filter boxes, and the center point coordinates of the Mth filter box are obtained in the extension direction of the fitting curve, where M is a positive integer not less than 3. Repeat the above operation until the constraints are met, and generate a vision fitting curve based on the coordinates of the center points of all the filter boxes obtained in the extension direction of the fitting curve. Repeat the above steps to obtain the prospect fitting curves for the left track and the right track.

5. The obstacle detection method according to claim 4, characterized in that, The constraint is that there is no point cloud data in any of the Q consecutive filter boxes, where Q is a positive integer not less than 2.

6. The obstacle monitoring method according to claim 1, 2, 4 or 5, characterized in that, The orbital boundary curve is obtained based on the near-field boundary curve and the far-field fitted curve, including: The boundary fitting curve on the left is obtained based on the near-view boundary curve and the far-view fitting curve on the left. The boundary fitting curve on the right is obtained based on the near-view boundary curve and the far-view fitting curve on the right. Calculate the confidence scores of the boundary fitting curve on the left and the boundary fitting curve on the right, and compare their magnitudes. The boundary fitting curve with high confidence is defined as the orbital boundary curve, and the orbital boundary curve on the other side is obtained by translation.

7. The method according to claim 6, characterized in that, The confidence level of the boundary fitting curve on the left is calculated based on the following formula: ； in, The confidence level of the fitted curve for the left boundary. This represents the total number of filter boxes corresponding to the left track. This is the sum of the total number of filter boxes corresponding to the left track and the total number of filter boxes corresponding to the right track. This is the residual term corresponding to the left track. This is the sum of the residual terms corresponding to the left track and the residual terms corresponding to the right track.

8. The method according to claim 1, 2, 4, 5 or 7, characterized in that, Obstacle information is determined based on the obstacle point cloud, including: Cluster the obstacle point cloud to obtain at least one point cloud set; Obstacle information is determined based on the position coordinates and parameter information of the bounding box corresponding to each point cloud set.

9. The method according to claim 8, characterized in that, Obstacle information is determined based on the position coordinates and parameter information of the bounding box corresponding to each point cloud set, including: Acquire the target location and feature information of obstacles in the image to be identified captured by the camera; The position coordinates of each bounding box are projected onto the image to be identified to obtain the first position information of each bounding box relative to the image to be identified; Based on the first location information and the target location information, the feature information of the obstacle and the parameter information of the bounding box are fused to obtain obstacle information.

10. An obstacle monitoring device, controlled by the obstacle monitoring method as described in any one of claims 1-9, characterized in that, include: The point cloud data acquisition module is used to acquire point cloud data collected by LiDAR, including at least two filter boxes; A near-field boundary curve generation module is used to generate at least one near-field boundary curve based on the at least two filter boxes; The vision fitting curve generation module is used to generate at least one vision fitting curve based on the at least two filter boxes through continuous iteration. The track boundary curve generation module is used to obtain the track boundary curve based on the near-view boundary curve and the far-view fitting curve. An obstacle point cloud filtering module is used to filter obstacle point clouds from the point cloud data based on at least one track boundary curve; An obstacle information determination module is used to determine obstacle information based on the obstacle point cloud.

11. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the obstacle detection method according to any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the obstacle detection method according to any one of claims 1-9.

Images