Forward looking obstacle detection system

Abstract

The application discloses a front obstacle detection system, comprising a data acquisition device and a control processing device. The data acquisition device comprises an infrared light source module, an infrared image acquisition module, a visible light image acquisition module and a radar. The infrared image acquisition module is used for acquiring an infrared image in a detection window in a train advancing direction. The visible light image acquisition module is used for acquiring a visible light image in the detection window in the train advancing direction. The radar is used for measuring distance information between an obstacle in a preset braking distance range along the train advancing direction and the train, and the preset braking distance range is determined by a train running speed and a braking deceleration. The control processing device is used for receiving data information sent by the data acquisition device and outputting an obstacle detection result in the train advancing direction. The system effectively improves the recognition accuracy of the obstacle in the train advancing direction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of target detection, in particular to a front obstacle detection system. BACKGROUND

[0002] With the rapid increase of urban population, urban rail transit vehicles have the advantages of large passenger capacity, fast running speed, accurate arrival time, green energy saving, etc., which effectively improve the urban traffic congestion phenomenon and have become one of the main public transportation tools for urban travel.

[0003] It can be understood that the rail transit vehicle such as subway has a fast driving speed and a long braking distance, and in the process of driving, efficiently and accurately identifying the obstacles appearing in front of the driving direction can effectively reduce the probability of traffic accidents. In the related art, the visible light camera is used as the basis for obstacle identification on the driving path. However, the front obstacle detection system based on the visible light camera is blurred in the imaging in the harsh environment, such as dark environment, heavy fog, heavy rain, heavy snow and other bad weather, which cannot provide a strong basis for the driver to judge the obstacles in front of the driving section. SUMMARY

[0004] The present application provides a front obstacle detection system to improve the identification accuracy of the obstacles in the train advancing direction.

[0005] To solve the above technical problems, the embodiment of the present application provides the following technical solutions:

[0006] The embodiment of the present application provides a front obstacle detection system, which comprises a data acquisition device and a control processing device.

[0007] The data acquisition device comprises an infrared light source module, an infrared image acquisition module, a visible light image acquisition module and a radar.

[0008] The control processing device is used for receiving the data information sent by the data acquisition device and outputting the obstacle detection result in the train advancing direction.

[0009] The infrared image acquisition module is used for acquiring the infrared image in the detection window in the train advancing direction, the visible light image acquisition module is used for acquiring the visible light image in the detection window in the train advancing direction, and the radar is used for measuring the distance information between the obstacles and the train in the preset braking distance range along the train advancing direction with the train as the starting point. The preset braking distance range is determined by the train running speed and the braking deceleration.

[0010] Optionally, the control processing device is used for calling the target recognition program instruction stored in the memory to perform the following operations:

[0011] identify a rail area in the visible light image, and generate a clearance window of a front view rail area based on a train geometric size, taking the rail area as a horizontal reference surface and a normal direction of a rail curve as a height direction;

[0012] fuse the visible light image and the infrared image to generate a target detection image;

[0013] identify a target object in the clearance window of the target detection image to generate an initial target identification result;

[0014] remove rail area inherent equipment from the initial target identification result based on pre-stored rail area inherent equipment information to generate an obstacle identification result; the obstacle identification result includes a number of obstacles and obstacle information in the target detection image, and the obstacle information includes a length value, a height value, and a distance value from a train front end;

[0015] match corresponding obstacles in the obstacle identification result according to obstacle distance information sent by the radar to generate an obstacle detection result.

[0016] Optionally, the control processing device is configured to invoke a rail identification program instruction stored in a memory to perform the following operations:

[0017] extract an ROI area containing a rail from the visible light image, and divide the ROI area into a plurality of sub-blocks;

[0018] calculate a spatial distance value of each pixel in the ROI area from a center of a frame image;

[0019] According to Ω = ∑R*α d +G*β d +B*γ d calculate a local color system feature value Ω of each sub-block of the ROI area; R, G, and B are RGB color space values of each pixel in each sub-block, and α d , β d , and γ d are weight values controlled by the spatial distance value of the pixel itself;

[0020] determine a probability that each sub-block contains a rail based on the local color system feature value of each sub-block using a pre-trained shallow convolutional neural network;

[0021] determine a target sub-block containing a rail according to a pre-set region judgment threshold and the probability that each sub-block contains a rail;

[0022] reconstruct each target sub-block based on continuous correlation between video frame image sequences to generate a rail area.

[0023] Optionally, the control processing device is configured to invoke the pre-warning program stored in the memory to perform the following operations:

[0024] The distance threshold ranges of a plurality of different pre-warning levels are preset, each distance threshold range corresponds to a unique level of alarm information, the alarm information is presented in the form of sound on the vehicle, including different sound decibels and sound types, and the alarm information is transmitted to the ground control center;

[0025] The relationship between the current distance between the obstacle in the obstacle detection result and the train and each distance threshold range is compared in real time, and the corresponding alarm instruction is generated according to the alarm information.

[0026] Optionally, the control processing device is configured to invoke the pre-warning program stored in the memory to perform the following operations:

[0027] When it is detected that the obstacle distance information sent by the radar cannot successfully match the corresponding obstacle in the obstacle identification result, the relationship between the current distance between the obstacle in the obstacle distance information and the train and each intrusion distance threshold range is compared in real time, and an alarm instruction is generated according to the corresponding alarm information.

[0028] Optionally, it further comprises an alarm;

[0029] The alarm comprises a plurality of alarm units, each alarm unit is configured to alarm and prompt according to the corresponding level of alarm instruction sent by the control processing device.

[0030] Optionally, it further comprises a full-automatic forced braking device;

[0031] The full-automatic forced braking device is configured to be automatically triggered when the distance between the train and the obstacle is less than a preset distance, so that the train travels forward in a forced braking mode.

[0032] Optionally, the contour size of the obstacle is not less than 50cm*50cm.

[0033] Optionally, it further comprises a storage device for storing obstacle image data and event log records.

[0034] Optionally, it further comprises a display for displaying the detected obstacle and the distance between the obstacle and the train head to the user in real time.

[0035] The technical scheme provided by the present application has the advantages that the data collected by multiple sensors are used to jointly realize detection of an obstacle in front of a train. The active infrared imaging technology based on an infrared image collection module can clearly image under various light conditions, is not disturbed by dark light / weak light / no light / strong light, has all-weather working capability, is not affected by bad weather environments such as rain, snow, and fog, and guarantees imaging quality under various working conditions; the visible light image collection module can provide images with higher resolution and richer details, can support identification of markers such as platform names, kilometer markers, and signal lights, and can be used for positioning and auxiliary driving; the combination of machine vision and radar can improve the distance measurement accuracy between an obstacle in the train's advancing direction and the front end of the train, thereby effectively improving the identification accuracy of the obstacle in the train's advancing direction, reducing the probability of train traffic accidents, and guaranteeing safe and stable operation of the train.

[0036] It should be understood that the foregoing general description and the following detailed description are only examples and are not limiting on the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the related art, the following will briefly introduce the drawings needed to be used in the embodiment or related art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0038] Figure 1 A specific embodiment structure diagram of the front obstacle detection system provided by the embodiment of the present application;

[0039] Figure 2 A flowchart of the front obstacle detection provided by the embodiment of the present application;

[0040] Figure 3 A network model diagram of monocular ranging provided by the embodiment of the present application;

[0041] Figure 4 Another specific embodiment structure diagram of the front obstacle detection system provided by the embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the person skilled in the art better understand the present application scheme, the following will further describe the present application in combination with the drawings and specific embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in 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.

[0043] The terms "first", "second", "third", "fourth" and the like in the description and in the claims of the present application and in the above FIGS are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. The terms "comprises", "comprising", "includes", "including" and the like are to be construed open- ended, meaning that they include the listed steps or elements, but not excluding others. For example, a process, method, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those listed steps or elements, but can include additional steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

[0044] Referring first to Figure 1 , Figure 1 A schematic diagram of the front obstacle detection system provided by the embodiments of the present application in a specific implementation, the embodiments of the present application can include the following content:

[0045] The front obstacle detection system can include a data acquisition device 1 and a control processing device 2, and the data acquisition device 1 and the control processing device 2 can transmit and communicate data through a switch 0. The data acquisition device 1 can include an infrared light source module 11, an infrared image acquisition module 12, a visible light image acquisition module 13, and a radar 14. The data acquisition device 1 uses various sensors to collect front view data in the forward direction of the train, and the front view data includes image data and distance data, and sends the collected data to the control processing device 2. The infrared image acquisition module 12 and the visible light image acquisition module 13 can be installed inside the cab, above the ceiling, behind the transparent windshield, for example, to image the front; the control processing device 2 can be installed in the vehicle electrical cabinet, for example.

[0046] The infrared light source module 11 serves as a fill light source for the infrared image acquisition module 12. The number of light sources in the infrared light source module 11 can be one, or the same as the number of infrared cameras in the infrared image acquisition module 12, which is not limited in the present application. The light source can be a near-infrared laser light source, or other types of light sources, which do not affect the implementation of the present application. The near-infrared laser light source can be installed on the roof of the train, for example. The infrared light source module 11 can be composed of a near-infrared (NIR) laser light source and a laser driver, and the illumination distance can reach more than 250 meters. Increasing the power of the light source can further improve the illumination distance, thereby improving the image quality of the images acquired by the infrared image acquisition module 12.

[0047] The imaging unit of the data acquisition device 1 of the present application is integrated by an infrared image acquisition module 12 and a visible light image acquisition module 13. The infrared image acquisition module 12 is used to acquire infrared images of the detection window in the train advancing direction. The visible light image acquisition module 13 is used to acquire visible light images of the detection window in the train advancing direction. The infrared image acquisition module 12 may, for example, include one or more infrared cameras, each of which may, for example, use a CMOS imaging sensor, and is integrated with a high-speed camera and a processing board. A laser light source is used for light supplementing, clear imaging can be achieved under various harsh working conditions, and all-weather work is ensured. The visible light image acquisition module 13 may, for example, include one or more visible light cameras, which can image the visible light band, provide higher resolution images and more detailed information, and identify markers on the line, such as signal lights, platform names, and kilometer markers.

[0048] In the embodiment of the present application, the radar 13 is used to measure the distance information between the obstacles in the preset braking distance range along the train advancing direction from the train as the starting point and the train. In order to improve the radar ranging distance and ranging accuracy, the radar 13 can be a millimeter wave radar. The working frequency band of the millimeter wave radar is 30-300 GHz, and the wavelength range is usually 1-10 mm. The millimeter wave radar has the advantages of strong ability to penetrate rain, snow, fog, smoke, and dust, can work all-weather and all-day, has strong anti-interference ability, and has high data reliability. Through the millimeter wave radar, obstacle detection of up to 250 meters of straight-line distance can be achieved. The preset braking distance range is determined by the train running speed and the braking deceleration. For example, when the braking deceleration is 1 m / s 2 , the train running speed is 80 km / h, and the preset braking distance is 246.9 m; when the braking deceleration is 1 m / s 2 , the train running speed is 100 km / h, and the preset braking distance is 385.8 m; and when the braking deceleration is 1 m / s 2 , the train running speed is 60 km / h, and the preset braking distance is 138.9 m. In the present application, the control processing device 2 is used to receive the data information sent by the data acquisition device and output the obstacle detection result in the train advancing direction. The control processing device 2 is responsible for acquiring infrared camera images, visible light camera images, and radar data. Based on the acquired sensor data, obstacle detection and identification, obstacle positioning, and other functions can be completed. The control processing device 2 may, for example, use a 19-inch standard 3U chassis design, and use high integration and modularization for design. Any module can be individually plugged in. The data acquired by the infrared image acquisition module 12, the visible light image acquisition module 13, and the radar 14 can be transmitted to the control processing device 2 through wireless transmission such as wifi, or can be transmitted to the control processing device 2 through wired transmission such as communication cable and transmission bus. The present application does not make any limitation thereto.

[0049] It should be further explained that the control processing device 2 can pre-store any image processing algorithm that can realize target detection and tracking in the related art, identify the target in the visible light image or the infrared image by using the pre-embedded image processing algorithm, and then obtain the distance information of the target from the train head by using the radar ranging, so that the obstacle existing in the train advancing direction can be accurately detected. In addition, the infrared image and the visible light image can also be fused, and the fused image can be used for target detection and tracking. Any infrared image and visible light image fusion technology in the related art can be used, and the present application does not make any limitation on this.

[0050] In the technical scheme provided in the embodiment of the present application, the data collected by the multiple sensors are used to realize the detection of the obstacle in front of the train. The active infrared imaging technology based on the infrared image acquisition module can clearly image under various lighting conditions, is not disturbed by dark light / weak light / no light / strong light, has all-weather working capability, is not affected by bad weather environments such as rain, snow and fog, and guarantees the imaging quality under various working conditions. The visible light image acquisition module can provide images with higher resolution and richer details, can support the recognition of platform names, kilometer markers, signal lights and other markers, and can be used for positioning and auxiliary driving. The combination of machine vision and radar can improve the distance measurement accuracy between the obstacle in the train advancing direction and the front end of the train, thereby effectively improving the recognition accuracy of the obstacle in the train advancing direction, reducing the probability of train traffic accidents, and guaranteeing the safe and stable operation of the train.

[0051] In the above embodiment, the control processing device 2 does not make any limitation on how to perform target detection and target tracking. In combination with the flowchart shown in FIG. 2, a target tracking detection method given in the present embodiment can include the following steps: Figure 2

[0052] The control processing device 2 can call the target recognition program instructions stored in the memory to perform the following operations:

[0053] The rail area is recognized in the visible light image, and based on the train geometric size, the rail area is taken as a horizontal reference surface, and the normal direction of the rail curve is taken as a height direction to generate a limit window of the front view rail area.

[0054] The visible light image and the infrared image are fused to generate a target detection image.

[0055] The target in the limit window of the target detection image is recognized to generate an initial target recognition result.

[0056] ​The track-specific equipment is removed from the initial target identification result based on pre-stored track-specific equipment information to generate an obstacle identification result. The obstacle identification result includes the number of obstacles contained in the target detection image and obstacle information, and the obstacle information includes a length value, a height value, and a distance value from the train head.

[0057] The corresponding obstacle in the obstacle identification result is matched according to the obstacle distance information sent by the radar to generate an obstacle detection result.

[0058] Optionally, the image captured by the camera of the visible light image acquisition module 1 can be pre-processed, such as denoising, normalization, etc., and then the rail contour line is extracted from the visible light image using the features of the rail. The detection of the rail utilizes the following features: the rail, whether straight or curved, is a parallel line in three-dimensional space; the rail plane is basically in the middle position in the image / video captured by the camera; the rail has a certain color differentiation degree (rgb or hsv color space) from the surrounding background; using the above geometric and color features of the rail, the detection of the rail can generally be realized by a straight line detection algorithm and a geometric position condition filtering.

[0059] The field of view in front of the train captured by the camera is a perspective view. According to the camera calibration program, the two-dimensional mapping straight line of the same physical three-dimensional height at different distances from the camera in the image / video captured by the camera can be determined. As long as the camera remains stable, or the camera has slight shaking but can be adjusted by the corresponding algorithm, this mapping relationship will not change with the running of the train. Therefore, on the basis of camera calibration, combined with the reference on the line, the width and height of the clearance detection frame can be determined.

[0060] According to the relevant industry standards, the track gauge of the rail is 1435mm. According to the basic principle of imaging, a close-range object occupies more pixels, and a long-distance object corresponds to fewer pixels. However, since the spacing of the rail remains unchanged, it can be used as a stable reference. After the track is identified, the rail can be used as a reference to extend to both sides at a fixed ratio, thereby generating a clearance detection frame.

[0061] It can be understood that the obstacles can include stationary obstacles and moving obstacles. For stationary obstacles, the track in the image is identified, and then the clearance is constructed with the track as a reference to form a clearance window, and the obstacles in the clearance window are detected. For moving obstacles, not only the obstacles between the rails can affect the train operation, but also the moving objects that have the tendency to invade the rails can affect the train operation. The system can detect the moving obstacles, and the radar ranging information can be used to judge the invading moving objects.

[0062] Generally, there are pre-installed devices in the track area, i.e., track area inherent devices, in order to prevent the existing devices from being mistakenly reported as obstacles, the system can pre-establish a white list. At the beginning of the system operation, the images of the installed devices in the line need to be collected, and the features are extracted as the matching basis. During the running of the train, the area in the clearance window can be detected, and when an object is detected, the system will retrieve the white list for feature matching. If the features match the objects in the white list, it is judged as an installed device; if the features do not match the white list, it is judged as an obstacle.

[0063] The relative position of the moving object in the image will change, and by tracking the relative position change of the feature points in the series of images, the moving obstacle detection can be realized. The moving obstacle includes the moving obstacle within the camera field of view, and also includes the moving obstacle outside the camera field of view. Here, the obstacle refers to the camera collectable area, such as the track running area. The moving obstacle in the camera field of view can be collected by the camera to form a visible light image or an infrared image, and then the distance of the object can be determined in real time through the calibration and solving algorithm of the monocular camera system. After the camera is calibrated, the three-dimensional coordinates of the physical space can be mapped to the two-dimensional space of the camera. After detecting the obstacle, the two-dimensional coordinates of the obstacle in the image can be locked, and then the two-dimensional coordinates on the image can be mapped to the real physical space through the mapping relationship, so that the distance information of the obstacle can be obtained. For the moving object that has the tendency to invade the track, i.e., the object that cannot be collected by the camera, the radar ranging information can be used to detect the distance information between the invading moving object and the train, so as to accurately determine whether there is an external object entering the track or having the tendency to enter the track.

[0064] Optionally, in the process of calling the target recognition program instruction to recognize and track the target, the control processing device 2 can perform real-time calculation of the contour information and the horizontal distance information of the object within the preset braking distance range, for example, within 250m, for example, the calculation time is not more than 40ms. The contour information refers to the horizontal length and the vertical height of the target object, which are two important factors affecting the safety of the subway train running. The contour size of the obstacle that can be recognized by the present application is not less than 50cm*50cm. Because the area of the far-end object passing through the optical sensing imaging area has only a few pixels, it is more similar to a light spot, so the approximate length and height of the far-end object can be estimated by designing a circular / elliptical approximate area matching, and then the length in the pixel coordinates is converted into the physical unit in the actual space through the calibration coefficient of the visible light camera:

[0065] (x, y) = Λ(ρ, v)

[0066] Wherein, x, y are the physical length and height of the target object, Lambda is the conversion relationship between the camera space and the actual space, rho, v are the approximate length and height of the pixels obtained by matching, and Lambda can be determined through the calibration process of the camera.

[0067] The horizontal distance of the target object is the distance between the remote object and the front end of the train, which is one of the important information for the driver. The horizontal distance information can be calculated and verified by the horizontal distance measurement of the monocular camera and the gating function of the infrared camera. The function of monocular distance measurement can be realized by the multi-layer connection neural network as shown in the formula (1), without prior information of the scene or explicit camera parameter information. Figure 3 The horizontal distance information between the matching region and the camera can be well predicted by inputting the coordinate information of multiple pixel points in the approximate matching imaging region and training a certain number of samples.

[0068] The above implementation is not limited to track identification, and the embodiment of the application further provides a track identification method, which can include the following contents:

[0069] The control processing device 2 is used to call the rail identification program instructions stored in the memory to perform the following operations:

[0070] The ROI region containing the rail is extracted from the visible light image, and the ROI region is divided into multiple subblocks.

[0071] The spatial distance value of each pixel in the ROI region from the center of the frame image is calculated; for example, the spatial distance value of each pixel from the center of the frame image can be calculated by using the Euclidean space distance calculation relationship, of course, other calculation methods can also be used, which does not affect the implementation of the application.

[0072] According to Omega = Sum R*alpha d + G*beta d + B*gamma d The local color system feature value Omega of each subblock of the ROI region is calculated; R, G, B are the RGB color space values of each pixel in each subblock, alpha d , beta d , gamma d are weight values controlled by the spatial distance value of the pixel. The relationship between the weight value and the spatial distance value of the pixel point can be calculated by using the pre-trained shallow full connection neural network, and the color system feature of each block is obtained by summing the features of the pixels.

[0073] Based on the local color system feature value of each sub-block, the probability of each sub-block containing the rail is determined by using a pre-trained shallow convolutional neural network. The shallow convolutional neural network can be trained by using sample images, in which the local color system feature value of each sub-block is known, and the probability of containing the rail is known. By inputting the local color system feature value of each sub-block of the present application into the shallow convolutional neural network, the probability of the sub-block containing the rail can be obtained.

[0074] The target sub-block containing the rail is determined according to the pre-set region judgment threshold and the probability of each sub-block containing the rail. For example, the judgment threshold and the prior Gamma probability of the far, middle and near three regions of the image frame can be pre-set, and the posterior distribution of the threshold can be determined by learning from samples. Finally, the calculated probability is distinguished by the three region judgment thresholds, and the region where the rail is located is finally determined.

[0075] Each target sub-block is reconstructed based on the continuity correlation between the sequence of video frame images to generate the rail region. It can be understood that the sequence of video frame images has a certain continuity correlation, and the characteristics are that the rail recognition region Θ n of the previous frame is limited to the rail existing region Θ n+1 of the next frame, i.e. Θ n+1 = Θ n + Gaussian(μ, σ), that is, the rail recognition region of the previous frame and the next frame has a random Gaussian migration; μ and σ are the mean and variance of the Gaussian determined by sample learning, and the subsequent several image frames are determined by the Gaussian time sequence determined by the continuity correlation.

[0076] As another optional embodiment, the present application can also perform hierarchical early warning based on the different distances between the obstacles and the train, which can include the following contents:

[0077] The control processing device 2 is used to call the pre-warning program instructions stored in the memory to perform the following operations:

[0078] A plurality of distance threshold ranges of different pre-warning levels are pre-set, each distance threshold range corresponds to a unique level of alarm information, the alarm information is presented in the form of sound on the vehicle, including different sound decibels and sound types, and the alarm information is transmitted to the ground control center.

[0079] The current distance between the obstacles in the obstacle detection result and the train is compared with the relationship between each distance threshold range in real time, and an alarm instruction is generated according to the corresponding alarm information.

[0080] In order to further improve the safety of the train, the application can also be pre-set two brake deceleration, one is the conventional brake deceleration, one is the emergency brake deceleration, the emergency brake deceleration is greater than the conventional brake deceleration, and the corresponding emergency brake distance is less than the conventional automatic distance. For example, when the conventional brake deceleration is 1m / s 2 , the train speed is 80km / h, and the conventional brake distance is 246.9m; when the emergency brake deceleration is 1.2m / s 2 , the train speed is 80km / h, and the emergency brake distance is 205.8m. Further, the system can display the safety distance in the image displayed to the driver or other staff according to the brake distance of the train, and give a hierarchical warning according to the distance between the obstacle and the train, and trigger the brake when necessary, to fully ensure the safety of train operation. In addition, in another embodiment, the control processing device 2 is also used to call the intrusion warning program instructions stored in the memory to perform the following operations:

[0081] When it is detected that the obstacle distance information sent by the radar cannot successfully match the corresponding obstacle in the obstacle identification result, the relationship between the current distance between the obstacle and the train in the obstacle distance information and each intrusion distance threshold range is compared in real time, and an alarm instruction is generated according to the corresponding alarm information. High-resolution visible light cameras can also be used to continuously monitor abnormal situations on short-distance lines, such as small foreign objects, trackside equipment status, platform area platform door abnormalities / advertising board abnormalities / gum strip shedding, etc.; when an anomaly is found, a log record is generated, and the anomaly point location is located, and the anomaly information is imported into the operation and maintenance department for line management and maintenance.

[0082] Correspondingly, please refer to Figure 4 , the forward obstacle detection system can also include an alarm 3. The alarm 3 can include a plurality of alarm units, each alarm unit being used to alarm and prompt according to the corresponding level of alarm instruction received from the control processing device. Among them, the alarm 3 can be installed on the left or right side of the driver's room main console.

[0083] As another optional embodiment, in order to further improve the safety of train operation and reduce the probability of accidents, the forward obstacle detection system can further include a full-automatic forced braking device 4. The full-automatic forced braking device 4 is used to be automatically triggered when the distance between the train and the obstacle is less than the preset safety distance, so that the train travels forward in the forced braking mode, for example, the train is forced to brake when the distance is less than 50m. The preset safety distance can be determined based on train operation parameters such as running speed and brake deceleration, and the minimum value of the safety distance is to make the train stop sliding when the safety distance is walked. The highest speed of the train is taken as the standard when determining the safety distance.

[0084] As another optional implementation, the front obstacle detection system can further comprise a storage device 5, which can be a solid state disk for example. The storage device 5 can be used to store obstacle image data and event log records. That is, when the obstacle detection result output by the control processing device 2 is that there is an obstacle, the collected image data or video data is automatically stored in the storage device 5. The event log records are used to record log information of pre-set events, such as parking events, fault events, etc., which can provide accurate basis for train event positioning according to trackside markers, such as 100-meter markers, station distance prompts, and station name markers recorded by numbers, Chinese characters, or characters, etc. The train event positioning function can be realized by recognizing information in the markers using any character recognition technology.

[0085] In addition, the front obstacle detection system can further comprise a display 6 used to display the detected obstacles and the distance between the obstacles and the train head to the user in real time. The information can be displayed to the driver in a visual form through the display, which provides rich information for the safe driving of the driver and ensures the safe driving of the driver.

[0086] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0087] The skilled person can further realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general terms in the above description. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0088] The above describes in detail a front obstacle detection system provided by the present application. The principles and implementation manners of the present application are described by using specific examples in the present application. The above description of the examples is only used to help understand the method of the present application and its core idea. It should be pointed out that, for the person skilled in the art, without departing from the principles of the present application, the present application can be improved and modified in several ways, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A forward-looking obstacle detection system, characterized in that: Including data acquisition device and control processing device; The data acquisition device includes an infrared light source module, an infrared image acquisition module, a visible light image acquisition module and a radar; The control processing device is configured to receive data information sent by the data acquisition device, identify the rail area in the visible light image, and generate a bounding window of the forward-looking track area based on the geometric dimensions of the train, with the rail area as a horizontal reference plane, and the normal direction along the rail curve as a height direction; perform dual-band information fusion on the visible light image and the infrared image to generate a target detection image; identify targets within the bounding window of the target detection image with a contour size of not less than 50 cm*50 cm to generate an initial target recognition result; and remove inherent equipment in the track area from the initial target recognition result based on pre-stored inherent equipment information in the track area to generate an obstacle recognition result; The obstacle recognition result includes the number of obstacles contained in the target detection image and obstacle information, wherein the obstacle information includes the length, height, and distance from the front end of the train. The obstacle recognition result is matched with the corresponding obstacle based on the obstacle distance information transmitted by the radar, an obstacle detection result is generated, and the obstacle detection result in the direction of the train's advance is output. The outline dimensions are the horizontal length and vertical height of the target object; the outline information of the remote object is determined by the relationship (x, y) = Λ(ρ, v), where x, y are the horizontal length and vertical height of the target object, Λ is the conversion relationship between camera space and real space, which is determined by the camera calibration process, and ρ, v are the pixels of the approximate length and height obtained by matching; The infrared image acquisition module is used to acquire infrared images within a detection window in the direction of the train's advance, and the visible light image acquisition module is used to acquire visible light images within a detection window in the direction of the train's advance. The radar is used to measure distance information between the train and obstacles within a preset braking distance range along the train's advance direction, with the train as the starting point. The preset braking distance range is determined by the train's speed and braking deceleration. The control processing device is used to call the target recognition program instructions stored in the memory to perform the following operations: identifying the rail area in the visible light image; extracting the ROI area containing the rail from the visible light image, and dividing the ROI area into multiple sub-blocks; calculating the spatial distance value of each pixel in the ROI area from the center of the frame image; Calculate the local color feature value of each sub-block in the ROI area ; R, G, B are the RGB color space values ​​of each pixel in each sub-block. is a weight value controlled by the spatial distance value of the pixel itself; based on the local color feature value of each sub-block, a pre-trained shallow convolutional neural network is used to determine the probability that each sub-block contains a rail; the judgment thresholds and prior Gamma probabilities of the far, middle and near areas of the image frame are pre-set, and the posterior distribution of the thresholds is determined through sample learning. The probability of each sub-block containing a rail is distinguished according to the judgment thresholds corresponding to the far, middle and near areas, and the target sub-block containing a rail is determined; the rail identification area of ​​the previous frame The area with the railroad in the next frame Satisfy between: , and This is determined by sample learning. The mean and variance of the railroad identification area based on the previous and next frames are random. Migrate,reconstruct each target sub-block and generate the rail area.

2. The forward-looking obstacle detection system according to claim 1, characterized in that: The control processing device is used to call the early warning program instructions stored in the memory to perform the following operations: Multiple distance threshold ranges with different warning levels are pre-set, and each distance threshold range corresponds to a unique level of alarm information. The alarm information is displayed in the form of sound on the vehicle, including different sound decibels and sound types, and the alarm information is transmitted to the ground control center at the same time; The relationship between the current distance between the obstacle and the train in the obstacle detection result and each distance threshold range is compared in real time, and a corresponding alarm instruction is generated according to the alarm information.

3. The forward-looking obstacle detection system according to claim 1, characterized in that: The control processing device is used to call the intrusion warning program instructions stored in the memory to perform the following operations: When it is detected that the obstacle distance information sent by the radar cannot successfully match the corresponding obstacle in the obstacle identification result, the relationship between the current distance between the obstacle and the train in the obstacle distance information and each intrusion distance threshold range is compared in real time, and an alarm instruction is generated according to the corresponding alarm information.

4. The forward obstacle detection system according to claim 2 or 3, characterized in that: Also includes alarm; The alarm device includes a plurality of alarm units, each of which is used to issue an alarm prompt according to an alarm instruction of a corresponding level sent by the control processing device.

5. The forward obstacle detection system according to any one of claims 1 to 3, characterized in that: Also included is a fully automatic mandatory brake; The fully automatic forced braking device is used to be automatically triggered when the distance between the train and the obstacle is less than a preset distance, so that the train moves forward in a forced braking mode.

6. The forward-looking obstacle detection system according to claim 1, characterized in that: Also included is a storage device for storing obstacle image data and event log records.

7. The forward obstacle detection system according to claim 6, characterized in that: The system also includes a display for displaying detected obstacles and the distance between the obstacles and the train head to the user in real time.

Images