Control method, device, electronic device and storage medium for target vehicle

Abstract

The present disclosure provides a control method, device, electronic device and storage medium for a target vehicle, wherein the control method comprises: acquiring multiple frames of point cloud images collected by a radar device during the driving process of the target vehicle; performing obstacle detection on each frame of the point cloud image to determine the current position and confidence of the target obstacle; and controlling the driving of the target vehicle based on the determined current position and confidence of the target obstacle and the current posture data of the target vehicle.

Description

Technical Field

[0001] The present disclosure relates to the field of autonomous driving technology, and in particular to a control method, device, electronic device and storage medium for a target vehicle. Background Art

[0002] In the field of assisted driving or autonomous driving, point cloud images can be obtained through radar, and whether there is a target obstacle can be determined based on the point cloud image. When the presence of a target obstacle is detected, the vehicle's driving can be controlled based on the position of the detected target obstacle, such as whether to slow down and avoid the obstacle. Summary of the invention

[0003] The disclosed embodiments at least provide a control scheme for a target vehicle.

[0004] In a first aspect, an embodiment of the present disclosure provides a method for controlling a target vehicle, the method comprising:

[0005] When the target vehicle is driving, a multi-frame point cloud image collected by a radar device is obtained;

[0006] Perform obstacle detection on each frame of point cloud image to determine the current position and confidence of the target obstacle;

[0007] Based on the determined current position and confidence of the target obstacle and the current position data of the target vehicle, the target vehicle is controlled to travel.

[0008] In the disclosed embodiment, the position changes of the target obstacle in the multi-frame point cloud images can be tracked together through the multi-frame point cloud images. In this way, the accuracy of the confidence that the target obstacle appears at the current position is improved, so that when the vehicle is controlled based on the confidence, effective control of the target vehicle is achieved. For example, frequent stops or collisions due to false detection of target obstacles can be avoided.

[0009] In a possible implementation, the confidence is determined based on at least two of the following parameters: average detection confidence, tracking matching confidence, tracking chain effective length, velocity smoothness, and acceleration smoothness;

[0010] Determine the confidence level of the target obstacle, including:

[0011] The confidence level of the target obstacle is obtained by weighted summing or multiplying the at least two parameters.

[0012] In the disclosed embodiment, it is proposed to jointly determine the confidence of the target obstacle at the current position by using multiple parameters. In this way, when the confidence of the target obstacle is determined from multiple angles, the accuracy of the confidence corresponding to the determined target obstacle at the current position can be improved.

[0013] In a possible implementation, the average detection confidence is determined in the following manner:

[0014] According to the detection confidence of the target obstacle appearing in each frame of the point cloud image, the average detection confidence corresponding to the target obstacle is determined.

[0015] In an embodiment of the present disclosure, it is proposed that the parameters for determining the confidence of a target obstacle include an average detection confidence, which can reflect the average reliability of the position of the target obstacle in multiple frame point cloud images. When the confidence of the target obstacle is determined based on the average detection confidence, the stability of the confidence of the determined target obstacle can be improved.

[0016] In a possible implementation manner, the tracking matching confidence is determined in the following manner:

[0017] Based on the position information of the target obstacle in each frame of the point cloud image, a tracking matching confidence that the target obstacle is a tracking object matched by the multiple frames of point cloud images is determined.

[0018] In the disclosed embodiment, the tracking matching confidence is used to represent the possibility of the target obstacle appearing in the continuous multi-frame point cloud images. If the possibility of the target obstacle appearing in the continuous multi-frame point cloud images is greater, it means that the possibility of the target obstacle being a false detection result is smaller. Based on this, the tracking matching confidence of the target obstacle and the tracking chain can be used as a parameter for determining the confidence of the target obstacle to improve the accuracy of the confidence.

[0019] In a possible implementation, determining the tracking matching confidence that the target obstacle is a tracking object matched by the multiple frames of point cloud images based on the position information of the target obstacle in each frame of the point cloud image includes:

[0020] For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image of the frame of point cloud image, the predicted position information of the target obstacle in the frame of point cloud image is determined; based on the predicted position information and the position information of the target obstacle in the frame of point cloud image, the displacement deviation information of the target obstacle in the frame of point cloud image is determined;

[0021] Determine detection frame difference information corresponding to the target obstacle based on the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image and the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image before the frame point cloud image;

[0022] Determine the orientation angle difference information corresponding to the target obstacle based on the orientation angle of the target obstacle in the frame point cloud image and the orientation angle of the target obstacle in the previous frame point cloud image;

[0023] Determine, based on the displacement deviation information, the detection frame difference information, and the orientation angle difference information, a single-frame tracking matching confidence that the target obstacle is a tracking object matched by the frame point cloud image;

[0024] According to the single-frame tracking matching confidence of the target obstacle being the tracking object matched by each frame of the multi-frame point cloud image, the tracking matching confidence of the target obstacle being the tracking object matched by the multi-frame point cloud image is determined.

[0025] In the disclosed embodiment, the parameters for determining the confidence of the target obstacle include tracking matching confidence, which can reflect the reliability of the target obstacle being a tracking object of the multi-frame point cloud image. In this way, when determining the confidence of the target obstacle based on the multi-frame point cloud image, taking this parameter into consideration can improve the accuracy of the confidence of the target obstacle.

[0026] In a possible implementation, for each frame of point cloud image, based on the position information of the target obstacle in a previous frame of point cloud image of the frame of point cloud image, determining the predicted position information of the target obstacle in the frame of point cloud image includes:

[0027] For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image, the position information of the target obstacle in the previous frame of point cloud image, and the acquisition time interval between two adjacent frames of point cloud image, determine the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image;

[0028] Based on the position information of the target obstacle in the previous frame of point cloud image, the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image, and the acquisition time interval between the current frame of point cloud image and the previous frame of point cloud image, the predicted position information of the target obstacle in the frame of point cloud image is determined.

[0029] In a possible implementation manner, the effective length of the tracking chain is determined in the following manner:

[0030] Based on the position information of the target obstacle in each frame of the point cloud image, the number of missed detection frames for the target obstacle in the multiple frames of point cloud images is determined; and based on the total number of frames corresponding to the multiple frames of point cloud images and the number of missed detection frames, the effective length of the tracking chain is determined.

[0031] In the disclosed embodiment, it is proposed to use the effective length of the tracking chain as a parameter for determining the confidence of the target obstacle. The accuracy of the neural network for detecting the target obstacle for each frame of the point cloud image is determined by the effective length of the tracking chain. Therefore, when the confidence of the target obstacle is determined based on the effective length of the tracking chain, the accuracy of the confidence can be improved.

[0032] In a possible implementation manner, the speed smoothness is determined in the following manner:

[0033] Determine a speed error of the target obstacle within a collection time corresponding to the multiple frames of point cloud images based on the speed of the target obstacle at the collection time corresponding to each frame of point cloud images;

[0034] Based on the speed error corresponding to the target obstacle and a pre-stored standard deviation preset value, the speed smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images is determined.

[0035] In the disclosed embodiment, the speed smoothness can reflect the smoothness of the change in the speed of the target obstacle, and can reflect the position change of the target obstacle in the continuous multi-frame point cloud image, so that the reliability of the detected position information of the target obstacle can be reflected. Based on this, the speed smoothness can be used as a parameter for determining the confidence of the target obstacle to improve the accuracy of the confidence.

[0036] In a possible implementation manner, the acceleration smoothness is determined in the following manner:

[0037] Determine the acceleration of the target obstacle at the acquisition time corresponding to each frame of the point cloud image based on the speed of the target obstacle at the acquisition time corresponding to each frame of the point cloud image and the acquisition time interval between two adjacent frames of the point cloud image;

[0038] Determine the acceleration error of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images based on the acceleration of the target obstacle at the acquisition time corresponding to each frame of point cloud images;

[0039] Based on the acceleration error corresponding to the target obstacle and a pre-stored standard deviation preset value, the acceleration smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images is determined.

[0040] In the disclosed embodiment, the acceleration smoothness can reflect the smoothness of the change of the acceleration of the target obstacle, and can reflect the speed change of the target obstacle within the acquisition time corresponding to the continuous multi-frame point cloud image. It can also reflect the position change of the target obstacle in the continuous multi-frame point cloud image. In this way, the reliability of the detected position information of the target obstacle can be reflected. Based on this, the acceleration smoothness can be used as a parameter for determining the confidence of the target obstacle to improve the accuracy of the confidence.

[0041] In a possible implementation, the controlling the target vehicle to travel based on the determined current position and confidence of the target obstacle and the current position data of the target vehicle includes:

[0042] When it is determined that the confidence level corresponding to the target obstacle is higher than a preset confidence threshold, determining the distance information between the target vehicle and the target obstacle based on the current position of the target obstacle and the current position data of the target vehicle;

[0043] The target vehicle is controlled to travel based on the distance information.

[0044] In a second aspect, an embodiment of the present disclosure provides a control device for a target vehicle, the control device comprising:

[0045] An acquisition module is used to acquire a multi-frame point cloud image collected by a radar device during the driving process of the target vehicle;

[0046] The determination module is used to perform obstacle detection on each frame of the point cloud image and determine the current position and confidence of the target obstacle;

[0047] A control module is used to control the target vehicle to travel based on the determined current position and confidence of the target obstacle and the current posture data of the target vehicle.

[0048] In a possible implementation, the confidence is determined based on at least two of the following parameters: average detection confidence, tracking matching confidence, tracking chain effective length, velocity smoothness, and acceleration smoothness;

[0049] The determination module is specifically used for:

[0050] The confidence level of the target obstacle is obtained by weighted summing or multiplying the at least two parameters.

[0051] In a possible implementation manner, the determination module is further configured to determine the average detection confidence in the following manner:

[0052] According to the detection confidence of the target obstacle appearing in each frame of the point cloud image, the average detection confidence corresponding to the target obstacle is determined.

[0053] In a possible implementation manner, the determination module is further configured to determine the tracking match confidence in the following manner:

[0054] Based on the position information of the target obstacle in each frame of the point cloud image, a tracking matching confidence that the target obstacle is a tracking object matched by the multiple frames of point cloud images is determined.

[0055] In a possible implementation manner, the determining module is specifically configured to:

[0056] For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image of the frame of point cloud image, the predicted position information of the target obstacle in the frame of point cloud image is determined; based on the predicted position information and the position information of the target obstacle in the frame of point cloud image, the displacement deviation information of the target obstacle in the frame of point cloud image is determined;

[0057] Determine detection frame difference information corresponding to the target obstacle based on the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image and the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image before the frame point cloud image;

[0058] Determine the orientation angle difference information corresponding to the target obstacle based on the orientation angle of the target obstacle in the frame point cloud image and the orientation angle of the target obstacle in the previous frame point cloud image;

[0059] Determine, based on the displacement deviation information, the detection frame difference information, and the orientation angle difference information, a single-frame tracking matching confidence that the target obstacle is a tracking object matched by the frame point cloud image;

[0060] According to the single-frame tracking matching confidence of the target obstacle being the tracking object matched by each frame of the multi-frame point cloud image, the tracking matching confidence of the target obstacle being the tracking object matched by the multi-frame point cloud image is determined.

[0061] In a possible implementation manner, the determining module is specifically configured to:

[0062] For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image, the position information of the target obstacle in the previous frame of point cloud image, and the acquisition time interval between two adjacent frames of point cloud image, determine the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image;

[0063] Based on the position information of the target obstacle in the previous frame of point cloud image, the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image, and the acquisition time interval between the current frame of point cloud image and the previous frame of point cloud image, the predicted position information of the target obstacle in the frame of point cloud image is determined.

[0064] In a possible implementation manner, the determination module is further configured to determine the effective length of the tracking chain in the following manner:

[0065] Based on the position information of the target obstacle in each frame of the point cloud image, the number of missed detection frames for the target obstacle in the multiple frames of point cloud images is determined; and based on the total number of frames corresponding to the multiple frames of point cloud images and the number of missed detection frames, the effective length of the tracking chain is determined.

[0066] In a possible implementation manner, the determining module is further configured to determine the speed smoothness in the following manner:

[0067] Determine a speed error of the target obstacle within a collection time corresponding to the multiple frames of point cloud images based on the speed of the target obstacle at the collection time corresponding to each frame of point cloud images;

[0068] Based on the speed error corresponding to the target obstacle and a pre-stored standard deviation preset value, the speed smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images is determined.

[0069] In a possible implementation manner, the determination module is further configured to determine the acceleration smoothness in the following manner:

[0070] Determine the acceleration of the target obstacle at the acquisition time corresponding to each frame of the point cloud image based on the speed of the target obstacle at the acquisition time corresponding to each frame of the point cloud image and the acquisition time interval between two adjacent frames of the point cloud image;

[0071] Determine the acceleration error of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images based on the acceleration of the target obstacle at the acquisition time corresponding to each frame of point cloud images;

[0072] Based on the acceleration error corresponding to the target obstacle and a pre-stored standard deviation preset value, the acceleration smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images is determined.

[0073] In a possible implementation manner, the control module is specifically used to:

[0074] When it is determined that the confidence level corresponding to the target obstacle is higher than a preset confidence threshold, determining the distance information between the target vehicle and the target obstacle based on the current position of the target obstacle and the current position data of the target vehicle;

[0075] The target vehicle is controlled to travel based on the distance information.

[0076] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the control method described in the first aspect are performed.

[0077] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the control method described in the first aspect are executed.

[0078] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.

[0080] Figure 1 A flow chart of a method for controlling a target vehicle provided by an embodiment of the present disclosure is shown;

[0081] Figure 2 A flow chart of a method for determining tracking matching confidence corresponding to a target obstacle provided by an embodiment of the present disclosure is shown;

[0082] Figure 3 A flow chart of a method for determining predicted position information of a target obstacle provided by an embodiment of the present disclosure is shown;

[0083] Figure 4 A flow chart of a method for determining a speed smoothing length provided by an embodiment of the present disclosure is shown;

[0084] Figure 5 A flow chart of a method for determining acceleration smoothing length provided by an embodiment of the present disclosure is shown;

[0085] Figure 6 A schematic structural diagram of a control device for a target vehicle provided by an embodiment of the present disclosure is shown;

[0086] Figure 7 A schematic diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0088] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0089] During the driving process of the target vehicle, a point cloud image within a set range of distance from the target vehicle can be collected at a set time interval, and further, the position information of the target obstacle within the set range of distance from the target vehicle can be detected based on the point cloud image. For example, the point cloud image can be input into a neural network for obstacle detection, and the target obstacle contained in the point cloud image and the position information of the target obstacle are output. Considering that due to various circumstances, such as detection errors of the neural network, detection problems of point cloud data, etc., the position information of the target obstacle detected in the point cloud image may not be accurate, so when the position information of the target obstacle is detected, the confidence of the position information of the target obstacle will be given, that is, the reliability of the accurate position information of the target obstacle. When the confidence is high, the vehicle can be controlled to decelerate and avoid obstacles based on the position information of the target obstacle. When the confidence is low, the vehicle can still be controlled to decelerate and avoid obstacles based on the position information of the target obstacle with a higher confidence previously detected. Therefore, how to improve the confidence of the detected target obstacle is more critical, and the embodiments of the present disclosure will discuss this.

[0090] Based on the above research, the present disclosure provides a control method for a target vehicle, which obtains multiple frames of point cloud images collected by a radar device, performs obstacle detection on each frame of the point cloud image, and determines the current position and confidence of the target obstacle. For example, each frame of the point cloud image can be detected to determine whether the frame of the point cloud image contains the target obstacle, as well as the position information of the target obstacle in the frame of the point cloud image. In this way, the position changes of the target obstacle in the multiple frames of the point cloud images can be tracked together through the multiple frames of the point cloud images. In this way, the accuracy of the confidence that the target obstacle appears at the current position is improved, so that when the vehicle is controlled based on the confidence, effective control of the target vehicle is achieved. For example, frequent stops or collisions due to false detection of target obstacles can be avoided.

[0091] To facilitate understanding of this embodiment, a control method for a target vehicle disclosed in an embodiment of the present disclosure is first introduced in detail. The execution subject of the control method provided in the embodiment of the present disclosure is generally a computer device with certain computing capabilities, and the computer device includes, for example: a terminal device or a server or other processing device, and the terminal device can be a user equipment (UE), a mobile device, a user terminal, a computing device, a vehicle-mounted device, etc. In some possible implementations, the control method can be implemented by a processor calling a computer-readable instruction stored in a memory.

[0092] See also Figure 1 FIG. 1 is a flow chart of a method for controlling a target vehicle provided by an embodiment of the present disclosure. The method for controlling a target vehicle includes steps S101 to S103, wherein:

[0093] S101, when the target vehicle is traveling, obtaining a plurality of point cloud images collected by a radar device.

[0094] Exemplarily, the radar device may include a laser radar device, a millimeter wave radar device, an ultrasonic radar device, etc., which are not specifically limited here.

[0095] For example, taking a laser radar device as an example, the laser radar device can obtain a frame of point cloud image by scanning 360 degrees. When the radar device is set on a target vehicle, as the target vehicle travels, the radar device can collect point cloud images at set time intervals. In this way, multiple frames of point cloud images can be obtained.

[0096] Exemplarily, the multi-frame point cloud images here may be continuous multi-frame point cloud images collected at set time intervals. For the current frame point cloud image, the continuous multi-frame point cloud images may include the current frame point cloud image and multi-frame point cloud images collected before and after the collection time of the current frame point cloud image within the set time length.

[0097] S102, performing obstacle detection on each frame of the point cloud image to determine the current position and confidence level of the target obstacle.

[0098] Exemplarily, obstacle detection is performed on each frame of point cloud graphics, which may include detecting the position and confidence of the target obstacle in each frame of point cloud image, or may also include detecting the speed of the target obstacle in each frame of point cloud image, or may also include detecting the acceleration of the target obstacle in each frame of point cloud image. A variety of detection methods can be used to jointly determine the current position and confidence of the target obstacle.

[0099] Among them, the current position of the target obstacle in each frame of the point cloud image can be the current position of the target obstacle in the coordinate system where the target vehicle is located, and the confidence is the possibility that the target obstacle appears at the current position. Here, when determining the possibility of the target obstacle appearing at the current position, it can be determined by performing obstacle detection on multiple frames of point cloud images collected within a set time period including the current moment and before the current moment.

[0100] Exemplarily, when obstacle detection is performed on each frame of point cloud image, the obstacles contained in the frame of point cloud image can be detected, and the obstacles in the driving direction of the target vehicle can be used as the target obstacles here. When a frame of point cloud image contains multiple obstacles, the target obstacles in the multiple frames of point cloud images can be determined based on the numbers corresponding to the obstacles in each frame of point cloud image. The embodiment of the present disclosure will be explained by determining the confidence of one of the target obstacles. When there are multiple obstacles, multiple target obstacles can be determined, and each target obstacle can be determined in the same way.

[0101] S103, based on the determined current position and confidence level of the target obstacle and the current position and posture data of the target vehicle, controlling the target vehicle to travel.

[0102] Furthermore, after determining the current position and confidence of the target vehicle, the possibility of the target vehicle appearing at the current position can be determined based on the confidence. For example, when it is determined that the possibility of the target vehicle appearing at the current position is high, the target vehicle can be controlled based on the current position of the target obstacle and the current posture data of the target vehicle. Conversely, when it is determined that the possibility of the target vehicle appearing at the current position is low, the current position of the target obstacle can be ignored when controlling the target vehicle, or the target vehicle can be controlled based on the previous position information of the target obstacle and the current posture data of the target vehicle.

[0103] Specifically, when controlling the target vehicle to travel based on the determined current position and confidence of the target obstacle and the current position and posture data of the target vehicle, the following steps may be included:

[0104] (1) when it is determined that the confidence level corresponding to the target obstacle is higher than a preset confidence threshold, the distance information between the target vehicle and the target obstacle is determined based on the current position of the target obstacle and the current position and posture data of the target vehicle;

[0105] (2) Control the target vehicle to move based on the distance information.

[0106] Specifically, the current posture data of the target vehicle may include the current position of the target vehicle and the current driving direction of the target vehicle. In this way, the current relative distance between the target obstacle and the target vehicle can be determined based on the current position of the target vehicle and the current position of the target obstacle. Combined with the current driving direction of the target vehicle, the distance information between the target vehicle and the target obstacle is determined. The distance information can be used to predict whether the target vehicle will collide with the target obstacle when it continues to travel in the original direction and at the original speed. In this way, the travel of the target vehicle can be controlled based on the distance information.

[0107] Exemplarily, the target vehicle can be controlled to travel according to the distance information and a preset safety level. For example, if the safety distance level to which the distance information belongs is low, emergency braking can be performed. If the safety distance level to which the distance information belongs is high, the vehicle can decelerate in the original direction.

[0108] In the disclosed embodiment, the position changes of the target obstacle in the multi-frame point cloud images can be tracked together through the multi-frame point cloud images. In this way, the accuracy of the confidence that the target obstacle appears at the current position is improved, so that when the vehicle is controlled based on the confidence, effective control of the target vehicle is achieved. For example, frequent stops or collisions due to false detection of target obstacles can be avoided.

[0109] In order to improve the accuracy of the confidence, the confidence proposed in the embodiment of the present disclosure is determined according to at least two of the following parameters: average detection confidence, tracking matching confidence, tracking chain effective length, velocity smoothness, and acceleration smoothness;

[0110] Among them, the average detection confidence indicates the average reliability of the position of the target obstacle detected in each frame of point cloud images during the detection process of multi-frame point cloud images; the tracking matching confidence can indicate the matching degree between the detected target obstacle and the tracking chain, and the tracking chain can be a continuous multi-frame point cloud image; the effective length of the tracking chain can indicate the number of frames in which the target obstacle is detected in the continuous multi-frame point cloud image; the speed smoothness can indicate the speed change degree of the target obstacle in the time period corresponding to the continuous multi-frame point cloud image; the acceleration smoothness can indicate the acceleration change degree of the target obstacle in the time period corresponding to the continuous multi-frame point cloud image.

[0111] When determining the confidence of the obstacle based on the above parameters, each parameter is positively correlated with the confidence. The embodiment of the present disclosure proposes to determine the confidence of the current position of the target obstacle based on at least two of the above parameters. The confidence of the target obstacle at the current position is determined by multiple parameters, thereby improving the accuracy of the confidence of the target obstacle at the current position.

[0112] Specifically, when determining the confidence level of the target obstacle, the following may be included:

[0113] After weighted summing or multiplying at least two parameters, the confidence level of the target obstacle is obtained.

[0114] When weighted summation is performed based on the above at least two parameters, the confidence of the target obstacle can be determined according to the following formula (1):

[0115]

[0116] Where i represents a variable, i∈(1,n); n represents the total number of parameters, w i represents the preset weight of the i-th parameter, represents the parameter value of the i-th parameter of the target obstacle numbered j; C j Indicates the confidence of the target obstacle numbered j. When the point cloud image contains only one target obstacle, j is 1.

[0117] Exemplarily, the preset weight corresponding to each parameter can be set in advance, for example, by using big data statistics to determine in advance the importance of each parameter on the confidence level.

[0118] In another embodiment, when the at least two parameters are multiplied, the confidence level of the target obstacle may be determined according to the following formula (2):

[0119]

[0120] In the disclosed embodiment, it is proposed to jointly determine the confidence of the target obstacle at the current position by using multiple parameters. In this way, when the confidence of the target obstacle is determined from multiple angles, the accuracy of the confidence corresponding to the determined target obstacle at the current position can be improved.

[0121] The determination processes of the above-mentioned various parameters are described below respectively.

[0122] In one implementation, the average detection confidence may be determined as follows:

[0123] According to the detection confidence of the target obstacle appearing in each frame of the point cloud image, the average detection confidence corresponding to the target obstacle is determined.

[0124] Specifically, each frame of point cloud image can be input into a pre-trained neural network for detecting and tracking obstacles. The neural network includes a first module for detecting the position of obstacles in each frame of point cloud image, and a second module for tracking target obstacles. After each frame of point cloud image is input into the neural network, a detection frame representing the position of the target obstacle in the frame of point cloud image and the detection confidence of the detection frame can be obtained through the first module, and the number of the obstacle contained in each frame of point cloud image can be determined through the second module, thereby determining the target obstacle.

[0125] Specifically, the second module in the neural network can perform similarity detection on obstacles contained in continuously input point cloud images, determine the same obstacles in different frames of point cloud images, and number the obstacles contained in each frame of point cloud images. The same obstacle has the same number in different frames of point cloud images, so that the target obstacle can be determined in different frames of point cloud images.

[0126] Furthermore, after obtaining the detection confidence corresponding to the target obstacle in each frame of the point cloud image, the average detection confidence corresponding to the target obstacle can be determined according to the following formula (3):

[0127]

[0128] in, represents the average detection confidence of the target obstacle numbered j; L represents the number of frames of the multi-frame point cloud image, Represents the detection confidence of the target obstacle numbered j in the tth frame point cloud image in the continuous multi-frame point cloud images.

[0129] L can be a set number of frames. For example, if L=10 is pre-set, it means that 10 consecutive frames of point cloud images are detected, and t=1 represents the first frame of the 10 frames of point cloud images. During the driving process of the target vehicle, as the number of collected point cloud images gradually increases, the 10 consecutive frames of point cloud images here also change dynamically. t=L is the current frame of the point cloud image, and t=1 represents the first frame of the 10 consecutive frames of point cloud images including the current frame of the point cloud image and the 9 frames of point cloud images collected in the historical stage.

[0130] In particular, when the number of frames of the point cloud image collected by the radar device during this working process does not reach the set number of frames, L is the total number of frames collected from the start of collection to the current moment. For example, if the set number of frames is 10, and the point cloud image collected at the current moment is the 7th frame of the point cloud image collected by the radar device during this working process, when determining the confidence of the target obstacle at the current position, L here is equal to 7; when the number of frames of the point cloud image collected by the radar device during this working process reaches the set number of frames, L here is always equal to the set number of frames. The current working process of the radar device refers to the process of the radar device starting to collect point cloud images this time.

[0131] In particular, when collecting point cloud images at set time intervals, each moment corresponds to a frame of point cloud image. Therefore, the above t=1 can also be expressed as the point cloud image corresponding to the first collection moment within the collection time corresponding to multiple consecutive frames of point cloud images. The first collection moment here changes dynamically and is not the starting moment of the radar device in this working process.

[0132] In an embodiment of the present disclosure, it is proposed that the parameters for determining the confidence of a target obstacle include an average detection confidence, which can reflect the average reliability of the position of the target obstacle in multiple frame point cloud images. When the confidence of the target obstacle is determined based on the average detection confidence, the stability of the confidence of the determined target obstacle can be improved.

[0133] In a possible implementation, the tracking match confidence is determined in the following manner:

[0134] Based on the position information of the target obstacle in each frame of the point cloud image, the tracking matching confidence that the target obstacle is a tracking object matched by multiple frames of point cloud images is determined.

[0135] The position information of the target obstacle in each frame of the point cloud image can be determined by a pre-trained neural network. After each frame of the point cloud image is input into the neural network, the position information of the detection box representing the target obstacle in the frame of the point cloud image can be detected.

[0136] Considering that the multi-frame point cloud images are collected by the radar device at a set time interval, the time interval between two adjacent frames of point cloud images in the multi-frame point cloud images is short. In a short time, the displacement change degree of the same target obstacle is generally less than a certain range. Based on this, the tracking matching confidence of the target obstacle as the tracking object matched by the multi-frame point cloud images can be determined.

[0137] Specifically, when multiple consecutive frames of point cloud images all contain the same tracking object, the multiple consecutive frames of point cloud graphics can be used as a tracking chain for the tracking object. The change in position information of the tracking object in two adjacent frames of point cloud graphics in the tracking chain should be less than a preset range. Based on this, it is possible to determine whether the tracked target obstacle is the tracking object matched by the tracking chain according to the position information of the target obstacle in each frame of point cloud image, or to determine whether the target obstacles in the tracking chain are the same target obstacle. For example, the tracking chain contains 10 frames of point cloud images. For the target obstacle numbered 1, it is possible to determine whether the target obstacle coded as 1 in the tracking chain is the same target obstacle according to the position information of the target obstacle numbered 1 in each frame of point cloud image, that is, to determine whether the target obstacle is the tracking object matched by the tracking chain. The tracking matching confidence here can be used to indicate the matching degree between the target obstacle numbered 1 and the tracking chain. The higher the matching degree, the greater the possibility that the target obstacle is the tracking object matched by the tracking chain. Conversely, the smaller the possibility that the target obstacle is the tracking object matched by the tracking chain.

[0138] In the disclosed embodiment, the tracking matching confidence is used to represent the possibility of the target obstacle appearing in the continuous multi-frame point cloud images. If the possibility of the target obstacle appearing in the continuous multi-frame point cloud images is greater, it means that the possibility of the target obstacle being a false detection result is smaller. Based on this, the tracking matching confidence of the target obstacle and the tracking chain can be used as a parameter for determining the confidence of the target obstacle to improve the accuracy of the confidence.

[0139] Specifically, when determining the tracking matching confidence of the target obstacle as the tracking object matched by the multiple frame point cloud images based on the position information of the target obstacle in each frame point cloud image, such as Figure 2 As shown, the following steps S201 to S205 may be included:

[0140] S201, for each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image of the frame of point cloud image, determine the predicted position information of the target obstacle in the frame of point cloud image; based on the predicted position information and the position information of the target obstacle in the frame of point cloud image, determine the displacement deviation information of the target obstacle in the frame of point cloud image.

[0141] According to the method of determining the position information of the target obstacle in each frame of the point cloud image mentioned above, the position information of the target obstacle in each frame of the point cloud image can be determined. Specifically, the position information of the center point of the detection box representing the target obstacle in each frame of the point cloud image can be used as the position information of the target obstacle in the frame of the point cloud image.

[0142] If two frames of point cloud images are known, such as the time interval between the nth frame and the n+1th frame of point cloud images, the speed of the target obstacle at the time of acquisition of the nth frame of point cloud image, and the position information of the target obstacle in the nth frame of point cloud image, the predicted position information of the target obstacle in the n+1th frame of point cloud image can be predicted.

[0143] Furthermore, based on the predicted position information of the target obstacle and the position information of the target obstacle in the frame point cloud image, the displacement deviation information of the target obstacle in the frame point cloud image can be determined, and the displacement deviation information can be used as one of the parameters to measure whether the target obstacle matches the tracking chain.

[0144] Specifically, with respect to the above S201, when determining the predicted position information of the target obstacle in the frame of point cloud image based on the position information of the target obstacle in the previous frame of point cloud image, such as Figure 3 As shown, the following S2011 to S2012 may be included:

[0145] S2011, for each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image of the frame of point cloud image, the position information of the target obstacle in the previous frame of point cloud image, and the acquisition time interval between two adjacent frames of point cloud image, determine the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image;

[0146] S2012, determining the predicted position information of the target obstacle in the frame of point cloud image based on the position information of the target obstacle in the previous frame of point cloud image, the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image, and the acquisition time interval between the current frame of point cloud image and the previous frame of point cloud image.

[0147] Specifically, for each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image of the frame of point cloud image (specifically referring to the position information of the center point of the detection frame), the position information of the target obstacle in the previous frame of point cloud image of the previous frame of point cloud image (specifically referring to the position information of the center point of the detection frame), and the acquisition time interval between two adjacent frames of point cloud images, the average speed of the target obstacle within the acquisition time interval between the two adjacent frames of point cloud images can be determined, and the average speed is used as the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image.

[0148] Furthermore, taking the acquisition time corresponding to the frame point cloud image as the time corresponding to the acquisition of the t-th frame point cloud image in the continuous multi-frame point cloud image as an example, when determining the predicted position information of the target obstacle in the frame point cloud image, it can be determined according to the following formula (4):

[0149]

[0150] in, Represents the predicted position information of the target obstacle numbered j in the tth frame point cloud image among the continuous multi-frame point cloud images; Represents the position information of the target obstacle numbered j in the t-1th frame of the point cloud image in the continuous multi-frame point cloud image. It represents the speed of the target obstacle numbered j when collecting the t-1th frame of point cloud images in the continuous multi-frame point cloud images; Δt represents the time interval between collecting the tth frame of point cloud image and collecting the t-1th frame of point cloud image.

[0151] Furthermore, the displacement deviation information of the target obstacle in the point cloud image of the frame can be determined based on the following formula (5):

[0152]

[0153] in, Indicates the displacement deviation information of the target obstacle numbered j corresponding to the t-th frame point cloud image in the continuous multi-frame point cloud images; It represents the position information of the target obstacle numbered j in the tth frame point cloud image in the continuous multi-frame point cloud images; T represents the preset parameter.

[0154] S202, determining detection frame difference information corresponding to the target obstacle based on the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image and the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image before the frame point cloud image.

[0155] Similarly, if the time interval between two frames of point cloud images is short, the position information of the same target obstacle in these two frames of point cloud images should be relatively close. Therefore, the difference information of the detection boxes corresponding to the target obstacle in the two frames of point cloud images can be used as one of the parameters to measure whether the target obstacle matches the tracking chain.

[0156] Specifically, the area of ​​the detection box corresponding to the target obstacle numbered j in the t-1th frame of the continuous multi-frame point cloud image can be determined according to the following formula (6), the area of ​​the detection box corresponding to the target obstacle numbered j in the tth frame of the continuous multi-frame point cloud image can be determined according to the following formula (7), and the detection box difference information corresponding to the target obstacle numbered j in the tth frame of the continuous multi-frame point cloud image can be determined according to the following formula (8):

[0157]

[0158]

[0159]

[0160] in, Indicates the area of ​​the detection box corresponding to the target obstacle numbered j in the t-1th frame of the point cloud image in the continuous multi-frame point cloud image; Indicates the width of the detection box corresponding to the target obstacle numbered j in the t-1th frame of the continuous multi-frame point cloud image; Indicates the height of the detection box corresponding to the target obstacle numbered j in the t-1th frame of the continuous multi-frame point cloud image; Indicates the area of ​​the detection box corresponding to the target obstacle numbered j in the t-th frame of the continuous multi-frame point cloud image; Indicates the width of the detection box corresponding to the target obstacle numbered j in the t-th frame of the continuous multi-frame point cloud image; Indicates the height of the detection box corresponding to the target obstacle numbered j in the t-th frame of the continuous multi-frame point cloud image; Represents the detection box difference information corresponding to the t-th frame point cloud image of the target obstacle numbered j in the continuous multi-frame point cloud image.

[0161] S203: Determine orientation angle difference information corresponding to the target obstacle based on the orientation angle of the target obstacle in the point cloud image of the frame and the orientation angle of the target obstacle in the point cloud image of the previous frame.

[0162] Similarly, if the time interval between two frames of point cloud images is short, the orientation angles of the same target obstacle in these two frames of point cloud images should be relatively close. Therefore, the orientation angle difference information corresponding to the target obstacle in the two frames of point cloud images can be used as one of the parameters to measure whether the target obstacle matches the tracking chain.

[0163] Specifically, the orientation angle difference information corresponding to the target obstacle can be determined according to the following formula (9):

[0164]

[0165] in, represents the orientation angle difference information corresponding to the t-th frame point cloud image of the target obstacle numbered j in the continuous multi-frame point cloud images; represents the orientation angle corresponding to the t-th frame point cloud image of the target obstacle numbered j in the continuous multi-frame point cloud images; It represents the orientation angle of the target obstacle numbered j in the t-1th frame of the point cloud image in the continuous multi-frame point cloud images.

[0166] For example, the orientation angle corresponding to the t-th frame point cloud image of the target obstacle in the continuous multi-frame point cloud image specifically refers to the orientation angle of the target obstacle when the t-th frame point cloud image is collected. The orientation angle of the target obstacle in the point cloud image can be determined in the following manner:

[0167] First, a positive direction is set in the three-dimensional space, for example, the direction perpendicular to the ground and pointing to the sky is the positive direction, and then the angle formed by the positive direction and the line connecting the center point of the detection frame corresponding to the target obstacle in the point cloud image and the vehicle is taken as the orientation angle of the target obstacle in the point cloud image of this frame.

[0168] S204, determining the single-frame tracking matching confidence that the target obstacle is the tracking object matched by the frame point cloud image based on the displacement deviation information, the detection frame difference information and the orientation angle difference information.

[0169] For example, a weighted summation may be performed based on the displacement deviation information, the detection frame difference information, and the orientation angle difference information. For example, the above-obtained and By performing weighted summation, the single-frame tracking matching confidence of the target obstacle being the tracking object matched by the t-th frame point cloud image in the continuous multi-frame point cloud images can be obtained.

[0170] Specifically, the single-frame tracking matching confidence of the target obstacle being the tracking object matched by the t-th frame point cloud image in the continuous multi-frame point cloud images can be determined according to the following formula (10):

[0171]

[0172] Among them, p t j ′ represents the single-frame tracking matching confidence that the target obstacle numbered j is the tracking object matched by the t-th frame point cloud image in the continuous multi-frame point cloud images; w ΔL Represents the preset weight of displacement deviation information, w ΔD Indicates the preset weight of the detection box difference information; w ΔH Indicates the preset weight of the heading angle difference information.

[0173] According to the above method, the single-frame tracking matching confidence of the target obstacle being the tracking object matched by each frame of the point cloud image can be obtained.

[0174] Specifically, the single-frame tracking matching confidence level of the tracking object matched by each frame of point cloud graphics, where the target obstacle is a target obstacle, can indicate the reliability that the target obstacle in the point cloud image of this frame and the target obstacle in the point cloud image of the previous frame are the same obstacle.

[0175] For example, the preset tracking chain is 10 consecutive frames of point cloud images. For the second frame of point cloud image, the single-frame tracking matching confidence of the target obstacle being the tracking object matched by the second frame of point cloud image indicates the reliability that the target obstacle in the second frame of point cloud image and the target obstacle in the first frame of point cloud image are the same target obstacle. Similarly, for the third frame of point cloud image, the single-frame tracking matching confidence of the target obstacle being the tracking object matched by the third frame of point cloud graphics can indicate the reliability that the target obstacle in the third frame of point cloud image and the target obstacle in the second frame of point cloud image are the same target obstacle.

[0176] S205 , determining a tracking matching confidence level of the target obstacle being the tracking object matched by the multiple point cloud images according to a single-frame tracking matching confidence level of the tracking object matched by each point cloud image in the multiple point cloud images.

[0177] Specifically, the tracking matching confidence of the target obstacle being the tracking object matched by the multi-frame point cloud images can be determined according to the following formula (11):

[0178]

[0179] in, Indicates the tracking matching confidence of the target obstacle numbered j as the tracking object matched by the multi-frame point cloud images.

[0180] It can be determined from formula (11) that the tracking matching confidence corresponding to the target obstacle can be obtained by averaging the single-frame tracking matching confidence corresponding to the target obstacle.

[0181] In the disclosed embodiment, the parameters for determining the confidence of the target obstacle include tracking matching confidence, which can reflect the reliability of the target obstacle being a tracking object of the multi-frame point cloud image. In this way, when determining the confidence of the target obstacle based on the multi-frame point cloud image, taking this parameter into consideration can improve the accuracy of the confidence of the target obstacle.

[0182] In a possible implementation manner, when the at least two parameters include the effective length of the tracking chain, the effective length of the tracking chain may be determined in the following manner:

[0183] Based on the position information of the target obstacle in each frame of the point cloud image, the number of missed frames for the target obstacle in the multi-frame point cloud image is determined; and based on the total number of frames and the number of missed frames corresponding to the multi-frame point cloud image, the effective length of the tracking chain is determined.

[0184] Each frame of point cloud image is input into a pre-trained neural network. When the neural network operates normally, the position information of the target obstacle contained in the frame of point cloud image can be output. If the position information of the target obstacle contained in the frame of point cloud image is not output, the frame of point cloud image can be determined to be a missed point cloud image. In the embodiment of the present disclosure, multiple frames of point cloud images are point cloud images collected continuously in a short period of time. For the tracking chain containing multiple consecutive frames of point cloud images corresponding to the same target obstacle, when the target obstacle is included in the first frame of point cloud image and the last frame of point cloud image, each frame of point cloud image between the first frame of point cloud image and the last frame of point cloud image will generally also include the target obstacle. Therefore, if the neural network outputs a point cloud image that does not contain the position information of the target obstacle, it can be regarded as a missed point cloud image.

[0185] Specifically, the effective length of the tracking chain can be determined according to the following formula (12):

[0186]

[0187] in, represents the effective length of the tracking chain for the target obstacle numbered j; η represents the preset weight coefficient; L represents the number of frames of the multi-frame point cloud image; NL represents the number of missed frames.

[0188] In the disclosed embodiment, it is proposed to use the effective length of the tracking chain as a parameter for determining the confidence of the target obstacle. The accuracy of the neural network for detecting the target obstacle for each frame of the point cloud image is determined by the effective length of the tracking chain. Therefore, when the confidence of the target obstacle is determined based on the effective length of the tracking chain, the accuracy of the confidence can be improved.

[0189] In another possible implementation, when the at least two parameters include speed smoothness, such as Figure 4 As shown, the speed smoothness can be determined in the following manner, specifically including the following S401 to S402:

[0190] S401, determining a speed error of the target obstacle within a collection time corresponding to multiple frames of point cloud images based on the speed of the target obstacle at the collection time corresponding to each frame of point cloud images;

[0191] S402, based on the speed error corresponding to the target obstacle and the pre-stored standard deviation preset value, determine the speed smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images.

[0192] Exemplarily, a method similar to the Kalman filter algorithm may be used to determine speed errors corresponding to multiple speeds, and the speed errors may represent the noise of the speed of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images.

[0193] Specifically, the speed smoothness of the target obstacle within the acquisition time corresponding to the multi-frame point cloud image can be determined by the following formula (13):

[0194]

[0195] in, It represents the speed smoothness of the target obstacle numbered j within the acquisition time corresponding to the multi-frame point cloud image; σ represents the pre-stored standard deviation preset value; δv represents the speed error of the target obstacle within the acquisition time corresponding to the multi-frame point cloud image.

[0196] The speed smoothness corresponding to the target obstacle can indicate the speed smoothness of the target obstacle within the acquisition time corresponding to the multi-frame point cloud image. Because the speed is determined based on the position information of the target obstacle in two adjacent frames of point cloud images, the higher the speed smoothness, the smaller the displacement deviation change of the target obstacle in two adjacent frames of point cloud images, and thus the more accurate the detected position of the target obstacle.

[0197] In the disclosed embodiment, the speed smoothness can reflect the smoothness of the change in the speed of the target obstacle, and can reflect the position change of the target obstacle in the continuous multi-frame point cloud image, so that the reliability of the detected position information of the target obstacle can be reflected. Based on this, the speed smoothness can be used as a parameter for determining the confidence of the target obstacle to improve the accuracy of the confidence.

[0198] In another possible implementation manner, when the at least two parameters include acceleration smoothness, such as Figure 5 As shown, the acceleration smoothness can be determined in the following manner, specifically including the following S501 to S503:

[0199] S501, determining the acceleration of the target obstacle at the acquisition time corresponding to each frame of point cloud image based on the speed of the target obstacle at the acquisition time corresponding to the frame of point cloud image and the acquisition time interval between two adjacent frames of point cloud image;

[0200] S502, determining an acceleration error of the target obstacle within a collection time corresponding to multiple frames of point cloud images based on the acceleration of the target obstacle at the collection time corresponding to each frame of point cloud images;

[0201] S503, based on the acceleration error corresponding to the target obstacle and the pre-stored standard deviation preset value, determine the acceleration smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images.

[0202] Exemplarily, the method for determining the speed of the target obstacle at the acquisition time corresponding to each frame of point cloud image is detailed above and will not be repeated here. Furthermore, the acceleration of the target obstacle at the acquisition time corresponding to each frame of point cloud image can be determined based on the acquisition time interval between two adjacent frames of point cloud image and the speed of the target obstacle at the acquisition time corresponding to each frame of point cloud image.

[0203] For example, a method similar to the Kalman filter algorithm can also be used to determine the acceleration errors corresponding to multiple accelerations, and the acceleration errors can represent the noise of the acceleration of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images.

[0204] Specifically, the speed smoothness of the target obstacle within the acquisition time corresponding to the multi-frame point cloud image can be determined by the following formula (14):

[0205]

[0206] in, It represents the acceleration smoothness of the target obstacle numbered j within the acquisition time corresponding to the multi-frame point cloud image; σ represents the pre-stored standard deviation preset value; δa represents the acceleration error of the target obstacle within the acquisition time corresponding to the multi-frame point cloud image.

[0207] The acceleration smoothness corresponding to the target obstacle can indicate the acceleration smoothness of the target obstacle during the acquisition time corresponding to the multi-frame point cloud image. The higher the acceleration smoothness, the smoother the speed change of the target obstacle during the acquisition time corresponding to the continuous multi-frame point cloud image, and further, the more accurate the detected position of the target obstacle.

[0208] In the disclosed embodiment, the acceleration smoothness can reflect the smoothness of the change of the acceleration of the target obstacle, and can reflect the speed change of the target obstacle within the acquisition time corresponding to the continuous multi-frame point cloud image. It can also reflect the position change of the target obstacle in the continuous multi-frame point cloud image. In this way, the reliability of the detected position information of the target obstacle can be reflected. Based on this, the acceleration smoothness can be used as a parameter for determining the confidence of the target obstacle to improve the accuracy of the confidence.

[0209] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0210] Based on the same technical concept, a control device corresponding to the control method of the target vehicle is also provided in the embodiment of the present disclosure. Since the principle of solving the problem by the device in the embodiment of the present disclosure is similar to the above-mentioned control method in the embodiment of the present disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0211] Reference Figure 6 FIG. 6 is a schematic diagram of a control device 600 for a target vehicle provided by an embodiment of the present disclosure, and the control device includes:

[0212] An acquisition module 601 is used to acquire a plurality of point cloud images collected by a radar device during the driving process of the target vehicle;

[0213] The determination module 602 is used to perform obstacle detection on each frame of the point cloud image to determine the current position and confidence of the target obstacle;

[0214] The control module 603 is used to control the target vehicle to travel based on the determined current position and confidence of the target obstacle and the current position data of the target vehicle.

[0215] In a possible implementation, the confidence is determined based on at least two of the following parameters: average detection confidence, tracking match confidence, tracking chain effective length, velocity smoothness, and acceleration smoothness;

[0216] The determination module 602 is specifically used for:

[0217] After weighted summing or multiplying at least two parameters, the confidence level of the target obstacle is obtained.

[0218] In a possible implementation, the determination module 602 is further configured to determine the average detection confidence in the following manner:

[0219] According to the confidence that the target obstacle appears in each frame of the point cloud image, the average detection confidence corresponding to the target obstacle is determined.

[0220] In a possible implementation, the determination module 602 is further configured to determine the tracking match confidence in the following manner:

[0221] Based on the position information of the target obstacle in each frame of the point cloud image, the tracking matching confidence that the target obstacle is a tracking object matched by multiple frames of point cloud images is determined.

[0222] In a possible implementation, the determination module 602 is specifically configured to:

[0223] For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image of the frame of point cloud image, the predicted position information of the target obstacle in the frame of point cloud image is determined; based on the predicted position information and the position information of the target obstacle in the frame of point cloud image, the displacement deviation information of the target obstacle in the frame of point cloud image is determined;

[0224] Determine detection frame difference information corresponding to the target obstacle based on the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image and the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image before the frame point cloud image;

[0225] Determine the orientation angle difference information corresponding to the target obstacle based on the orientation angle of the target obstacle in the point cloud image of the frame and the orientation angle of the target obstacle in the point cloud image of the previous frame;

[0226] Based on the displacement deviation information, the detection frame difference information and the orientation angle difference information, the single frame tracking matching confidence level of the target obstacle being the tracking object matched by the frame point cloud image is determined;

[0227] According to the single-frame tracking matching confidence of the tracking object matched by each frame of the multi-frame point cloud image, the tracking matching confidence of the target obstacle as the tracking object matched by the multi-frame point cloud image is determined.

[0228] In a possible implementation, the determination module 602 is specifically configured to:

[0229] For each point cloud image, based on the position information of the target obstacle in the previous point cloud image, the position information of the target obstacle in the previous point cloud image, and the acquisition time interval between two adjacent point cloud images, determine the speed of the target obstacle at the acquisition time corresponding to the previous point cloud image.

[0230] Based on the position information of the target obstacle in the previous frame of point cloud image, the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image, and the acquisition time interval between the current frame of point cloud image and the previous frame of point cloud image, the predicted position information of the target obstacle in the current frame of point cloud image is determined.

[0231] In a possible implementation, the determination module 602 is further configured to determine the effective length of the tracking chain in the following manner:

[0232] Based on the position information of the target obstacle in each frame of the point cloud image, the number of missed frames for the target obstacle in the multi-frame point cloud image is determined; and based on the total number of frames and the number of missed frames corresponding to the multi-frame point cloud image, the effective length of the tracking chain is determined.

[0233] In a possible implementation, the determination module 602 is further configured to determine the speed smoothness in the following manner:

[0234] Based on the speed of the target obstacle at the acquisition time corresponding to each frame of the point cloud image, determine the speed error of the target obstacle within the acquisition time corresponding to the multiple frames of the point cloud image;

[0235] Based on the speed error corresponding to the target obstacle and the pre-stored standard deviation preset value, the speed smoothness of the target obstacle within the acquisition time corresponding to the multi-frame point cloud image is determined.

[0236] In a possible implementation, the determination module 602 is further configured to determine the acceleration smoothness in the following manner:

[0237] Based on the speed of the target obstacle at the acquisition time corresponding to each frame of point cloud image and the acquisition time interval between two adjacent frames of point cloud image, the acceleration of the target obstacle at the acquisition time corresponding to the frame of point cloud image is determined;

[0238] Based on the acceleration of the target obstacle at the acquisition time corresponding to each frame of the point cloud image, determine the acceleration error of the target obstacle within the acquisition time corresponding to the multiple frames of the point cloud image;

[0239] Based on the acceleration error corresponding to the target obstacle and the pre-stored standard deviation preset value, the acceleration smoothness of the target obstacle within the acquisition time corresponding to the multi-frame point cloud image is determined.

[0240] In a possible implementation, the control module 603 is specifically configured to:

[0241] When it is determined that the confidence level corresponding to the target obstacle is higher than a preset confidence threshold, the distance information between the target vehicle and the target obstacle is determined based on the current position of the target obstacle and the current position and posture data of the target vehicle;

[0242] The target vehicle is controlled to travel based on the distance information.

[0243] For descriptions of the processing flow of each module in the device and the interaction flow between each module, reference may be made to the relevant descriptions in the above method embodiment, which will not be described in detail here.

[0244] Corresponds to Figure 1 The control method of the target vehicle in the present disclosure also provides an electronic device 700, such as Figure 7 FIG. 7 is a schematic diagram of the structure of an electronic device 700 provided in an embodiment of the present disclosure, including:

[0245] Processor 71, memory 72, and bus 73; memory 72 is used to store execution instructions, including internal memory 721 and external memory 722; the internal memory 721 here is also called internal memory, which is used to temporarily store the calculation data in the processor 71, and the data exchanged with the external memory 722 such as a hard disk. The processor 71 exchanges data with the external memory 722 through the internal memory 721. When the electronic device 700 is running, the processor 71 communicates with the memory 72 through the bus 73, so that the processor 71 executes the following instructions: during the driving process of the target vehicle, obtain multiple frames of point cloud images collected by the radar device; perform obstacle detection on each frame of the point cloud image to determine the current position and confidence of the target obstacle; based on the determined current position and confidence of the target obstacle, and the current posture data of the target vehicle, control the driving of the target vehicle.

[0246] The present disclosure also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the target vehicle control method described in the above method embodiment are executed. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0247] The computer program product of the target vehicle control method provided in the embodiment of the present disclosure includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the steps of the target vehicle control method described in the above method embodiment. Please refer to the above method embodiment for details, which will not be repeated here.

[0248] The present disclosure also provides a computer program, which implements any one of the methods of the aforementioned embodiments when executed by a processor. The computer program product can be implemented in hardware, software, or a combination thereof. In an optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK), etc.

[0249] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the device or unit can be electrical, mechanical or other forms.

[0250] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0251] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0252] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0253] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for controlling a target vehicle, It is characterized in that The control method comprises: When the target vehicle is driving, a multi-frame point cloud image collected by a radar device is obtained; Obstacle detection is performed on each frame of point cloud image to determine the current position of the target obstacle, and the confidence of the target obstacle is obtained by weighted summing or multiplying at least two of the following parameters; the parameters include: average detection confidence, tracking matching confidence, effective length of tracking chain, speed smoothness and acceleration smoothness; the average detection confidence represents the average reliability of the position of the target obstacle corresponding to each frame of point cloud image detected during the detection process of the multiple frames of point cloud images; the tracking matching confidence represents the matching degree between the detected target obstacle and the tracking chain, and the tracking chain is a continuous multiple-frame point cloud image; the effective length of the tracking chain represents the number of frames in which the target obstacle is detected in the continuous multiple-frame point cloud image; the speed smoothness represents the speed change degree of the target obstacle in the time period corresponding to the continuous multiple-frame point cloud image; the acceleration smoothness represents the acceleration change degree of the target obstacle in the time period corresponding to the continuous multiple-frame point cloud image; Based on the determined current position and confidence of the target obstacle and the current position data of the target vehicle, the target vehicle is controlled to travel.

2. The control method according to claim 1, It is characterized in that The average detection confidence is determined as follows: According to the detection confidence of the target obstacle appearing in each frame of the point cloud image, the average detection confidence corresponding to the target obstacle is determined.

3. The control method according to claim 1, It is characterized in that The tracking match confidence is determined as follows: Based on the position information of the target obstacle in each frame of the point cloud image, a tracking matching confidence that the target obstacle is a tracking object matched by the multiple frames of point cloud images is determined.

4. The control method according to claim 3, It is characterized in that The step of determining the tracking matching confidence that the target obstacle is a tracking object matched by the multiple point cloud images based on the position information of the target obstacle in each frame of the point cloud image comprises: For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image of the frame of point cloud image, the predicted position information of the target obstacle in the frame of point cloud image is determined; based on the predicted position information and the position information of the target obstacle in the frame of point cloud image, the displacement deviation information of the target obstacle in the frame of point cloud image is determined; Determine detection frame difference information corresponding to the target obstacle based on the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image and the area of ​​the detection frame representing the position information of the target obstacle in the frame point cloud image before the frame point cloud image; Determine the orientation angle difference information corresponding to the target obstacle based on the orientation angle of the target obstacle in the frame point cloud image and the orientation angle of the target obstacle in the previous frame point cloud image; Determine, based on the displacement deviation information, the detection frame difference information, and the orientation angle difference information, a single-frame tracking matching confidence that the target obstacle is a tracking object matched by the frame point cloud image; According to the single-frame tracking matching confidence of the target obstacle being the tracking object matched by each frame of the multi-frame point cloud image, the tracking matching confidence of the target obstacle being the tracking object matched by the multi-frame point cloud image is determined.

5. The control method according to claim 4, It is characterized in that For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image, determining the predicted position information of the target obstacle in the frame of point cloud image, including: For each frame of point cloud image, based on the position information of the target obstacle in the previous frame of point cloud image, the position information of the target obstacle in the previous frame of point cloud image, and the acquisition time interval between two adjacent frames of point cloud image, determine the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image; Based on the position information of the target obstacle in the previous frame of point cloud image, the speed of the target obstacle at the acquisition time corresponding to the previous frame of point cloud image, and the acquisition time interval between the current frame of point cloud image and the previous frame of point cloud image, the predicted position information of the target obstacle in the frame of point cloud image is determined.

6. The control method according to claim 1, It is characterized in that The effective length of the tracking chain is determined as follows: Based on the position information of the target obstacle in each frame of the point cloud image, the number of missed detection frames for the target obstacle in the multiple frames of point cloud images is determined; and based on the total number of frames corresponding to the multiple frames of point cloud images and the number of missed detection frames, the effective length of the tracking chain is determined.

7. The control method according to claim 1, It is characterized in that The velocity smoothness is determined as follows: Determine a speed error of the target obstacle within a collection time corresponding to the multiple frames of point cloud images based on the speed of the target obstacle at the collection time corresponding to each frame of point cloud images; Based on the speed error corresponding to the target obstacle and a pre-stored standard deviation preset value, the speed smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images is determined.

8. The control method according to claim 1, It is characterized in that The acceleration smoothness is determined as follows: Determine the acceleration of the target obstacle at the acquisition time corresponding to each frame of the point cloud image based on the speed of the target obstacle at the acquisition time corresponding to each frame of the point cloud image and the acquisition time interval between two adjacent frames of the point cloud image; Determine the acceleration error of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images based on the acceleration of the target obstacle at the acquisition time corresponding to each frame of point cloud images; Based on the acceleration error corresponding to the target obstacle and a pre-stored standard deviation preset value, the acceleration smoothness of the target obstacle within the acquisition time corresponding to the multiple frames of point cloud images is determined.

9. The control method according to any one of claims 1 to 8, It is characterized in that The controlling the target vehicle to travel based on the determined current position and confidence of the target obstacle and the current position data of the target vehicle comprises: When it is determined that the confidence level corresponding to the target obstacle is higher than a preset confidence threshold, determining the distance information between the target vehicle and the target obstacle based on the current position of the target obstacle and the current position data of the target vehicle; The target vehicle is controlled to travel based on the distance information.

10. A control device for a target vehicle, It is characterized in that The control device comprises: An acquisition module is used to acquire a multi-frame point cloud image collected by a radar device during the driving process of the target vehicle; A determination module, used to perform obstacle detection on each frame of point cloud image, determine the current position of the target obstacle, and obtain the confidence of the target obstacle by weighted summing or multiplying at least two of the following parameters; the parameters include: average detection confidence, tracking matching confidence, effective length of tracking chain, speed smoothness and acceleration smoothness; the average detection confidence represents the average reliability of the position of the target obstacle corresponding to each frame of point cloud image detected during the detection process of the multiple frames of point cloud images; the tracking matching confidence represents the matching degree between the detected target obstacle and the tracking chain, and the tracking chain is a continuous multiple-frame point cloud image; the effective length of the tracking chain represents the number of frames in which the target obstacle is detected in the continuous multiple-frame point cloud image; the speed smoothness represents the speed change degree of the target obstacle in the time period corresponding to the continuous multiple-frame point cloud image; the acceleration smoothness represents the acceleration change degree of the target obstacle in the time period corresponding to the continuous multiple-frame point cloud image; A control module is used to control the target vehicle to travel based on the determined current position and confidence of the target obstacle and the current posture data of the target vehicle.

11. An electronic device, It is characterized in that include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the control method as described in any one of claims 1 to 9 are performed.

12. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the control method according to any one of claims 1 to 9 are executed.

Images