A method and device for processing vehicle sensor parameters

Abstract

An embodiment of the present application provides a method and device for processing vehicle sensor parameters, which relates to the field of autonomous driving technology. The method can be applied to business scenarios such as ports, ports, road freight, urban distribution, mines, airports, etc. The method includes: obtaining first data collected by a first sensor and second data collected by at least two second sensors, obtaining target data according to the at least two second data, obtaining a cost value according to the target data and true value data, the cost value being used to represent the matching relationship between the target data and the true value data, and processing the first parameter corresponding to each second data according to the cost value. By collecting data on the same scenario through the first sensor and the second sensor, and using the true value data as a standard, the parameters are processed according to the matching cost between the target data and the true value data, so that the adjustment and calibration of the parameters can be completed quickly and accurately without relying on the testing of testers.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of intelligent driving technology, and in particular, to a method and device for processing vehicle sensor parameters. Background Art

[0002] The planning and control of an autonomous driving system rely on the results provided by the perception system. The process of the perception system mainly includes that each sensor provides raw data, the perception algorithm provides perception results, and the fusion algorithm fuses multiple perception results to provide the perception fusion results.

[0003] Currently, for the adjustment of the fusion algorithm parameters, the initial parameters are mainly set manually, and through actual on-vehicle testing, the initial parameters are adjusted according to the test results, and the test process is repeated until the test results are accurate.

[0004] However, the prior art has disadvantages such as a large workload in the whole iteration process, low iteration efficiency, and strong dependence on the test scenario. Summary of the Invention

[0005] The embodiments of the present application provide a method and device for processing vehicle sensor parameters, which can improve the efficiency of adjusting vehicle sensor parameters and reduce the dependence on the test scenario.

[0006] In a first aspect, the embodiments of the present application provide a method for processing vehicle sensor parameters, including:

[0007] Obtain first data collected by a first sensor and second data collected by at least two second sensors;

[0008] Obtain target data according to the at least two second data, where the target data is data obtained by performing a fusion process on the at least two second data;

[0009] Obtain a cost value according to the target data and the true value data, where the cost value is used to represent the matching relationship between the target data and the true value data, and the true value data is data obtained by processing the first data;

[0010] Process a first parameter corresponding to each second data according to the cost value, where the first parameter is the fusion weight of each second data in the fusion algorithm. By collecting data of the same scene through the first sensor and the second sensors, taking the true value data as the standard, and processing the parameters of the fusion algorithm according to the matching cost between the target data and the true value data, the adjustment and calibration of the parameters of the fusion algorithm can be completed quickly and accurately without relying on the testing of testers.

[0011] Optionally, both the first sensor and the second sensor are used to detect obstacles. The obtaining of the cost value according to the target data and the true value data includes:

[0012] Matching the first obstacle in the target data with the second obstacle in the true value data;

[0013] Obtaining the matching information between the first obstacle and the second obstacle, where the matching information includes at least one of the number of matches, the matching cost, or the matching weight;

[0014] Obtaining the non-matching information between the first obstacle and the second obstacle, where the non-matching information includes at least one of the number of non-matches, the non-matching cost, or the non-matching weight;

[0015] Obtaining the cost value according to the matching information and the non-matching information.

[0016] Optionally, if the matching information includes the matching cost, the obtaining of the matching information between the first obstacle and the second obstacle includes:

[0017] Obtaining the initial weight of each second data;

[0018] Performing normalization processing on the initial weight to obtain the weight after normalization processing;

[0019] Obtaining the matching cost according to the true value data, the target data, and the weight after normalization processing;

[0020] If the non-matching information includes the non-matching cost, the obtaining of the non-matching information between the first obstacle and the second obstacle includes:

[0021] Obtaining the non-matching cost according to the target data, the weight after normalization processing, and the second data.

[0022] Optionally, the obtaining of the target data according to at least two second data includes:

[0023] Obtaining the initial weight of each second data;

[0024] Obtaining the target data through an attribute processing function according to the initial weight and the second data, where the attribute processing function is used to represent the constraint relationship between the target data and the second data.

[0025] Optionally, the processing of the first parameter corresponding to each second data according to the cost value includes:

[0026] If the cost value meets the preset condition, the parameter processing of the fusion algorithm is completed, where the preset condition is that the cost value is less than the preset value, or after N adjustments, the corresponding P cost values are all greater than M% of the optimal value within the N adjustment periods, and the optimal value is the value with the smallest difference from the preset value, and M, N, and P are integers greater than 1;

[0027] If the cost value is greater than the preset value, set a new first parameter, re-obtain the target data, and obtain the cost value according to the re-obtained target data and the true value data.

[0028] Optionally, the setting of a new first parameter, re-obtaining the target data, and obtaining the cost value according to the re-obtained target data and the true value data includes:

[0029] Based on the initial weight, taking the range of increasing or decreasing by K% as the value range and L% as the step size for value taking, to obtain corresponding multiple groups of weights, where K and L are integers greater than 1;

[0030] Take each group of weights in the multiple groups of weights as a new first parameter, and obtain the cost value corresponding to each new first parameter;

[0031] If the optimal value among the cost values corresponding to each new first parameter does not meet the preset condition, take the weight corresponding to the optimal value as the new initial weight;

[0032] Repeat the process of obtaining multiple cost values according to the new initial weight until the optimal value among the multiple cost values meets the preset condition.

[0033] Optionally, after obtaining the first data collected by the first sensor and the second data collected by at least two second sensors, the method further includes:

[0034] If the collection times of the first data and the second data are not synchronized, perform time synchronization on the first data and the second data by the difference method. This can improve the accuracy of matching the first data and the second data.

[0035] In a second aspect, an embodiment of the present application provides a vehicle sensor parameter processing device, including:

[0036] A first acquisition module, configured to acquire first data collected by a first sensor and second data collected by at least two second sensors;

[0037] A second acquisition module, configured to obtain target data according to at least two second data, where the target data is data obtained by performing fusion processing on the at least two second data;

[0038] A third acquisition module, configured to acquire a cost value according to the target data and the true value data, where the cost value is used to represent the matching relationship between the target data and the true value data, and the true value data is data obtained by processing the first data;

[0039] A processing module, configured to process the first parameter corresponding to each second data according to the cost value, where the first parameter is the fusion weight of each second data in the fusion algorithm.

[0040] Optionally, the vehicle sensor parameter processing device is used to implement any possible parameter processing method in the first aspect.

[0041] In a third aspect, the present application provides an electronic device, including: a memory and a processor;

[0042] The memory is used to store computer instructions; the processor is used to run the computer instructions stored in the memory to implement the method in any item of the first aspect.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the method in any item of the first aspect.

[0044] In a fifth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the method in any item of the first aspect.

[0045] In a sixth aspect, the present application provides a chip or a chip system, where the chip or the chip system includes at least one processor and a communication interface, the communication interface and the at least one processor are interconnected by a line, and the at least one processor is used to run a computer program or instruction to execute the parameter processing method of the fusion algorithm described in the possible implementation manners of the first aspect. Among them, the communication interface in the chip may be an input / output interface, a pin, a circuit, etc.

[0046] In a possible implementation, the chip or the chip system described above in the present application further includes at least one memory, and instructions are stored in the at least one memory. The memory may be an internal storage unit of the chip, such as a register, a cache, etc., or a storage unit of the chip (such as a read-only memory, a random access memory, etc.). Description of the Drawings

[0047] Figure 1 It is a schematic diagram of the scenario provided by the embodiment of the present application;

[0048] Figure 2 It is a schematic flow chart of the vehicle sensor parameter processing method provided by the embodiment of the present application Figure 1 ;

[0049] Figure 3 Flow schematic of the vehicle sensor parameter processing method provided by the embodiment of the present application Figure 2 ;

[0050] Figure 4 Structural schematic diagram of the vehicle sensor parameter processing device provided by the embodiment of the present application;

[0051] Figure 5 Structural schematic diagram of the vehicle sensor parameter processing electronic device provided by the embodiment of the present application. Specific embodiments

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application.

[0053] First, a brief introduction to the technical terms in the embodiments of the present application is given:

[0054] 1) The true value system is a system that can obtain a large amount of data by arranging high-precision and high-performance sensors on a test vehicle, and through complex and time-consuming neural network algorithms, it can obtain true data values close to the actual road obstacle situation.

[0055] 2) The operational design domain (ODD) refers to the operating conditions specifically designed for a specific driving automation system or its functions, including but not limited to environmental, geographical, and time restrictions, and / or the presence or absence of certain traffic or road characteristics. Simply put, ODD is to define under which working conditions autonomous driving can be achieved. Without these working conditions, autonomous driving cannot guarantee normal operation. Any autonomous driving vehicle must have certain limited working conditions. And this working condition can be very broad or very precise, and determines what scenarios the autonomous driving vehicle can handle. For example, the autonomous driving system of a vehicle can only be used on the highway, and it can automatically maintain the lane, overtake automatically, follow the vehicle automatically, give way automatically, enter and exit the ramp automatically, etc., but it cannot fully perform autonomous driving in the city. At the same time, to ensure the integrity of autonomous driving testing and verification, at least it is necessary to ensure that all aspects of ODD have been passed by ensuring the safe operation of the system, or by ensuring that the system can identify the range beyond ODD.

[0056] 3) Other terms

[0057] In the embodiments of the present application, terms such as "first" and "second" are used to distinguish identical or similar items with basically the same functions and effects, and their sequence is not limited. Those skilled in the art can understand that terms such as "first" and "second" do not limit the quantity and execution order, and "first", "second", etc. do not necessarily mean different.

[0058] It should be noted that in the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or having more advantages than other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific manner.

[0059] The vehicle sensor parameter processing method provided by the embodiments of the present application will be introduced in detail below with reference to the accompanying drawings. It should be noted that "when... " in the embodiments of the present application can be at the instant when a certain situation occurs, or within a period of time after a certain situation occurs. The embodiments of the present application do not make specific limitations on this.

[0060] The planning and control of the autonomous driving system rely on the results provided by the perception system. The process of the perception system mainly involves each sensor providing raw data. After the fusion algorithm processes and fuses the raw data, it provides the results of perception fusion.

[0061] Since each sensor cannot completely and correctly perceive the obstacle information in the traffic environment, it is very crucial to perform multi-sensor fusion to provide complementary fusion results by leveraging the strengths and compensating for the weaknesses.

[0062] Currently, the main method in the industry is to manually set the parameters of the fusion algorithm based on the results provided by the perception system, and then install this fusion algorithm on the vehicle for debugging. The overall usage effect is evaluated according to the results of actual road tests. When the perception results of each sensor are correct, if incorrect planning and control occur, it is usually due to some incorrect parameters in the fusion algorithm. The tester adjusts the parameters in the fusion algorithm according to the test results, and then gets back in the vehicle for testing, repeating the above process until the test results are fully accepted.

[0063] However, in the above process, the iterative process of algorithm adjustment has a large workload, a long iteration cycle, low efficiency, a strong dependence on test scenarios, and weak support for the adjustment of multiple parameters.

[0064] In view of this, the embodiments of the present application provide a vehicle sensor parameter processing method and device, which can quickly and accurately complete the adjustment and calibration of the parameters of the fusion algorithm without relying on the tests of testers by combining ground truth data.

[0065] The following uses specific embodiments to elaborate in detail on the technical solutions of this application and how the technical solutions of this application solve the above technical problems. These several specific embodiments below can be implemented independently or in combination with each other. For the same or similar concepts or processes, they may not be elaborated in some embodiments.

[0066] Figure 1 It is a schematic diagram of the application scenario of the embodiment of this application, including vehicle 10, ground truth system 11, perception system 12, and obstacle 13. Among them, vehicle 10 can be an autonomous vehicle. The ground truth system 11 and the perception system 12 are installed in vehicle 10, and the ground truth system 11 and the perception system 12 can include multiple sensors.

[0067] During the driving process of vehicle 10, multiple sensors in the perception system 12 can collect information about the obstacle 13 on the road. After being processed through a fusion algorithm, specific information about the obstacle 13 can be obtained, including but not limited to information such as position, speed, type, and size. The control system of vehicle 10 controls vehicle 10 and plans the driving path according to the specific information of obstacle 13.

[0068] Before the vehicle 10 leaves the factory, it is necessary to calibrate the parameters of the fusion algorithm in the perception system 12. In the embodiment of this application, the ground truth system 11 can be installed in vehicle 10, and the parameters of the fusion algorithm in the perception system 12 can be processed through the ground truth data of obstacle 13 output by the ground truth system 11 and the perception data of obstacle 13 output by the perception system 12.

[0069] The above briefly introduced the application scenario of the embodiment of this application. The following takes the vehicle applied to Figure 1 as an example to elaborate in detail on the vehicle sensor parameter processing method provided by the embodiment of this application.

[0070] Figure 2 It is a flowchart of the vehicle sensor parameter processing method provided by the embodiment of this application Figure 1 , including the following steps:

[0071] S201. Obtain the first data collected by the first sensor and the second data collected by at least two second sensors.

[0072] In the embodiment of this application, the first sensor is a ground truth system sensor, the second sensor is a perception system sensor, and the first data and the second data are data collected in the same scenario.

[0073] The ground truth system sensors include multiple high-precision and high-performance sensors. For example, radar sensors with an accuracy of 1 micron, etc. The perception system sensors are sensors for conventional use, including at least two of millimeter-wave radars, lidars, and cameras.

[0074] The perception system refers to a system used to provide specific information about obstacles to the vehicle control center, including multiple second sensors and a fusion algorithm.

[0075] A scenario refers to a scenario that may occur during the driving of a vehicle, such as scenarios of overtaking, lane changing, emergency braking, etc.

[0076] The first data and the second data refer to data obtained by respectively collecting obstacles in the same scenario through the first sensor and the second sensor. The first data and the second data may include data of multiple obstacles.

[0077] Optionally, when the vehicle collects obstacles, different data can be obtained according to different sensors. For example, point cloud data of obstacles can be obtained through a lidar, and image data of obstacles can be obtained through a camera. The embodiments of the present application do not limit the type of the collected data.

[0078] Optionally, during the data collection process, data of multiple traffic scenarios can be collected at one time to cover as many ODD scenarios as possible. Through tests in multiple different scenarios, the accuracy of parameter processing of the fusion algorithm can be improved.

[0079] The vehicle can obtain the first data and the second data of obstacles in the same scenario through the mounted first sensor and second sensor.

[0080] S202. Obtain target data according to at least two second data, where the target data is data obtained by performing fusion processing on at least two second data.

[0081] In the embodiments of the present application, a single second sensor cannot obtain all the information of an obstacle. To ensure that the obtained obstacle information is not incorrect, the second data collected by multiple second sensors needs to be fused to obtain all the information of the obstacle, that is, the target data. That is to say, the target data is a data set including all the information of the obstacles that the vehicle can collect. Among them, the fusion processing can be to perform superposition processing on multiple second data.

[0082] The fusion algorithm refers to an algorithm that can fuse multiple second data of an obstacle collected by multiple second sensors to obtain specific information of an obstacle, including but not limited to weighted average method, Kalman filtering method, neural network model method, etc. The embodiments of the present application do not limit the specific fusion algorithm.

[0083] The vehicle can obtain multiple second data of the obstacle through multiple second sensors in the perception system, and use a fusion algorithm to perform fusion processing on the multiple second data to obtain target data.

[0084] S203. Obtain a cost value according to the target data and the ground truth data, where the cost value is used to represent the matching relationship between the target data and the ground truth data.

[0085] In the embodiment of the present application, the ground truth data is the ground truth result obtained by processing the first data by the ground truth system.

[0086] The ground truth system is used to process the obstacle information collected by the ground truth system sensors through a neural network algorithm, and a ground truth result close to the actual obstacle situation can be obtained.

[0087] Matching refers to the process of finding data belonging to the same obstacle in the target data and the ground truth data. For the same obstacle, the relationship between the target data and the ground truth data can be divided into the following three cases:

[0088] There is data of the obstacle in both the target data and the ground truth data (match), there is data of the obstacle in the target data but there is no data of the obstacle in the ground truth data (redundant), and there is no data of the obstacle in the target data but there is data of the obstacle in the ground truth data (miss).

[0089] The cost value refers to the matching cost in the process of matching the target data and the ground truth data. For example, the time required to obtain the match, the resources required for the match, etc. The smaller the cost value, the higher the matching degree between the target data and the ground truth data, that is, the higher the similarity.

[0090] In the embodiment of the present application, matching algorithms such as a stereo matching algorithm and a disparity map method can be used to match the target data and the ground truth data to obtain the cost value of the matching between the target data and the ground truth data. The embodiment of the present application does not limit the matching algorithm.

[0091] The vehicle can match the target data and the ground truth data to obtain a cost value.

[0092] S204. Process the first parameter corresponding to each second data according to the cost value, where the first parameter is the fusion weight of each second data in the fusion algorithm.

[0093] In the embodiments of the present application, the data collected by each second sensor is usually error-free. Each second sensor describes the same obstacle at a different angle. In order to ensure that the target data does not deviate from the expectation during the process of obtaining the target data from multiple second data, when fusing the second data collected by multiple second sensors, a weight needs to be set for each second data to ensure that the obstacle can be described as accurately as possible through data fusion.

[0094] The weight corresponding to each second data can represent the importance of the second data when used to describe the obstacle. The higher the weight, the higher the importance of the corresponding second data.

[0095] Processing the parameters of the fusion algorithm means adjusting the parameters, that is, processing the first parameter corresponding to each second data respectively, such as decreasing or increasing the parameter, so that the target data obtained through the fusion algorithm is closer to the true value data.

[0096] Exemplarily, after obtaining the cost value based on the target data and the true value data, it can be compared with a pre-set cost threshold. If the cost value is less than the cost threshold, it means that the matching degree between the target data and the true value data is relatively high and meets the requirements. If the cost value is greater than the cost threshold, it means that the matching degree between the target data and the true value data is relatively low, and a parameter of the fusion algorithm needs to be re-set. New target data is obtained according to the new parameter of the fusion algorithm, and the cost value is obtained based on the new target data and the true value data. The above process is repeated until the cost value is less than the cost threshold.

[0097] Among them, the new parameter of the fusion algorithm can be obtained based on the initial parameter. For example, the initial parameter is increased or decreased to obtain the new parameter of the fusion algorithm.

[0098] The vehicle sensor parameter processing method provided by the embodiments of the present application obtains the first data collected by the first sensor and the second data collected by at least two second sensors, and obtains target data according to the at least two second data. The target data is the fusion data obtained by fusing the at least two second data. The cost value is obtained based on the target data and the true value data, and the cost value is used to represent the matching relationship between the target data and the true value data. The first parameter corresponding to each second data is processed according to the cost value. By collecting data of the same scene through the first sensor and the second sensor, and taking the true value data as the standard, processing the parameters of the fusion algorithm according to the matching cost between the target data and the true value data can quickly and accurately complete the adjustment and calibration of the vehicle sensor parameters without relying on the testing of testers.

[0099] Figure 3 It is a flow schematic of the vehicle sensor parameter processing method provided by the embodiments of the present application Figure 2 , inFigure 2 Based on the embodiments shown, the parameter processing method of the fusion algorithm will be further described. For example, Figure 3 as shown, it includes the following steps:

[0100] S301. Obtain the first data collected by the first sensor and the second data collected by at least two second sensors.

[0101] In the embodiments of the present application, the specific implementation manner shown in S301 is similar to the specific implementation manner of the embodiments shown Figure 2 and will not be elaborated here.

[0102] S302. Obtain the initial weight of each second data, and obtain the target data according to the initial weight and the second data.

[0103] In the embodiments of the present application, the second data includes the data of the obstacles collected by each sensor in terms of distance and speed from the vehicle.

[0104] Exemplarily, for a common 1R1V1L perception system in autonomous driving, the sensors of the perception system (second sensors) include a forward millimeter-wave radar, a forward camera, and a forward lidar.

[0105] Correspondingly, the second data collected by the second sensors includes four types of data: Dx_i, Dy_i, Vx_i, and Vx_i.

[0106] Among them, Dx_i represents the distance data of the obstacle collected by any sensor in the x direction following the x direction of the sensor, Dy_i represents the distance data of the obstacle collected by any sensor in the y direction following the y direction of the sensor, Vx_i represents the speed data of the obstacle collected by any sensor in the x direction following the x direction of the sensor, and Vx_i represents the speed data of the obstacle collected by any sensor in the y direction following the y direction of the sensor.

[0107] In the embodiments of the present application, after obtaining the second data, it is necessary to obtain the initial weight of each second data, and the initial weight can be set according to actual requirements. The initial weights of the second data include: W R_Dx 、W V_Dx 、W L_Dx 、W R_Dy 、W V_Dy 、W L_Dy 、W R_Vx 、W V_Vx 、W L_Vx 、W R_Vy 、W V_Vy 、W L_Vy 。

[0108] Among them, W R_Dxis the weight of the x-direction distance of the obstacle following the x-direction distance of the forward millimeter-wave radar, W V_Dx is the weight of the x-direction distance of the obstacle following the x-direction distance of the forward camera, W L_Dx is the weight of the x-direction distance of the obstacle following the x-direction distance of the forward lidar, W R_Dy is the weight of the y-direction distance of the obstacle following the y-direction distance of the forward millimeter-wave radar, W V_Dy is the weight of the y-direction distance of the obstacle following the y-direction distance of the forward camera, W L_Dy is the weight of the y-direction distance of the obstacle following the y-direction distance of the forward lidar, W R_Vx is the weight of the x-direction speed of the obstacle following the x-direction speed of the forward millimeter-wave radar, W V_Vx is the weight of the x-direction speed of the obstacle following the x-direction speed of the forward camera, W L_Vx is the weight of the x-direction speed of the obstacle following the x-direction speed of the forward lidar, W R_Vy is the weight of the x-direction speed of the obstacle following the y-direction speed of the forward millimeter-wave radar, W V_Vy is the weight of the x-direction speed of the obstacle following the y-direction distance of the forward camera, W L_Vy The x-direction speed of the obstacle follows the y-direction distance of the forward lidar.

[0109] Among them, the initial weights of the respective second data have the following relationships:

[0110] W R_Dx +W V_Dx +W L_Dx = 1

[0111] W R_Dy +W V_Dy +W L_Dy = 1

[0112] W R_Vx +W V_Vx +W L_Vx = 1

[0113] W R_Vy +W V_Vy +W L_Vy = 1

[0114] Therefore, the weights that need to be parameter-processed can be simplified to the following 8, W R_Dx 、W V_Dx 、W R_Dy 、W V_Dy 、W R_Vx 、W V_Vx 、W R_Vy 、W V_Vy , and the above weights can be briefly recorded as Wi_Dx , W i_Dy , W i_Vx , W i_Vy .

[0115] After obtaining the initial weights of the second data, the target data can be obtained through an attribute processing function based on the initial weights and the second data.

[0116] Exemplarily, the attribute processing function is as follows:

[0117]

[0118]

[0119]

[0120]

[0121] where is the target data, and N is the total number of frames of the second data.

[0122] S303. Obtain the matching information and non-matching information of the obstacle according to the target data and the ground truth data.

[0123] In the embodiments of the present application, the matching information includes at least one of the matching quantity, the matching cost, or the matching weight, and the non-matching information includes at least one of the non-matching quantity, the non-matching cost, or the non-matching weight.

[0124] Among them, the matching quantity refers to the number of the same obstacles existing in the target data and the ground truth data, the matching cost refers to the cost required to successfully match an obstacle, for example, the time required in the matching process, the computing power consumed in the matching process, etc., and the matching weight is the weight of the obstacles successfully matched in all obstacles during the matching process. The non-matching information includes two cases of miss and redundant, then the non-matching quantity, the non-matching cost, or the non-matching weight also correspond to two cases, and their specific meanings are similar to those in the matching information, which will not be elaborated here.

[0125] The similarity can be used to determine whether an obstacle is successfully matched. For example, if the similarity of any obstacle in the target data and the ground truth data is greater than 90%, it can be considered a successful match. The matching weight and the non-matching weight can be set according to actual needs, or according to the precision required by the parameters to be processed. The embodiments of the present application do not limit this.

[0126] In the embodiments of the present application, the matching cost and the non-matching cost of the obstacle can be obtained through the matching cost calculation formula and the non-matching cost calculation formula.

[0127] The matching cost calculation formula can be as follows:

[0128]

[0129] Wherein, is the true value result output after the true value system processes the first data, W Dx , W Dy , W Vx , W Vy are the weights after normalizing each initial weight.

[0130] The non - matching cost calculation formula can be as follows:

[0131]

[0132] Wherein, C Dx , C Dy , C Vx , C Vy are empirical constants, and their values mainly depend on the tolerance for miss targets and redundant targets. The tolerance refers to the number of allowable miss targets and redundant targets. The more allowable the number, the higher the tolerance.

[0133] S304. Obtain the cost value according to the matching information and the non - matching information.

[0134] In the embodiments of the present application, the corresponding cost value can be obtained according to the matching information and the non - matching information by using the cost value calculation formula.

[0135] Exemplarily, the calculation method of the cost value can satisfy the following formula:

[0136]

[0137] Wherein, N is the total number of frames of the first data, M is the number of matches in each frame, R is the number of misses in each frame, T is the number of redundancies in each frame, P is the number of terms required to calculate the cost of the target obstacle, w iM , w iR , w iP are the corresponding weights, f(x iM ), f(x iR ), f(x iP ) are the corresponding matching cost values.

[0138] Optionally, usually we allow higher miss targets and penalize redundant targets, so we can set smaller w iR and larger w iP values, w iR and wiP The value can also be set according to actual requirements, and the embodiments of the present application do not limit this.

[0139] Optionally, when obtaining the cost value, the first data and the second data selected need to be data collected at the same time. If the sensor includes a lidar, the lidar timestamp is used as the standard for selecting the first data and the second data.

[0140] If the collection times of the first data and the second data are not based on the lidar cycle, the first data and the second data need to be time-synchronized before subsequent calculation of the cost value. In the embodiments of the present application, interpolation methods such as piecewise linear interpolation and strip interpolation can be used to time-synchronize the first data and the second data.

[0141] S305. Determine whether the cost value is less than a preset value. If not, select the parameters of the new fusion algorithm as the initial weight of the second data to obtain the target data. If so, the processing of the parameters of the fusion algorithm is completed.

[0142] In the embodiments of the present application, after obtaining the cost value, the parameters of the fusion algorithm can be processed according to the size relationship between the cost value and the preset value.

[0143] Specifically, if the cost value meets the preset conditions, the processing of the parameters of the fusion algorithm is completed, where the preset conditions are that the cost value is less than the preset value, or after N adjustments, the corresponding P cost values are all greater than M% of the optimal value within the N adjustment cycles, and the optimal value is the value with the smallest difference from the preset value, and M, N, and P are integers greater than 1.

[0144] If the cost value is greater than the preset value, set a new first parameter, re-obtain the target data, and obtain the cost value according to the re-obtained target data and the true value data.

[0145] For example, based on the initial weight, take the range of increasing or decreasing by K% as the value range, and take L% as the step size for value taking to obtain corresponding multiple groups of weights, where K and L are integers greater than 1.

[0146] Take each group of weights in the multiple groups of weights as a new first parameter, and obtain the cost value corresponding to each new first parameter; if the optimal value among the cost values corresponding to each new first parameter does not meet the preset conditions, take the weight corresponding to the optimal value as the new initial weight; repeat the process of obtaining multiple cost values according to the new initial weight until the optimal value among the multiple cost values meets the preset conditions.

[0147] Exemplarily, if the cost value is greater than the preset value, based on the initial weight, with an increase or decrease of 10% as the value range of the new weight, within this value range, taking 5% as the step size, multiple groups of new weights are obtained. Each group of new weights and the second data are input into the fusion algorithm to obtain multiple new target data. Then, based on the multiple new target data and the true data, multiple cost values are obtained.

[0148] Compare the multiple cost values with the preset value. If there is a cost value among the multiple cost values that is less than the preset value, then use the weight corresponding to this cost value as the parameter of the fusion algorithm, and the processing is completed.

[0149] If there is no cost value among the multiple cost values that is less than the preset value, then use the weight corresponding to the optimal value among the multiple cost values as the parameter of the new fusion algorithm, and perform the next parameter processing.

[0150] Repeat the above process until there is a cost value among the multiple obtained cost values that is less than the preset value, or after N times of parameter processing, the corresponding P cost values are all greater than 2% of the optimal value within the N adjustment cycles, it can be considered that the parameter calculation has converged, the current parameter meets the requirements, and the processing is completed.

[0151] The vehicle sensor parameter processing method provided by the embodiments of the present application obtains the first data collected by the first sensor and the second data collected by at least two second sensors, obtains the initial weights of the second data, obtains the target data according to the initial weights and the second data, obtains the matching information and non-matching information of the obstacle according to the target data and the true value data, obtains the cost value according to the matching information and non-matching information, and processes the parameters of the fusion algorithm according to the cost value. By collecting data on the same scene through the true value system and the perception system, and processing the parameters of the fusion algorithm according to the matching result of the first data and the second data, the adjustment and calibration of the vehicle sensor parameters can be quickly and accurately completed without relying on the testing of testers.

[0152] Figure 4 FIG. 40 is a schematic structural diagram of a vehicle sensor parameter processing device 40 provided by an embodiment of the present application, including: a first acquisition module 401, a second acquisition module 402, a third acquisition module 403, and a processing module 404.

[0153] The first acquisition module 401 is configured to acquire the first data collected by the first sensor and the second data collected by at least two second sensors.

[0154] The second acquisition module 402 is configured to obtain target data according to at least two second data, where the target data is data obtained by performing a fusion process on at least two second data.

[0155] A third acquisition module 403, configured to obtain a cost value according to the target data and the ground truth data, where the cost value is used to represent the matching relationship between the target data and the ground truth data, and the ground truth data is data obtained by processing the first data.

[0156] A processing module 404, configured to process the first parameter corresponding to each second data according to the cost value, where the first parameter is the fusion weight of each second data in the fusion algorithm.

[0157] The vehicle sensor parameter processing device provided by the embodiments of the present application can execute Figure 2 or Figure 3 the technical solutions of the method embodiments shown, which will not be elaborated here.

[0158] Figure 5 It is a schematic structural diagram of an electronic device for processing vehicle sensor parameters provided by the embodiments of the present application. As Figure 5 shown, the electronic device 50 for processing vehicle sensor parameters provided in this embodiment may include:

[0159] A processor 501.

[0160] A memory 502, configured to store executable instructions of the terminal device.

[0161] Wherein, the processor is configured to execute the technical solutions of the above vehicle sensor parameter processing method embodiments by executing the executable instructions, and the implementation principles and technical effects are similar, which will not be elaborated here.

[0162] In the embodiments of the present application, the vehicle sensor parameters may also be referred to as vehicle sensor fusion algorithms, and the embodiments of the present application do not limit this.

[0163] The embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the technical solutions of the above vehicle sensor parameter processing method embodiments, and the implementation principles and technical effects are similar, which will not be elaborated here.

[0164] In one possible implementation, the computer-readable medium may include a random access memory (RAM), a read-only memory (ROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, a magnetic disk storage, or any other medium targeted to carry or store the required program code in the form of instructions or data structures and accessible by a computer. Moreover, any connection is properly termed a computer-readable medium. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of the medium. As used herein, disk and optical disc include optical disc, laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc, where disks typically reproduce data magnetically, while optical discs utilize lasers to optically reproduce data. Combinations of the above should also be included within the scope of computer-readable media.

[0165] In an embodiment of the present application, a computer program product is further provided, including a computer program, which implements the technical solutions of the above-mentioned embodiment of the vehicle sensor parameter processing method when executed by a processor. The implementation principle and technical effects are similar and will not be elaborated herein.

[0166] In the specific implementation of the above terminal device or server, it should be understood that the processor may be a central processing unit (CPU for short), or other general-purpose processors, digital signal processors (DSP for short), application-specific integrated circuits (ASIC for short), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application may be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0167] Those skilled in the art can understand that all or part of the steps of any of the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, all or part of the steps of the above method embodiments are executed.

[0168] If the technical solution of the present application is implemented in the form of software and sold or used as a product, it can be stored in a computer-readable storage medium. Based on such an understanding, all or part of the technical solution of the present application can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a computer program or several instructions. The computer software product enables a computer device (which may be a personal computer, a server, a network device or a similar electronic device) to execute all or part of the steps of the method of the embodiments of the present application.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for processing vehicle sensor parameters, characterized in that, it includes: obtaining first data collected by a first sensor and second data collected by at least two second sensors; obtaining target data according to the at least two second data, wherein the target data is data obtained by performing fusion processing on the at least two second data; obtaining a cost value according to the target data and true value data, the cost value being used to represent the matching relationship between the target data and the true value data, wherein the true value data is data obtained by processing the first data; processing the first parameter corresponding to each second data according to the cost value, wherein the first parameter is the fusion weight of each second data in the fusion algorithm; both the first sensor and the second sensor are used to detect obstacles, and the obtaining of the cost value according to the target data and the true value data includes: matching a first obstacle in the target data with a second obstacle in the true value data; obtaining matching information between the first obstacle and the second obstacle, the matching information including at least one of the number of matches, the matching cost, or the matching weight; obtaining mismatch information between the first obstacle and the second obstacle, the mismatch information including at least one of the number of mismatches, the mismatch cost, or the mismatch weight; obtaining the cost value according to the matching information and the mismatch information.

2. The method according to claim 1, characterized in that, if the matching information includes a matching cost, the obtaining of the matching information between the first obstacle and the second obstacle includes: obtaining an initial weight of each of the second data; performing normalization processing on the initial weight to obtain a normalized weight; obtaining the matching cost according to the true value data, the target data, and the normalized weight; if the mismatch information includes a mismatch cost, the obtaining of the mismatch information between the first obstacle and the second obstacle includes: obtaining the mismatch cost according to the target data, the normalized weight, and the second data.

3. The method according to claim 1, characterized in that, the obtaining of the target data according to the at least two second data includes: obtaining an initial weight of each of the second data; obtaining the target data according to the initial weight and the second data through an attribute processing function, wherein the attribute processing function is used to represent the constraint relationship between the target data and the second data.

4. The method according to claim 1, characterized in that, the processing of the first parameter corresponding to each second data according to the cost value includes: if the cost value meets a preset condition, the parameter processing of the fusion algorithm is completed, wherein the preset condition is that the cost value is less than a preset value, or after N adjustments, the corresponding P cost values are all greater than M% of the optimal value within the N adjustment periods, the optimal value being the value with the smallest difference from the preset value, and M, N, and P are integers greater than 1; If the cost value is greater than a preset value, set a new first parameter, re-acquire the target data, and obtain the cost value according to the re-acquired target data and the true value data.

5. The method according to claim 4, wherein, the setting of the new first parameter, re-acquiring the target data, and obtaining the cost value according to the re-acquired target data and the true value data includes: Based on the initial weight, taking the range of increasing or decreasing by K% as the value range and L% as the step size for value taking to obtain corresponding multiple groups of weights, where K and L are integers greater than 1; Regarding each group of weights in the multiple groups of weights as one of the new first parameters, and obtaining the cost value corresponding to each new first parameter; If the optimal value among the cost values corresponding to each new first parameter does not meet the preset condition, take the weight corresponding to the optimal value as the new initial weight; Repeat the process of obtaining multiple cost values according to the new initial weight until the optimal value among the multiple cost values meets the preset condition.

6. The method according to claim 1, wherein, after obtaining the first data collected by the first sensor and the second data collected by at least two second sensors, the method further includes: If the acquisition times of the first data and the second data are not synchronized, perform time synchronization on the first data and the second data by the difference method.

7. A vehicle sensor parameter processing device, wherein, comprising: A first acquisition module for acquiring the first data collected by the first sensor and the second data collected by at least two second sensors; A second acquisition module for obtaining target data according to at least two second data, where the target data is data obtained by performing fusion processing on the at least two second data; A third acquisition module for obtaining a cost value according to the target data and the true value data, where the cost value is used to represent the matching relationship between the target data and the true value data, and the true value data is data obtained by processing the first data; A processing module for processing the first parameter corresponding to each second data according to the cost value, where the first parameter is the fusion weight of each second data in the fusion algorithm; Both the first sensor and the second sensor are used to detect obstacles. The third acquisition module is specifically used to match the first obstacle in the target data with the second obstacle in the true value data; obtain the matching information between the first obstacle and the second obstacle, where the matching information includes at least one of the matching quantity, the matching cost, or the matching weight; obtain the non-matching information between the first obstacle and the second obstacle, where the non-matching information includes at least one of the non-matching quantity, the non-matching cost, or the non-matching weight; and obtain the cost value according to the matching information and the non-matching information.

8. An electronic device, wherein, comprising: A memory for storing a computer program; A processor for executing the computer program to implement the method according to any one of claims 1-6.

9. A computer-readable storage medium, wherein, A computer program is stored thereon, and the computer program is executed by a processor to implement the method according to any one of claims 1-6.

Images