CN115047439A - Data processing method and device for vehicle-based detection system and storage medium - Google Patents
Data processing method and device for vehicle-based detection system and storage medium Download PDFInfo
- Publication number
- CN115047439A CN115047439A CN202210590767.9A CN202210590767A CN115047439A CN 115047439 A CN115047439 A CN 115047439A CN 202210590767 A CN202210590767 A CN 202210590767A CN 115047439 A CN115047439 A CN 115047439A
- Authority
- CN
- China
- Prior art keywords
- information
- obstacle
- vehicle
- detection system
- trajectory
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 128
- 238000003672 processing method Methods 0.000 title claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 61
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000003062 neural network model Methods 0.000 claims description 10
- 238000011156 evaluation Methods 0.000 description 20
- 230000006870 function Effects 0.000 description 12
- 230000000694 effects Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 238000005315 distribution function Methods 0.000 description 4
- 230000008447 perception Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000002372 labelling Methods 0.000 description 3
- 238000012886 linear function Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011144 upstream manufacturing Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
- G01S7/497—Means for monitoring or calibrating
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
- G01S17/931—Lidar systems specially adapted for specific applications for anti-collision purposes of land vehicles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Abstract
The invention discloses a data processing method, a data processing device and a storage medium of a vehicle-based detection system. Wherein, the method comprises the following steps: acquiring point cloud information of obstacles acquired by a vehicle on a current road section; determining target data of a detection system, which is obtained by first track curve information of a detection value of the obstacle and second track curve information of a truth value of the obstacle, based on point cloud information of the obstacle, wherein the target data is used for representing a difference between the first track curve information and the second track curve information; performance data of the detection system is determined based on the driving parameters and the target data, wherein the driving parameters are used for representing preset values of the driving track of the vehicle, and the performance data are used for representing the performance of the detection system.
Description
Technical Field
The invention relates to the field of vehicles, in particular to a data processing method, a data processing device and a storage medium of a vehicle-based detection system.
Background
At present, when the performance of the vehicle detector is evaluated, the index of average precision is generally adopted, but the method is treated as the same for all errors in the driving process, so that the problem of evaluating the performance of the vehicle detector for a single object is caused.
No effective solution has been proposed to the above-mentioned problem of single evaluation of vehicle detector performance.
Disclosure of Invention
The embodiment of the invention provides a data processing method and device of a vehicle-based detection system and a storage medium, which are used for at least solving the technical problem of singly evaluating the performance of a vehicle detector.
According to an aspect of an embodiment of the present invention, there is provided a data processing method, apparatus, and storage medium for a vehicle-based detection system. Wherein, the method comprises the following steps: acquiring point cloud information of obstacles acquired by a vehicle on a current road section; determining target data obtained by first track curve information of a detected value of the obstacle and second track curve information of a true value of the obstacle based on point cloud information of the obstacle, wherein the target data is used for representing a difference between the first track curve information and the second track curve information; performance data of the detection system is determined based on the driving parameters and the target data, wherein the driving parameters are used for representing preset values of the driving track of the vehicle, and the performance data are used for representing the performance of the detection system.
Optionally, determining target data obtained by first trajectory curve information of the detected value of the obstacle and second trajectory curve information of the true value of the obstacle based on the point cloud information of the obstacle includes: performing linear processing on the detected value of the obstacle to obtain first track curve information, wherein the first track curve information comprises a first track curve of vehicle running and the position and the direction of a point on the first track curve; performing linear processing on the true value of the obstacle to obtain second track curve information, wherein the second track curve information comprises a second track curve of vehicle running and the position and the direction of a point on the second track curve; target data is determined based on the first trajectory profile information and the second trajectory profile information.
Optionally, the obtaining the first trajectory curve information based on linear processing of the detected value of the obstacle includes: processing point cloud information of the obstacle based on a neural network model to obtain a detection value of the obstacle, wherein the neural network model is a model generated based on array feature processing, and the detection value of the obstacle is used for representing the information of the obstacle; and performing linear processing on the detection value of the obstacle to obtain first track curve information.
Optionally, a true value of the obstacle is obtained based on point cloud information of the obstacle, wherein the true value of the point cloud information is used for representing information of the obstacle; and carrying out linear processing on the true value of the point cloud information to obtain second track curve information.
Optionally, determining the target data based on the first trajectory profile information and the second trajectory profile information comprises: carrying out nonlinear processing on the first track curve information and the second track curve information to obtain difference information between the first track curve information and the second track curve information; the difference information is determined as target data.
Optionally, determining performance data of the detection system based on the driving parameters and the target data comprises: and performing linear processing on the driving parameters and the target data to obtain performance data according to the driving parameters.
Optionally, after determining performance data of the detection system based on the driving parameters and the target data, the method further comprises: and determining performance data of the detection system based on the running parameters of different time periods and the target data of different time periods.
According to another aspect of the embodiment of the present invention, there is also provided a data processing apparatus of a vehicle-based detection system, including an obtaining unit, configured to obtain point cloud information of an obstacle acquired by a vehicle for a current road segment; the device comprises a first determining unit, a second determining unit and a third determining unit, wherein the first determining unit is used for determining target data of a detection system obtained by first track curve information of a detection value of an obstacle and second track curve information of a truth value of the point obstacle based on point cloud information of the obstacle, and the target data is used for representing the difference between the first track curve information and the second track curve information; and the second determination unit is used for determining performance data of the detection system based on the running parameters and the target data, wherein the running parameters are used for representing the preset value of the running track of the vehicle, and the performance data are used for representing the performance of the detection system.
According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium. The computer-readable storage medium includes a stored program, wherein the apparatus in which the computer-readable storage medium is stored is controlled to perform the data processing method of the vehicle-based detection system of the embodiment of the present invention when the program is executed.
According to another aspect of the embodiments of the present invention, there is also provided a processor. The processor is configured to run a program, wherein the program when executed performs the data processing method of the vehicle-based detection system of the embodiment of the present invention.
In the embodiment of the invention, the point cloud information of the obstacles collected by the vehicle on the current road section is obtained; determining target data obtained by first track curve information of a detection value of the obstacle and second track curve information of a true value of the obstacle based on point cloud information of the obstacle, wherein the target data is used for representing a difference between the first track curve information and the second track curve information; performance data of the detection system is determined based on the driving parameters representing a predetermined value being a driving trajectory of the vehicle and the target data representing performance of the detection system. That is to say, in the embodiment of the present invention, first, a detection value of an obstacle is obtained by a vehicle detection system, a true value of the obstacle is obtained by an operation instruction, then, a first trajectory information is obtained by performing linear processing on the detection value of the obstacle, a second trajectory information is obtained by performing linear processing on the true value of the obstacle, and the first trajectory curve information and the second trajectory curve information are subjected to nonlinear processing to obtain difference information between the first trajectory curve information and the second trajectory curve information; and determining the difference information as target data, and finally determining the performance data of the detection system based on the driving parameters and the target data. Therefore, the purpose of evaluating the vehicle detector based on the planned indexes is achieved, the technical problem of single evaluation of the performance of the vehicle detector is solved, and the technical effect of evaluating the performance of the vehicle detector in multiple aspects is achieved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
FIG. 1 is a flow chart of a data processing method for a vehicle-based detection system according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a vehicle driving trajectory prediction according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a data processing device of a vehicle-based detection system according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other sequences than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example 1
In accordance with an embodiment of the present invention, there is provided a data processing method for a vehicle-based detection system, it should be noted that the steps illustrated in the flowchart of the accompanying drawings may be executed in a computer system such as a set of computer-executable instructions and that, although a logical order is illustrated in the flowchart, in some cases, the steps illustrated or described may be executed in an order different from that described herein.
Fig. 1 is a flowchart of a data processing method of a vehicle-based detection system according to an embodiment of the present invention, which may include the steps of, as shown in fig. 1:
and step S101, acquiring point cloud information of the obstacles acquired by the vehicle on the current road section.
In the technical scheme provided by the above step S101 of the present invention, point cloud information of an obstacle acquired by a vehicle on a current road segment is obtained, where the vehicle may be a vehicle of the own vehicle, and the point cloud information of the obstacle may be point cloud information on the obstacle of the current road segment acquired by an information acquisition system in the vehicle of the own vehicle.
Optionally, the point cloud information of the obstacle acquired by the vehicle for the current road segment may be acquired by a perception sensor, which is not limited herein.
Step S102, determining target data obtained by first track curve information of a detection value of an obstacle and second track curve information of a truth value of the point obstacle based on point cloud information of the obstacle, wherein the target data is used for representing a difference between the first track curve information and the second track curve information.
In the technical solution provided in step S102 of the present invention, based on the point cloud information of the obstacle, first trajectory curve information of the detected value of the obstacle and second trajectory curve information of the true value of the obstacle are processed to obtain vehicle target data. Wherein the detected value of the obstacle may be a pedestrian, an object, or the like, the true value of the obstacle may be a pedestrian, a bicycle, or the like, the first trajectory curve information of the detected value of the obstacle may be vehicle travel estimated curve information avoiding the detected value of the obstacle, the second trajectory curve information of the true value of the obstacle may be vehicle travel estimated curve information avoiding the true value of the obstacle, and the target data may be the first evaluation value for the vehicle detection system.
And S103, determining performance data of the detection system based on the driving parameters and the target data, wherein the driving parameters are used for representing the preset value of the driving track of the vehicle, and the performance data are used for representing the performance of the detection system.
In the technical solution provided by the above step S103 of the present invention, a product between the driving parameter and the target data is determined as performance data of the detection system, wherein the driving parameter may be a predetermined value during the driving of the vehicle, and the performance data may be a second evaluation value for the vehicle detection system.
The method comprises the following steps of S101 to S103, acquiring point cloud information of obstacles acquired by a vehicle on a current road section; determining target data obtained by first track curve information of a detection value of the obstacle and second track curve information of a true value of the obstacle based on point cloud information of the obstacle, wherein the target data is used for representing a difference between the first track curve information and the second track curve information; performance data of the detection system is determined based on the driving parameters and the target data, wherein the driving parameters are used for representing preset values of the driving track of the vehicle, and the performance data are used for representing the performance of the detection system. That is to say, in the embodiment of the present invention, first, a detection value of an obstacle is obtained by a vehicle detection system, a true value of the obstacle is obtained by an operation instruction, then, a first trajectory information is obtained by performing linear processing on the detection value of the obstacle, a second trajectory information is obtained by performing linear processing on the true value of the obstacle, and the first trajectory curve information and the second trajectory curve information are subjected to nonlinear processing to obtain difference information between the first trajectory curve information and the second trajectory curve information; and determining the difference information as target data, and finally determining the performance data of the detection system based on the driving parameters and the target data. Therefore, the purpose of evaluating the vehicle detector based on the planned track is achieved, the technical problem of single evaluation of the performance of the vehicle detector is solved, and the technical effect of evaluating the performance of the vehicle detector in multiple aspects is achieved.
The above-described method of this embodiment is further described below.
As an alternative embodiment, the step S102 of determining target data obtained by first trajectory curve information of the detected value of the obstacle and second trajectory curve information of the true value of the obstacle based on the point cloud information of the obstacle includes: performing linear processing on the detection value of the obstacle to obtain first track curve information, wherein the first track curve information comprises a first track curve of vehicle running and the position and the direction of a point on the first track curve; performing linear processing on the true value of the obstacle to obtain second track curve information, wherein the second track curve information comprises a second track curve of vehicle running and the position and the direction of a point on the second track curve; target data is determined based on the first trajectory profile information and the second trajectory profile information.
In this embodiment, a decision planning algorithm is adopted to process a detection value of an obstacle to obtain first trajectory curve information, where the first trajectory curve information may be a vehicle driving trajectory point obtained according to the detection value of the obstacle, and a vehicle drives according to the vehicle driving trajectory point to obtain a first trajectory curve, where the first estimation curve information includes the vehicle driving trajectory point and the first trajectory curve; processing the true value of the obstacle by adopting a decision planning algorithm to obtain second track curve information, wherein the second track curve information can be vehicle running track points obtained according to the true value of the obstacle, and the vehicle runs according to the vehicle running track points to obtain a second track curve, wherein the second estimation curve information comprises the vehicle running track points and the second track curve; and carrying out non-calculation on the information of the first estimation curve and the information of the second estimation curve to obtain target data.
As an alternative embodiment, the step S102 of obtaining the first trajectory graph information based on the linear processing of the detected value of the obstacle includes: and processing the point cloud information based on a neural network model to obtain a detection value of the obstacle, wherein the neural network model is a model generated by array features, the detection value of the obstacle is used for representing the information of the obstacle, and the detection value of the obstacle is subjected to linear processing to obtain first track curve information.
In this embodiment, point cloud information of an obstacle is identified based on a neural network model in a detection system of a vehicle, so as to obtain a detection value of the obstacle, where the detection value of the obstacle may be obstacle information in a current road detected by the detection system of the vehicle, where the obstacle information includes a type, a position, and a size of the obstacle, and the obstacle information is processed by using a decision planning algorithm, so as to obtain first driving trajectory information of the vehicle, where the first driving trajectory information of the vehicle may be a driving trajectory of the vehicle determined according to each obstacle information.
For example, point cloud information of 10 groups of obstacles on a current road section of a neural network in a vehicle detection system is adopted for identification, 8 groups of obstacle information are identified, the 8 groups of obstacle information are processed through a decision planning algorithm, and curve information based on the 8 groups of obstacle information is obtained, wherein the curve information can be represented by a function.
As an alternative embodiment, the step S102, performing linear processing on the true value of the obstacle to obtain the second trajectory curve information, includes: acquiring a truth value of the obstacle based on the point cloud information of the obstacle, wherein the truth value of the obstacle is used for representing the current obstacle information; and carrying out linear processing on the true value of the obstacle to obtain second track curve information.
In this embodiment, a truth value of an obstacle is obtained by manually labeling point cloud information of the obstacle, where the truth value of the obstacle may be obtained by manually labeling the point cloud information of the obstacle on a current road, and the truth value of the obstacle may be obstacle information on the current road, where the obstacle information includes a type, a position, and a size of the obstacle, and a decision planning algorithm is used to process the truth value of the obstacle to obtain second travel track information of the vehicle, where the second travel track information of the vehicle may be a travel track determined according to each obstacle information.
For example, point cloud information of 10 groups of obstacles on the current road is manually labeled to obtain 10 groups of obstacle information, and the 10 groups of obstacle information are processed through a decision planning algorithm to obtain curve information based on the 10 groups of obstacle information, wherein the curve information can be represented by a function.
As an alternative embodiment, in step S102, determining the target data based on the first trajectory profile information and the second trajectory profile information includes: carrying out nonlinear processing on the first track curve information and the second track curve information to obtain difference information between the first track curve information and the second track curve information; determining the difference information as the target data.
In this embodiment, the target data is obtained by performing nonlinear calculation on the first trajectory curve information and the second trajectory curve information, where the target data may be a first evaluation value of the detection system, and the nonlinear calculation is KL divergence (KL divergence).
For example, a function of the first trajectory graph information and a function of the second trajectory graph information are calculated by using a first calculation model to obtain a first evaluation value of the detection system, where the function of the first trajectory graph information may be represented by P, the function of the second trajectory graph information may be represented by Q, and the first evaluation value of the detection system may be represented by R.
As an alternative embodiment, step S103, determining performance data of the detection system based on the driving parameters and the target data, includes: and performing linear processing on the driving parameters and the target data to obtain performance data according to the driving parameters.
In this embodiment, the running parameter and the target data may be multiplied to obtain performance data of the detection system. The driving parameter may be an empirical value of a driving estimation offset of the vehicle during driving.
For example, the first evaluation value R is multiplied by the driving parameter D to obtain the performance data E of the system.
As an alternative embodiment, in step S103, after determining the performance data of the detection system based on the driving parameters and the target data, the method further includes: and determining performance data of the detection system based on the running parameters of different time periods and the target data of different time periods.
In this embodiment, the running parameters and the target data of different time periods may be multiplied and added, and then averaged to obtain the final performance data of the detection system, where the running parameters of each time period are different.
For example, the running parameter R of the first time period is multiplied by the running parameter D to obtain the performance data E of the first time period, the running parameter a of the second time period is multiplied by the running parameter B to obtain the performance data C of the second time period, and the performance data E of the first time period and the performance data C of the second time period are added and divided by 2 to obtain the final performance data, and so on for multiple time periods.
In the embodiment, the point cloud information of the obstacle is processed by adopting a neural network model based on the point cloud information of the obstacle of the current road section to obtain the detection value of the obstacle, and the detection value of the obstacle is linearly processed to obtain first track curve information; based on the point cloud information of the obstacles on the current road section, manually processing the point cloud information of the obstacles to obtain a true value of the obstacles, linearly processing the true value of the obstacles to obtain second track curve information, and performing nonlinear processing on the first track curve information and the second track curve information to obtain difference information between the first track curve information and the second track curve information; determining the difference information as target data; and performing linear processing on the driving parameters and the target data to obtain performance data according to the driving parameters. When a plurality of time periods exist, the performance data of the detection system is determined based on the driving parameters of different time periods and the target data of different time periods, the technical problem of single evaluation of the performance of the vehicle detector is solved, and the technical effect of evaluating the performance of the vehicle detector in multiple aspects is achieved.
Example 2
The technical solutions of the embodiments of the present invention will be illustrated below with reference to preferred embodiments.
In autonomous driving, the detection performance of an on-board system for objects around a vehicle is crucial to the safety of the vehicle. In research, development and testing of automatic driving radar perception target recognition, the radar target recognition effect is evaluated according to the scheme, and therefore researchers are helped to improve the recognition method in a targeted mode.
In the field of the autonomous vehicle, an index such as a Mean Average Precision (MAP) has been proposed as an index for ranking the detectors. The indexes are all treated in the same way aiming at all errors, some errors seriously influence planning decision in the actual driving process, and some errors do not influence the normal running of the actual automatic driving vehicle.
In order to meet the actual requirements of automatic driving, the scheme provides a competition object detector based on planning indexes to evaluate, and the perception performance analysis is connected with the performance of downstream driving tasks. This would be superior to those carefully manually designed evaluation metrics.
Currently, the existing mainstream target detection effect evaluation is to calculate the matching degree between the detection result and the real obstacle information. 3D recognition is carried out in a three-dimensional space, and the attributes of the obstacle are as follows: category, position (x, y, z), size (w, l, h), orientation angle (r) and velocity (Vx, Vy), state attributes, and the like. The system calculates the center distance or the overlapping rate (IOU), then matches according to a set threshold value, finds out the correct matching number, the false detection number and the missed detection number, and then calculates the accuracy rate and the recall rate to evaluate the indexes. The disadvantage of this method is that, without distinguishing between the correct and wrong type, often different mistakes may have a greatly different effect on the actual driving of the autonomous vehicle. For example, false detection or missed detection at a long distance may not affect driving, and a small error at a close distance may threaten driving safety.
Therefore, in order to overcome the above problems, in one related art, an online path planner is proposed that queries each of one or more automated vehicles for possible solutions for at least one executable task; checking the result of the query; determining a coordinated path plan for each vehicle; and communicating the coordinated path plan to a traffic manager, wherein the traffic manager ensures that the one or more automated vehicles perform each executable task according to the coordinated path plan.
In another related technology, a navigation path planning method based on strategy reuse and reinforcement learning, a one-way KL divergence method and a two-way KL divergence method are provided, wherein the two methods both represent the dissimilarity degree between strategies by means of KL divergence (relative entropy) of C function distribution corresponding to each strategy, the one-way KL divergence method adds a new strategy which is dissimilar to all source strategies into a strategy library, and the two-way KL divergence method further judges whether to add some strategies with weak representativeness of the source strategy library into the strategy library according to the dissimilarity of the KL divergence.
However, the proposed method of the present invention is to train a planned network and then use the network to measure the performance of the upstream detector. The key idea is to evaluate the detection by a planner that plans the driving trajectory based on semantic observations (i.e. detection). Depending on the design, the evaluation system will return the best score if the perception system is perfect. It learns that those detections are important to the driving task and then uses this network to evaluate the upstream detectors.
Firstly, data preparation is carried out, the frame rate of a data set is required to be 10 frames/s or more, each section of data is more than 15s, the data contains truth-value frame information, and a real driving track under the condition of the truth-value frame information is provided; and meanwhile, a plurality of target detection models to be evaluated can output the detected truth-value frame information, and a planning track based on the detection values is given under the condition. Calculating how much difference exists between the planned path of the vehicle and the path of the real vehicle in the detection value state through the KL divergence to measure the quality of the detection effect, wherein the calculation process is as follows:
noise is introduced (data to be detected),
representing the position information of the own vehicle at the time t, the obstacle is represented by i e {1, …, N },
information representative of the obstacles detected by the detection system,
the position information of the own vehicle, which represents the ideal detected value (i.e., true value) in the T time, in the case of the predicted value of the obstacle, can be represented by a joint probability density function,
we want to calculate the change in this distribution given our specified noise observation, which can be measured by KL divergence as follows: d KL (P | Q), in which the position information of the own vehicle in the case of the ideal detection value of the obstacle, can be expressed by a joint probability density function,
first, assuming that all the obstacle prediction values at the time t-1 are observed values, the detection performance of the detector at the time t-1 can be represented by a joint probability P:
since they are independent of each other, the joint probability becomes each obstacleProduct of edge probabilities
Then, assuming that the predictions at each time are independent of each other, there are
Thus, the joint distribution Q is decomposed into:
substituting P and Q into KL divergence formula, and obtaining the result of evaluating the detection performance of the detector at the time t-1, the formula can be as follows:
optionally, when S 1 ，S 2 …S t E.S represents a laser radar original point cloud data sequence,
representing the sequence of truth values for detecting obstacles, x 1 ，x 2 …x t Represents the self-parking position information o t Represents s t When the obstacle information is output by the detector a, the performance evaluation result is detected by the detector at the time T ═ 1, so the evaluation index of the T period can be defined as follows:
in the above formula (1), p θ (x t+Δ |o ≤t ) Can be used for representing the trajectory distribution of the data set D under the condition of truth frame strips
The asymmetry of the predicted results in the real world cannot be reflected, and the prediction with the same measurement precision may result in different results, such as the influence of the tracks deviating from the same scale to different directions on the own vehicle is greatly different, but the asymmetry cannot be reflected in the index. On the basis, the influence of the deviated track predicted by the target vehicle on the driving route of the self vehicle is added with the weight:
the track deviated towards the direction of the self-vehicle has greater threat to the self-vehicle and is endowed with a greater weight value; the track which seriously influences the driving of the vehicle by the original vehicle prediction result without intersection with the track of the vehicle is endowed with a larger weight value; under the condition that the track prediction has deviation but does not influence the self vehicle, a smaller weight value is given; trajectory deviations farther from the host vehicle are given smaller weight values, and vice versa;
the sensitivity is represented by calculating the gradient of each own vehicle position by using a linear function, and the formula can be as follows:
in the above equation (2), s is a state vector (position, velocity, acceleration), u is a running vector of the vehicle, s is a predicted trajectory, θ T Is a weight vector of a plurality of feature vectors. All the characteristics can be considered by using a linear function, and the decoupling is easy. We consider the trajectory of a human operated vehicle as a true value, i.e., an assessment indicator of overall time can be defined as follows:
in the above-mentioned formula (3),
the gradient can be expressed as a linear function, a is how many segments the total time is divided into, f (a, g) can be expressed as a calculation weight, a normalization function or a softmax function can be used, or an empirical value can be used. PKL is a prediction trajectory and GT similarity index calculated previously, and the Score value is 0 when matching is optimal, as Score is closer to perfect state.
Vehicle trajectory prediction as shown in fig. 2, a five-pointed star can represent the driving trajectory of the vehicle after the obstacle information is manually marked, a six-pointed star is used for representing that the driving direction of the vehicle is dangerous after the PKL is calculated, and a seven-pointed star is used for representing that the driving direction of the vehicle is safe after the PKL is calculated, so that the driving trajectory of the vehicle is related to a detection system, and the detection system can better drive the vehicle after the obstacle is detected.
In the embodiment, a point cloud data sequence acquired by a radar outputs a detected obstacle information sequence through a detector A, the point cloud data sequence acquired by the radar outputs real information of an obstacle through manual labeling, a track distribution function P is obtained based on the detected obstacle information sequence, a track distribution function Q is obtained based on the real information of the obstacle, the track distribution function P and the track distribution function Q are substituted into a KL divergence formula to obtain an evaluation index of a period of time, driving estimation experience values of different time periods and the evaluation indexes of different time periods are calculated to obtain a final evaluation index, the technical problem of single evaluation of the performance of a vehicle detector is solved, and the technical effect of evaluating the performance of the vehicle detector in multiple aspects is achieved.
Example 3
According to the embodiment of the invention, a device for processing data of the vehicle-based detection system is also provided. It should be noted that the apparatus for determining data processing of the vehicle-based detection system may be used to execute the data processing method of the vehicle-based detection system in embodiment 1.
FIG. 3 is a schematic diagram of a data processing device of a vehicle-based detection system according to an embodiment of the present invention. As shown in fig. 3, the data processing apparatus 300 for determining a vehicle-based detection system may include: an acquisition unit 301, a first determination unit 302, and a second determination unit 303.
The acquiring unit 301 is configured to acquire point cloud information of an obstacle acquired by a vehicle on a current road segment.
A first determining unit 302, configured to determine target data obtained by first trajectory curve information of a detected value of an obstacle and second trajectory curve information of a true value of the obstacle based on point cloud information of the obstacle, where the target data is used to represent a difference between the first trajectory curve information and the second trajectory curve information.
A second determining unit 303, configured to determine performance data of the detection system based on a driving parameter and target data, where the driving parameter is used to represent a predetermined value of a driving trajectory of the vehicle, and the performance data is used to represent performance of the detection system.
Alternatively, the first determining unit 302 may include: the first processing module is used for carrying out linear processing on the detection value of the obstacle to obtain first track curve information, wherein the first track curve information comprises a first track curve of vehicle running and the position and the direction of a point on the first track curve.
Alternatively, the first determining unit 302 may include: and the second processing module is used for carrying out linear processing on the true value of the obstacle to obtain second track curve information, wherein the second track curve information comprises a second track curve of the vehicle running and the position and the direction of a point on the second track curve.
Alternatively, the first determining unit 302 may include: a first determination module to determine target data based on the first trajectory profile information and the second trajectory profile information.
Alternatively, the first determining unit 302 may include: the first processing submodule is used for processing point cloud information of the obstacle based on a neural network model to obtain a detection value of the obstacle, wherein the neural network model is a model generated based on array feature processing, and the detection value of the obstacle is used for performing linear processing on the detection value of the obstacle by using information representing the obstacle to obtain first track curve information.
Alternatively, the first determining unit 302 may include: the second processing submodule is used for acquiring a true value of the obstacle based on the point cloud information of the obstacle, wherein the true value of the point cloud information is used for representing the information of the obstacle; and carrying out linear processing on the true value of the obstacle to obtain second track curve information.
Optionally, the first determining module 302 may include: the first integration submodule is used for carrying out nonlinear processing on the first track curve information and the second track curve information to obtain difference information between the first track curve information and the second track curve information; the difference information is determined as target data.
Alternatively, the second determining unit 303 may include: and the first determining module is used for carrying out linear processing on the driving parameters and the target data to obtain performance data according to the driving parameters.
Wherein the apparatus further comprises: a third determination unit 304 for determining performance data of the detection system based on the driving parameters and the target data, the method further comprising: determining performance data of the detection system based on the driving parameters for different time periods and the target data for different time periods.
In the embodiment, the point cloud information of the obstacles collected by the vehicle on the current road section is acquired; the first determining unit determines target data obtained by first track curve information of a detection value of the obstacle and second track curve information of a true value of the obstacle based on point cloud information of the obstacle, wherein the target data is used for representing a difference between the first track curve information and the second track curve information; and the second determining unit is used for determining the performance data of the detection system based on the driving parameters and the target data, wherein the driving parameters are used for representing the preset value of the driving track of the vehicle, and the performance data are used for representing the performance of the detection system, so that the technical problem of single evaluation of the performance of the vehicle detector is solved, and the technical effect of evaluating the performance of the vehicle detector in multiple aspects is achieved.
Example 4
According to an embodiment of the present invention, there is also provided a computer-readable storage medium including a stored program, wherein the program executes the data processing method of the vehicle-based detection system in embodiment 1.
Example 5
According to an embodiment of the present invention, there is also provided a processor for executing a program, wherein the program executes a data processing method of the vehicle-based detection system in embodiment 1 when running.
Example 6
According to an embodiment of the present invention, there is also provided a vehicle for performing the data processing method of any one of claims 1 to 7 of the vehicle-based detection system.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present invention, the description of each embodiment has its own emphasis, and reference may be made to the related description of other embodiments for parts that are not described in detail in a certain embodiment.
In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, a division of a unit may be a division of a logic function, and an actual implementation may have another division, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or may not be executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that it is obvious to those skilled in the art that various modifications and improvements can be made without departing from the principle of the present invention, and these modifications and improvements should also be considered as the protection scope of the present invention.
Claims (10)
1. A data processing method for a vehicle-based detection system, comprising:
acquiring point cloud information of obstacles acquired by a vehicle on a current road section;
determining target data of the detection system, which is obtained by first trajectory curve information of a detection value of the obstacle and second trajectory curve information of a true value of the obstacle, based on the point cloud information of the obstacle, wherein the target data is used for representing a difference between the first trajectory curve information and the second trajectory curve information;
determining performance data of the detection system based on a driving parameter and the target data, wherein the driving parameter is used for representing a preset value of a driving track of the vehicle, and the performance data is used for representing the performance of the detection system.
2. The method of claim 1, wherein determining target data of the detection system based on the point cloud information of the obstacle from first trajectory curve information of the detected values of the obstacle and second trajectory curve information of the true values of the obstacle comprises:
performing linear processing on the detected value of the obstacle to obtain first track curve information, wherein the first track curve information comprises a first track curve of vehicle running and the position and direction of a point on the first track curve;
performing the linear processing on the true value of the obstacle to obtain second trajectory curve information, wherein the second trajectory curve information comprises a second trajectory curve on which the vehicle runs and the position and direction of a point on the second trajectory curve;
determining the target data based on the first trajectory profile information and the second trajectory profile information.
3. The method according to claim 2, wherein obtaining the first trajectory profile information based on linear processing of the detected value of the obstacle comprises:
processing the point cloud information of the obstacle based on a neural network model to obtain a detection value of the obstacle, wherein the neural network model is a model generated based on array feature processing, and the detection value of the obstacle is used for representing the information of the obstacle;
and performing the linear processing on the detection value of the obstacle to obtain the first track curve information.
4. The method of claim 2, wherein the deriving the second trajectory profile information based on the linear processing of the true values of the obstacle comprises:
acquiring a true value of the obstacle based on point cloud information of the obstacle, wherein the true value of the point cloud information is used for representing information of the obstacle;
and carrying out linear processing on the true value of the obstacle to obtain the second track curve information.
5. The method of claim 2, wherein determining the target data based on the first trajectory profile information and the second trajectory profile information comprises:
carrying out nonlinear processing on the first track curve information and the second track curve information to obtain difference information between the first track curve information and the second track curve information; determining the difference information as the target data.
6. The method of claim 1, wherein determining performance data for the detection system based on the driving parameters and the target data comprises:
and performing linear processing on the driving parameters and the target data to obtain the performance data according to the driving parameters.
7. The method of claim 1, wherein after determining performance data for the detection system based on the driving parameters and the target data, the method further comprises:
determining performance data of the detection system based on the driving parameters for different time periods and the target data for different time periods.
8. A data processing apparatus of a vehicle-based detection system, comprising:
the acquisition unit is used for acquiring the point cloud information of the obstacles acquired by the vehicle on the current road section;
a first determination unit configured to determine target data of the detection system, which is obtained from first trajectory curve information of a detection value of the obstacle and second trajectory curve information of a true value of the obstacle, based on point cloud information of the obstacle, wherein the target data is used to represent a difference between the first trajectory curve information and the second trajectory curve information;
a second determination unit for determining performance data of the detection system based on a driving parameter and the target data, wherein the driving parameter is used for representing a preset value of a driving track of the vehicle, and the performance data is used for representing the performance of the detection system.
9. A computer-readable storage medium, comprising a stored program, wherein the program, when executed, controls an apparatus in which the computer-readable storage medium is located to perform the method of any one of claims 1 to 7.
10. A vehicle for carrying out the data processing method of the vehicle-based detection system according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210590767.9A CN115047439A (en) | 2022-05-27 | 2022-05-27 | Data processing method and device for vehicle-based detection system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210590767.9A CN115047439A (en) | 2022-05-27 | 2022-05-27 | Data processing method and device for vehicle-based detection system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115047439A true CN115047439A (en) | 2022-09-13 |
Family
ID=83159840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210590767.9A Pending CN115047439A (en) | 2022-05-27 | 2022-05-27 | Data processing method and device for vehicle-based detection system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115047439A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200023842A1 (en) * | 2019-09-27 | 2020-01-23 | David Gomez Gutierrez | Potential collision warning system based on road user intent prediction |
| CN113753077A (en) * | 2021-08-17 | 2021-12-07 | 北京百度网讯科技有限公司 | Method and device for predicting movement locus of obstacle and automatic driving vehicle |
| CN114355950A (en) * | 2022-01-25 | 2022-04-15 | 苏州挚途科技有限公司 | Planning method and device for U-turn trajectory |
| WO2022090040A1 (en) * | 2020-10-29 | 2022-05-05 | Zf Friedrichshafen Ag | Method and device for controlling a vehicle along a journey trajectory |
| CN114475593A (en) * | 2022-01-18 | 2022-05-13 | 上汽通用五菱汽车股份有限公司 | Travel track prediction method, vehicle, and computer-readable storage medium |
-
2022
- 2022-05-27 CN CN202210590767.9A patent/CN115047439A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200023842A1 (en) * | 2019-09-27 | 2020-01-23 | David Gomez Gutierrez | Potential collision warning system based on road user intent prediction |
| WO2022090040A1 (en) * | 2020-10-29 | 2022-05-05 | Zf Friedrichshafen Ag | Method and device for controlling a vehicle along a journey trajectory |
| CN113753077A (en) * | 2021-08-17 | 2021-12-07 | 北京百度网讯科技有限公司 | Method and device for predicting movement locus of obstacle and automatic driving vehicle |
| CN114475593A (en) * | 2022-01-18 | 2022-05-13 | 上汽通用五菱汽车股份有限公司 | Travel track prediction method, vehicle, and computer-readable storage medium |
| CN114355950A (en) * | 2022-01-25 | 2022-04-15 | 苏州挚途科技有限公司 | Planning method and device for U-turn trajectory |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107341819B (en) | Target tracking method and storage medium | |
| CN102800207B (en) | Traffic signals detection system and method | |
| CN112781887B (en) | Method, device and system for testing vehicle performance | |
| US9053433B2 (en) | Assisting vehicle guidance over terrain | |
| US20130052614A1 (en) | Driver Performance Metric | |
| RU2742213C1 (en) | Method to control information on lanes, method of traffic control and device for control of information on lanes | |
| Huang et al. | Task-specific performance evaluation of UGVs: Case studies at the IVFC | |
| Li et al. | Map-aided dead-reckoning with lane-level maps and integrity monitoring | |
| CN112833903B (en) | Track prediction method, device, equipment and computer readable storage medium | |
| Völz et al. | Predicting pedestrian crossing using quantile regression forests | |
| CN114005280A (en) | Vehicle track prediction method based on uncertainty estimation | |
| CN113933858A (en) | Abnormal detection method and device of positioning sensor and terminal equipment | |
| WO2019073024A1 (en) | Lane sensing method | |
| EP2405383A1 (en) | Assisting with guiding a vehicle over terrain | |
| CN110909907A (en) | Method and device for predicting fuel consumption of truck and storage medium | |
| Augustin et al. | Prediction of highway lane changes based on prototype trajectories | |
| Jeong | Predictive lane change decision making using bidirectional long shot-term memory for autonomous driving on highways | |
| Dey et al. | Robust perception architecture design for automotive cyber-physical systems | |
| Nguyen et al. | Learning of lane information reliability for intelligent vehicles | |
| CN111612818A (en) | Novel binocular vision multi-target tracking method and system | |
| CN115047439A (en) | Data processing method and device for vehicle-based detection system and storage medium | |
| Nguyen et al. | Multisource fusion for robust road detection using online estimated reliabilities | |
| US20230314599A1 (en) | Multi-Scan Sensor Fusion for Object Tracking | |
| Lee et al. | Fail-safe multi-modal localization framework using heterogeneous map-matching sources | |
| Nguyen | A Reliability-Aware Fusion Concept Toward Robust Ego-Lane Estimation Incorporating Multiple Sources |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |