CN117490710A - Parameter adjusting method and device for target fusion and target fusion system - Google Patents
Parameter adjusting method and device for target fusion and target fusion system Download PDFInfo
- Publication number
- CN117490710A CN117490710A CN202210880327.7A CN202210880327A CN117490710A CN 117490710 A CN117490710 A CN 117490710A CN 202210880327 A CN202210880327 A CN 202210880327A CN 117490710 A CN117490710 A CN 117490710A
- Authority
- CN
- China
- Prior art keywords
- association
- target
- data
- parameter
- detection data
- 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
- 230000004927 fusion Effects 0.000 title claims abstract description 84
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000001514 detection method Methods 0.000 claims abstract description 99
- 230000008859 change Effects 0.000 claims abstract description 23
- 238000002372 labelling Methods 0.000 claims abstract description 20
- 239000000523 sample Substances 0.000 claims abstract description 15
- 230000006870 function Effects 0.000 claims description 34
- 230000008685 targeting Effects 0.000 claims description 6
- 230000033001 locomotion Effects 0.000 claims description 4
- 230000000875 corresponding effect Effects 0.000 description 42
- 238000010586 diagram Methods 0.000 description 10
- 238000004590 computer program Methods 0.000 description 9
- 238000012360 testing method Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 230000002596 correlated effect Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 4
- 238000012886 linear function Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000003247 decreasing effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000011895 specific detection Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
- G01C21/34—Route searching; Route guidance
- G01C21/3407—Route searching; Route guidance specially adapted for specific applications
- G01C21/3415—Dynamic re-routing, e.g. recalculating the route when the user deviates from calculated route or after detecting real-time traffic data or accidents
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C25/00—Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M17/00—Testing of vehicles
- G01M17/007—Wheeled or endless-tracked vehicles
-
- 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
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/86—Combinations of radar systems with non-radar systems, e.g. sonar, direction finder
- G01S13/867—Combination of radar systems with cameras
-
- 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
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
- G01S13/931—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of land vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Automation & Control Theory (AREA)
- Manufacturing & Machinery (AREA)
- Electromagnetism (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The embodiment of the invention provides a parameter adjusting method and device aiming at target fusion and a target fusion system, and belongs to the technical field of automatic driving. The parameter adjusting method comprises the following steps: acquiring scene data of an automatic driving vehicle under the current working condition; labeling detection data with the same attribute from a plurality of sensors in the scene data; determining a plurality of associated parameters affecting the accuracy of the probe data, wherein each associated parameter is configured with a weight for showing how much it affects the accuracy of the probe data under different conditions; determining an association threshold corresponding to the current working condition based on preset corresponding relations between each association parameter and weight thereof and the association threshold; and adjusting the weight according to the working condition change so as to change the association threshold value to enable the association success rate of the detection data which is associated for the target fusion to reach a preset requirement. The scheme of the invention realizes the dynamic adjustment of the association threshold values among different sensors under complex working conditions.
Description
Technical Field
The invention relates to the technical field of automatic driving, in particular to a parameter adjusting method and device aiming at target fusion and a target fusion system.
Background
The target fusion system is a part of an automatic driving system, and has the main functions of processing target data detected by various vehicle-mounted sensing devices and mutually fusing data with the same attribute to obtain a method which is superior to the detection performance of a single sensor. The fusion mode can enhance the system function or increase the system safety characteristic, for example, two sensors of the millimeter wave radar and the laser radar work simultaneously to detect the object targets in front, the object fusion system fuses the object targets detected by the two sensors, and the output fused object targets not only improve the measurement precision, but also can ensure that the automatic driving system can still operate under the condition that one sensor fails, thereby meeting the functional safety requirement of failure operability.
However, because the characteristic differences of different types of sensors are larger, and the characteristic differences of similar sensors of different manufacturers are also larger, manual intervention is often required to adjust thresholds reflecting the characteristic differences of different sensors, so that the sensor association in the target fusion can be realized. For example, the millimeter wave radar and the laser radar are adopted to detect the obstacle, the distance between the detected targets is 10m and 7m respectively, and the set threshold value for the target fusion is 2m, so that the situation that the targets detected by the millimeter wave radar and the laser radar can be correlated together to perform data fusion only by manually adjusting the threshold value to be 3m or more is easily known. When the automatic driving vehicle detects at a middle distance or a long distance, the existing errors are possibly larger, once the errors exceed the set threshold, the two sensor data cannot be used for target fusion, in addition, the vehicle speed also has an influence on the sensor errors, the original set threshold is manually adjusted to realize successful association of different sensors in the current working conditions, the workload is greatly increased, and the complex and variable working conditions are difficult to apply by simply relying on manual work.
Therefore, the method for manually adjusting the threshold has the defects of low efficiency, large workload, insufficient refinement and the like, and is not suitable for the current automatic driving development requirement.
Disclosure of Invention
The embodiment of the invention aims to provide a parameter adjusting method and device aiming at target fusion and a target fusion system, which are used for at least partially solving the technical problems.
In order to achieve the above object, an embodiment of the present invention provides a parameter tuning method for target fusion, including: acquiring scene data of an automatic driving vehicle under the current working condition; labeling detection data with the same attribute from a plurality of sensors in the scene data; determining a plurality of associated parameters affecting the accuracy of the probe data, wherein each associated parameter is configured with a weight for showing how much it affects the accuracy of the probe data under different conditions; determining an association threshold corresponding to the current working condition based on preset corresponding relation between each association parameter and weight thereof and the association threshold, wherein the association threshold is a critical value of whether detection errors of the detection data among the sensors are accepted by the target fusion; and adjusting the weight according to the working condition change so as to change the association threshold value to enable the association success rate of the detection data which is associated for the target fusion to reach a preset requirement, wherein the association success rate is the ratio of the number of frames of the detection data which is finally used for the target fusion to the total number of frames of the marked detection data.
Optionally, the labeling the detection data with the same attribute from the plurality of sensors in the scene data includes: marking the identification numbers (ID) of the sensors of the detection data, which are determined by the user and are associated with the same attribute of the same detection target, according to the scene data, wherein the IDs of the different sensors when detecting the same target are different; and inquiring detection data of a corresponding sensor from the scene data based on the noted sensor ID, and marking.
Optionally, after labeling of the detection data of all the sensors detecting the same target is completed, labeling content is presented based on the table of sensor IDs.
Optionally, the plurality of associated parameters includes a plurality of detection distances of the respective sensors, detection data obtained by the respective sensors, a vehicle speed of the host vehicle, a vehicle speed of the front target, a type of the front target, and a motion state of the front target.
Optionally, the preset correspondence is represented by the following formula:
THR＝∑A i ×f i (B)
wherein A is i Representing the value of the ith associated parameter, f i (B) And (3) representing a weight function corresponding to the ith association parameter, wherein B represents a coefficient of the weight function, and THR represents an association threshold.
Optionally, said adjusting said weight to adapt to a change in operating conditions includes: and adjusting coefficients in the weight function so that the weight is adapted to the driving working condition to change.
Optionally, the adjusting the coefficients in the weight function includes: traversing all coefficients in a set step size within a set parameter tuning range from a given initial coefficient; determining corresponding association success rates according to each traversed coefficient; and determining the corresponding coefficient when the association success rate is highest as the optimal coefficient.
In another aspect, the present invention provides a machine-readable storage medium having stored thereon instructions for causing a machine to perform any of the above-described targeting fusion method of the present application.
In another aspect, the present invention provides a parameter adjusting device for target fusion, including: the data acquisition module is configured to acquire scene data of the automatic driving vehicle under the current working condition; the labeling module is configured to label detection data with the same attribute from a plurality of sensors in the scene data; and a parameter adjusting module. The parameter adjustment module is configured to: determining a plurality of associated parameters affecting the accuracy of the probe data, wherein each associated parameter is configured with a weight for showing how much it affects the accuracy of the probe data under different conditions; determining an association threshold corresponding to the current working condition based on preset corresponding relation between each association parameter and weight thereof and the association threshold, wherein the association threshold is a critical value of whether detection errors of the detection data among the sensors are accepted by the target fusion; and adjusting the weight according to the working condition change so as to change the association threshold value to enable the association success rate of the detection data which is associated for the target fusion to reach a preset requirement, wherein the association success rate is the ratio of the number of frames of the detection data which is finally used for the target fusion to the total number of frames of the marked detection data.
In another aspect, the present invention provides a parameter adjusting device for target fusion, including: a memory storing a program capable of running on a processor; and the processor is configured to implement any of the above-described parameter tuning methods for target fusion when the program is executed.
On the other hand, the invention provides a target fusion system of an automatic driving vehicle, which comprises any parameter adjusting device aiming at target fusion.
Through the technical scheme, the parameter adjusting scheme aiming at target fusion has strong universality, can be suitable for adjusting the association threshold values among different sensors under complex working conditions of different detection distances and different vehicle speeds, greatly shortens the time of adjusting the association threshold values completely by manpower, is beneficial to solving the practical problems of short time and heavy tasks of vehicle road test projects, reduces a large amount of manpower and material resources, and ensures the road test quality of automatic driving vehicles.
Additional features and advantages of embodiments of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings are included to provide a further understanding of embodiments of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain, without limitation, the embodiments of the invention. In the drawings:
FIG. 1 is a flow chart of a method for tuning a target fusion according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a device for adjusting parameters for target fusion according to an embodiment of the present invention; and
fig. 3 is a schematic structural diagram of another parameter adjusting device for target fusion according to an embodiment of the present invention.
Description of the reference numerals
210. A data acquisition module; 220. a labeling module; 230. a parameter adjusting module; 240. and an evaluation module.
Detailed Description
The following describes the detailed implementation of the embodiments of the present invention with reference to the drawings. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
Fig. 1 is a flow chart of a parameter tuning method for target fusion according to an embodiment of the invention. Here, the "parameter adjustment" refers to adjusting a parameter related to an association threshold of sensor detection data for target fusion, and the association threshold refers to a critical value of whether or not detection errors between sensors for detection data with the same attribute are accepted by the target fusion. Taking the example that the camera, the front radar and the right angle radar of the vehicle jointly detect the longitudinal distance of the front target relative to the vehicle, if the longitudinal distance of the front target detected by the camera is 10m, the longitudinal distance detected by the front radar is 7m, and the longitudinal distance detected by the right angle radar is 6m, the maximum detection error between every two sensors is 4m, so that if the correlation threshold corresponding to the longitudinal distance is set to be greater than or equal to 4m, the longitudinal distances detected by the camera, the front radar and the right angle radar can be correlated for target fusion, and if the correlation threshold corresponding to the longitudinal distance is set to be less than 4m, the longitudinal distance detected by the camera and the longitudinal distance detected by the right angle radar can not be successfully correlated.
As shown in fig. 1, the method for adjusting parameters for target fusion according to the embodiment of the present invention may include the following steps S100 to S500:
step S100, scene data of the automatic driving vehicle under the current working condition is obtained.
The scene data includes, for example, target attribute information, vehicle attribute information, environment information and the like corresponding to the current working condition, and the information includes detection data detected by a sensor on the vehicle, for example, longitudinal distance information and vehicle speed information of a front target detected by the sensor relative to the vehicle.
The target attribute information includes, specifically, a target type (e.g., pedestrian or vehicle), a target position (e.g., longitudinal distance with respect to a host vehicle), a target motion state (stationary or moving), and the like; the attribute information of the vehicle comprises acceleration of the vehicle, speed of the vehicle, steering wheel angle of the vehicle and the like; the environmental information includes road type, barrier information, surrounding weather information, and the like.
According to the embodiment of the invention, the target attribute information, the vehicle attribute information and the environment information are combined to obtain complete and comprehensive multi-dimensional scene data aiming at the current working condition, and the sensor detection data is marked based on the scene data in the next step.
And step S200, labeling detection data with the same attribute from a plurality of sensors in the scene data.
The "same attribute" refers to detection data belonging to the same type, for example, information that the front targets detected by the cameras and the radar respectively belong to the same attribute with respect to the longitudinal distance of the own vehicle, and information that the front target vehicle speed detected by the cameras and the longitudinal distance detected by the radar belong to different attributes.
In a preferred embodiment, the step S200 further comprises: marking the identification numbers (ID) of the sensors which are determined by the user and are associated with the detection data with the same attribute for the same detection target according to the scene data, wherein the IDs of the different sensors when detecting the same target are different; and inquiring detection data of a corresponding sensor from the scene data based on the noted sensor ID, and marking.
In a more preferred embodiment, after the labeling of the detection data of all the sensors detecting the same target is completed, labeling content is displayed based on the table of the sensor IDs, so that query efficiency is improved accordingly, and verification of detection data labeling is facilitated for a user.
For example, for the scene data, the association relationship between the detection data observed by human eyes and the sensor identification number is labeled in a predetermined period (the relevant period is determined during labeling), so as to obtain the labeling content shown in the following table 1:
TABLE 1
Taking the first row of table 1 as an example, it shows one target in the real world, and based on the setting of "different sensor detects that the ID of the target is different", it is known that the first row represents "the target with the camera ID of 37 is the same obstacle in the real world as the target with the front radar of 16 and the target with the left-angle radar of 11". Accordingly, for the first row of table 1, the obstacle detected by the three different sensors may be selected (i.e., marked) by, for example, manually clicking, so that the ID information of the three sensors of the obstacle exists in the mark file in this regard, and the table 1 will be automatically generated after the file mark is completed. It should be noted that the meanings of other rows in table 1 are correspondingly understood, and are not described herein.
Further for example, if table 1 is labeled with a camera ID of 17, a front radar ID of 16, and a left radar ID of 11, the corresponding sensor detection data can be directly obtained from the scene data according to the corresponding ID number and labeled for subsequent association. The ID numbers recorded by each sensor when detecting the detection data with different attributes are different, so that each sensor ID has the corresponding specific detection data, and corresponding detection data can be obtained from the scene data by inquiring the ID.
Step S300, determining a plurality of associated parameters affecting the accuracy of the detection data, wherein each associated parameter is configured with a weight for showing the extent of its effect on the accuracy of the detection data under different conditions.
The sensor is limited by various factors such as self precision, environmental change and the like, and detected detection data is often inaccurate relative to the real world. Accordingly, an associated parameter is introduced to describe the impact on the accuracy of the probe data. In particular, the correlation parameter is a parameter that is correlated with the accuracy of the probe data, and that indicates the extent of its impact on the accuracy of the probe data by different weights.
In a preferred embodiment, the plurality of associated parameters may include a plurality of detection distances of the respective sensors, detection data obtained by the respective sensors, a vehicle speed of the own vehicle, a vehicle speed of the front target, a type of the front target, a movement state of the front target, and the like. The weights corresponding to the associated parameters may be determined based on weight functions adapted to different working conditions, which will be described in detail below in conjunction with a weight adjustment scheme, which will not be described here.
Step S400, determining an association threshold corresponding to the current working condition based on preset corresponding relations between each association parameter and the weight thereof and the association threshold.
As an example of the above definition of the association threshold, it is currently common to set the association threshold according to a detection error between sensors, for example, the detection error of the sensors defined in the target fusion is 3m, that is, the association threshold may be set to be greater than or equal to 3m, so long as the detection error between the sensors does not exceed 3m, the corresponding sensor detection data may be finally used in the target fusion. However, this set correlation threshold remains constant under different conditions, which ultimately does not guarantee optimal target fusion results. For example, for the case where the camera detects a front target with a longitudinal distance of 10m, the front radar detects a distance of 7m, and the right-angle radar detects a distance of 6m, if the correlation threshold is maintained for an initial 3m, the right-angle radar detection result (whose detection error with the camera is greater than 3 m) cannot participate in the target fusion. The actual working conditions are far more complex than this, for example, the allowed detection errors at the near, middle and far distances of the automatic driving vehicle are different (for example, the near distance is only allowed to be 3m, and the far distance is allowed to be 6 m), and for example, as the speed of the self-vehicle increases, the detection error of the sensor acceptable by the target fusion also gradually increases. Therefore, the fixed association threshold cannot adapt to the complex working conditions faced by the automatic driving, and in step S400, the preset correspondence between the association threshold and the association parameter and the weight thereof is given first, so as to study the scheme of dynamically adjusting the association threshold.
For example, the preset correspondence is represented by the following formula:
THR＝∑A i ×f i (B) (1)
wherein A is i Representing the value of the ith associated parameter, f i (B) And (3) representing a weight function corresponding to the ith association parameter, wherein B represents a coefficient of the weight function, and THR represents an association threshold.
Taking further the longitudinal distance of the front target as the detection data as an example, the above formula (1) can be further described as:
longitudinal distance associated threshold = longitudinal distance of target + weight 1+ longitudinal speed of target + weight 2+ type of target + weight 3+ other dimension + weight n.
The weight 1, the weight 2 and the weight 3 … … are weight values obtained based on a weight function, and the weight function may be a linear function or a nonlinear function corresponding to the complex condition of the working condition.
In this way, it is readily apparent that the associated threshold value may be changed by changing the weight, and that the particular value of the weight may in turn be adjusted by adjusting the coefficient of the weight function. The weight function is a function preset to adapt to different working conditions, so that the weight value is adapted to the working condition change to accurately reflect different influence of corresponding associated parameters on the accuracy of the detection data under different working conditions. Based on the complexity of the working condition, the weight function can be a linear function or a nonlinear function, so that the coefficient adjustment of the weight function is relatively complex, and the weight function is not always adjusted up or down, and different change states are presented according to different vehicle speeds, different distances, different target types, different target running states and the like. That is, the coefficients correspond to the possible time sizes of different working conditions, and the optimal coefficient is determined based on the association success rate, the association threshold with the highest association success rate is optimal, and the corresponding coefficient is also optimal. The following step S500 will specifically describe the association success rate.
Here, the longitudinal distance is exemplary, and an associated threshold value, such as the target longitudinal speed, may similarly be obtained. The above expression (1) is also exemplary, and the calculation expression for obtaining the correlation threshold may be a linear function or a multiple function.
And S500, adjusting the weight according to the working condition change so as to change the association threshold value to enable the association success rate of the detection data which is associated for the target fusion to reach a preset requirement.
The association success rate is the ratio of the number of frames of the detection data finally used for the target fusion to the total number of frames of the marked detection data.
With respect to equation (1) above, adjusting the set association threshold may include: and adjusting coefficients in the weight function so that the weight is adapted to the change of the working condition.
In a preferred embodiment, adjusting the coefficients in the weighting function may comprise the steps of:
1) Traversing all coefficients in a set step size within a set parameter tuning range from a given initial coefficient;
2) Determining corresponding association success rates according to each traversed coefficient;
3) And determining the corresponding coefficient when the association success rate is highest as the optimal coefficient.
The initial coefficients may be set according to sensor characteristics, or may be set based on experience, for example, according to the association success rates obtained by performing experiments under different working conditions, a group of coefficients with higher association success rates is set as association coefficients for performing target fusion again later.
The parameter adjusting range refers to a parameter adjusting range formed between a maximum value and a minimum value of the coefficient to be adjusted, the range influences the length of adjusting time, and the longer the range is, the longer the parameter adjusting time is.
The step length is the embodiment of the refinement degree of coefficient adjustment, and the shorter the step length is, the finer the parameter adjustment is. Accordingly, the number of times of overall adjustment can be determined according to the parameter adjustment range and the parameter step length.
Wherein, the parameter adjusting range and the step length can be set empirically. The larger the parameter adjusting range is, the more the sensor detection data which can be suitable for the target fusion can be, but the parameter adjusting range is not suitable to be set too large, on one hand, the adjustment efficiency is affected, on the other hand, the detection error is too large, which can mean that the obviously wrong detection data exist, and the obviously wrong detection data are not needed for the subsequent target fusion. It is generally preferred to set the step size to 0.1m, the larger the step size, the faster the associated threshold is adjusted, but the relatively worse the adjustment effect.
For the above 1) -3), in an example, for the case that the foregoing camera detection result is 10m, the distance detected by the front radar is 7m, and the distance detected by the right-angle radar is 6m, the initial association threshold determined based on the initial coefficient is 2m, at this time, only the front radar and the right-angle radar can be associated, the camera with the highest test precision cannot be associated with the other two, so that all coefficients can be traversed within a set range in a step size of, for example, 0.1m, wherein when the first set of coefficients enables the association threshold to be adjusted to 3m, the camera can be associated with the front radar, and when the second set of coefficients enables the association threshold to be adjusted to 4m, the camera, the front radar and the right-angle radar can be associated, so that the detection data of all three sensors can be used for target fusion under the current working condition; continuing to adjust to the third set of coefficients, at this time, the correlation threshold is adjusted to 5m, so that some error data other than the detection data of the camera, the front radar and the right angle radar are also correlated, and further, the correlation success rate is lowered. Therefore, when the association threshold is adjusted to 4m in this example, the corresponding association success rate is above 90%, which is the optimal association success rate in this adjustment, so the corresponding second set of coefficients should be the optimal coefficients.
In a further example, the sensor has three conditions of short distance, medium distance and long distance for the detection of the front object, in the case of short distance, the error of the radar and the camera can be within 3m, but in the case of medium distance, the error can be within 6m, and in the case of long distance, the error can be within 20m, so that the sensor is suitable for the increase of the detection distance, and the correlation threshold can be increased by an adjustable coefficient so as to successfully correlate the radar and the camera at different distances.
In a further example, the allowable errors in the upper paragraph relating to 3m, 6m and 20m are for a vehicle speed of 40km/h, whereas when the vehicle speed is 60km/h, the short distance becomes within 4m, the medium distance becomes within 5m, and the long distance becomes within 30m, so as to accommodate an increase in vehicle speed, the coefficient may also be adjusted so that the association threshold increases so that the radar and the camera can be successfully associated at different vehicle speeds.
The success rate of the association is further explained below, for example, sensor detection data of the same attribute for the same target in the relevant scene data is manually marked, so that the marked result is essentially an ideal result expected by the user to be used for target fusion; in the association threshold adjustment, each time, a corresponding association result is obtained by a different association threshold. Therefore, the coefficient of the weighting function corresponding to all the marked data (or marked as case) can be traversed through the program, and the association success rate of each group of coefficients is evaluated. From the above, it can be known whether the calculated association success rate is related to the quality of the corresponding association threshold, and the corresponding association threshold when the association success rate is highest can be determined as the optimal association threshold. However, the higher the association threshold value is, the higher the probability of correct association is, because as the association threshold value is increased, the probability of incorrect association may be increased, resulting in a decrease in the overall association success rate. For example, in the above example regarding the longitudinal distance association threshold, if the front target is a pedestrian, if the association success rate of the pedestrian target is to be ensured, the weight of the longitudinal speed of the pedestrian target needs to be reduced, and the weight of the longitudinal distance needs to be increased, and the weight of the target type needs to be increased, so that the association success rate of the pedestrian target can be improved; however, the vehicle targets may be opposite, the weight of the longitudinal speed needs to be increased, the weight of the longitudinal distance needs to be decreased, so that different association success rates can be realized, and if the vehicle targets are not regulated according to similar trends, the overall association success rate becomes low. That is, the correlation threshold corresponding to the optimal correlation success rate is different for the pedestrian target and the vehicle target, and it is further indicated that the larger the non-correlation threshold is, the higher the correlation success rate is, which is related to various aspects such as the type of the detection target.
In addition, it should be noted that, at present, the association success rate cannot be guaranteed to be 100%, because the adjustment of the weight of a certain association parameter is not necessarily proved to be a correspondingly larger or smaller threshold value in the whole association process, but the association threshold value is likely to be larger with the weight at certain speeds and certain distances, but in other speeds and distances, the association threshold value is reduced according to the weight, so that the association threshold value which is dynamically larger or smaller can guarantee the improvement of the association success rate of the whole working condition.
In this way, by analyzing the association success rate, the evaluation of the adjustment of the association threshold is substantially completed, and the optimal association threshold is determined. Further, an actual road test may be performed based on the optimal association threshold to determine its corresponding real vehicle effect.
Furthermore, based on the optimal association success rate shown by the actual road test result, each weight function coefficient corresponding to the optimal association success rate can be determined as the initial coefficient of the next target fusion. For example, when the weight coefficient is adjusted, each coefficient is traversed in a set interval and a specified step length, the optimal coefficient corresponding to the optimal correlation success rate is obtained, and the optimal coefficient is used as the initial coefficient of the next target fusion. In this way, the correlation success rate becomes an important index for adjusting the correlation threshold, and the coefficient with high correlation success rate is the optimal coefficient.
In summary, the embodiment of the invention marks all sensor IDs corresponding to the detection data with the same attribute aiming at scene data corresponding to complex working conditions such as different detection distances, different vehicle speeds and the like, so that detection data (such as distance, speed, classification and other attribute information) corresponding to each sensor can be queried through marking results, then related parameters related to the accuracy of the detection data are determined, and related thresholds are indirectly and dynamically adjusted through coefficient adjustment of a weight function of the related parameters, so that the overall target fusion effect is optimal (such as an initial related threshold is 3m and meets the basic requirement of target fusion, but 3.1m is found to be more applicable after adjustment, and then the target fusion effect is adjusted to be 3.1 m). The embodiment of the invention realizes the association threshold value of the automatic parameter adjustment sensor, so that the method is better suitable for target fusion under different working conditions.
Specifically, the parameter adjusting method aiming at target fusion in the embodiment of the invention has the following advantages:
1) The method does not depend on manual adjustment completely, solves the efficiency problem of the adjustment of the associated threshold value, and saves a great amount of parameter adjustment time. In addition, the automatic adjustment scheme of the embodiment of the invention is performed based on a large amount of multi-dimensional scene data, and compared with the dependence of a manual adjustment scheme on subjectivity of a person, the automatic adjustment scheme of the embodiment of the invention enables an adjustment mode to be more scientific. For example, the manual adjustment is highly subjective, misjudgment on the scene may occur, and the size adjustment is more random.
2) The method has strong universality, can be suitable for adjusting fusion parameters between different sensors under complex working conditions of different detection distances and different vehicle speeds, greatly shortens the time of completely manually adjusting the association threshold, is beneficial to solving the practical problems of short time and heavy tasks of vehicle road test projects, reduces a large amount of manpower and material resources, and ensures the road test quality of the automatic driving vehicle.
3) The association threshold can be increased or decreased by a smaller step size, so that the association success rate of the sensor is improved, more refined association threshold adjustment can be realized after adjustment of a large amount of data, and higher association success rate is realized. The existing scheme which completely relies on manual adjustment of the association threshold is difficult to grasp in step setting, and particularly fine adjustment based on smaller step is difficult to achieve.
4) The automatic driving typical problems such as false braking and the like caused by the wrong association can be avoided under certain complex working conditions because the association threshold is unreasonable.
5) At most, only manual data labeling is needed, and then the association threshold adjustment is completely automatic.
In addition, it should be noted that the parameter adjusting method according to the embodiment of the present invention is exemplified by automatic driving of a vehicle, but is not limited to automatic driving of a vehicle, and is equally applicable in other scenarios involving an adjusting portion of target fusion.
Fig. 2 is a schematic structural diagram of a parameter adjusting device for target fusion according to an embodiment of the present invention. As shown in fig. 2, the parameter adjusting device includes: a data acquisition module 210 configured to acquire scene data of the autonomous vehicle under the current working condition; the labeling module 220 is configured to label detection data of all sensors for detecting the same target according to the scene data; and a parameter adjustment module 230. Wherein the parameter adjustment module is configured to: determining a plurality of associated parameters affecting the accuracy of the probe data, wherein each associated parameter is configured with a weight for showing how much it affects the accuracy of the probe data under different conditions; determining an association threshold corresponding to the current working condition based on preset corresponding relation between each association parameter and weight thereof and the association threshold, wherein the association threshold is a critical value of whether detection errors of the detection data among the sensors are accepted by the target fusion; and adjusting the weight according to the working condition change so as to change the association threshold value to enable the association success rate of the detection data which is associated for the target fusion to reach a preset requirement, wherein the association success rate is the ratio of the number of frames of the detection data which is finally used for the target fusion to the total number of frames of the marked detection data.
Wherein the data acquisition module 210 may be coupled with, for example, in-vehicle sensors and associated vehicle subsystems to acquire various scene data.
Preferably, the parameter adjusting device may further include: and the evaluation module 240 is configured to evaluate the corresponding association threshold according to the association success rate. For example, an optimal correlation threshold is determined based on the optimal correlation success rate, thereby determining an optimal set of weight function parameters.
In addition, for more implementation details of the data acquisition module 210, the labeling module 220, the parameter adjustment module 230 and the evaluation module 240, and for the effect of the parameter adjustment device, reference may be made to the above-mentioned embodiments of the parameter adjustment method, which are not described herein.
Fig. 3 is a schematic structural diagram of another parameter tuning device for target fusion according to an embodiment of the present invention, where the parameter tuning device includes: a memory storing a program capable of running on a processor; and the processor is configured to execute the parameter tuning method for target fusion according to any embodiment. It should be noted that the parameter adjusting device is, for example, a computer device.
Another embodiment of the present invention further provides a target fusion system for an autonomous vehicle, which includes the parameter tuning device for target fusion according to any of the above embodiments, for example, the parameter tuning device shown in fig. 2 or fig. 3.
Another embodiment of the present invention also provides a machine-readable storage medium having stored thereon instructions for causing a machine to perform a method of tuning an autonomous vehicle as described above.
The method for adjusting parameters in the computer device and the machine-readable storage medium can be understood with reference to the above embodiments, and will not be described herein. The computer device and machine-readable storage medium are described further below primarily in conjunction with application scenarios.
It will be apparent to those skilled in the art that embodiments of the present invention may be provided as a method, apparatus (device or system), or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the invention may take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (devices or systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In a typical configuration, a computer device includes one or more processors (CPUs), memory, input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the invention may take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The foregoing details of the optional implementation of the embodiment of the present invention have been described in detail with reference to the accompanying drawings, but the embodiment of the present invention is not limited to the specific details of the foregoing implementation, and within the scope of the technical concept of the embodiment of the present invention, various simple modifications may be made to the technical solution of the embodiment of the present invention, for example, the execution sequence of the steps is changed, and these simple modifications all fall within the protection scope of the embodiment of the present invention.
In addition, the specific features described in the above embodiments may be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, various possible combinations of embodiments of the present invention are not described in detail.
In addition, any combination of various embodiments of the present invention may be performed, so long as the concept of the embodiments of the present invention is not violated, and the disclosure of the embodiments of the present invention should also be considered.
Claims (10)
1. A method for adjusting parameters for target fusion, which is characterized by comprising the following steps:
acquiring scene data of an automatic driving vehicle under the current working condition;
labeling detection data with the same attribute from a plurality of sensors in the scene data;
determining a plurality of associated parameters affecting the accuracy of the probe data, wherein each associated parameter is configured with a weight for showing how much it affects the accuracy of the probe data under different conditions;
determining an association threshold corresponding to the current working condition based on preset corresponding relation between each association parameter and weight thereof and the association threshold, wherein the association threshold is a critical value of whether detection errors of the detection data among the sensors are accepted by the target fusion; and
and adjusting the weight according to the working condition change so as to change the association threshold value to enable the association success rate of the detection data which is associated for the target fusion to reach a preset requirement, wherein the association success rate is the ratio of the number of frames of the detection data which is finally used for the target fusion to the total number of frames of the marked detection data.
2. The method of claim 1, wherein annotating detection data in the scene data that is identical in attribute from a plurality of sensors comprises:
marking the identification numbers (ID) of the sensors of the detection data, which are determined by the user and are associated with the same attribute of the same detection target, according to the scene data, wherein the IDs of the different sensors when detecting the same target are different; and
inquiring detection data of a corresponding sensor from the scene data based on the marked sensor ID and marking;
preferably, after the labeling of the detection data of all the sensors detecting the same target is completed, the labeling content is presented based on the table of the sensor IDs.
3. The method of claim 1, wherein the plurality of associated parameters includes a plurality of sensors of a detection distance, detection data obtained by the respective sensors, a vehicle speed of a front target, a type of the front target, and a motion state of the front target.
4. The method for adjusting parameters for target fusion according to claim 1, wherein the preset correspondence is represented by the following formula:
THR＝∑A i ×f i (B)
wherein A is i Representing the value of the ith associated parameter, f i () And (3) representing a weight function corresponding to the ith association parameter, wherein B represents a coefficient of the weight function, and THR represents an association threshold.
5. The method of claim 4, wherein adjusting the weights to adapt to a change in operating conditions comprises: and adjusting coefficients in the weight function so that the weight is adapted to the driving working condition to change.
6. The method of claim 5, wherein adjusting coefficients in the weighting function comprises:
traversing all coefficients in a set step size within a set parameter tuning range from a given initial coefficient;
determining corresponding association success rates according to each traversed coefficient; and
and determining the corresponding coefficient when the association success rate is highest as the optimal coefficient.
7. A machine-readable storage medium having instructions stored thereon for causing a machine to perform the method for targeting fusion according to any of the claims 1 to 6.
8. A targeting fusion parameter tuning device, characterized in that the targeting fusion parameter tuning device comprises:
the data acquisition module is configured to acquire scene data of the automatic driving vehicle under the current working condition;
the labeling module is configured to label detection data with the same attribute from a plurality of sensors in the scene data; and
a parameter adjustment module configured to:
determining a plurality of associated parameters affecting the accuracy of the probe data, wherein each associated parameter is configured with a weight for showing how much it affects the accuracy of the probe data under different conditions;
determining an association threshold corresponding to the current working condition based on preset corresponding relation between each association parameter and weight thereof and the association threshold, wherein the association threshold is a critical value of whether detection errors of the detection data among the sensors are accepted by the target fusion; and
and adjusting the weight according to the working condition change so as to change the association threshold value to enable the association success rate of the detection data which is associated for the target fusion to reach a preset requirement, wherein the association success rate is the ratio of the number of frames of the detection data which is finally used for the target fusion to the total number of frames of the marked detection data.
9. A targeting fusion parameter tuning device, characterized in that the targeting fusion parameter tuning device comprises:
a memory storing a program capable of running on a processor; and
the processor configured to implement the parametric approach to target fusion of any one of claims 1 to 6 when executing the program.
10. A target fusion system for an autonomous vehicle, comprising a parameter tuning device for target fusion according to claim 8 or 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210880327.7A CN117490710A (en) | 2022-07-25 | 2022-07-25 | Parameter adjusting method and device for target fusion and target fusion system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210880327.7A CN117490710A (en) | 2022-07-25 | 2022-07-25 | Parameter adjusting method and device for target fusion and target fusion system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117490710A true CN117490710A (en) | 2024-02-02 |
Family
ID=89674961
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210880327.7A Pending CN117490710A (en) | 2022-07-25 | 2022-07-25 | Parameter adjusting method and device for target fusion and target fusion system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117490710A (en) |
-
2022
- 2022-07-25 CN CN202210880327.7A patent/CN117490710A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112256590B (en) | Virtual scene effectiveness judgment method and device and automatic driving system | |
| CN112540365B (en) | Evaluation method, device, equipment and storage medium | |
| US11080544B1 (en) | Method and device for calibrating pitch of camera on vehicle and method and device for continual learning of vanishing point estimation model to be used for calibrating the pitch | |
| CN114973198A (en) | Course angle prediction method and device of target vehicle, electronic equipment and storage medium | |
| CN112309004B (en) | Virtual deceleration strip generation method and device, vehicle and storage medium | |
| CN116543271A (en) | Method, device, electronic equipment and medium for determining target detection evaluation index | |
| CN117490710A (en) | Parameter adjusting method and device for target fusion and target fusion system | |
| CN116148821A (en) | Laser radar external parameter correction method and device, electronic equipment and storage medium | |
| CN112346046B (en) | Single-target tracking method and system based on vehicle-mounted millimeter wave radar | |
| CN115140029A (en) | Safety capability detection method and device for automatic driving automobile | |
| CN115031755A (en) | Automatic driving vehicle positioning method and device, electronic equipment and storage medium | |
| CN112784705A (en) | Vehicle side edge determining method and device | |
| CN112798293B (en) | Test method and device of lane departure early warning system | |
| CN114705121B (en) | Vehicle pose measurement method and device, electronic equipment and storage medium | |
| CN116626630B (en) | Object classification method and device, electronic equipment and storage medium | |
| CN117268424B (en) | Multi-sensor fusion automatic driving hunting method and device | |
| CN113763483B (en) | Method and device for calibrating pitch angle of automobile data recorder | |
| CN116381698B (en) | Road remains detection method and device and electronic equipment | |
| US11166003B1 (en) | Dynamic vibration sensor optics distortion prediction | |
| CN117572335B (en) | Updating method and device for laser positioning confidence coefficient and electronic equipment | |
| CN115284809B (en) | Intelligent internet fleet active suspension control method and system and computer equipment | |
| CN117854058A (en) | Target perception method, device, equipment and storage medium | |
| CN117665796A (en) | Target association method, target identification method, corresponding medium and system | |
| CN116843888A (en) | Target positioning method and device and electronic equipment | |
| CN117079236A (en) | Target detection method and device, electronic equipment and storage medium |
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 |