Foreign-object intrusion identification method based on engineering vehicle operation safety monitoring system

Abstract

Disclosed in the present invention is a foreign-object intrusion identification method based on an engineering vehicle operation safety monitoring system. By using LiDAR and a multi-modal camera in combination with basic data of an engineering vehicle operation safety monitoring system and artificial intelligence algorithm technology, category and position information of key targets such as a person, a vehicle and a general common foreign object in front of an engineering vehicle during operation can be accurately identified, so that a driver is prompted to take a corresponding measure in a timely manner, thereby achieving the functions of early-warning and protection against foreign-object intrusion in an operation process of an engineering vehicle during driving.

Description

An alien intrusion identification method based on an engineering vehicle operation safety monitoring system TECHNICAL FIELD

[0001] The present application relates to the field of rail transit safety protection, and in particular to an alien intrusion identification method based on an engineering vehicle operation safety monitoring system. BACKGROUND

[0002] Alien intrusion refers to the phenomenon that obstacles such as sand and rock falls, animals and pedestrians around the track invade the track and endanger the safety of train operation. At present, the track safety monitoring system in China generally uses video monitoring method to monitor whether there is alien intrusion in the track and initiate an alarm. However, in the traditional monitoring method, the alarm generated when any area of the track line has alien intrusion is the same, so that the relevant staff cannot take targeted measures according to the alarm, and it is difficult to better guarantee the safety of rail transit operation.

[0003] In recent years, accidents such as personnel staying on the track and alien falling into the line often occur on urban rail transit lines, which seriously threaten the safety of metro operation. Since the engineering vehicles of urban rail transit often operate at night and the working environment is complex, relying only on the human eye to identify in the case of alien intrusion is easy to misjudge and may not be able to accurately perform corresponding operations, thus there is a great safety risk.

[0004] According to the intelligent operation and safety construction requirements of China Urban Rail Transit Smart City Rail Development Outline, the standardization system of key system safety protection and risk assessment for train operation will be completed by 2025, and the indicators such as operation safety and equipment guarantee of the whole industry will reach the world leading level by 2035, so it is urgent to develop an alien intrusion identification system for the operation of engineering vehicles. Machine vision is used instead of human eyes, and multiple sensor detection data are fused to detect and process personnel, vehicles and common alien objects on the line in real time, minimize human factors, solve the problem of alien intrusion during the operation of engineering vehicles, and further improve the standardization system of key system safety protection and risk assessment for train operation of urban rail transit.

[0005] Chinese patent document CN107671414A discloses a "track alien intrusion monitoring method, device, system and monitoring host equipment". The method comprises: acquiring a monitoring image of the track; identifying and detecting the monitoring image to generate alien intrusion information; the alien intrusion information includes alien position information; determining a preset prompt area where the alien is located according to the alien position information; and outputting alarm information of a corresponding level according to the preset prompt area where the alien is located. The above technical solution has a single detection method, limited identification accuracy and a single response measure.

[0006] SUMMARY

[0007] The present application mainly solves the technical problems of the original technical solution, such as single detection method, limited recognition accuracy and single response measure, and provides an alien intrusion recognition method based on an engineering vehicle running safety monitoring system, which adopts a laser radar, a multi-modal camera, combines engineering vehicle running safety monitoring system basic data and artificial intelligence algorithm technology, can accurately identify the category and position information of key targets in front of the engineering vehicle running, such as personnel, vehicles and general common alien objects, so as to remind the driver to take corresponding measures in time, and realize the early warning and protection function of alien intrusion in the engineering vehicle driving operation process.

[0008] The above technical problems of the present application are mainly solved by the following technical solution: the present application comprises the following steps:

[0009] S1 collects point cloud information and multi-modal image information in the receptive field;

[0010] S2 outputs line and vehicle information in the engineering vehicle driving operation process;

[0011] S3 obtains three-dimensional point cloud segmentation of the track area and three-dimensional coordinates of the alien intrusion target;

[0012] S4 outputs alien intrusion information and takes response measures.

[0013] The laser radar, multi-modal camera and basic data of the engineering vehicle running safety monitoring system are input into the multi-modal information, the image information and point cloud information of the track area are obtained through registration and post-processing, the multi-modal information is sent into the edge computing system, the multi-modal information is processed by the edge computing module, the intrusion early warning information in the engineering vehicle driving operation process is obtained, and is transmitted to the engineering vehicle running safety monitoring system through gigabit Ethernet, and finally the voice broadcast or vehicle emergency braking measures are taken in the man-machine interface of the engineering vehicle running safety monitoring system.

[0014] As a preferred, the step S1 adopts a laser radar to collect point cloud information in the receptive field, and adopts a multi-modal camera to collect multi-modal image information. The short-focus visible light camera obtains image information in a large range at a short distance, and the long-focus visible light camera obtains image information at a long distance.

[0015] As a preferred, the step S2 specifically comprises that the engineering vehicle running safety monitoring system obtains the accurate position of the engineering vehicle and the position of the ground prevention and control point, combines the vehicle-mounted basic data and station data, outputs the position information of the line in front of the engineering vehicle running and the position information of the passing landmark terrain, and simultaneously outputs the phase and speed information of the current running vehicle. In the engineering vehicle driving operation process, the engineering vehicle running safety monitoring system outputs line basic data, vehicle positioning, vehicle speed and other information.

[0016] Preferably, the step S3 comprises collating the laser radar point cloud information and the engineering vehicle running safety monitoring system basic data information at the same time based on unified timestamp management, and obtaining multi-modal information in the same coordinate system through coordinate transformation.

[0017] The midpoint of the track directly in front of and below the train is taken as the origin of the track coordinate system, the driving direction of the train is taken as the z-axis, the axis perpendicular to the track plane upward is the y-axis, and the direction of the x-axis points to the left side of the z-axis. The coordinate system of the camera and the laser radar is set similarly. The normal vector of the imaging plane is aligned with the z-axis outward; the upward vector is aligned with the y-axis; and the x-axis is located on the left side of the z-axis.

[0018] The transformation relationship between the camera coordinate system and the track coordinate system is shown in formula (1), wherein Z c is the depth of the object pointing to the image plane, K is the internal parameter of the camera, T r_c is the external parameter of the camera relative to the track coordinate system. In our system, the pixel coordinates of the image are represented by u and v, wherein u represents the horizontal pixel, and v represents the vertical pixel. The railway track coordinates are represented by X r , Y r and Z r .

[0019] In the medium distance (for example, within 1 km), the slope change of the track is slow enough, so it can be abstractly assumed that the track within the distance is located on a single plane. The y value of the points within the track area in the track coordinate system is zero, a linear equation set can be established under the known parameters, and the coordinates of each pixel point in the camera coordinate system in the track coordinate system can be calculated.

[0020] The transformation process is shown in formula (2), wherein M is the transformation matrix of the camera coordinate to the track coordinate. The orthogonal projection can be considered as the scaling and rotation of the points in the track coordinate system, and this process can be represented by formula (3), wherein S is the transformation matrix of the two coordinate systems, and u0 and v0 are the pixel points of the orthogonal image coordinate system. The laser radar point cloud is projected in the track coordinate system using a similar method.

[0021] The vehicle position information and the position information of the line, turnout, curve, bridge, tunnel and the like in front of the running of the engineering vehicle running safety monitoring system are changed in the corresponding coordinates relative to the track coordinate system, taking the midpoint of the track directly in front of and below the train as the origin of the track coordinate system and taking the driving direction of the train as the z-axis, as shown in formula (4): X g = 0, Y g = 0, Z g = T r_g (4)

[0022] X, which outputs location information from the engineering vehicle operation safety monitoring system. g and Y g All are 0; T r_g The absolute difference between the kilometer marker output by the engineering vehicle operation safety monitoring system and the kilometer marker at the midpoint of the track directly in front of and below the train.

[0023] Subsequent detection algorithms and corresponding output results all use the orbital coordinate system as the reference coordinate system.

[0024] Preferably, step S3 includes training a track semantic segmentation algorithm model using massive publicly available track datasets and track image data stored in urban rail transit safety monitoring cameras as samples to identify the position coordinates and direction trend of track lines in the two-dimensional image. Multimodal information under the same reference coordinate system at the same time is input into the edge computing module. The track semantic segmentation algorithm model BisenetV2 is used to identify the position coordinates, direction, and trend of track lines in the two-dimensional image.

[0025] Preferably, step S3 includes: inferring the track area where the engineering vehicle is located based on the installed external reference; calculating the back-projection point cloud information of the track area boundary by means of the internal reference of the camera and the external reference of the relative position of the radar and the camera, combined with the position information of the engineering vehicle operation safety monitoring system.

[0026] As a preferred method, by combining the original point cloud and the back-projected point cloud, a precise track point cloud is segmented, and a multi-segment polynomial of the track curve is fitted to determine the track area and its plane. The acquired multimodal image information is processed by a track semantic segmentation algorithm, fused with basic data from the engineering vehicle operation safety monitoring system and LiDAR point cloud information, to obtain a three-dimensional point cloud segmentation of the track area. Furthermore, the specific three-dimensional coordinates of the foreign object intrusion target in three-dimensional space are located, and finally, the foreign object intrusion information is output, and corresponding measures are taken.

[0027] Preferably, step S3 involves performing motion compensation processing on the point cloud based on the speed information output by the engineering vehicle operation safety monitoring system to eliminate motion errors.

[0028] A uniform acceleration model is used to calibrate the motion of the point cloud. It is assumed that the initial pose of the lidar scan is... We assume that during the scanning process, the vehicle can be considered to accelerate at a constant speed (or move at a constant speed), and that the engineering vehicle only undergoes displacement changes. Therefore, we can use the velocity integral method to calculate the displacement of each point relative to the target at each moment. arrive The displacement is then calculated, and the motion error is eliminated through translation transformation, as shown in equations (5) and (6):

[0029] As preferred, the step S4 searches the intrusion point cloud in the determined track area three-dimensional space, determines whether there is an intrusion in the track through the pre-set height and point cloud density threshold, and simultaneously alarms and takes measures according to the intrusion result. After determining the boundary of the intrusion object, the alarm information is output, and finally the voice broadcast or vehicle emergency braking and other measures are taken on the man-machine interface of the engineering vehicle running safety monitoring system. If it is judged according to the pre-set height and point cloud density threshold that there is no intrusion in the track, the engineering vehicle running safety monitoring system does not issue an alarm, and continues to travel according to the original path; if it is judged according to the pre-set height and point cloud density threshold that there is an intrusion in the track, the intrusion object is accurately judged, and it is judged whether there is a safety hazard in the continuous travel according to the original path. If there is no safety hazard in the continuous travel according to the original path, a bumping reminder is issued and the continuous travel according to the original path is continued; if there is a safety hazard in the continuous travel according to the original path, an alarm is issued and emergency braking countdown is performed, and if the safety hazard is not removed and the emergency braking is not performed after the countdown is over, the emergency braking is automatically performed.

[0030] The beneficial effects of the present application are: the use of laser radar, multi-modal camera, combined with engineering vehicle running safety monitoring system basic data and artificial intelligence algorithm technology, can accurately identify the category and position information of the key target in front of the engineering vehicle running, such as personnel, vehicles and general common foreign objects, so as to remind the driver to take corresponding measures in time, and realize the early warning and protection function of foreign object intrusion in the engineering vehicle driving process. BRIEF DESCRIPTION OF DRAWINGS

[0031] Fig. 1 is a flowchart of the present application.

[0032] Fig. 2 is a working principle diagram of the present application.

[0033] Fig. 3 is a coordinate transformation schematic diagram of the present application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the technical scheme of the present application is further described in detail below through examples, and combined with the drawings. It should be understood that the specific embodiments described here are only one of the best embodiments of the present application, which are used to explain the present application and do not limit the protection scope of the present application. All other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0035] Before discussing the example embodiments in more detail, it should be mentioned that some of the example embodiments are described as processes or methods depicted as flow charts. Although the processes are described in a certain order, many of the operations (or steps) can be performed concurrently, in parallel, or simultaneously. In addition, the order of the operations can be re-arranged. The processes can be terminated when their operations are completed, but can also have additional steps not included in the figure(s); the processes can correspond to methods, functions, procedures, subroutines, etc.

[0036] The technical solutions of the present application will be further described in detail below through examples and in combination with the drawings.

[0037] Embodiment: An alien intrusion recognition method based on the operation safety monitoring system of the engineering vehicle, as shown in FIG. 1, the scheme inputs the laser radar, multi-modal camera and the basic data of the engineering vehicle operation safety monitoring system into the multi-modal information, obtains the image information and point cloud information of the track area through registration and post-processing, sends the multi-modal information into the edge computing system, processes the multi-modal information by the edge computing module, obtains the intrusion early warning information in the engineering vehicle driving process, and transmits it to the engineering vehicle operation safety monitoring system through the gigabit Ethernet, and finally performs voice broadcast or takes vehicle emergency braking and other measures on the human-computer interface of the engineering vehicle operation safety monitoring system. It includes the following steps:

[0038] S1: Collecting point cloud information and multi-modal image information in the receptive field. Laser radar is used to collect point cloud information in the receptive field, and multi-modal camera is used to collect multi-modal image information. The short-focus visible light camera obtains image information in a wide range at a short distance, and the long-focus visible light camera obtains image information at a long distance.

[0039] S2: Outputting line and vehicle information in the engineering vehicle driving process, specifically including that the engineering vehicle operation safety monitoring system outputs the position information of the line in front of the engineering vehicle and the position information of the landmark terrain passed by the engineering vehicle by obtaining the accurate position of the engineering vehicle and the position of the ground prevention and control point, combining the vehicle basic data and the station data, and outputting the phase and speed information of the current running vehicle. The engineering vehicle operation safety monitoring system outputs line basic data, vehicle positioning, vehicle speed and other information in the engineering vehicle driving process.

[0040] S3: Obtaining three-dimensional point cloud segmentation of the track area and three-dimensional coordinates of the alien intrusion target.

[0041] Based on the unified timestamp management, the laser radar point cloud information and the engineering vehicle operation safety monitoring system basic data information at the same time are sorted out, and the multi-modal information in the same coordinate system is obtained through coordinate transformation.

[0042] The midpoint of the track directly in front of and below the train is taken as the origin of the track coordinate system, the direction of travel of the train is taken as the z-axis, the axis perpendicular to the track plane upward is the y-axis, and the direction of the x-axis points to the left of the z-axis. The coordinate system of the camera and the laser radar is set similarly. The normal vector of the imaging plane is aligned with the z-axis, outward; the upward vector is aligned with the y-axis; and the x-axis is to the left of the z-axis.

[0043] The transformation relationship between the camera coordinate system and the track coordinate system is shown in formula (1), where Z c is the depth of the object pointing to the image plane, K is the intrinsic parameter of the camera, and T r_c is the extrinsic parameter of the camera relative to the track coordinate system. In our system, the pixel coordinates of the image are represented by u and v, where u represents the horizontal pixel and v represents the vertical pixel. The railway track coordinates are represented by X r , Y r , and Z r .

[0044] At medium distances (e.g., within 1 km), the slope of the track changes slowly enough, so it can be abstractly assumed that the track within the distance lies on a single plane. The points within the track area in the track coordinate system have a y value of zero, and a linear equation system can be established under known parameters to calculate the coordinates of each pixel point in the camera coordinate system in the track coordinate system.

[0045] The transformation process is shown in formula (2), where M is the transformation matrix of the camera coordinate to the track coordinate. The orthogonal projection can be considered as the scaling and rotation of the points in the track coordinate system, which can be represented by formula (3), where S is the transformation matrix of the two coordinate systems, and u0 and v0 are the pixel points of the orthogonal image coordinate system. The laser radar point cloud is projected in the track coordinate system using a similar method.

[0046] The vehicle position information output by the engineering vehicle running safety monitoring system and the position information of the line, turnout, curve, bridge, tunnel, etc. in front of the running, relative to the track coordinate system, take the midpoint of the track directly in front of and below the train as the origin of the track coordinate system and take the direction of travel of the train as the z-axis, and perform corresponding coordinate changes, as shown in formula (4): X g = 0, Y g = 0, Z g = T r_g (4)

[0047] Where X g and Y g of the position information output by the engineering vehicle running safety monitoring system are both 0; T r_gThe absolute difference between the kilometer marker output by the engineering vehicle operation safety monitoring system and the kilometer marker of the midpoint of the track directly in front of and below the train.

[0048] The subsequent detection algorithm and the corresponding output result are both based on the track coordinate system as the reference coordinate system.

[0049] Through the massive public track data set and the track image data stored by the urban rail transit safety monitoring camera as samples, a track semantic segmentation algorithm model is trained to identify the track line position coordinates and trend in the two-dimensional image. The multi-modal information in the same time and the same reference coordinate system is input into the edge computing module. Through the track semantic segmentation algorithm model BisenetV2, the track line position coordinates, trend, and other information in the two-dimensional image are identified.

[0050] According to the installed external reference, the track area where the engineering vehicle is located is inferred, and the track area boundary back-projection point cloud information is calculated by combining the position information of the engineering vehicle operation safety monitoring system via the internal reference of the camera and the relative position external reference of the radar and the camera.

[0051] Combined with the original point cloud and the back-projection point cloud, the accurate track point cloud is segmented, the track curve polynomial is fitted, and the track area and the track area plane are determined. The collected multi-modal image information is subjected to track semantic segmentation algorithm, and the engineering vehicle operation safety monitoring system basic data and laser radar point cloud information are fused to obtain the three-dimensional point cloud segmentation of the track area, and the specific three-dimensional coordinates of the foreign matter intrusion target in the three-dimensional space are found. Finally, the foreign matter intrusion information is output and corresponding measures are taken.

[0052] According to the speed information output by the engineering vehicle operation safety monitoring system, the point cloud is subjected to motion compensation processing to eliminate motion errors. A uniform acceleration model is used to calibrate the point cloud, and the initial pose of the laser radar scanning is assumed to be We assume that during the scanning process, the vehicle can be considered to be uniformly accelerated (or uniformly moving), and that the engineering vehicle only has displacement changes, which can be calculated by the speed integration method to calculate the displacement of each point at each time relative to to Then, through the translation transformation, the motion error is eliminated, as shown in equations (5) and (6):

[0053] S4 outputs the foreign matter intrusion information and takes countermeasures. In the determined three-dimensional space of the track area, the sliding window searches the intrusion point cloud, determines whether there is intrusion in the track according to the height and point cloud density threshold set in advance, and simultaneously alarms and takes countermeasures according to the intrusion result. After determining the boundary of the intrusion object, the alarm information is output, and finally the voice broadcast or vehicle emergency braking and other measures are taken on the man-machine interface of the engineering vehicle running safety monitoring system. If it is judged according to the height and point cloud density threshold set in advance that there is no intrusion in the track, the engineering vehicle running safety monitoring system does not issue an alarm, and continues to travel according to the original path; if it is judged according to the height and point cloud density threshold set in advance that there is intrusion in the track, the intrusion object is accurately judged, and it is judged whether there is a safety hazard in the original path. If there is no safety hazard in the original path, a bumping reminder is issued and the original path is continued to travel; if there is a safety hazard in the original path, an alarm is issued and emergency braking is performed, and if the safety hazard is not removed and the emergency braking is not performed after the countdown, the emergency braking is automatically performed.

[0054] The laser radar, multi-modal camera and basic data of the engineering vehicle running safety monitoring system are input into multi-modal information, and image information and point cloud information of the track area are obtained through registration and post-processing. The multi-modal information is sent to the edge computing system, the multi-modal information is processed by the edge computing module, the intrusion early warning information in the engineering vehicle driving operation process is obtained, and the information is transmitted to the engineering vehicle running safety monitoring system through gigabit Ethernet. Finally, the voice broadcast or vehicle emergency braking and other measures are taken on the man-machine interface of the engineering vehicle running safety monitoring system.

[0055] Embodiment

[0056] As shown in FIG. 3, it is a foreign matter intrusion identification scheme based on the engineering vehicle running safety monitoring system. The scheme inputs the laser radar, multi-modal camera and basic data of the engineering vehicle running safety monitoring system into multi-modal information, obtains image information and point cloud information of the track area through registration and post-processing, sends the multi-modal information to the edge computing system, processes the multi-modal information by the edge computing module, obtains the intrusion early warning information in the engineering vehicle driving operation process, and transmits the information to the engineering vehicle running safety monitoring system through gigabit Ethernet. Finally, the voice broadcast or vehicle emergency braking and other measures are taken on the man-machine interface of the engineering vehicle running safety monitoring system. The identification method includes the following steps:

[0057] Step 100, during the engineering vehicle driving operation process, the laser radar deployed on the vehicle collects the point cloud information in the receptive field;

[0058] In step 200, during the operation of the engineering vehicle, the multi-modal camera deployed on the vehicle collects multi-modal image information, in which the short-focus visible light camera acquires image information in a large range at a short distance, and the long-focus visible light camera acquires image information at a long distance.

[0059] In step 300, during the operation of the engineering vehicle, the engineering vehicle running safety monitoring system outputs line basic data, vehicle positioning, vehicle speed and other information.

[0060] In step 400, the collected laser radar point cloud information, multi-modal image information and engineering vehicle running safety monitoring system basic data information are input into a multi-modal information receiving and calculating unit.

[0061] In step 401, based on unified timestamp management, the laser radar point cloud information, multi-modal image information and engineering vehicle running safety monitoring system basic data information at the same time are sorted out, and multi-modal information in the same coordinate system is obtained through coordinate transformation.

[0062] In step 402, the midpoint of the track directly in front of and below the train is taken as the origin of the track coordinate system, the driving direction of the train is taken as the z-axis, the axis perpendicular to the track plane upward is taken as the y-axis, and the direction of the x-axis points to the left side of the z-axis. The coordinate system of the camera and the laser radar is set similarly. The normal vector of the imaging plane is aligned with the z-axis outward; the upward vector is aligned with the y-axis; and the x-axis is located on the left side of the z-axis.

[0063] The transformation relationship between the camera coordinate system and the track coordinate system is shown in formula (1), wherein Z c is the depth of the object pointing to the image plane, K is the intrinsic parameter of the camera, T r_c is the extrinsic parameter of the camera relative to the track coordinate system. In our system, the pixel coordinates of the image are represented by u and v, wherein u represents the horizontal pixel, and v represents the vertical pixel. The railway track coordinates are represented by X r , Y r and Z r .

[0064] In the medium distance (for example, within 1 kilometer), the slope change of the track is slow enough, so it can be abstractly assumed that the track within the distance is located on a single plane. The y values of the points in the track region in the track coordinate system are all zero, a linear equation set can be established under the known parameters, and the coordinates of each pixel point in the camera coordinate system in the track coordinate system are calculated.

[0065] The transformation process is shown in equation (2), where M is the transformation matrix from camera coordinates to track coordinates. The orthogonal projection can be considered as a scaling and rotation of the point in the track coordinate system, which can be represented by equation (3), where S is the transformation matrix of the two coordinate systems, and u0 and v0 are the pixel points of the orthogonal image coordinate system. The laser radar point cloud is projected in the track coordinate system using a similar method.

[0066] The vehicle position information output by the engineering vehicle running safety monitoring system and the position information of the front line, turnout, curve, bridge, tunnel, etc. relative to the track coordinate system, with the track midpoint directly in front of and below the train as the origin of the track coordinate system and the direction of travel of the train as the z-axis, are subjected to corresponding coordinate changes, as shown in equation (4): g = 0, Y g = 0, Z g = T r_g (4)

[0067] where X g and Y g of the position information output by the engineering vehicle running safety monitoring system are both 0; T r_g is the absolute difference between the kilometer marker of the position information output by the engineering vehicle running safety monitoring system and the kilometer marker of the track midpoint directly in front of and below the train.

[0068] The subsequent detection algorithms and corresponding output results are all based on the track coordinate system as the reference coordinate system.

[0069] Step 403, input the multi-modal information under the same reference coordinate system at the same time into the edge computing module. Through the track semantic segmentation algorithm model BisenetV2, identify the track line position coordinates, trend, etc. in the two-dimensional image; among the detected track lines, according to the installation external parameters, infer the track area where the vehicle route is located, calculate the track area boundary back projection point cloud information through the camera's internal parameters, the relative position external parameters of the radar and the camera and the position information of the engineering vehicle running safety monitoring system; combine the original point cloud and the back projection point cloud to segment the accurate track point cloud, fit the track curve polynomial, and determine the track area and the track area plane;

[0070] Step 404, according to the speed information output by the engineering vehicle running safety monitoring system, do motion compensation processing on the point cloud to eliminate motion error. Adopt a uniform acceleration model to calibrate the point cloud, assuming that the initial pose of the laser radar scanning is We assume that during the scanning process, the vehicle can be considered as uniform acceleration (or uniform motion), and assuming that the engineering vehicle only has displacement change, the speed integral method can be used to calculate the position of each point at each time relative to to The displacement of the object is then removed by a translation transformation, as shown in equations (5) and (6):

[0071] Step 405, in the determined three-dimensional space of the track area, a sliding window searches the intrusion point cloud, and whether there is an intrusion in the track is determined by the pre-set height and point cloud density threshold. After determining the boundary of the intrusion object, an alarm information is output, and finally a voice broadcast or vehicle emergency braking and the like is taken on the man-machine interface of the engineering vehicle running safety monitoring system. If it is judged according to the pre-set height and point cloud density threshold that there is no intrusion in the track, the engineering vehicle running safety monitoring system does not issue an alarm, and continues to travel according to the original path; if it is judged according to the pre-set height and point cloud density threshold that there is an intrusion in the track, the intrusion object is accurately judged, and it is judged whether there is a safety hazard in the continuous travel according to the original path. If there is no safety hazard in the continuous travel according to the original path, a bumping reminder is issued and the continuous travel according to the original path is continued; if there is a safety hazard in the continuous travel according to the original path, an alarm is issued and emergency braking is performed with a countdown, and if the safety hazard is not removed and the emergency braking is not performed when the countdown is over, the emergency braking is automatically performed.

[0072] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by programs instructing related hardware, and the programs can be stored in a computer readable storage medium, which can include a read only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0073] The specific embodiments described herein are merely illustrative of the spirit of the present application. The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be noted that those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, without departing from the spirit of the present application or exceeding the scope defined by the appended claims. For those skilled in the art, various modifications and improvements can be made without departing from the concept of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A foreign matter intrusion recognition method based on an engineering vehicle operation safety monitoring system, characterized by, The method comprises the following steps: S1: collecting point cloud information and multi-modal image information in the receptive field; S2: outputting line and vehicle information during the operation of the engineering vehicle; S3: obtaining three-dimensional point cloud segmentation of the track area and three-dimensional coordinates of the foreign object intrusion target; S4: outputting foreign object intrusion information and taking countermeasures.

2. The foreign object intrusion identification method of the operation safety monitoring system for engineering vehicles according to claim 1, characterized in that, The step S1 uses a laser radar to collect point cloud information in the receptive field and uses a multi-modal camera to collect multi-modal image information.

3. The foreign object intrusion identification method of the operation safety monitoring system for engineering vehicles according to claim 1, characterized in that, The step S2 specifically comprises: an engineering vehicle operation safety monitoring system obtains the accurate position of the engineering vehicle and the position of the ground prevention and control point, combines vehicle-based data and station data, outputs the position information of the line in front of the engineering vehicle and the position information of the landmark terrain passed through, and simultaneously outputs the phase and speed information of the current operation of the vehicle.

4. The foreign object intrusion identification method of the operation safety monitoring system for engineering vehicles according to claim 1, characterized in that, The step S3 comprises: based on unified timestamp management, the laser radar point cloud information and the engineering vehicle operation safety monitoring system basic data information at the same time are sorted out, and multi-modal information in the same coordinate system is obtained through coordinate transformation.

5. The foreign object intrusion identification method of claim 4, wherein, The step S3 comprises: a track semantic segmentation algorithm model is trained by using mass public track data sets and track image data stored by a city rail transit safety monitoring camera as samples to identify the track line position coordinates and trend in a two-dimensional image.

6. The foreign object intrusion identification method of the operation safety monitoring system for engineering vehicles according to claim 5, characterized in that, The step S3 comprises: the track area where the engineering vehicle is located is inferred according to the installed external reference, the boundary back projection point cloud information of the track area is calculated by combining the position information of the engineering vehicle operation safety monitoring system via the internal reference of the camera and the relative position external reference of the radar and the camera.

7. The foreign object intrusion identification method of the operation safety monitoring system for engineering vehicles according to claim 6, characterized in that, Combined with the original point cloud and the back projection point cloud, the accurate track point cloud is segmented, the track curve polynomial is fitted, and the track area and the track area plane are determined.

8. The foreign object intrusion identification method of the operation safety monitoring system for engineering vehicles according to claim 5 or 6 or 7, characterized in that, The step S3 performs motion compensation processing on the point cloud according to the speed information output by the engineering vehicle operation safety monitoring system to eliminate motion errors.

9. The foreign object intrusion identification method of the operation safety monitoring system for engineering vehicles according to claim 1, characterized in that, The step S4 searches for intrusion point clouds in the determined three-dimensional space of the track area by using a sliding window, determines whether there is intrusion in the track according to the height and point cloud density threshold values set in advance, and simultaneously alarms and takes countermeasures according to the intrusion result.

Images