Blind area pre-collision target detection method, vehicle-mounted millimeter wave radar, product and vehicle

Abstract

The invention provides a blind area pre-collision target detection method, a vehicle-mounted millimeter wave radar, a product and a vehicle, and the method comprises the steps: receiving a radar echo signal, and processing the radar echo signal to obtain point cloud data; performing target clustering on the point cloud data, and determining a target and information of the target; determining a dynamic target shielded by a static target as a potential early warning target according to the information of the target in the detection area range; and determining a blind area pre-collision target from the potential early warning targets according to the multipath reflection point cloud of the potential early warning targets and the motion direction of the potential early warning targets. According to the method, the pre-collision target in the blind area can be accurately detected by utilizing the multipath effect of the radar under the condition of not additionally increasing radar hardware setting, so that the occurrence of a ghost probe accident can be reduced, and the auxiliary driving performance is improved.

Description

Technical Field

[0001] The present application relates to the field of radar technology, and in particular to a blind spot pre-collision target detection method, a vehicle-mounted millimeter-wave radar, a computer program product and a vehicle. Background Art

[0002] Millimeter-wave radar detects and tracks targets by emitting and receiving electromagnetic waves, and can simultaneously measure the distance, speed, and angle of the target. Compared with cameras and lidars, millimeter-wave radars have good environmental and weather adaptability, and can adapt to application scenarios such as strong light, night, heavy rain, fog, and dust. They have been widely used in vehicle-mounted ADAS (Advanced Driving Assistance System) assisted driving systems and vehicle-mounted AD (Autonomous Driving) automatic driving systems. Through the mutual integration of cameras, lidars, and millimeter-wave radars, the redundancy of intelligent decision-making can be improved, and the high reliability and safety of ADAS assisted driving or AD automatic driving can be achieved.

[0003] In vehicle-mounted application scenarios, with the increase in the number of cars and traffic participants, "ghost-peek" traffic accidents often occur on the road. "Ghost-peek" usually occurs when there is a vehicle or obstacle blocking the line of sight in front, forming a blind spot. If a non-motor vehicle or pedestrian suddenly jumps out from the roadside and the driver fails to avoid it in time, it is very easy to cause property damage and casualties. Therefore, the vehicle-mounted assisted driving system can effectively improve vehicle safety by detecting and warning "ghost-peek" targets around the vehicle. Summary of the invention

[0004] In order to solve the existing technical problems, the present application provides a blind spot pre-collision target detection method, a vehicle-mounted millimeter-wave radar and a computer program product that can detect "ghost-peek" targets around a vehicle to improve vehicle safety.

[0005] In a first aspect, a blind spot pre-collision target detection is provided, comprising:

[0006] Receiving radar echo signals, and processing the radar echo signals to obtain point cloud data;

[0007] Performing target clustering on the point cloud data to determine targets and information about the targets;

[0008] According to the information of the target within the detection area, the dynamic target blocked by the static target is determined as a potential warning target;

[0009] According to the multipath reflection point cloud of the potential warning target and the moving direction of the potential warning target, a blind area pre-collision target is determined from the potential warning targets.

[0010] In a second aspect, a vehicle-mounted millimeter-wave radar is provided, comprising a millimeter-wave radar processor and a memory, wherein the memory stores a computer program executable by the millimeter-wave radar processor;

[0011] When the computer program is executed by the millimeter wave radar processor, the steps of the blind spot pre-collision target detection method described in any embodiment of the present application are implemented.

[0012] In a third aspect, a vehicle is provided, including an on-board central control system and an on-board millimeter-wave radar interactively connected to the on-board central control system;

[0013] The vehicle-mounted millimeter-wave radar is used to perform the steps of the blind spot pre-collision target detection method described in any embodiment of the present application;

[0014] The vehicle-mounted central control system is used to output corresponding auxiliary driving information according to the detection result of the millimeter-wave radar on the blind spot pre-collision target.

[0015] In a fourth aspect, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps of blind spot pre-collision target detection described in any embodiment of the present application.

[0016] The above-mentioned blind spot pre-collision target detection method obtains point cloud data based on radar echo signal processing, clusters targets based on point cloud data, obtains target and target information, finds dynamic targets blocked by static targets based on the information of targets within the detection area, preliminarily determines potential warning targets, and further eliminates false targets based on multipath reflection point clouds and movement directions of potential warning targets, and finds real blind spot pre-collision targets detected through multipath reflections and whose movement directions may collide with the direction of travel of the vehicle. This method utilizes the multipath effect of radar and can accurately detect blind spot pre-collision targets without the need for additional radar hardware settings, which can reduce the occurrence of "ghost probe" accidents and improve assisted driving performance.

[0017] The vehicle-mounted millimeter-wave radar, computer program product, and vehicle provided in the above embodiments belong to the same concept as the corresponding blind spot pre-collision target detection method embodiments, and thus have the same technical effects as the corresponding blind spot pre-collision target detection method embodiments, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 1 is a diagram of optional application scenarios of a blind spot pre-collision target detection method in one embodiment.

[0019] Figure 2 Schematic diagram of the working principle of a vehicle-mounted millimeter-wave radar in one embodiment.

[0020] Figure 3 Flow chart of a blind spot pre-collision target detection method in one embodiment.

[0021] Figure 4 Schematic diagram of a millimeter wave radar multipath reflection scenario in one embodiment.

[0022] Figure 5 A schematic diagram of a scene of a detection area range in an embodiment.

[0023] Figure 6 Flow chart of a blind spot pre-collision target detection method in one embodiment.

[0024] Figure 7 A flowchart of the operation of a vehicle-mounted millimeter-wave radar in one embodiment.

[0025] Figure 8 Schematic diagram of the structure of a vehicle-mounted millimeter-wave radar in one embodiment.

[0026] Fig. 9 FIG. 1 is a schematic diagram of the structure of a vehicle in one embodiment. DETAILED DESCRIPTION

[0027] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0028] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0029] In the following description, the expression "some embodiments" is involved, which describes a subset of all possible embodiments. It should be noted that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

[0030] In the following description, the terms "first, second, third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first, second, third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0031] At present, there are no similar products at home and abroad to monitor and prevent "ghosting" traffic accidents. Existing vehicle-mounted assisted driving systems can usually only detect the movement status of vehicles in sight, and cannot detect obscured targets or vehicles. The "ghosting" target and the driver are both in each other's visual blind spot, appearing without warning, at close range, and in a short time, leaving neither party with time and space to react or evade, and ultimately causing a traffic accident.

[0032] Therefore, it is of great significance to improve vehicle safety by detecting and warning "ghostly" targets around the vehicle through the on-board assisted driving system.

[0033] Millimeter-wave radar signals have a multipath effect. The signals can detect obscured objects through reflection, diffuse reflection, diffraction, and other methods. The inventors of this application have proposed that this characteristic of millimeter-wave radar signals can be used to detect and warn blind spot pre-collision targets in vehicle-mounted driving scenarios. Millimeter-wave radar has irreplaceable advantages in addressing these problems that are difficult for other sensors to handle and are prone to accidents.

[0034] Specifically, the present application provides a blind spot pre-collision target detection method. The blind spot pre-collision target is a target in the blind spot that is about to collide with the vehicle, and is an important hidden danger to the safe driving of the vehicle. A typical blind spot pre-collision target is the well-known "ghost head" target, which usually refers to a target or vehicle that is in the blind spot due to being obscured. The target or vehicle is in motion, and its direction of movement is close to the vehicle. From the driver's perspective, it is a target that suddenly breaks in. If the brakes are not applied in time, it is likely to cause a traffic accident. The blind spot pre-collision target detection of the present application can utilize the unique characteristic of the multipath effect of the millimeter-wave radar signal without the need to add millimeter-wave radar hardware equipment to realize the detection of the position and speed of the "ghost head" target that is obscured in the driver's blind spot, and issue an early warning to reduce the occurrence of collision accidents.

[0035] See also Figure 1 , which is an optional application scenario diagram of the blind spot pre-collision target detection method provided by an embodiment of the present application applied to a vehicle. The vehicle-mounted millimeter wave radar 200 is usually installed on both sides of the front of the vehicle 100. Common installation locations are not limited to Figure 1 The forward position shown also includes the front corner, rear corner and rear corner positions, which are used to detect and perceive the surrounding environment targets (target 1, target 2, target 3).

[0036] See also Figure 2 , is a schematic diagram of the structure of the vehicle-mounted millimeter-wave radar. The vehicle-mounted millimeter-wave radar mainly consists of four functional parts:

[0037] 1) Antenna array, which consists of a transmitting array and a receiving array. Vehicle-mounted millimeter-wave radar generally adopts the MIMO (Multiple-Input Multiple-Output) working mode, and realizes the antenna array required for direction finding through the MIMO working mode between the unit antennas of the transmitting array.

[0038] 2) MRTR (Millimeter-wave Radar Transceiver, millimeter-wave radar transceiver module or chip), millimeter-wave radar transceiver functional module, mainly composed of transmitting components, receiving components and linear frequency modulation local oscillator source components. MRTR generates and transmits the working waveform signal of the millimeter-wave radar, receives and discretely samples the reflected waveform signal of the target, and sends the sampled reflected waveform signal to the MRP (Millimeter-wave Radar Processor) millimeter-wave radar processor.

[0039] 3) MRP (Millimeter-wave Radar Processor), millimeter-wave radar processor, receives the reflected waveform signal from MRTR and performs digital processing to obtain information such as the distance, speed and angle of environmental targets, clusters, identifies and tracks targets, and outputs point cloud-level detection results or target-level detection results. From the main functional division, MRP is mainly composed of distance measurement module, speed measurement module, angle measurement module and target clustering identification and tracking module. The above modules can be implemented through embedded software, FPGA (Field Programmable Gate Array) or AISC (Application Specific Integrated Circuit) application-specific integrated circuits.

[0040] 4) Readable and writable storage, used to store the working data of the vehicle-mounted millimeter-wave radar, including working parameter data and process data stored in advance, such as, in the embodiment of the present application, in the process of implementing the blind spot pre-collision target detection method, identifying target parameters such as the blind spot pre-collision target and parameters of the motion state, as well as threshold parameters for distinguishing the blind spot pre-collision target.

[0041] The blind spot pre-collision target detection method provided in the embodiment of the present application can be mainly completed in the target clustering identification and tracking module. The implementation form, composition mode or module division mode of MRP does not affect the use of the blind spot pre-collision target detection method provided in the embodiment of the present application.

[0042] 5) A read-write memory used to store the working data of the millimeter-wave radar, including working parameter data and process data stored in advance.

[0043] It should be noted that the configuration form or composition method of the functional parts of the above-mentioned vehicle-mounted millimeter-wave radar does not affect the use of the blind spot pre-collision target detection method provided in the embodiment of the present application.

[0044] See also Figure 3 , a blind spot pre-collision target detection method provided in an embodiment of the present application comprises the following steps:

[0045] S302, receiving radar echo signals, and processing the radar echo signals to obtain point cloud data.

[0046] Radar echo signal refers to the signal received by the radar's receiving end after the electromagnetic wave emitted by the radar is reflected or scattered back after encountering the target. The radar echo signal usually contains radial distance, azimuth, pitch, radial velocity and other characteristic information about the target. By analyzing and processing these signals, the radar can obtain the spatial point cloud data at the current moment, and based on the point cloud data, it can realize target detection, tracking and identification functions. Point cloud data refers to the data set of spatial points obtained by radar scanning. Each point in the radar point cloud contains the polar coordinate information, velocity information and other attribute information of a certain position in the space, such as Rcs. By analyzing the spatial correlation between point clouds, more features and attributes about the target can be obtained.

[0047] S304, performing target clustering on the point cloud data to determine the target and target information.

[0048] Target clustering refers to the process of grouping a large amount of point data generated by radar during target detection. These data points may represent multiple reflection points of the same target. Through clustering algorithms, these point data can be grouped into different groups to represent an independent target. Target clustering can also further include grouping multiple target points detected by radar according to their characteristics and positions for better target identification and tracking.

[0049] The target information may include the target's distance, angle, speed, target state, etc. The target state is used to indicate whether the target is a static target or a dynamic target.

[0050] S306, based on the information of the targets within the detection area, the dynamic targets blocked by the static targets are determined as potential warning targets.

[0051] The radar can be installed in the front, front corner, rear and rear corner of the vehicle to detect in the corresponding direction, and each radar has a corresponding radar detection range. The detection area range in this embodiment is a part of the radar detection range, which can be used to detect blind spot pre-collision targets.

[0052] In one embodiment, the detection area range may be a preset empirical value, which is a preset area with a fixed range size.

[0053] In one embodiment, the detection area range can be flexibly set in consideration of different driving environments to meet the detection requirements of different driving environments, such as being set to be related to the direction of travel of the vehicle and the speed of the vehicle. In one embodiment, when the direction of travel of the vehicle is forward, the detection area range is in front of the vehicle. When the direction of travel of the vehicle is backward, the detection area range is behind the vehicle. In one embodiment, when the vehicle is traveling too fast, the detection range can be set larger so that the blind spot pre-collision target can be found in advance within a larger detection range, so that the vehicle traveling at a faster speed has sufficient time to brake and other measures to reduce the occurrence of collision accidents. When the vehicle is traveling at a lower speed, the detection area range can be set smaller.

[0054] For each target within the detection area, further analysis is performed to identify dynamic targets blocked by static targets to obtain potential warning targets. In one embodiment, dynamic targets blocked by static targets can be determined based on the radial distance and azimuth between the static targets and the dynamic targets.

[0055] Step 308 , determining a blind area pre-collision target from the potential warning targets according to the multipath reflection point cloud of the potential warning targets and the movement direction of the potential warning targets.

[0056] Among them, the multipath reflection point cloud refers to the point cloud data formed after multiple reflections occur after the millimeter wave is transmitted by the radar.

[0057] Blind spot pre-collision targets are usually not detected by direct millimeter waves because they are in the blind spot of direct detection. Radar has a multipath effect, which means that during the propagation process of radar signals, due to the influence of obstacles (such as the ground, buildings, and other objects), multiple reflections or scatterings are generated, resulting in the weakening of the received signal strength, phase distortion, pulse broadening, and other phenomena. These reflected or scattered signals will propagate to the receiving end along different paths, forming multipath propagation. This also allows targets in the blind spot to be detected by the radar through multipath reflections.

[0058] An example of multipath reflection is Figure 4 As shown, radar 200 emits millimeter waves for detection, and target 102 is located behind target 101, and target 101 may be a vehicle, etc. Radar 200 cannot detect target 102 from a conventional perspective. However, there is a gap in the vehicle chassis of target 101, and the millimeter waves of radar 200 can reach target 102 through reflection, so that target 102 can be detected.

[0059] Based on the above theory, the present application can determine whether the target is a target detected by multipath reflection according to the information of the multipath reflection point cloud of the potential warning target, so as to eliminate the false blind spot pre-collision target from the potential warning target. Among them, the false blind spot pre-collision target can be eliminated from the potential warning target according to any one of the number, proportion, concentration and divergence of the multipath reflection point cloud of the potential warning target. For example, if the large area point cloud of the target is a direct point cloud instead of a multipath reflection point cloud, it means that the target is detected by direct millimeter-wave radar and may be a false blind spot target, which can be eliminated to improve the accuracy of blind spot target detection.

[0060] In addition, by eliminating targets whose movement direction will not collide with the vehicle's direction of travel based on the movement direction of potential warning targets, the accuracy of blind spot pre-collision targets can be improved.

[0061] The above-mentioned blind spot pre-collision target detection method, after obtaining point cloud data based on radar echo signal processing, performs target clustering based on the point cloud data to obtain target and target information, finds dynamic targets blocked by static targets based on the information of targets within the detection area, preliminarily determines potential warning targets, and further eliminates false targets based on the multipath reflection point cloud and movement direction of potential warning targets, and finds real blind spot pre-collision targets detected through multipath reflection and whose movement direction may collide with the direction of travel of the vehicle. This method utilizes the multipath effect of radar and can accurately detect blind spot pre-collision targets without the need for additional radar hardware settings, which can reduce the occurrence of "ghost probe" accidents and improve assisted driving performance.

[0062] In one embodiment, target clustering is performed on point cloud data to determine targets and target information, including: converting the point cloud data from a radar coordinate system to a radar Cartesian coordinate system, compensating the converted point cloud for radar installation angle error and installation position; restoring the speed of the point cloud data based on the vehicle speed and steering angle in the current frame, and performing dynamic and static target separation processing on the point cloud; performing target clustering on the processed point cloud data to obtain targets and target information.

[0063] According to Formula 1, the point cloud is converted from the radar sphere coordinate system to the radar Cartesian coordinate system.

[0064]

[0065] Among them, x t ,y t 、z t represents the point cloud coordinates in the radar Cartesian coordinate system, r represents the radial distance of the point cloud, represents the azimuth angle, and θ represents the elevation angle.

[0066] According to Formula 2, the point cloud is compensated for the radar installation angle error and installation position.

[0067]

[0068] Among them, x w ,y w 、z w Represents the point cloud coordinates in the world Cartesian coordinate system, R(α,β,γ) is:

[0069]

[0070] Where c1 = cosα, c2 = cosβ, c3 = cosγ; s1 = sinα, s2 = sinβ, s3 = sinγ, α represents the radar installation yaw angle, β represents the pitch angle, γ represents the roll angle, offsetX, offsetY, offsetZ represent the radar installation position deviation in the X direction, Y direction and Z direction respectively.

[0071] Among them, according to formula three, the velocity of the point cloud data is restored, and the dynamic and static target separation processing is performed on the point cloud.

[0072]

[0073] Among them, satisfying abs(VV n )>VelThreshold is marked as a moving point cloud, and the rest are marked as a stationary point cloud. Where n is the point cloud index, V is the actual velocity of the point cloud, and V n is the radial component of the vehicle velocity at the point cloud position, is the point cloud azimuth, V0 is the current vehicle speed, is the current vehicle steering angle. VelThreshold is the moving and static target separation speed threshold.

[0074] In this embodiment, by converting the coordinate system, compensating for installation position and installation angle errors, and dynamically separating the point cloud data obtained by millimeter wave radar detection, target clustering is performed on the point cloud data, which can improve the accuracy of target clustering.

[0075] In one embodiment, based on the information of the targets within the detection area, the dynamic targets blocked by the static targets are determined as potential warning targets, and only the position factors of the dynamic targets and the static targets are considered, among which there may be false blind spot pre-collision targets. In order to accurately screen out the blind spot pre-collision targets from the potential warning targets, the present application further determines the blind spot pre-collision targets from the potential warning targets based on the multipath reflection point cloud and the movement direction of the potential warning targets.

[0076] Specifically, according to the multipath reflection point cloud of the potential warning target and the movement direction of the potential warning target, a blind spot pre-collision target is determined from the potential warning target, including: according to the multipath reflection point cloud ratio of the potential warning target, a candidate target is determined from the potential warning target; if the movement direction of the candidate target is close to the travel direction of the vehicle, the candidate target is determined as a blind spot pre-collision target.

[0077] Specifically, as mentioned above, although the target in the blind spot cannot be detected by the direct millimeter wave, the millimeter wave can reach the target in the blind spot through multiple reflections by using the multipath effect of the radar and be detected by the radar. Based on this characteristic of the blind spot target, in this embodiment, the candidate target is first determined from the potential warning target according to the multipath reflection point cloud ratio of the potential warning target. In this way, the candidate target is the potential warning target in the blind spot that is detected by the multipath effect.

[0078] Among them, the multipath reflection point cloud ratio of the potential warning target refers to the ratio of the multipath reflection point cloud in the potential warning target to its total point cloud. The inventor of the present application found in an experiment in a traffic environment that if the multipath reflection point cloud ratio of a potential warning target is greater than 20%, then the potential warning target is usually a blind spot target. Therefore, the multipath reflection point cloud ratio of the potential warning target can be compared with a threshold to determine whether the potential warning target is a candidate target. Among them, the threshold can be determined after experiments based on a variety of traffic environments, and is not limited to the 20% mentioned in this embodiment.

[0079] Further combined with the movement characteristics of the "ghost poking" target, in this embodiment, for the determined candidate targets, the candidate targets are further accurately screened according to the movement direction of the candidate targets. If the movement direction of the candidate target is far away from the driving direction of the vehicle, it will usually not collide with the vehicle. If the movement direction of the candidate target is close to the traveling direction of the vehicle, the candidate target is very likely to collide with the vehicle if it continues to move. Therefore, the movement direction of the candidate target is further judged. If the movement direction of the candidate target is close to the traveling direction of the vehicle, the candidate target is determined as a blind spot pre-collision target.

[0080] In this embodiment, the accuracy of blind spot pre-collision detection is improved by eliminating targets that are not in the blind spot or whose movement direction does not pose a safety threat to the safe driving of the vehicle from the multipath reflection point cloud and the target's movement direction.

[0081] In one embodiment, candidate targets are determined from potential warning targets based on the multipath reflection point cloud ratio of potential warning targets, including: obtaining the multipath reflection point cloud ratio of the potential warning target based on the ratio of the number of point clouds with height below 0 in the point cloud of the potential warning target to the total number of point clouds of the potential warning target; and determining the potential warning target whose multipath reflection point cloud ratio meets the requirements as a candidate target.

[0082] The blind spot pre-collision target detection method of the present application utilizes the technical principle that targets in the blind spot can be detected by radar through multipath reflection. Therefore, it is possible to determine whether the target is a target detected through multipath reflection based on the multipath reflection point cloud of the potential warning target, so that false blind spot pre-collision targets can be eliminated from the potential warning targets. In terms of specific implementation methods, false blind spot targets that are not detected through multipath reflection can be screened out based on the proportion of the multipath reflection point cloud of the warning target.

[0083] like Figure 4 As for the blind spot target 102 shown, the radar 200 cannot detect the target 102 from the conventional perspective. However, there is a gap in the vehicle chassis of the target 101, and the millimeter wave of the radar 200 can reach the target 102 through reflection, so that the target 102 is detected. In the radar perspective, the image of the person is below the ground, so the height of the point cloud of the detected target will be less than 0, and at the same time, a part of the electromagnetic waves may reach the feet or other parts of the target 102 without reflection, so the height of a part of the point cloud of the detected target will also be greater than 0. Based on this, in this embodiment, the multipath reflection point cloud ratio of the potential warning target is obtained according to the ratio of the number of point clouds with a height lower than 0 in the point cloud of the potential warning target to the total number of point clouds of the potential warning target; the potential warning target whose multipath reflection point cloud ratio meets the requirements is determined as a candidate target. In this way, it can be accurately determined whether the target is a blind spot target detected by electromagnetic waves through the multipath effect.

[0084] In one embodiment, before determining whether the direction of movement of the candidate target is close to the direction of travel of the vehicle, the blind spot pre-collision target detection method of the present application also includes: obtaining at least N frames of track of the candidate target; determining whether the speed of the candidate target converges based on at least N frames of track of the candidate target; if so, obtaining the direction of movement of the candidate target.

[0085] Specifically, the current frame radar echo signal is processed to obtain a point cloud, and the point cloud is clustered and analyzed to obtain the point track of the target. However, the point track can only represent the state of the current frame radar echo signal, and the target determined based on the point track may be unstable. In order to accurately detect the blind spot pre-collision target, in this embodiment, detection is performed based on at least N frames of tracks of the candidate target. Among them, the track refers to the continuous representation of the motion trajectory of the target in the airspace. By correlating and filtering multiple radar point tracks, a track can be formed. The track not only includes the motion trajectory of the target in space, but also provides more information about the target state and characteristics, such as speed, acceleration, etc.

[0086] In order to obtain the track of the candidate target, in this embodiment, track data processing is also performed on the candidate target obtained by real-time detection of the radar echo signal.

[0087] Specifically, for the candidate targets obtained by real-time detection of radar echo signals, if the track of the candidate target currently exists, the track of the candidate target is associated with the clustering result with the candidate target identification. If the association conditions are met, the target track is updated; other target tracks are associated with the clustering results without candidate target identification.

[0088] If the track of the current candidate target is the candidate target, the candidate target is used as the start of the new track. The tracks of other targets are associated with the clustering results without the candidate target identification, and the track update is performed.

[0089] After performing the track association and update operations, it is determined whether the frame number of the candidate target's track meets the threshold. In one embodiment, the frame number threshold can be set to N, and then it is determined whether at least N frames of the candidate target's track can be obtained. If so, it is determined whether the speed of the candidate target converges based on at least N frames of the candidate target's track; if so, the movement direction of the candidate target is obtained.

[0090] Specifically, in millimeter-wave radar target tracking, the millimeter-wave radar can only detect the component of the target's true speed in the radar radial direction, so tracking filtering is required to allow its speed to converge to near the true speed. If the speed of the candidate target converges, it can be considered that the speed of the target detected by the current radar is close to its true moving speed. In this embodiment, after determining that the target speed has converged, the movement direction of the candidate target is obtained to determine whether the movement direction of the candidate target is close to the direction of travel of the vehicle, which can eliminate unstable candidate targets, reduce the probability of false detection of blind spot pre-collision targets, and improve accuracy.

[0091] Among them, after determining the blind spot pre-collision target according to the track, in this application, the target track that meets the conditions of elimination or start will also be executed for track elimination and start, and track merging and separation processing. Thereby, it helps to improve the accuracy and reliability of radar target tracking and provide strong support for subsequent decision-making and analysis.

[0092] In the above embodiment, after obtaining the potential warning target, the false blind spot targets obtained by non-multipath effect detection are further eliminated according to the multipath reflection point cloud of the potential warning target and the direction of movement, and the blind spot pre-collision targets whose movement direction is close to the direction of vehicle travel and may collide are determined, thereby improving the accuracy of blind spot pre-collision target detection. At the same time, the present application also improves the accuracy of blind spot pre-collision target detection by improving the accuracy of potential warning target detection.

[0093] In one embodiment, the target information includes radial distance, azimuth, and dynamic and static status.

[0094] Based on the information of the targets within the detection area, the dynamic targets blocked by the static targets are determined as potential warning targets, including: identifying the targets within the detection area; determining the dynamic targets and static targets within the detection area according to the dynamic and static states of the targets; determining the dynamic targets blocked by the static targets according to the azimuth and radial distances of the dynamic targets and the static targets within the detection area, and determining the dynamic targets blocked by the static targets as potential warning targets.

[0095] Specifically, the detection area is determined, the targets within the detection area are identified, and dynamic targets and static targets are distinguished. According to analysis, "ghost" targets usually have two characteristics: the first characteristic is that they are hidden behind static targets and are in the driver's visual blind spot; the second characteristic is that they are in motion in the visual blind spot and are likely to affect the safe driving of the vehicle.

[0096] According to the above two characteristics of the "ghosting" target, the information of the detected target is used in this embodiment to determine the dynamic targets and static targets that are close in position and have a chronological relationship in distance according to the azimuth and radial distance of the dynamic targets and static targets within the detection area. The dynamic targets obscured by the static targets are thereby determined as potential warning targets.

[0097] In this embodiment, it is taken into consideration that "ghosting" targets usually have two characteristics, and potential warning targets can be quickly screened out based on the azimuth and radial distance of dynamic targets and static targets within the detection area.

[0098] In one embodiment, determining the dynamic target blocked by the static target according to the azimuth angle and radial distance of the dynamic target and the static target within the detection area includes:

[0099] If the azimuth angle and radial distance of the dynamic target within the detection area and the static target within the detection range meet the preset relationship, the dynamic target is determined to be a dynamic target blocked by the static target. The preset relationship includes:

[0100] The radial distance difference between the dynamic target and the static target is less than a first distance threshold;

[0101] The radial distance of a dynamic target is greater than that of a static target;

[0102] The absolute value of the azimuth angle difference between the dynamic target and the static target is less than the first azimuth angle threshold.

[0103] By limiting the distance between the dynamic target and the static target to be less than the first distance threshold, it can be ensured that the distance between the dynamic target and the static target in the radial direction is not too large.

[0104] By limiting the distance of the dynamic target to be greater than the distance of the static target, it can be ensured that the dynamic target is located behind the static target.

[0105] By limiting the absolute value of the azimuth angle difference between the dynamic target and the static target to be smaller than the first azimuth angle threshold, it can be determined that the lateral distance between the dynamic target and the static target is not too large.

[0106] For easier description, the following relationship can be used to detect targets within the area and find dynamic targets blocked by static targets.

[0107] Among them, the relationship is:

[0108] abs(clusterOut[i].Azimuth-clusterOut[j].Azimuth)<AzimuthThreshold

[0109] clusterOut[i].Range<clusterOut[j].Range

[0110] clusterOut[j].Range-clusterOut[i].Range<RangeThreshold

[0111] clusterOut[i].Doppler=0

[0112] clusterOut[j].Doppler≠0

[0113] Among them, clusterOut[j].Doppler≠0 means that target j is a dynamic target. Indicates that target i is a static target, represents the radial distance of target i, represents the radial distance of target j, represents the azimuth of target i, represents the azimuth of target j, AzimuthThreshold represents the first azimuth threshold, and RangeThreshold represents the first distance threshold.

[0114] In one embodiment, if Figure 5 As shown, there is a static target 101 and a dynamic target 102 in the scene. The absolute value of the azimuth angle between the dynamic target 102 and the static target 101 is less than the first azimuth angle threshold. It can be determined that the lateral distance between the dynamic target 102 and the static target 101 is relatively close, which meets the requirements. The distance between the dynamic target 102 and the static target 101 is less than the first distance threshold. It can be determined that the radial distance between the dynamic target 102 and the static target 101 is relatively close, which meets the requirements. The distance of the static target 101 is less than the distance of the dynamic target 102. It can be determined that the static target 101 is located in front of the dynamic target 102, resulting in the dynamic target 102 being blocked by the static target 101. In this way, through the above relationship, according to the relationship between the radial and azimuth angles between the dynamic target and the static target, potential warning targets can be quickly screened out.

[0115] In one embodiment, identifying a target within a detection area includes: obtaining an azimuth range and a radial distance range; determining the detection area range according to the azimuth range and the radial distance range; if the azimuth of the target is within the azimuth range and the distance of the target is within the radial distance range, determining that the target is a target within the detection area range.

[0116] Among them, the azimuth range is used to determine the lateral range of the detection area, that is, the width, which includes the azimuth upper limit value and the azimuth lower limit value. The interval range between the azimuth upper limit value and the azimuth lower limit value is the width range of the detection area. The radial distance range is used to determine the radial range of the detection area, that is, the length, which includes the distance upper limit value and the distance lower limit value. The interval range between the distance upper limit value and the distance lower limit value is the length range of the detection area.

[0117] like Figure 5 As shown, in one embodiment, the detection area range 502 can be determined by the azimuth angle range and the radial angle range.

[0118] In one embodiment, the azimuth angle range and the radial distance range may be preset empirical values.

[0119] In one embodiment, the radial distance range is determined according to the vehicle speed. The vehicle speed is proportional to the radial distance range. The faster the vehicle speed, the larger the radial distance range should be, so that the blind spot pre-collision target can be detected within a longer detection range, so that the vehicle with a faster speed has sufficient time to brake and other measures, thereby reducing the occurrence of collision accidents.

[0120] After determining the detection area, the target's azimuth is compared with the boundary values ​​of the azimuth range, i.e., the upper and lower limits, and the target's radial distance is compared with the boundary values ​​of the radial distance range to determine whether the target is within the detection area.

[0121] In one embodiment, the following formula can be used to traverse the target detected by the point cloud to determine whether it is within the detection area.

[0122] minAzimuth<abs(clusterOut[i].Azimuth)<maxAzimuth

[0123] minRange<clusterOut[i].Range<maxRange

[0124] Among them, minAzimuth is the lower limit of the azimuth range, maxAzimuth is the upper limit of the azimuth range, minRange is the lower limit of the radial distance range, and maxRange is the upper limit of the radial distance range. is the azimuth of target i, is the radial distance of target i.

[0125] In this embodiment, the detection area range can be determined according to preset parameters, or the detection area range can be adaptively set according to the vehicle speed, and the target within the detection area range can be determined according to the azimuth angle and radial distance of the target.

[0126] In one embodiment, based on the multipath reflection point cloud of the potential warning target and the movement direction of the potential warning target, the blind spot pre-collision target is determined from the potential warning target, and the blind spot pre-collision target detection method also includes: warning the blind spot pre-collision target. Warning the blind spot pre-collision target can be that the auxiliary driving system sends a reminder message indicating that the blind spot pre-collision target text and / or picture is detected in the display interface of the vehicle-mounted central control system, and the reminder message may include the position and location of the blind spot pre-collision target; it can also refer to sending a reminder message that the blind spot pre-collision target is detected through the indicator light and / or speaker in the auxiliary driving system, and the reminder message may include the position and location of the blind spot pre-collision target. In some embodiments, warning the blind spot pre-collision target can also include various reminder methods that can remind the driver to know and identify the blind spot pre-collision target in time, such as sending a warning prompt message to the associated terminal device, a reminder message for enabling the maximum speed limit control for the vehicle, etc.

[0127] It should be noted that the vehicle-mounted central control system can detect the vehicle driving environment through various sensor elements such as vehicle-mounted cameras and speed sensors. In this embodiment, the vehicle-mounted millimeter-wave radar can be used as a sensor element of the vehicle-mounted central control system to accurately detect whether there are blind spot pre-collision targets that may affect the driving safety of the vehicle. The vehicle-mounted central control of the vehicle-mounted system predicts the driving safety risks based on the detection results of the vehicle-mounted millimeter-wave radar to improve the assisted driving performance.

[0128] To have a more comprehensive understanding of the blind spot pre-collision target detection method, please refer to Figure 6 and Figure 7 ,by Figure 2 Taking the structure of the vehicle-mounted millimeter-wave radar shown in the figure as an example, the main working process of the vehicle-mounted millimeter-wave radar and the process of the blind spot pre-collision target detection method are explained.

[0129] S11, millimeter wave radar will receive The echo signal is processed to obtain multiple detection points.

[0130] S12, preprocessing the detection points. The preprocessing includes invalid elimination of point cloud data, compensation of installation errors, coordinate system conversion, and dynamic and static separation.

[0131] S13, performing target clustering on the point cloud data to determine the target and target information.

[0132] S14, according to the information of the targets within the detection area, the dynamic targets blocked by the static targets are determined as potential warning targets, wherein the target information includes distance information, azimuth information, and static and dynamic status.

[0133] Among them, this step includes:

[0134] S141, obtaining an azimuth angle range and a radial distance range, wherein the radial distance range is determined according to the vehicle speed.

[0135] S142, determining the detection area range according to the azimuth angle range and the radial distance range.

[0136] S143: If the azimuth angle of the target is within the azimuth angle range and the radial distance of the target is within the radial distance range, the target is determined to be a target within the detection area range.

[0137] S144, determining the dynamic targets and static targets within the detection area according to the dynamic and static states of the targets.

[0138] S145: If the azimuth angle and radial distance of the dynamic target within the detection area and the static target within the detection area satisfy a preset relationship, it is determined that the dynamic target is a dynamic target blocked by the static target. The preset relationship includes:

[0139] The radial distance difference between the dynamic target and the static target is less than a first distance threshold;

[0140] The radial distance of a dynamic target is greater than that of a static target;

[0141] The absolute value of the azimuth difference between the dynamic target and the static target is less than the first azimuth threshold

[0142] According to the azimuth information and distance information of dynamic targets and static targets within the detection area, the dynamic targets blocked by static targets are determined, and the dynamic targets blocked by static targets are determined as potential warning targets.

[0143] After step S14, the method further includes:

[0144] Step S15, determining the blind spot pre-collision target from the potential warning targets according to the multipath reflection point cloud of the potential warning targets and the movement direction of the potential warning targets.

[0145] Wherein, step S15 comprises:

[0146] S151, obtaining the multipath reflection point cloud ratio of the potential warning target according to the ratio of the number of point clouds with heights lower than 0 in the point cloud of the potential warning target to the total number of point clouds of the potential warning target.

[0147] S152, determine the potential warning targets whose multipath reflection point cloud ratios meet the requirements as candidate targets.

[0148] S153, obtaining at least N frames of track of the candidate target.

[0149] S154, judging whether the speed of the candidate target is converged according to at least N frames of track of the candidate target, and if so, obtaining the moving direction of the candidate target.

[0150] S155: If the moving direction of the candidate target is close to the traveling direction of the vehicle, the candidate target is determined as a blind spot pre-collision target.

[0151] After step S15, the method further includes:

[0152] Step S16: issuing an early warning to the blind spot pre-collision target.

[0153] The blind spot pre-collision target detection method provided in the above embodiment is based on the ground multipath reflection echo signal of the vehicle-mounted millimeter-wave radar, and the point cloud information of the scene is obtained through signal processing. Through the clustering, association, tracking and target recognition methods in data processing, the position and speed of the blind spot pre-collision target can be accurately calculated, and an early warning can be issued. Without adding millimeter-wave radar hardware equipment, the unique characteristic of the millimeter-wave radar signal with multipath effect is utilized to realize the detection and early warning of the blind spot pre-collision target.

[0154] Among them, the typical workflow of the vehicle-mounted millimeter-wave radar using the blind spot pre-collision target detection method proposed in the embodiment of the present application is as follows: Figure 7 As shown, the following steps are included:

[0155] S211, the vehicle-mounted millimeter-wave radar is powered on and initialized, and the working parameters are read and sent down.

[0156] S212, the vehicle-mounted millimeter-wave radar receives an instruction to start working.

[0157] S213, the MRTR chip linear frequency modulation local oscillator source component generates a linear frequency modulation waveform local oscillator signal according to the waveform parameters, and transmits it to the transmitting component and the receiving component.

[0158] S214, the MRTR chip transmitting component generates a transmitting signal according to the linear frequency modulation waveform local oscillator signal, and radiates the transmitting signal through the transmitting antenna array.

[0159] S215, the receiving antenna array receives the waveform signal reflected by the target and transmits it to the MRTR chip receiving component.

[0160] S216, the MRTR chip performs discrete ADC sampling on the waveform signal reflected by the received target.

[0161] S217, the MRP chip ranging module performs ranging processing on the discrete ADC signal from the MRTR chip.

[0162] S218, the MRP chip speed measurement module performs speed measurement processing.

[0163] S219, the MRP chip angle measurement module performs angle measurement processing.

[0164] S220, the MRP chip target clustering association tracking module performs target-level data processing.

[0165] S221, MRP chip target recognition and event recognition processing.

[0166] S222, the MRP chip outputs point cloud-level detection results or target-level detection results.

[0167] S223, completing scene target detection in a loop until a condition for stopping operation is met.

[0168] Among them, the blind spot pre-collision target detection method proposed in the embodiment of the present application can be executed by a millimeter wave radar chip, and can be specifically completed in step S220.

[0169] On the other hand, please refer to Figure 2 and Figure 8 , a vehicle-mounted millimeter-wave radar is also provided, including a millimeter-wave radar processor 201 and a memory 202. The memory 202 stores a computer program that can be executed by the millimeter-wave radar processor 201; when the computer program is executed by the millimeter-wave radar processor 201, the blind spot pre-collision target detection method described in any embodiment of the present application is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0170] On the other hand, please refer to Fig. 9 , provides a vehicle, including an on-board central control system and an on-board millimeter-wave radar interactively connected to the on-board central control system; the on-board millimeter-wave radar is used to execute the blind spot pre-collision target detection method described in any embodiment of the present application; the on-board central control system is used to output corresponding auxiliary driving information according to the detection result of the blind spot pre-collision target.

[0171] Among them, the assisted driving information can be the detection result of the blind spot pre-collision target by the vehicle-mounted central control system using the vehicle-mounted millimeter-wave radar, and the assisted driving information output after calculating the driving safety risk prediction, such as including but not limited to real-time updating and displaying the driving trajectory of the blind spot pre-collision target, the position and speed of the blind spot pre-collision target in the vehicle-mounted assisted driving screen.

[0172] On the other hand, an embodiment of the present application further provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the blind spot pre-collision target detection method of any embodiment of the present application and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0173] On the other hand, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of the above-mentioned blind spot pre-collision target detection method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0174] The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0175] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0176] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0177] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A blind spot pre-collision target detection method, characterized in that: include: Receiving radar echo signals, and processing the radar echo signals to obtain point cloud data; Performing target clustering on the point cloud data to determine targets and information about the targets; According to the information of the target within the detection area, the dynamic target blocked by the static target is determined as a potential warning target; According to the multipath reflection point cloud of the potential warning target and the moving direction of the potential warning target, a blind area pre-collision target is determined from the potential warning targets.

2. The blind spot pre-collision target detection method according to claim 1, characterized in that: The step of determining a blind spot pre-collision target from the potential warning target according to the multipath reflection point cloud of the potential warning target and the moving direction of the potential warning target comprises: Determining candidate targets from the potential warning targets according to the multipath reflection point cloud ratio of the potential warning targets; If the moving direction of the candidate target is close to the traveling direction of the vehicle, the candidate target is determined as a blind spot pre-collision target.

3. The blind spot pre-collision target detection method according to claim 2, characterized in that: Determining candidate targets from the potential warning targets according to the multipath reflection point cloud ratio of the potential warning targets includes: Obtaining a multipath reflection point cloud ratio of the potential warning target according to a ratio of the number of point clouds with heights below 0 in the point cloud of the potential warning target to the total number of point clouds of the potential warning target; The potential warning target whose multipath reflection point cloud ratio meets the requirements is determined as the candidate target.

4. The blind spot pre-collision target detection method according to claim 2, characterized in that: Before determining whether the moving direction of the candidate target is close to the traveling direction of the vehicle, the method includes: Acquire at least N frames of track of the candidate target; Determining whether the speed of the candidate target converges according to at least N frames of track of the candidate target; If so, the moving direction of the candidate target is obtained.

5. The blind spot pre-collision target detection method according to claim 1, characterized in that: The target information includes radial distance, azimuth, and dynamic and static status; The step of determining, based on the information of the target within the detection area, the dynamic target blocked by the static target as a potential warning target includes: Identify targets within the detection area; Determine the dynamic targets and static targets within the detection area according to the dynamic and static states of the targets; The dynamic target blocked by the static target is determined according to the azimuth angle and the radial distance of the dynamic target and the static target within the detection area, and the dynamic target blocked by the static target is determined as a potential warning target.

6. The blind spot pre-collision target detection method according to claim 5, characterized in that: The determining the dynamic target blocked by the static target according to the azimuth angle and the radial distance of the dynamic target and the static target within the detection area includes: If the azimuth angle and the radial distance of the dynamic target within the detection area satisfy a preset relationship with the static target within the detection range, it is determined that the dynamic target belongs to the dynamic target blocked by the static target: the preset relationship includes: The radial distance difference between the dynamic target and the static target is less than a first distance threshold; The radial distance of the dynamic target is greater than the radial distance of the static target; An absolute value of an azimuth angle difference between the dynamic target and the static target is smaller than a first azimuth angle threshold.

7. The blind spot pre-collision target detection method according to claim 5, characterized in that: The identifying of the target within the detection area includes: Get the azimuth range and radial distance range; Determining the detection area range according to the azimuth angle range and the radial distance range; If the azimuth angle of the target is within the azimuth angle range, and the radial distance of the target is within the radial distance range, the target is determined to be a target within the detection area.

8. The blind spot pre-collision target detection method according to claim 7, characterized in that: The method of obtaining the radial distance range includes: The radial distance range is determined based on the vehicle speed.

9. A vehicle-mounted millimeter wave radar, characterized in that: The invention comprises a millimeter wave radar processor and a memory, wherein the memory stores a computer program executable by the millimeter wave radar processor; When the computer program is executed by the millimeter wave radar processor, the blind spot pre-collision target detection method according to any one of claims 1 to 8 is implemented.

10. A vehicle, characterized in that: It includes a vehicle-mounted central control system and a vehicle-mounted millimeter-wave radar interactively connected to the vehicle-mounted central control system; The vehicle-mounted millimeter-wave radar is used to perform the blind spot pre-collision target detection method according to any one of claims 1 to 8; The vehicle-mounted central control system is used to output corresponding auxiliary driving information according to the detection result of the millimeter-wave radar on the blind spot pre-collision target.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the blind spot pre-collision target detection method according to any one of claims 1 to 8 is implemented.

Images