A method for dividing regions of interest of underground locomotives and a method for extracting obstacles

Abstract

The present invention relates to a method for dividing an area of interest of an underground locomotive. A method for dividing an area of interest of an underground locomotive comprises the following steps: S1. Establish an orbital coordinate system, which is used to characterize the mapping relationship between the correction amount of the locomotive and the road curvature of the underground track. S2. Convert the orbital coordinate system into a lidar coordinate system, fit a planar path Path in the lidar coordinate system, and form an extended region ROI by expanding the two sides of the planar path Path. S3. Grid process the extended region ROI into an area of interest ROI2. By fusing wireless positioning and lidar to extract accurate track features, the present invention replaces the traditional way of extracting track features, generates an area of interest by combining the method of drawing curvature with real-time positioning, has good stability, is not affected by the sensor's field of view, and divides the generated area of interest, laying a foundation for subsequent acquisition of track peripheral features or obstacle information.

Description

Technical Field

[0001] The present invention relates to the technical field of underground locomotives, and in particular to a method for dividing an area of ​​interest of an underground locomotive, a coordinate conversion method of a positioning beacon based on the division of the area of ​​interest of the underground locomotive, and a method for extracting obstacles of the underground locomotive. Background Art

[0002] The driving environment of underground locomotives is harsh, and manual driving needs to maintain a high degree of concentration to avoid collisions and safety accidents. At the same time, manual operation of underground locomotives is often accompanied by certain uncontrollable factors, which can easily cause transportation accidents. As autonomous driving technology gradually matures and begins to gradually replace manual driving, locomotives have obvious track boundary characteristics when running. Accurately extracting obstacles that intrude into the boundary is of great significance to ensuring the safe driving of autonomous locomotives.

[0003] Usually, laser radar or camera is used to extract track features. The area of ​​interest divided based on this is unstable and easily affected by the sensor's field of view. In the weak light environment underground, the track range that can be extracted is small, and when the sensor's field of view is blocked, it is difficult to obtain track features stably. Summary of the invention

[0004] Based on this, it is necessary to provide a method for dividing the area of ​​interest of an underground locomotive, a coordinate transformation method for a positioning beacon based on the division of the area of ​​interest of an underground locomotive, and a method for extracting obstacles for an underground locomotive in order to address the problem of instability in the divided area of ​​interest caused by the traditional method that is difficult to stably obtain track features.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for dividing an area of ​​interest of an underground locomotive, the method comprising the following steps:

[0007] S1. Establishing a track coordinate system, which is used to characterize the mapping relationship between the correction amount of the locomotive and the road curvature of the underground track;

[0008] S2. Convert the orbital coordinate system into the laser radar coordinate system, fit the plane path Path in the laser radar coordinate system, and expand the two side boundaries of the plane path Path to form an extended area ROI. The method for obtaining the extended area ROI is as follows:

[0009] S2.1 Taking the position of the laser radar of the locomotive as the position point, the coordinates of the positioning beacon in the track coordinate system are converted into laser radar coordinates with the center of the laser radar coordinates as the origin;

[0010] S2.2 obtains the laser radar coordinates of the location and multiple positioning beacons in front of it, and fits the adjacent laser radar coordinates in sequence to obtain the plane path Path;

[0011] S2.3 Extend the plane path Path along the positive and negative directions of the horizontal coordinate by l move , get the extension boundary Path on both sides of the underground track left 、Path right , expand the boundary Path on both sides left 、Path right Enclose the expanded area ROI;

[0012] S3. Rasterize the expanded area ROI into the region of interest ROI 2 , the rasterization method is as follows:

[0013] S3.1 Extending Boundary Path left and expand the boundary Path right Encrypt and expand the boundary Path left and expand the boundary Path right The boundary projection of obtains continuous grid cell one;

[0014] S3.2 sets the value of grid cell one to a fixed value PATH, and searches all grid cells one in the expanded area ROI with the location as the origin;

[0015] S3.3 When the value of grid cell 1 is found to be the fixed value PATH, stop searching in this direction until all grid cells 1 in the expanded area ROI are filled with the fixed value PATH, and the preliminary area ROI is obtained. 1 ;

[0016] S3.4 obtains the original point cloud data of the expanded area ROI, segments the original point cloud data, and rasterizes the non-ground point cloud data obtained after segmentation, and obtains that the value of the grid unit 2 of the non-ground point cloud data is a fixed value OBSTACLE.

[0017] S3.5 Project the grid cells of the non-ground point cloud data to the preliminary region ROI 1 , get the region of interest ROI 2 , where the region of interest ROI 2 The value of grid cell three that overlaps with the projection of grid cell two is increased by a fixed value of OBSTACLE.

[0018] Furthermore, the orbital coordinate system is established as follows:

[0019] S1.1 Preset multiple positioning beacons along the underground track, and obtain the road curvature of the underground track corresponding to the positioning beacons and the displacement offset of the locomotive relative to the starting point of the section of the underground track;

[0020] S1.2 corrects the displacement offset of the locomotive to obtain the correction value, and establishes a track coordinate system with the correction value as the horizontal coordinate, the road curvature as the vertical coordinate, and the starting point of the underground track section as the origin.

[0021] Furthermore, the calculation method of correcting the displacement offset includes the following steps:

[0022] Get the displacement offset P of the locomotive 0 Longitudinal deviation Δy and lateral deviation Δx from the laser radar coordinate center;

[0023] Calculate the corrected displacement offset P offset :

[0024] Furthermore, the search method for the expanded region ROI includes a depth-first search algorithm.

[0025] The present invention also includes a method for transforming the coordinates of a positioning beacon, which adopts the aforementioned method for dividing the region of interest of an underground locomotive. The method for transforming the coordinates of a positioning beacon includes the following steps:

[0026] Let the position be C, and obtain the coordinates of C in the orbital coordinate system (P offset1 , k 1 ), and two adjacent positioning beacons C i-1 and Ci in the orbital coordinate system, where C i-1 The coordinates are (offset i-1 , k i-1 ), C i The coordinates are (offset i , k i );

[0027] Calculate C and C i The arc length Δl 1 :Δl 1 =offset i -P offset1 ;

[0028] According to C i-1 The road curvature k i-1 Calculate the Δl 1 The central angle θ 1 :θ 1 =Δl 1 ×k i-1 ;

[0029] Ci The coordinates in the orbital coordinate system (offset i , k i ) into lidar coordinates (xi, yi), where

[0030] The present invention also includes a method for extracting an underground locomotive obstacle, and the method for extracting an underground locomotive obstacle includes the following steps:

[0031] S100. Taking the locomotive position as the origin, the region of interest (ROI) is mapped according to a preset path. 2 The grid cells are used for obstacle search;

[0032] S200. Determine whether there is an obstacle in the searched grid unit three, and continue searching along a preset path after determination;

[0033] S300. When searching for an end point of a preset search path, return to the origin and search again, and three-cluster the grid cells that are determined to have obstacles to obtain an obstacle list;

[0034] In the region of interest ROI 2 Before performing obstacle search, you need to search for the region of interest (ROI). 2 The division is performed, and the division method adopts the aforementioned underground locomotive area of ​​interest division method.

[0035] Furthermore, the obstacle determination method for grid unit three is as follows:

[0036] S201. Determine whether the value of grid cell three is a fixed value PATH+OBSTACLE;

[0037] S202. If yes, it is determined that there is an obstacle in grid unit 3, and the obstacle is intrusive;

[0038] S203. Otherwise, it is determined that there is no intrusive obstacle in grid cell three.

[0039] Furthermore, the method for obtaining the position of a locomotive includes the following steps:

[0040] The laser radar of the locomotive is used as a detection point, and the detection point is controlled to send an initial detection signal to all positioning beacons. The positioning beacons adjacent to the locomotive receive the initial detection signal as feedback points, and send a cutoff data signal to the detection point, and the detection point receives the cutoff data signal;

[0041] The time for the detection point to send the initial detection signal and receive the cut-off data signal and the time for the feedback point to receive the initial signal and send the cut-off data signal are used to obtain the flight time error t2 through the TOF function;

[0042] Controlling the detection point to send a re-detection signal to the feedback point;

[0043] The flight error time t2 is adjusted according to the time when the detection point sends the re-detection signal and the time when the feedback point receives the re-detection signal to obtain the actual error time. The method for obtaining the actual error time is as follows:

[0044] (1) collecting the time a3 of the detection point sending the re-detection signal and the time b3 of the feedback point receiving the re-detection signal;

[0045] (2) The time a3 at which the detection point sends the re-detection signal and the time a2 at which the detection point receives the cut-off data signal are calculated by difference to obtain the feedback time reply of the detection point. A ;

[0046] (3) Feedback time reply through detection point A Adjust the flight error time t2 to obtain the actual error time t4:

[0047] The error actual time t4 and the true value tof are added to calculate the estimated value of the signal flight time Estimated value The relative distance between the detection point and the feedback point is calculated by a speed formula.

[0048] Furthermore, the calibration method of the feedback point includes the following steps:

[0049] Preset a calibration time range;

[0050] Obtain the cycle time from the detection point transmitting the signal to receiving the feedback signal;

[0051] Determine whether the cycle time is within the calibration time range

[0052] If yes, the positioning beacon that sends the feedback signal is determined to be the measurement point.

[0053] The technical solution provided by the present invention has the following beneficial effects:

[0054] 1. The present invention extracts precise track features by fusing wireless positioning with lidar, replacing the traditional method of extracting track features. It combines real-time positioning with curvature drawing to generate regions of interest. It has good stability and is not affected by the sensor's field of view. It also divides the generated regions of interest, laying the foundation for subsequent acquisition of track perimeter features or obstacle information.

[0055] 2. The present invention establishes a track coordinate system through the road curvature and the displacement offset of the locomotive, and combines this with the real-time positioning of the locomotive to generate an area of ​​interest in the passage area in the direction of locomotive travel, and accurately generates and divides the area of ​​interest.

[0056] 3. The present invention extracts obstacle information based on the divided regions of interest, obtains an obstacle list by clustering, and accurately obtains obstacle information in the direction of the locomotive's advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flowchart of a method for dividing an area of ​​interest of an underground locomotive according to the present invention;

[0058] Figure 2 Based on Figure 1 Schematic diagram of the ranging principle;

[0059] Figure 3 Based on Figure 1 Schematic diagram of wireless positioning layout;

[0060] Figure 4 Based on Figure 1 Schematic diagram of boundary expansion;

[0061] Figure 5 Based on Figure 1 A flow chart of an underground locomotive obstacle extraction method;

[0062] Figure 6 Based on Figure 1 The actual effect diagram of boundary trajectory expansion;

[0063] Figure 7 Based on Figure 1 The actual effect of rasterization of the area of ​​interest;

[0064] Figure 8 Based on Figure 1 Obstacle search clustering effect diagram in the area of ​​interest; DETAILED DESCRIPTION

[0065] 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 only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] This embodiment provides a method for dividing the region of interest of an underground locomotive to address the problem in the prior art that it is difficult to stably obtain track features, resulting in unstable division of the region of interest. By fusing wireless positioning and lidar to extract accurate track features, this embodiment replaces the traditional method of extracting track features, generates the region of interest by combining real-time positioning and curvature mapping, has good stability, is not affected by the sensor's field of view, and divides the generated region of interest, laying a foundation for subsequent acquisition of track perimeter features or obstacle information.

[0067] As Figure 1 shown, the method for dividing the region of interest of the underground locomotive in this embodiment includes the following steps:

[0068] S1. Establish a track coordinate system, which is used to represent the mapping relationship between the correction amount of the locomotive and the road curvature of the underground track.

[0069] Preset multiple positioning beacons along the underground track, and obtain the road curvature of the underground track corresponding to the positioning beacons and the displacement offset of the locomotive relative to the starting point of the track section.

[0070] Correct the displacement offset of the locomotive to obtain a correction amount. Taking the correction amount as the abscissa, the road curvature as the ordinate, and the starting point of the track section of the underground track as the origin, establish a track coordinate system. The calculation method of the correction amount is where P 0 is the displacement offset, Δx is the lateral deviation between P 0 and the lidar coordinate center, and Δy is the longitudinal deviation between P 0 and the lidar coordinate center.

[0071] S2. Convert the track coordinate system into the lidar coordinate system, fit a plane path Path in the lidar coordinate system, and expand to form an extended region ROI based on the two sides of the plane path Path. The method for obtaining the extended region ROI is as follows:

[0072] S2.1 Taking the position where the lidar of the locomotive is located as a locus, convert the coordinates of the positioning beacon in the track coordinate system into lidar coordinates with the lidar coordinate center as the origin. The coordinate conversion method of the positioning beacon includes marking the locus as C, obtaining the coordinates (P offset1 , k i ) of C in the track coordinate system, and the coordinates of two adjacent positioning beacons C i-1 and Ci of C in the track coordinate system. Among them, the coordinates of C i-1 are (offset i-1 , k i-1 ), and the coordinates of C i are (offset i , ki ). Calculate C and C i The arc length Δl 1 :

[0073] Δl 1 =offset i -P offset1 .

[0074] According to C i-1 The road curvature k i-1 Calculate Δl 1 The central angle θ 1 :

[0075] θ 1 =Δl 1 ×k i-1 .

[0076] C i The coordinates in the orbital coordinate system (offset i ,k i ) into lidar coordinates (xi, yi), where

[0077] S2.2 obtains the laser radar coordinates of the site and multiple positioning beacons in front of it, and fits the adjacent laser radar coordinates in sequence to obtain the plane path Path. The basis for determining the laser radar coordinates of the site and multiple positioning beacons in front of it is to determine the position of the locomotive on the track, that is, the position of the site on the track. Based on ultra-wideband UWB technology, real-time monitoring of the position, driving direction and distribution of the locomotive on the track can be achieved. By setting positioning beacons along the track and equipping the mobile locomotive with on-board detectors, that is, on-board laser radars, the locomotive is positioned. Reasonably increasing the number of positioning beacons can effectively improve the positioning accuracy of the vehicle.

[0078] For the positioning method of the locomotive, the TOF method can be used for ranging. When positioning the locomotive underground, in addition to removing the positioning error caused by the time asynchrony between different beacons, it is also necessary to reduce the time asynchrony between the locomotive and the beacon and the influence of clock offset on the positioning accuracy. Asymmetric two-way ranging is used to further reduce the ranging error.

[0079] like Figure 2 As shown in the figure, the principle of ranging is as follows: the vehicle-mounted detector A and the positioning beacon B are represented by node A and node B respectively. 1 Send data to node B at b 1 The data is received at the same time. Since the time of node B and node A is not synchronized, it cannot be directly received through b 1 -a 1Obtain the flight time between the two nodes. It is necessary to eliminate the time synchronization error, so the B node is introduced through b 2 -b 1 Feedback time reply B , sends the cutoff data to node A, node A is in a 2 The cut-off data is received at all times, and the time difference between sending and receiving messages by the same A node is used to eliminate the time synchronization error.

[0080] But in Figure 1 The entire ranging cycle of node A is round 1 The time offset e between nodes A and B caused by the crystal oscillator is included. a 、e b , then round 1 Cycle time:

[0081] round l (e a +1)=2tof+reply B (e b +I).

[0082] We can further get an estimate of the flight time

[0083]

[0084] Estimated value The flight time difference error between the actual tof can be expressed as:

[0085]

[0086] Since the feedback time of node B is B is much larger than the flight time, so the flight time error can be further expressed as:

[0087]

[0088] The above formula can eliminate the influence of time asynchrony of different nodes on ranging accuracy, but it cannot reduce the time drift caused by the crystal oscillator. It is necessary to further reduce the influence of the crystal oscillator on ranging and introduce the feedback time reply from node A to node B. A Continue to send the cutoff data to node B to obtain the ranging cycle round of node B 2 for:

[0089] round 2 (e h + I) = 2tof + reply A (e a +I).

[0090] By and round i Combined with the cycle time, we can further get an estimate of the flight time

[0091]

[0092] Further get the ranging error:

[0093]

[0094] Similarly, the feedback time of nodes A and B is much longer than the flight time, and the actual error time t4 is further obtained:

[0095]

[0096] When the locomotive is running underground, the asymmetric two-way ranging method can reduce the ranging error caused by time asynchrony and time drift between different nodes, and improve the positioning accuracy of the locomotive.

[0097] like Figure 3 As shown in the figure, the underground iron ore transport locomotive runs along the track, the operating range is determined, and the positioning beacon arrangement is relatively convenient. The positioning process is simplified to positioning in a one-dimensional linear space along the track. Positioning beacons are arranged along the locomotive running track. When the locomotive is near the beacon position, its own position is calibrated. Between the two positioning beacons, the number of wheel rotations is counted with the help of an encoder to measure the vehicle travel distance and obtain the precise position of the locomotive on the track.

[0098] The following method can be used to determine the positioning beacon B: preset a calibration time range. Obtain the cycle time from the detection point transmitting the signal to receiving the feedback signal. Determine whether the cycle time is within the calibration time range. If yes, determine that the positioning beacon that sends the feedback signal is the measurement point.

[0099] For the positioning of locomotives, wireless wifi can also be used to locate locomotives, obtain the MAC address and coordinates of base stations near the track, and the wireless communication module of the locomotive will report the MAC address and signal strength of base stations around the locomotive to the host computer in real time. The host computer performs Gaussian filtering on the signal strength uploaded by the locomotive to filter out abnormal signal strength values, and then uses the distance loss model to convert the signal strength value into distance. Finally, the centroid method, neighbor matching algorithm, least squares method and other methods are used to locate the current location of the locomotive and display it on the map.

[0100] For the positioning of locomotives, RFID technology can also be used to identify special terrains, arrange RFID electronic tags on the locomotive running track, and install a tag collector on the locomotive. During the operation of the locomotive, the tag collector reads the electronic tags to obtain the running direction and position information of the locomotive on the track. The tag collector has an identification range during the identification process. Since the arrangement distance of the tags is greater than the RFID reader distance, the locomotive cannot determine the real-time position between the two tags. In order to improve the positioning accuracy of the locomotive and obtain the real-time running speed of the locomotive, a Hall-type speed sensor is used to monitor the running speed of the locomotive. The accurate position of the locomotive is obtained through the fusion algorithm of speed and position based on the Kalman filter.

[0101] For the positioning of locomotives, wireless wifi technology and RFID technology can be combined for positioning, or wireless wifi technology can be combined with UWB positioning to obtain accurate positioning of locomotives. At the same time, sensors, such as speed sensors, can be used. When the position between two positioning points cannot be accurately determined, the travel distance of the locomotive can be calculated based on the speed and time, and then the position of the previous positioning point can be matched to obtain the position of the locomotive on the track. For errors, multiple sensors or other methods can be used to obtain more accurate positioning information through comparison, which also lays the foundation for the subsequent division of areas of interest.

[0102] S2.3 Extend the plane path Path along the positive and negative directions of the horizontal coordinate by l move , get the extension boundary Path on both sides of the underground track left 、Path right , expand the boundary Path on both sides left 、Path right Enclose the expansion area ROI.

[0103] like Figure 4 As shown, the path boundaries on both sides can be generated in the following way: according to the position coordinates of two adjacent points (Px i , Py i ) and (Px j , Py j ) The normal direction of the connecting line is extended to one side. move Get the translated point (Px i ′,Py i ′). All (Px i ′,Py i ′) to get the boundary on one side. move When it is a positive value, the left boundary Path can be obtained. left On the contrary move When it is a negative value, the right boundary Path can be obtained right .

[0104]

[0105]

[0106] Here move The value of needs to be combined with the locomotive width to ensure that the generated area of ​​interest can completely include the locomotive's travel range. Here l move Expand 1.2m to the left and right sides. So far, by expanding the path to both sides, the boundary of the locomotive is obtained. left and Path right It can be used as the two side boundaries of ROI, laying the foundation for the subsequent extraction of intrusive obstacles.

[0107] S3. Rasterize the expanded region ROI into the region of interest ROI2. The rasterization method is as follows:

[0108] S3.1 Extending Boundary Path left and expand the boundary Path right Encrypt and expand the boundary Path left and expand the boundary Path right The boundary projection obtains continuous grid cell one.

[0109] S3.2 sets the value of grid cell 1 to a fixed value PATH, takes the site as the origin, and searches all grid cells 1 in the expanded area ROI. The search method adopts a depth-first search method, and searches all grid cells 1 according to the optimal conditions. When searching, if a grid cell 1 does not meet the conditions, it will go back one step and reselect, and the whole process is repeated until all grid cells 1 are visited.

[0110] S3.3 When the value of the searched grid cell 1 is the fixed value PATH, stop searching in this direction until all grid cells 1 in the expanded area ROI are filled with the fixed value PATH, and obtain the preliminary area ROI1.

[0111] S3.4 obtains the original point cloud data of the expanded area ROI, segments the original point cloud data, rasterizes the non-ground point cloud data obtained after segmentation, and obtains that the value of the grid unit 2 of the non-ground point cloud data is a fixed value OBSTACLE.

[0112] The segmentation of the original point cloud data is to filter the collected laser point cloud by setting a height threshold, and to obtain the spatial plane equation by fitting the point cloud below the height threshold as a seed point into an initial plane; based on the distance from the point to the plane, the point cloud within a certain range near the plane is secondary screened, and the plane equation is updated using the screened point cloud; the plane is fitted through multiple iterations to obtain a plane equation that is more in line with the real ground, and the fitted plane equation is used to filter the point cloud, marking the point cloud below the plane as a ground point cloud, and the point cloud above the plane as a non-ground point cloud. This method significantly improves the accuracy of segmentation when dealing with complex underground scenes.

[0113] The main steps of the multi-iteration plane fitting ground segmentation method based on height initial screening are as follows:

[0114] Step 1: By setting the height threshold Th hight Initially screen the point cloud collected by the laser radar, Th hight The setting of the value is related to the installation height of the radar. The point cloud set P collected by the laser radar is calculated by the height. i ,y i , z i )|i=1,2,...,n-1} i No more than Th hight The point cloud is stored in the set P ground The remaining point clouds are stored in P notground middle:

[0115]

[0116] Step 2: Through the ground point set P ground The ground plane Ground(x, y, z) is fitted for the first time.

[0117] Step 3: Set the distance threshold Th from the point to the plane dist , clear P ground and P notground , traverse all points and calculate the distance Dist from the point to the ground plane Ground (x, y, z) i ≤Th dist , or when it is below the Ground (x, y, z) plane, store the point cloud in P ground The remaining points are stored in P notground middle.

[0118]

[0119] Where Dist i is the distance from the point to the plane.

[0120] Step 4: Set the number of iterations T_iterat based on the requirements of real-time performance and segmentation accuracy, and fit the plane after the number of iterations T_iterat.

[0121] S3.5 projects the grid cell 2 of the non-ground point cloud data to the preliminary region ROI1 to obtain the region of interest ROI2, wherein the value of the grid cell 3 in the region of interest ROI2 that overlaps with the projection of the grid cell 2 is increased by a fixed value OBSTACLE.

[0122] By integrating wireless positioning and laser radar to extract track features, the precise positioning information of the locomotive is obtained, and the region of interest is generated in front of the locomotive positioning information. When the expanded area is segmented, a more fitting two-dimensional plane is obtained through multiple iterations, and the non-ground point cloud is screened out. The obtained non-ground point cloud is more accurate, and the divided region of interest is also more accurate.

[0123] like Figure 5 As shown, based on the aforementioned division of the underground locomotive area of ​​interest, obstacle extraction is performed, and the obstacle extraction includes the following steps:

[0124] S100. Taking the locomotive position as the origin, perform obstacle search on the grid cells of the region of interest ROI2 according to a preset path.

[0125] S200. Determine whether there is an obstacle in the searched grid cell 3, and continue searching along a preset path after determination. The specific steps of obstacle determination are: determine whether the value of grid cell 3 is a fixed value PATH+OBSTACLE. If yes, it is determined that there is an obstacle in grid cell 3, and this obstacle has an intrusive behavior. Otherwise, it is determined that there is no intrusive obstacle in grid cell 3;

[0126] S300. When searching for the end point of a preset search path, return to the origin and search again, and cluster the grid cells that are determined to have obstacles into three groups to obtain an obstacle list. Clustering can be performed by Euclidean clustering. Clustering can also be performed by combining scan line and Euclidean clustering. Using the scan line structural features in the original point cloud data, first make preliminary category markings for the point clouds on the same scan line, and then cluster the marked point clouds between different line bundles. The distance threshold can be set dynamically to improve the clustering effect of laser point clouds within different distance ranges. In the process of clustering non-ground point clouds, the Euclidean clustering is improved by clustering the scan line distribution features, which can effectively reduce the calculation time and improve the execution efficiency of the program.

[0127] After clustering, the original point cloud data will be divided into multiple different point cloud clusters, and different point cloud clusters represent different obstacles. The positional relationship between the point cloud cluster and the laser radar body is the positional relationship between the obstacle and the laser radar. At the same time, the shape characteristics of the obstacle can be expressed through the point cloud cluster. The external expansion size characteristics of the point cloud cluster can be expressed by the rectangular model. The rectangular model uses the method of establishing the minimum envelope rectangle (calculating the maximum difference of all points in the same point cloud cluster on the three axes) to express the external expansion size and distance information of the obstacle.

[0128] Using VRB-6200 industrial computer, track features are extracted from the raw data collected by the laser radar carried by the underground locomotive. During the feature extraction process, there are many steps in the tunnel that are similar to the track geometry. The method of combining laser radar with wireless positioning can stably divide the area of ​​interest without being affected by the sensor field of view. The generated area of ​​interest can completely include the passage area in the direction of the locomotive's advance.

[0129] Comparison of the number of obstacle targets output by the perception system

[0130]

[0131] The table above is a comparison of the number of obstacle targets output by the perception system before and after the area of ​​interest is divided. By artificially setting a moving obstacle target on the locomotive's driving path, there is only one obstacle target within the locomotive's travel range. Using a VRB-6200 industrial computer, the 455 frames of valid point cloud data collected were processed, and the number of obstacle targets output by the perception system before and after the area of ​​interest was divided was counted.

[0132] As shown in the table above, before the fusion positioning is used to divide the region of interest, the average number of obstacle targets output per frame is 7.2, and there is an obstacle target with intrusive behavior in each frame of data, which affects the normal driving of the locomotive by 13.89%. After dividing the region of interest, the average number of obstacle targets output per frame is 1.04, and the proportion of intrusive obstacle targets is 96.15%. By integrating indoor wireless positioning to divide the region of interest, it is possible to stably obtain intrusive obstacle targets, greatly improve the pertinence of the perception system, and reduce the decision complexity of the subsequent autonomous driving decision system.

[0133] By dividing the area of ​​interest and clustering obstacle information, it is possible to stably acquire intrusive obstacle targets, greatly improving the targeting of the perception system in autonomous driving.

[0134] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0135] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for dividing an area of ​​interest of an underground locomotive, which is used to divide the surrounding area of ​​the locomotive traveling on an underground track into areas of interest, It is characterized in that The division method comprises the following steps: S1. Establishing a track coordinate system, which is used to characterize the mapping relationship between the correction amount of the locomotive and the road curvature of the underground track; S2. The orbital coordinate system is converted into a laser radar coordinate system, a plane path Path is fitted in the laser radar coordinate system, and an extended region ROI is formed according to the boundaries of both sides of the plane path Path. The method for obtaining the extended region ROI is as follows: S2.1 Taking the laser radar position of the locomotive as the position point, convert the coordinates of the positioning beacon in the track coordinate system into laser radar coordinates with the laser radar coordinate center as the origin; S2.2 obtains the laser radar coordinates of the location and multiple positioning beacons in front of the location, and fits the adjacent laser radar coordinates in sequence to obtain a planar path Path; S2.3 Extend the plane path Path along the positive direction and negative direction of the horizontal coordinate by l move , get the extension boundary Path on both sides of the underground track left 、Path right , expand the boundary Path on both sides left 、Path right Enclose the expanded area ROI; S3. Rasterize the expanded area ROI into the region of interest ROI 2 , the rasterization method is as follows: S3.1 Extending Boundary Path left and expand the boundary Path right Encryption is performed to make the extension boundary Path left and the extended boundary Path right The boundary projection of obtains continuous grid cell one; S3.2 sets the value of grid cell one to a fixed value PATH, takes the location as the origin, and searches all grid cells one in the expanded area ROI; S3.3 When the value of the grid cell 1 is found to be the fixed value PATH, the search in this direction is stopped until all grid cells 1 in the expanded area ROI are filled with the fixed value PATH, and the preliminary area ROI is obtained. 1 ; S3.4 obtains the original point cloud data of the expanded area ROI, segments the original point cloud data, and rasterizes the non-ground point cloud data obtained after segmentation, so that the value of the grid unit 2 of the non-ground point cloud data is a fixed value OBSTACLE. S3.5 Projecting the grid unit of the non-ground point cloud data to the preliminary region ROI 1 , get the region of interest ROI 2 , where the region of interest ROI 2 The value of grid cell three that overlaps with the projection of grid cell two is increased by a fixed value OBSTACLE.

2. The method for dividing the region of interest of an underground locomotive according to claim 1, It is characterized in that The method for establishing the orbital coordinate system is as follows: S1.1 Preset multiple positioning beacons along the underground track, and obtain the road curvature of the underground track corresponding to the positioning beacons and the displacement offset of the locomotive relative to the starting point of the section of the underground track; S1.2 corrects the displacement offset of the locomotive to obtain a correction amount, and establishes a track coordinate system with the correction amount as the horizontal coordinate, the road curvature as the vertical coordinate, and the starting point of the underground track section as the origin.

3. The method for dividing the region of interest of an underground locomotive according to claim 2, It is characterized in that The method for calculating the corrected displacement offset comprises the following steps: Get the displacement offset P of the locomotive 0 Longitudinal deviation Δy and lateral deviation Δx from the laser radar coordinate center; Calculate the corrected displacement offset P offset :

4. The method for dividing the region of interest of an underground locomotive according to claim 1, It is characterized in that The search method for the expanded region ROI includes a depth-first search algorithm.

5. A method for transforming coordinates of a positioning beacon, It is characterized in that The method for dividing the area of ​​interest of an underground locomotive according to any one of claims 1 to 4 is adopted, and the coordinate conversion method of the positioning beacon comprises the following steps: The position is recorded as C, and the coordinates of C in the orbital coordinate system are obtained (P offset1 , k 1 ), and two adjacent positioning beacons C i-1 and C i The coordinates in the orbital coordinate system, where C i-1 The coordinates of (offset i-1 ,k i-1 ), C i The coordinates of (offset i ,k i ); Calculate C and C i The arc length Δl 1 :Δl 1 =offset i -P offset1 ; According to C i-1 The road curvature k i-1 Calculate the Δl 1 The central angle θ 1 :θ 1 =Δl 1 ×k i-1 ; C i The coordinates in the orbital coordinate system (offset i ,k i ) into lidar coordinates (xi, yi), where 6. A method for extracting obstacles of underground locomotives, the method for extracting obstacles of underground locomotives The following steps are involved: S100. Taking the locomotive position as the origin, the region of interest ROI is mapped according to a preset path. 2 The grid cells are used for obstacle search; S200. Determine whether there is an obstacle in the searched grid unit three, and continue searching along a preset path after determination; S300. When searching for an end point of a preset search path, return to the origin and search again, and three-cluster the grid cells that are determined to have obstacles to obtain an obstacle list; It is characterized in that In the region of interest ROI 2 Before performing obstacle search, you need to search for the region of interest (ROI). 2 The division is performed, and the division method adopts the underground locomotive area of ​​interest division method as described in any one of claims 1-4.

7. The underground vehicle obstacle extraction method according to claim 6, It is characterized in that The obstacle determination method of the grid unit 3 is as follows: S201. Determine whether the value of the grid cell three is a fixed value PATH+OBSTACLE; S202. If yes, it is determined that there is an obstacle in grid unit 3, and the obstacle is intrusive; S203. Otherwise, it is determined that there is no intrusive obstacle in grid cell three.

8. The underground vehicle obstacle extraction method according to claim 6, It is characterized in that The laser radar coordinates of the location and multiple positioning beacons in front of the location are obtained according to the position of the locomotive on the underground track.

9. The underground vehicle obstacle extraction method according to claim 8, It is characterized in that The method for obtaining the position of a locomotive comprises the following steps: Taking the laser radar of the locomotive as a detection point, controlling the detection point to send an initial detection signal to all positioning beacons, the positioning beacons adjacent to the locomotive receive the initial detection signal as feedback points, and send a cutoff data signal to the detection point, and the detection point receives the cutoff data signal; The time when the detection point sends the initial detection signal and receives the cut-off data signal and the time when the feedback point receives the initial signal and sends the cut-off data signal are used to obtain the flight time error t2 through the TOF function; Controlling the detection point to send a re-detection signal to the feedback point; The flight error time t2 is adjusted according to the time when the detection point sends the re-detection signal and the time when the feedback point receives the re-detection signal to obtain the actual error time. The method for obtaining the actual error time is as follows: (1) collecting the time a3 when the detection point sends the re-detection signal and the time b3 when the feedback point receives the re-detection signal; (2) The feedback time of the detection point is calculated by calculating the difference between the time a3 when the detection point sends the re-detection signal and the time a2 when the detection point receives the cut-off data signal. A ; (3) Feedback time reply through detection point A Adjust the flight error time t2 to obtain the actual error time t4: Among them, reply A is the feedback time of the detection point, reply B is the feedback time of the feedback point, e a is the time offset of the detection point caused by the crystal oscillator, e b The time offset of the feedback point caused by the crystal oscillator; The error actual time t4 and the true value tof are added to calculate the estimated value of the signal flight time The estimated value The relative distance between the detection point and the feedback point is calculated by a speed formula.

10. The underground vehicle obstacle extraction method according to claim 9, It is characterized in that The feedback point calibration method comprises the following steps: Preset a calibration time range; Obtaining the cycle time from when the detection point transmits a signal to when it receives a feedback signal; Determine whether the cycle time is within the calibration time range If yes, it is determined that the positioning beacon that sends the feedback signal is the measurement point.

Images