Perimeter intrusion identification method and system based on laser radar detection

Abstract

The present application provides a perimeter intrusion identification method and system based on laser radar detection, which relates to the field of railway safety technology. The method comprises: reading detection constraints of a preset laser radar; obtaining a radar deployment position, and setting a preset laser radar in a predetermined railway perimeter area according to the radar deployment position; obtaining a background point cloud data set, and constructing a first reference coordinate system; generating a perimeter detection point cloud data set; obtaining real-time intrusion point cloud data; obtaining multiple point cloud aggregation data; calling an intrusion feature portrait database for intrusion feature matching, and obtaining a perimeter intrusion identification result, which solves the technical problem that the denoising effect of the laser radar in harsh environments is poor, which leads to insufficient anti-interference performance in harsh environments and insufficient accuracy of perimeter intrusion identification, and achieves the technical effect of effectively improving the anti-interference performance of the laser radar in harsh environments, and thus improving the detection accuracy and the accuracy of perimeter intrusion identification.

Description

Technical Field

[0001] The present application relates to the field of railway safety technology, and in particular to a perimeter intrusion identification method and system based on laser radar detection. Background Art

[0002] Railway perimeter intrusion occurs when an external target, such as a person or foreign object, enters the train's operating area without authorization. Railways rely on standardized operating procedures and enclosed physical barriers to regulate passenger boarding and alighting. Unauthorized entry into the train's operating area poses a safety risk to normal train operations. Therefore, timely identification of perimeter intrusions is crucial for maintaining normal train operations.

[0003] LiDAR can quickly and accurately identify perimeter intrusions by emitting laser signals. However, the detection signal of LiDAR will attenuate in harsh environments. The existing technology does not perform well in denoising processing of LiDAR in harsh environments, which leads to insufficient anti-interference performance in harsh environments and insufficient accuracy in perimeter intrusion identification. Summary of the Invention

[0004] The present application provides a perimeter intrusion identification method and system based on lidar detection, which is used to solve the technical problems existing in the prior art, such as poor denoising effect of lidar in harsh environments, resulting in insufficient anti-interference performance in harsh environments and insufficient accuracy of perimeter intrusion identification.

[0005] According to the first aspect of the present application, a perimeter intrusion identification method based on laser radar detection is provided, comprising: reading the detection constraint conditions of a preset laser radar, wherein the detection constraint conditions include detection distance, horizontal field of view angle, and vertical field of view angle; obtaining the regional spatial characteristics of a predetermined railway perimeter area, and planning the layout position of the preset laser radar in combination with the detection constraint conditions to obtain a radar layout position, and setting a preset laser radar in the predetermined railway perimeter area according to the radar layout position; performing a perimeter background scan of the predetermined railway perimeter area using the preset laser radar under the first preset condition to obtain a background point cloud dataset, and constructing a background point cloud dataset based on the background point cloud dataset. A first reference coordinate system; starting the preset laser radar in a predetermined time period to perform continuous laser scanning, receiving echo signals and performing lost signal compensation to generate a perimeter detection point cloud data set; performing background removal on the perimeter detection point cloud data set according to the background point cloud data set to obtain real-time intrusion point cloud data; denoising the real-time intrusion point cloud data through noise simulation, clustering the denoised real-time intrusion point cloud data according to point cloud intensive features to obtain a plurality of point cloud aggregated data; calling the intrusion feature portrait database to perform intrusion feature matching on the plurality of point cloud aggregated data, and performing position identification in the first reference coordinate system to obtain a perimeter intrusion identification result.

[0006] According to the second aspect of the present application, a perimeter intrusion identification system based on laser radar detection is provided, including: a detection constraint reading module, the detection constraint reading module is used to read the detection constraint of a preset laser radar, the detection constraint including the detection distance, the horizontal field of view angle and the vertical field of view angle; a radar deployment position planning module, the radar deployment position planning module is used to obtain the regional spatial characteristics of a predetermined railway perimeter area, and plan the deployment position of the preset laser radar in combination with the detection constraint to obtain the radar deployment position, and set the preset laser radar in the predetermined railway perimeter area according to the radar deployment position; a perimeter background scanning module, the perimeter background scanning module is used to perform a perimeter background scan of the predetermined railway perimeter area using the preset laser radar under a first preset condition to obtain a background point cloud data set, and construct a background point cloud data set based on the background point cloud data set. A first reference coordinate system; a laser detection module, the laser detection module is used to start the preset laser radar to perform continuous laser scanning in a predetermined time period, receive echo signals and perform lost signal compensation, and generate a perimeter detection point cloud data set; a background removal module, the background removal module is used to perform background removal on the perimeter detection point cloud data set according to the background point cloud data set to obtain real-time intrusion point cloud data; a clustering module, the clustering module is used to denoise the real-time intrusion point cloud data through noise simulation, cluster the denoised real-time intrusion point cloud data according to the point cloud dense features, and obtain multiple point cloud aggregated data; a perimeter intrusion identification module, the perimeter intrusion identification module is used to call the intrusion feature portrait database to perform intrusion feature matching on the multiple point cloud aggregated data, and perform position identification in the first reference coordinate system to obtain a perimeter intrusion identification result.

[0007] The beneficial effects that can be achieved by one or more technical solutions adopted in this application are as follows:

[0008] By reading the detection constraints of the preset laser radar, the regional spatial characteristics of the predetermined railway perimeter area are obtained, and the preset laser radar is planned to be deployed in combination with the detection constraints to obtain the radar deployment position. The preset laser radar is set in the predetermined railway perimeter area according to the radar deployment position, and the preset laser radar is used to perform perimeter background scanning of the predetermined railway perimeter area under the first preset conditions to obtain a background point cloud dataset, and a first reference coordinate system is constructed based on the background point cloud dataset. The preset laser radar is started to perform laser continuous scanning in a predetermined time period, and the echo signal is received and the lost signal compensation is performed to generate a perimeter detection point cloud dataset. The background of the perimeter detection point cloud dataset is removed according to the background point cloud dataset to obtain real-time intrusion point cloud data, and the real-time intrusion point cloud data is denoised by noise simulation. The denoised real-time intrusion point cloud data is clustered according to the point cloud density characteristics to obtain multiple point cloud aggregation data, and the intrusion feature portrait database is called to perform intrusion feature matching on the multiple point cloud aggregation data. Position recognition is performed in the first reference coordinate system to obtain perimeter intrusion recognition results. Therefore, by compensating for signal loss of the laser radar's detection signal and denoising it in combination with the meteorological environment, perimeter intrusion identification is performed using the denoised point cloud data, which effectively improves the laser radar's anti-interference performance in harsh environments, thereby improving the detection accuracy and perimeter intrusion identification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in this application or the prior art, the following briefly introduces the drawings required for the embodiments or the prior art description. The drawings that constitute a part of this application are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without making any creative efforts.

[0010] Figure 1 A schematic diagram of a process flow of a perimeter intrusion identification method based on laser radar detection provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the structure of a perimeter intrusion identification system based on lidar detection provided in an embodiment of the present application.

[0012] Explanation of the accompanying drawings: detection constraint reading module 11, radar deployment position planning module 12, perimeter background scanning module 13, laser detection module 14, background removal module 15, clustering module 16, perimeter intrusion identification module 17. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solutions and advantages of this application more apparent, the following exemplary embodiments of this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this application, rather than all the embodiments of this application, and it should be understood that this application is not limited to the exemplary embodiments described herein.

[0014] The terms used in the specification are used to describe the embodiments rather than to limit the present application. As used in the specification, the singular terms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. When used in the specification, the terms "comprise" and / or "include" specify the presence of steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other steps, operations, elements, components, and / or groups thereof.

[0015] Unless otherwise defined, all terms (including technical and scientific terms) used in this specification shall have the same meaning as commonly understood by those skilled in the art to which this application belongs. Terms, such as those defined in commonly used dictionaries, should not be interpreted in an idealized or overly formal sense unless explicitly defined herein. Throughout the specification, the same reference numerals represent the same elements.

[0016] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0017] Example 1

[0018] Figure 1 A diagram of a perimeter intrusion identification method based on laser radar detection provided in an embodiment of the present application, the method comprising:

[0019] Reading the detection constraints of the preset laser radar, wherein the detection constraints include detection range, horizontal field of view angle, and vertical field of view angle;

[0020] Specifically, the preset laser radar refers to a three-dimensional laser radar to be installed for perimeter intrusion identification. This embodiment does not limit the model of the three-dimensional laser radar, and an existing three-dimensional laser radar can be selected as the preset laser radar according to actual conditions. The detection constraints include detection distance, horizontal field of view angle, and vertical field of view angle. The detection distance refers to the maximum distance that the preset laser radar can detect by emitting a laser beam. The horizontal field of view angle refers to the detection angle range in the horizontal direction, such as 40° horizontally. The vertical field of view angle refers to the detection angle range in the vertical direction. Among them, the detection distance, horizontal field of view angle, and vertical field of view angle can all be directly read according to the user manual of the preset laser radar. The detection distance, horizontal field of view angle, and vertical field of view angle constitute the detection constraints, which facilitate the determination of the detection range of a single preset laser radar, thereby performing radar deployment and ensuring the comprehensiveness of detection.

[0021] Acquire regional spatial characteristics of a predetermined railway perimeter area, and plan the layout position of the preset laser radar in combination with the detection constraint conditions to obtain a radar layout position, and set the preset laser radar in the predetermined railway perimeter area according to the radar layout position;

[0022] A designated railway perimeter area refers to any area to be identified for perimeter intrusion, such as a section of railway. This involves timely identification of humans, animals, flash floods, and other intrusions into the designated railway perimeter area, facilitating prompting trains to respond with emergency measures and preventing major safety incidents. Regional spatial characteristics refer to the spatial shape, size, and other features of the designated railway perimeter area. These features can be uploaded directly by personnel working in the designated railway perimeter area through a user terminal, or by using existing drones to capture multi-angle images of the designated railway perimeter area. Based on existing technologies, a regional map is constructed based on the image capture results to obtain regional spatial characteristics. Furthermore, based on the regional spatial characteristics and the detection constraints, the locations of the designated laser radars are planned to ensure comprehensive monitoring of the designated railway perimeter area using the designated laser radars while avoiding excessive radar deployment and resource waste. Radar deployment locations are determined, and designated laser radars are deployed within the designated railway perimeter area according to the radar deployment locations.

[0023] In a preferred embodiment, it also includes:

[0024] A unit detection space is established according to the detection distance, the horizontal field of view angle, and the vertical field of view angle; spatial area coverage fitting is performed on the regional spatial features using the unit detection space, and the coverage quantity of the unit detection space is extracted using the coverage fitting result; overlapping coverage areas are identified on the coverage fitting result to obtain the overlapping space size; the minimum overlapping space size and the minimum coverage quantity are used as deployment targets, and the coverage fitting result is optimized to obtain the radar deployment position.

[0025] Specifically, a unit detection space is established based on the detection distance, the horizontal field of view angle, and the vertical field of view angle. A unit detection space refers to the range that a single preset laser radar can monitor. A fixed point is set, and the detection distance, the horizontal field of view angle, and the vertical field of view angle are combined to obtain the unit detection space. Spatial coverage fitting is performed on the regional spatial feature using the unit detection space. That is, multiple unit detection spaces are fitted to the regional spatial feature to ensure complete coverage of the regional spatial feature. Multiple combinations of unit detection spaces at different locations are obtained as coverage fitting results. The coverage fitting results are used to extract the coverage quantity of the unit detection space. Overlapping coverage areas are then identified on the coverage fitting results. This involves determining the position of each unit detection space within the regional spatial feature. Because the unit detection spaces and the predetermined railway perimeter are irregular shapes, to ensure complete coverage of the predetermined railway perimeter, some unit detection spaces will overlap at the same location. The overlapping areas are then identified and the size of the overlapping spaces is determined. The minimum overlapping space size and the minimum coverage number are taken as the layout goals, and the coverage fitting results are optimized. That is to say, the multiple combination results of the unit detection space at different positions obtained above are used as the coverage fitting results, and the combination result with the minimum overlapping space size and the minimum coverage number is screened out from the multiple combination results as the radar layout position, so as to ensure the accuracy of the radar layout and ensure that the predetermined railway perimeter area can be fully monitored.

[0026] Under a first preset condition, using the preset laser radar to perform a perimeter background scan of the predetermined railway perimeter area to obtain a background point cloud dataset, and constructing a first reference coordinate system based on the background point cloud dataset;

[0027] The first preset condition refers to an environment that does not cause significant attenuation of the preset laser radar, namely, an environment free of smoke, fog, rain, snow, etc. that may affect laser emission, and an environment in which there are no unusual intruders within the predetermined railway perimeter. Under the first preset condition, the preset laser radar is used to perform a perimeter background scan of the predetermined railway perimeter, specifically scanning the predetermined railway perimeter without any unusual intruders, receiving a background point cloud dataset (which can be understood as point cloud data of normal structures such as railways and platforms within the predetermined railway perimeter). A first reference coordinate system is constructed based on the background point cloud dataset, specifically, a coordinate system for the predetermined railway perimeter, to facilitate precise location of the intrusion during subsequent perimeter intrusion identification.

[0028] Start the preset laser radar to perform continuous laser scanning in a predetermined time period, receive echo signals and perform lost signal compensation to generate a perimeter detection point cloud data set;

[0029] The predetermined time period refers to the time period for perimeter intrusion identification, which can be determined by professionals in this field. Generally, continuous monitoring will be performed. The predetermined time period generally refers to the time period after the preset laser radar is installed. During the predetermined time period, the preset laser radar is activated to perform continuous laser scanning, receive echo signals, and perform lost signal compensation. The echo signal refers to the echo signal received by the preset laser radar after emitting a laser beam. During the laser beam emission process, if there is interference from rain or snow in the environment, the laser will be scattered, resulting in signal point loss. Therefore, lost signal compensation is required. Based on the compensated echo signal, the echo signal is processed using existing technology to obtain the signal detection distance, thereby generating a perimeter detection point cloud dataset.

[0030] In a preferred embodiment, it also includes:

[0031] When the preset laser radar is started to perform laser scanning in a predetermined time period, the laser scanning points are recorded; real-time environmental particle positions are identified based on the image sensor to obtain real-time environmental particle points; intersection points of the environmental particle points and the laser scanning points are identified to obtain intersection points and the number of intersection points; if the number of intersection points exceeds a predetermined intersection point threshold, a lost point compensation instruction is generated; and signal loss compensation is performed on the echo signal at the intersection point according to the lost point compensation instruction.

[0032] When the preset laser radar is activated for laser scanning at a predetermined time period, the laser scanning points are recorded. Simply put, the positions of the lasers emitted by the preset laser radar during the emission process are recorded to obtain the laser scanning points. Real-time environmental particle positions are identified using an image sensor to obtain real-time environmental particle points. The image sensor, which is a conventional high-precision camera, captures the real-time environment using the image sensor to obtain a real-time environmental image. Particles such as smoke, rain, and snow are identified in the real-time environmental image according to existing technologies, and the locations of these particles are obtained as the environmental particle points. In layman's terms, when a laser beam encounters particles such as smoke, rain, and snow, it scatters, causing the laser point to disappear, meaning that part of the return signal may be lost. Therefore, intersection identification is performed between the environmental particle points and the laser scanning points. This involves identifying whether there are any overlaps between the environmental particle points and the laser scanning points, thereby obtaining the locations and number of intersections. If the number of intersection points exceeds the predetermined intersection point threshold, a lost point compensation instruction is generated. That is to say, if the number of points that overlap between the environmental particle points and the laser scanning points is relatively small, the number of signals lost during laser detection is also small, and the impact on detection accuracy can be ignored; if the number of points that overlap between the environmental particle points and the laser scanning points is relatively large, the number of signals lost during laser detection is also large, which has a greater impact on detection accuracy, and lost signal compensation is required to ensure detection accuracy. The predetermined intersection point threshold is set by professionals in this field and can be the minimum number of laser signal points lost when the detection result is abnormal. The lost point compensation instruction refers to a control instruction for controlling lost signal compensation. According to the lost point compensation instruction, signal loss compensation is performed on the echo signal at the intersection position. Simply put, the laser emission path is simulated according to the intersection position, and the simulation result is superimposed on the echo signal to achieve compensation for the echo signal, thereby improving the accuracy of laser detection and thus improving the accuracy of perimeter intrusion identification.

[0033] Performing background removal on the perimeter detection point cloud dataset according to the background point cloud dataset to obtain real-time intrusion point cloud data;

[0034] The background of the perimeter detection point cloud dataset is removed according to the background point cloud dataset, that is, the point cloud data in the background point cloud dataset is removed from the perimeter detection point cloud dataset to obtain real-time intrusion point cloud data, which is the point cloud data of the intruder.

[0035] De-noising the real-time intrusion point cloud data through noise simulation, clustering the de-noised real-time intrusion point cloud data according to point cloud density features to obtain a plurality of point cloud cluster data;

[0036] The real-time intrusion point cloud data is denoised through noise simulation. This means that under different meteorological environments, the detection accuracy of the preset lidar will attenuate to varying degrees, meaning that the real-time intrusion point cloud data may contain inaccurate point cloud data and require denoising. The denoised real-time intrusion point cloud data is further clustered according to point cloud density characteristics to obtain multiple point cloud aggregation data. This means that there may be multiple intruders within the predetermined railway perimeter area at the same time, and the distance between each intruder's point cloud must be very close. Therefore, clustering the denoised real-time intrusion point cloud data according to cloud density characteristics can effectively identify different intruders. Specifically, the point cloud distribution density in the denoised real-time intrusion point cloud data can be analyzed, and point clouds with the same or similar point cloud density can be clustered together to obtain point cloud aggregation data.

[0037] In a preferred embodiment, it also includes:

[0038] Acquire attenuation meteorological environment samples of the predetermined railway perimeter area; perform point cloud noise simulation based on the attenuation meteorological environment samples to obtain simulated noise, wherein the attenuation meteorological environment samples and the simulated noise have a one-to-one correspondence; train a point cloud noise simulation channel with the attenuation meteorological environment samples and the simulated noise; collect meteorological environment information for the predetermined railway perimeter area during the predetermined time period to obtain real-time environmental information; perform noise simulation on the real-time environmental information through the point cloud noise simulation channel to obtain a real-time simulated noise point cloud; and denoise the real-time intrusion point cloud data with the real-time simulated noise point cloud.

[0039] In a preferred embodiment, it also includes:

[0040] A laser radar scanning test is performed based on the attenuation meteorological environment sample to obtain real point cloud data and attenuation point cloud data; point cloud attenuation modeling is performed based on the real point cloud data and the attenuation point cloud data, and attenuation point cloud data is generated according to the modeling results.

[0041] The process of denoising the real-time intrusion point cloud data through noise simulation is as follows:

[0042] First, an attenuation meteorological environment sample is collected for the predetermined railway perimeter. The attenuation meteorological environment sample refers to meteorological conditions that will cause detection signal attenuation for a preset lidar, such as varying degrees of rain or snow. Specifically, the historical meteorological conditions of the predetermined railway perimeter can be retrieved as the attenuation meteorological environment sample. A point cloud noise simulation is performed based on the attenuation meteorological environment sample to obtain simulated noise. The attenuation meteorological environment sample and the simulated noise have a one-to-one correspondence.

[0043] Among them, the point cloud noise simulation is performed based on the attenuation meteorological environment sample, and the process of obtaining the simulated noise is as follows: a laser radar scanning test is performed based on the attenuation meteorological environment sample, that is, an environmental simulation is performed on site according to the attenuation meteorological environment sample, a scanning test is performed through a preset laser radar, and the point cloud data obtained by the scanning test is used as the attenuation point cloud data. At the same time, a scanning test of the preset laser radar is also performed under the first preset condition, and the point cloud data obtained by the scanning test is used as the real point cloud data. Point cloud attenuation modeling is performed based on the real point cloud data and the attenuation point cloud data, that is, the difference between the real point cloud data and the attenuation point cloud data is analyzed as the modeling result, and the point cloud deviation direction and deviation distance are obtained, and the point cloud deviation direction and deviation distance are used as the attenuation point cloud data. In this way, point cloud attenuation analysis under different meteorological environments is realized, which provides support for subsequent denoising processing, improves the accuracy of point cloud data, and reduces interference from harsh environments.

[0044] Furthermore, the attenuated meteorological environment samples and the simulated noise are used to train a point cloud noise simulation channel. The point cloud noise simulation channel is a machine learning model in the prior art, such as a neural network. The attenuated meteorological environment samples are used as the input of the point cloud noise simulation channel, and the simulated noise is used to supervise and adjust the output of the point cloud noise simulation channel so that the output result is the same as the simulated noise, thereby training the point cloud noise simulation channel to convergence.

[0045] During the predetermined time period, existing weather stations are used to collect meteorological information about the predetermined railway perimeter area to obtain real-time environmental information. This real-time environmental information includes weather conditions, such as rainy, snowy, rainfall, and snowfall. Noise simulation is performed on this real-time environmental information through the point cloud noise simulation channel. This involves inputting the real-time environmental information into the point cloud noise simulation channel and outputting a real-time simulated noise point cloud. The real-time simulated noise point cloud includes the point cloud deviation direction and distance. De-noising is performed on the real-time intrusion point cloud data using the real-time simulated noise point cloud. This involves performing inverse compensation processing on the real-time intrusion point cloud data based on the point cloud deviation direction and distance in the simulated noise point cloud. This involves correcting the point cloud data in the opposite direction of the deviation direction, thereby achieving de-noising of the point cloud data and reducing the interference of inclement weather on the accuracy of lidar detection.

[0046] The intrusion feature portrait database is called to perform intrusion feature matching on the multiple point cloud aggregation data, and position identification is performed in the first reference coordinate system to obtain a perimeter intrusion identification result.

[0047] The intrusion feature portrait database is called to perform intrusion feature matching on the multiple point cloud aggregation data. The intrusion feature portrait database is constructed based on data mining technology and contains three-dimensional models corresponding to different intrusion types. Three-dimensional point cloud modeling is performed based on the multiple point cloud aggregation data. Position identification is then performed in the first reference coordinate system to obtain a perimeter intrusion identification result. The three-dimensional model is matched against the intrusion feature portrait database. The intrusion type is obtained, and position identification is performed in the first reference coordinate system. Security risk identification is performed based on the intrusion location. The intrusion type and security risk identification results are used as the perimeter intrusion identification result. The specific process is detailed below.

[0048] In a preferred embodiment, it also includes:

[0049] Based on data mining technology, perimeter intrusion type samples and three-dimensional model samples are obtained to establish an intrusion feature portrait database; three-dimensional modeling is performed using the multiple point cloud aggregation data, three-dimensional model matching is performed in the intrusion feature portrait database based on the three-dimensional modeling results, and the intrusion type is generated based on the matching result; three-dimensional coordinate conversion is performed in the first reference coordinate system based on the multiple point cloud aggregation data to obtain first position coordinates; dynamic identification of intruders is performed based on the first position coordinates to obtain a dynamic intrusion trajectory; intersection areas of the intrusion dynamic trajectory and the train tracks in the predetermined railway perimeter area are extracted, safety risk identification is performed based on the intersection areas to obtain a safety risk coefficient; the perimeter intrusion identification result is generated based on the intrusion type and the safety risk coefficient.

[0050] Based on data mining technology, perimeter intrusion type samples and three-dimensional model samples are obtained to establish an intrusion feature portrait database. Simply put, the existing large amount of data is searched for records of railway perimeter intrusions that have occurred. From these records, intrusion type samples, such as pedestrians climbing over railings and animal intrusions, are extracted, along with corresponding intrusion images. Point cloud extraction is performed on the intrusion images, and three-dimensional model samples are constructed to form the intrusion feature portrait database. Three-dimensional modeling is further performed using the multiple point cloud aggregation data to obtain multiple three-dimensional modeling results. Based on the three-dimensional modeling results, three-dimensional model matching is performed in the intrusion feature portrait database. In other words, intrusion type samples corresponding to three-dimensional model samples with different coverage are matched and obtained as matching results, i.e., intrusion types. Simultaneously, three-dimensional coordinate conversion is performed in the first reference coordinate system based on the multiple point cloud aggregation data. In other words, the multiple point cloud aggregation data are displayed in the first reference coordinate system to obtain first position coordinates. The intruder is dynamically identified based on the first position coordinates to obtain the dynamic trajectory of the intruder. That is, the trajectory of the intruder may change dynamically, and the preset laser radar also performs detection continuously. Therefore, in a continuous time, multiple continuously changing point cloud aggregation data can be obtained, that is, the first position coordinates also change dynamically. The dynamically changing first position coordinates are connected, and the connection result is the dynamic trajectory of the intruder.

[0051] The intersection area of ​​the intrusion dynamic trajectory and the train track in the predetermined railway perimeter area is extracted. This is to determine whether there is an overlapping area between the intrusion dynamic trajectory and the train track, that is, the intersection area. The existence of the intersection area indicates that the intruder is on the track, and the safety risk after the train is running is high. The train may hit the intruder. Therefore, the safety risk is identified based on the intersection area to obtain a safety risk coefficient. That is, if there is an intersection area, the safety risk coefficient is 1. If there is no intersection area, the safety risk coefficient can be generated according to the distance between the intrusion dynamic trajectory and the train track. At this time, the safety risk coefficient is less than 1. The distance between the intrusion dynamic trajectory and the train track and the ratio of the boundary of the predetermined railway perimeter area to the train track are used as the safety risk coefficient. Finally, the perimeter intrusion identification result is generated based on the intrusion type and the safety risk coefficient. When the safety risk coefficient is 1, the staff can be reminded to quickly take measures against the intruder to prevent a safety accident.

[0052] In a preferred embodiment, it also includes:

[0053] A calibrated dynamic feature portrait library is established for the predetermined railway perimeter area during the predetermined time period; the dynamic feature portrait library includes a calibrated dynamic position and a calibrated three-dimensional model feature; based on the calibrated dynamic position and the calibrated three-dimensional model feature, intrusion identification correction is performed on the first position coordinates and the intrusion type, and the perimeter intrusion identification result is updated with the correction result.

[0054] A calibrated dynamic feature profile library is established for the predetermined railway perimeter area during the predetermined time period. The calibrated dynamic feature profile library includes personnel who normally work in the predetermined railway perimeter area, such as those performing railway maintenance and inspection. These personnel are not fixed. Therefore, when railway facilities require maintenance or personnel need to enter the predetermined railway perimeter area, the personnel's maintenance location and corresponding three-dimensional model features are extracted in real time to form the calibrated dynamic feature profile library. Based on the calibrated dynamic location and the calibrated three-dimensional model features, intrusion identification correction is performed on the first location coordinates and the intrusion type. This involves determining whether the three-dimensional model corresponding to the intrusion type is identical to the three-dimensional model features in the calibrated dynamic feature profile library, and comparing the first location coordinates with the calibrated dynamic location to determine whether they are identical. If both comparisons are identical, it indicates that the identified intruder is a staff member. In this case, the perimeter intrusion identification result needs to be corrected. Specifically, the corresponding intrusion type and security risk factor in the perimeter intrusion identification result are deleted, and the deleted result is used as the corrected result, which replaces the perimeter intrusion identification result. This achieves correction and update of the perimeter intrusion identification result, improving identification accuracy and reducing false alarm rate.

[0055] Based on the above analysis, it can be seen that the beneficial effects that can be achieved by one or more technical solutions provided in this application are as follows:

[0056] By reading the detection constraints of the preset laser radar, the regional spatial characteristics of the predetermined railway perimeter area are obtained, and the preset laser radar is planned to be deployed in combination with the detection constraints to obtain the radar deployment position. The preset laser radar is set in the predetermined railway perimeter area according to the radar deployment position, and the preset laser radar is used to perform perimeter background scanning of the predetermined railway perimeter area under the first preset conditions to obtain a background point cloud dataset, and a first reference coordinate system is constructed based on the background point cloud dataset. The preset laser radar is started to perform laser continuous scanning in a predetermined time period, and the echo signal is received and the lost signal compensation is performed to generate a perimeter detection point cloud dataset. The background of the perimeter detection point cloud dataset is removed according to the background point cloud dataset to obtain real-time intrusion point cloud data, and the real-time intrusion point cloud data is denoised by noise simulation. The denoised real-time intrusion point cloud data is clustered according to the point cloud density characteristics to obtain multiple point cloud aggregation data, and the intrusion feature portrait database is called to perform intrusion feature matching on the multiple point cloud aggregation data. Position recognition is performed in the first reference coordinate system to obtain perimeter intrusion recognition results. Therefore, by compensating for signal loss of the lidar's detection signal and denoising it in combination with the meteorological environment, perimeter intrusion identification is performed using the denoised point cloud data, thereby effectively improving the lidar's anti-interference performance in harsh environments, thereby improving the detection accuracy and perimeter intrusion identification accuracy.

[0057] Example 2

[0058] Based on the same inventive concept as the perimeter intrusion identification method based on laser radar detection in the aforementioned embodiment, Figure 2 As shown, the present application also provides a perimeter intrusion identification system based on laser radar detection, the system comprising:

[0059] A detection constraint reading module 11 is used to read the detection constraint of the preset laser radar, wherein the detection constraint includes the detection range, the horizontal field of view angle, and the vertical field of view angle;

[0060] a radar placement location planning module 12, the radar placement location planning module 12 being configured to obtain regional spatial characteristics of a predetermined railway perimeter area, and perform placement location planning for the preset laser radar in combination with the detection constraint conditions to obtain a radar placement location, and to place the preset laser radar in the predetermined railway perimeter area according to the radar placement location;

[0061] a perimeter background scanning module 13, configured to perform a perimeter background scan of the predetermined railway perimeter area using the preset laser radar under a first preset condition, obtain a background point cloud data set, and construct a first reference coordinate system based on the background point cloud data set;

[0062] A laser detection module 14 is configured to activate the preset laser radar at a predetermined time period to perform continuous laser scanning, receive echo signals, perform lost signal compensation, and generate a perimeter detection point cloud dataset;

[0063] A background removal module 15 is used to remove the background of the perimeter detection point cloud data set according to the background point cloud data set to obtain real-time intrusion point cloud data;

[0064] A clustering module 16 is configured to perform denoising on the real-time intrusion point cloud data through noise simulation, and cluster the denoised real-time intrusion point cloud data according to point cloud density features to obtain a plurality of point cloud cluster data;

[0065] The perimeter intrusion identification module 17 is used to call the intrusion feature portrait database to perform intrusion feature matching on the multiple point cloud aggregation data, and perform position identification in the first reference coordinate system to obtain a perimeter intrusion identification result.

[0066] Furthermore, the clustering module 16 further includes:

[0067] Collecting and obtaining attenuation meteorological environment samples of the predetermined railway perimeter area;

[0068] Performing point cloud noise simulation based on the attenuation meteorological environment samples to obtain simulated noise, wherein the attenuation meteorological environment samples and the simulated noise have a one-to-one correspondence;

[0069] Training a point cloud noise simulation channel using the attenuated meteorological environment sample and the simulated noise;

[0070] Collecting meteorological environment data for the predetermined railway perimeter area during the predetermined time period to obtain real-time environmental information;

[0071] Performing noise simulation on the real-time environmental information through the point cloud noise simulation channel to obtain a real-time simulated noise point cloud;

[0072] The real-time simulated noise point cloud is used to perform denoising processing on the real-time intrusion point cloud data.

[0073] Furthermore, the clustering module 16 further includes:

[0074] Performing a laser radar scanning test based on the attenuation meteorological environment sample to obtain real point cloud data and attenuation point cloud data;

[0075] Point cloud attenuation modeling is performed based on the real point cloud data and the attenuated point cloud data, and attenuated point cloud data is generated according to the modeling result.

[0076] Furthermore, the laser detection module 14 further includes:

[0077] When the preset laser radar is started to perform laser scanning during a predetermined time period, recording the laser scanning points;

[0078] Real-time environmental particle position recognition based on image sensors to obtain real-time environmental particle points;

[0079] Performing intersection recognition on the environmental particle points and the laser scanning points to obtain the intersection positions and the number of intersections;

[0080] If the number of intersection points exceeds a predetermined intersection point threshold, generating a lost point compensation instruction;

[0081] Signal loss compensation is performed on the echo signal at the intersection point according to the loss point compensation instruction.

[0082] Furthermore, the radar deployment location planning module 12 further includes:

[0083] Establishing a unit detection space according to the detection distance, the horizontal field of view angle, and the vertical field of view angle;

[0084] Performing spatial region coverage fitting on the regional spatial features using the unit detection space, and extracting the coverage quantity of the unit detection space using the coverage fitting result;

[0085] Identifying overlapping coverage areas on the coverage fitting results to obtain overlapping space dimensions;

[0086] The minimum overlap space size and the minimum coverage quantity are taken as the deployment targets, and the coverage fitting results are optimized to obtain the radar deployment position.

[0087] Furthermore, the perimeter intrusion identification module 17 further includes:

[0088] Based on data mining technology, perimeter intrusion type samples and three-dimensional model samples are obtained to establish an intrusion feature portrait database;

[0089] Performing three-dimensional modeling with the plurality of point cloud aggregation data, performing three-dimensional model matching in an intrusion feature portrait database according to the three-dimensional modeling results, and generating an intrusion type according to the matching results;

[0090] Performing a three-dimensional coordinate transformation in the first reference coordinate system based on the plurality of point cloud aggregation data to obtain a first position coordinate;

[0091] Performing dynamic identification of the intruder according to the first position coordinates to obtain a dynamic intrusion trajectory;

[0092] Extracting the intersection area between the intrusion dynamic trajectory and the train track in the predetermined railway perimeter area, identifying the security risk based on the intersection area, and obtaining a security risk coefficient;

[0093] The perimeter intrusion identification result is generated based on the intrusion type and the security risk coefficient.

[0094] Furthermore, the perimeter intrusion identification module 17 further includes:

[0095] Establishing a calibration dynamic feature portrait library for the predetermined railway perimeter area during the predetermined time period;

[0096] The dynamic feature portrait library includes calibrated dynamic positions and calibrated three-dimensional model features;

[0097] Intrusion identification correction is performed on the first position coordinates and the intrusion type based on the calibrated dynamic position and the calibrated three-dimensional model features, and the perimeter intrusion identification result is updated with the correction result.

[0098] The specific example of the perimeter intrusion identification method based on laser radar detection in the aforementioned embodiment 1 is also applicable to the perimeter intrusion identification system based on laser radar detection in this embodiment. Through the aforementioned detailed description of the perimeter intrusion identification method based on laser radar detection, those skilled in the art can clearly understand the perimeter intrusion identification system based on laser radar detection in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here.

[0099] It should be understood that various forms of processes shown above can be used to reorder, add or delete steps, as long as the expected results of the technical solutions disclosed in this application can be achieved, and this document does not limit them here.

[0100] Note that the above are only preferred embodiments of the present application and the technical principles employed. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, readjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments and may include many other equivalent embodiments without departing from the scope of the present application. The scope of the present application is determined by the scope of the appended claims.

Claims

1. A perimeter intrusion identification method based on laser radar detection, characterized in that: The method comprises: Reading the detection constraint conditions of the preset laser radar, wherein the detection constraint conditions include detection distance, horizontal field of view angle and vertical field of view angle; Acquire regional spatial characteristics of a predetermined railway perimeter area, and plan the layout position of the preset laser radar in combination with the detection constraint conditions to obtain a radar layout position, and set a preset laser radar in the predetermined railway perimeter area according to the radar layout position; Under a first preset condition, using the preset laser radar to perform a perimeter background scan of the predetermined railway perimeter area to obtain a background point cloud data set, and constructing a first reference coordinate system based on the background point cloud data set; Start the preset laser radar to perform laser continuous scanning in a predetermined time period, receive echo signals and perform lost signal compensation, and generate a perimeter detection point cloud data set; Performing background removal on the perimeter detection point cloud data set according to the background point cloud data set to obtain real-time intrusion point cloud data; De-noising the real-time intrusion point cloud data through noise simulation, clustering the de-noised real-time intrusion point cloud data according to point cloud density features, and obtaining a plurality of point cloud cluster data; Calling the intrusion feature portrait database to perform intrusion feature matching on the plurality of point cloud aggregation data, and performing position recognition in the first reference coordinate system to obtain a perimeter intrusion recognition result; The denoising process of the real-time intrusion point cloud data by noise simulation includes: Collecting and obtaining attenuation meteorological environment samples of the predetermined railway perimeter area; Performing point cloud noise simulation based on the attenuation meteorological environment samples to obtain simulated noise, wherein the attenuation meteorological environment samples and the simulated noise have a one-to-one correspondence; Training a point cloud noise simulation channel with the attenuated meteorological environment sample and the simulated noise; Collecting meteorological environment data for the predetermined railway perimeter area during the predetermined time period to obtain real-time environmental information; Performing noise simulation on the real-time environmental information through the point cloud noise simulation channel to obtain a real-time simulated noise point cloud; De-noising the real-time intrusion point cloud data using the real-time simulated noise point cloud; The step of performing point cloud noise simulation based on the attenuated meteorological environment sample to obtain simulated noise includes: Performing a laser radar scanning test based on the attenuation meteorological environment sample to obtain real point cloud data and attenuation point cloud data; Point cloud attenuation modeling is performed based on the real point cloud data and the attenuated point cloud data, and attenuated point cloud data is generated according to the modeling result.

2. The method according to claim 1, characterized in that The method of starting the preset laser radar to perform laser continuous scanning in a predetermined time period, receiving echo signals and performing signal compensation also includes: When the preset laser radar is started to perform laser scanning in a predetermined time period, the laser scanning points are recorded; Based on the image sensor, the real-time environmental particle position recognition is performed to obtain the real-time environmental particle points; Performing intersection recognition on the environmental particle points and the laser scanning points to obtain the intersection positions and the number of intersections; If the number of intersection points exceeds a predetermined intersection point threshold, generating a lost point compensation instruction; The signal loss compensation is performed on the echo signal at the intersection point according to the loss point compensation instruction.

3. The method according to claim 1, characterized in that The obtaining of the regional spatial characteristics of the predetermined railway perimeter area and planning the layout of the preset laser radar in combination with the detection constraint conditions includes: Establishing a unit detection space according to the detection distance, the horizontal field of view angle, and the vertical field of view angle; Performing spatial area coverage fitting on the regional spatial features with the unit detection space, and extracting the coverage quantity of the unit detection space with the coverage fitting result; Identify the overlapping coverage area of ​​the coverage fitting result to obtain the overlapping space size; The minimum overlap space size and the minimum coverage quantity are taken as the deployment goals, and the coverage fitting results are optimized to obtain the radar deployment position.

4. The method according to claim 1, characterized in that The calling of the intrusion feature portrait database to perform intrusion feature matching on the plurality of point cloud aggregation data and performing position recognition in the first reference coordinate system to obtain a perimeter intrusion recognition result includes: Based on data mining technology, perimeter intrusion type samples and three-dimensional model samples are obtained to establish an intrusion feature portrait database; Performing three-dimensional modeling with the plurality of point cloud aggregation data, performing three-dimensional model matching in an intrusion feature portrait database according to the three-dimensional modeling result, and generating an intrusion type according to the matching result; Performing three-dimensional coordinate transformation in the first reference coordinate system based on the plurality of point cloud aggregation data to obtain first position coordinates; Perform dynamic identification of the intruder according to the first position coordinates to obtain a dynamic intrusion trajectory; Extracting the intersection area between the intrusion dynamic trajectory and the train track in the predetermined railway perimeter area, identifying the safety risk according to the intersection area, and obtaining the safety risk coefficient; The perimeter intrusion identification result is generated based on the intrusion type and the security risk coefficient.

5. The method according to claim 4, characterized in that The method further comprises: Establishing a calibration dynamic feature image library for the predetermined railway perimeter area during the predetermined time period; The dynamic feature image library includes calibrated dynamic positions and calibrated three-dimensional model features; Based on the calibrated dynamic position and the calibrated three-dimensional model features, intrusion identification correction is performed on the first position coordinates and the intrusion type, and the perimeter intrusion identification result is updated with the correction result.

6. The perimeter intrusion identification system based on laser radar detection is characterized by: The system for performing the steps of the method according to any one of claims 1 to 5 comprises: A detection constraint reading module, wherein the detection constraint reading module is used to read the detection constraint of a preset laser radar, wherein the detection constraint includes a detection distance, a horizontal field of view angle, and a vertical field of view angle; A radar deployment location planning module, the radar deployment location planning module is used to obtain the regional spatial characteristics of the predetermined railway perimeter area, and plan the deployment location of the preset laser radar in combination with the detection constraint conditions to obtain the radar deployment location, and set the preset laser radar in the predetermined railway perimeter area according to the radar deployment location; A perimeter background scanning module, the perimeter background scanning module is used to perform a perimeter background scan of the predetermined railway perimeter area using the preset laser radar under a first preset condition, obtain a background point cloud data set, and construct a first reference coordinate system based on the background point cloud data set; A laser detection module, the laser detection module is used to start the preset laser radar to perform laser continuous scanning in a predetermined time period, receive echo signals and perform lost signal compensation, and generate a perimeter detection point cloud data set; A background removal module, the background removal module is used to remove the background of the perimeter detection point cloud data set according to the background point cloud data set to obtain real-time intrusion point cloud data; A clustering module, wherein the clustering module is used to perform denoising on the real-time intrusion point cloud data through noise simulation, and to perform clustering on the denoised real-time intrusion point cloud data according to point cloud density features to obtain a plurality of point cloud clustering data; The perimeter intrusion identification module is used to call the intrusion feature portrait database to perform intrusion feature matching on the multiple point cloud aggregation data, and perform position identification in the first reference coordinate system to obtain a perimeter intrusion identification result.

Images