A comprehensive quality measurement method for a radar-vision fusion system

Abstract

The invention relates to the field of urban traffic technology, and specifically to a comprehensive quality determination method for a radar-vision fusion system, comprising: S1, acquiring radar-vision fusion data, calibrating parameters of the radar-vision fusion data, and determining the position of an identified vehicle; S2, judging the position of the vehicle, and determining the positioning integrity and trajectory integrity of the vehicle; S3, when the positioning integrity and trajectory integrity of the vehicle are greater than a preset integrity threshold, extracting vehicle characteristics of the vehicle; S4, acquiring traffic parameters corresponding to the radar-vision fusion data when the vehicle is moving, and extracting traffic characteristics corresponding to the traffic parameters; S5, acquiring traffic events corresponding to intersections when the vehicle is moving, and acquiring event characteristics; S6, based on the acquired vehicle characteristics, traffic characteristics, and event characteristics, fusing the vehicle characteristics, traffic characteristics, and event characteristics, and determining the comprehensive evaluation effect of the fused radar-vision fusion data; thereby improving the accuracy and reliability of the quality evaluation of the radar-vision fusion system.

Description

Technical Field

[0001] The invention relates to the field of urban traffic technology, and in particular to a comprehensive quality determination method for a radar-vision fusion system. Background Art

[0002] The reliability and efficiency of the highway roadside intelligent fusion system technology are crucial and related to road traffic safety. It is necessary to use a fast, accurate, simple and reliable evaluation solution to evaluate its performance. The radar-vision fusion system is an intelligent sensor designed for vehicle-road collaboration in the 5G era. It is a traffic sensor that integrates cameras, millimeter-wave radars and high-performance processors. The radar-vision fusion system integrates video and radar equipment. The video equipment can detect the type of traffic participants and has a long detection distance. The radar can accurately detect the target location, speed and other information and is not affected by weather conditions. Therefore, the radar-vision fusion system has significant advantages as a new means of intelligent highway roadside perception equipment.

[0003] However, the existing technology for real-time positioning of vehicles has equipment errors and feature point drift during vehicle movement, resulting in errors in the position coordinates of the same vehicle. Summary of the invention

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A comprehensive quality determination method for a radar-vision fusion system, comprising:

[0006] S1, obtaining radar-visual fusion data, calibrating parameters of the radar-visual fusion data, and determining the position of the identified vehicle.

[0007] S2, judging the position of the vehicle and determining the positioning integrity and trajectory integrity of the vehicle.

[0008] S3, when the positioning integrity and trajectory integrity of the vehicle are greater than a preset integrity threshold, extracting vehicle features of the vehicle.

[0009] S4, obtaining the traffic parameters corresponding to the radar and visual fusion data when the vehicle is moving, and extracting the traffic features corresponding to the traffic parameters.

[0010] S5, obtaining a traffic event corresponding to the intersection when the vehicle is moving, and obtaining event features.

[0011] S6, based on the acquired vehicle characteristics, traffic characteristics, and event characteristics, the vehicle characteristics, traffic characteristics, and event characteristics are fused and processed to determine the comprehensive evaluation effect of the fused radar-vision fusion data.

[0012] The beneficial effects of the present invention are as follows: 1. The present invention uses the method of judging the position of the vehicle to determine the positioning integrity and trajectory integrity of the vehicle, thereby improving the positioning effect of the vehicle and extracting the vehicle characteristics, event characteristics and traffic characteristics of the vehicle at the same time; constructing a judgment matrix based on the acquired vehicle characteristics, traffic characteristics and event characteristics, and calculating the relative weights between the various elements; determining the priority of the elements of vehicle characteristics, traffic characteristics and event characteristics, and performing consistency checks; improving the quality and reliability of the radar-visual fusion data, standardizing the judgment matrix, constructing a fuzzy evaluation matrix, and performing consistency checks to confirm the optimal consistency judgment matrix; determining the comprehensive evaluation effect of the fused radar-visual fusion data based on the fuzzy evaluation matrix and the optimal consistency judgment matrix; making the final radar-visual data able to reflect various factors, making the evaluation results more comprehensive and objective. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0014] Figure 1 The present invention is a flow chart of a comprehensive quality determination method of a radar-vision fusion system.

[0015] Figure 2 The present invention is a flow chart of step S6 of a method for comprehensively determining the quality of a radar-vision fusion system. DETAILED DESCRIPTION

[0016] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. If no specific techniques or conditions are specified in the embodiments, the techniques or conditions described in the literature in the art or the product specifications are used.

[0017] See also Figure 1 , a comprehensive quality determination method for a radar-vision fusion system, comprising:

[0018] S1, obtaining radar-visual fusion data, calibrating parameters of the radar-visual fusion data, and determining the position of the identified vehicle.

[0019] S2, judging the position of the vehicle and determining the positioning integrity and trajectory integrity of the vehicle.

[0020] S3, when the positioning integrity and trajectory integrity of the vehicle are greater than a preset integrity threshold, extracting vehicle features of the vehicle.

[0021] S4, obtaining the traffic parameters corresponding to the radar and visual fusion data when the vehicle is moving, and extracting the traffic features corresponding to the traffic parameters.

[0022] S5, obtaining a traffic event corresponding to the intersection when the vehicle is moving, and obtaining event features.

[0023] S6, based on the acquired vehicle characteristics, traffic characteristics, and event characteristics, the vehicle characteristics, traffic characteristics, and event characteristics are fused and processed to determine the comprehensive evaluation effect of the fused radar-vision fusion data.

[0024] First, two video acquisition devices and a millimeter microwave radar are set up at the intersection to obtain the video data and radar data of the vehicles at the intersection respectively, and the radar data and video data are fused according to the calibrated position points to identify the vehicles appearing at the intersection and the trajectory of the vehicles moving. At the same time, the lane corresponding to the vehicle on the moving trajectory is determined to determine the accuracy of lane recognition. Based on the identified vehicles, the vehicle features are extracted from the vehicles, and the various elements in the vehicle features are processed to determine the correlation between the various elements in the vehicle features and the corresponding priorities between the vehicle features. At the same time, it is necessary to identify the event features at the corresponding positions of the vehicle features. The event features are expressed as The features corresponding to wrong-way events, line-crossing events, lane-changing events, parking events, vehicle-to-vehicle crossing events, low-speed driving events, continuous lane-changing events, congestion events, pedestrian events, emergency lane occupation events, speeding events, and spilled objects events are mainly used to identify events appearing in the current video so that they can be handled in a timely manner when they occur; at the same time, the traffic features corresponding to the traffic parameters at this time are determined. The traffic features represent the features corresponding to traffic indicators such as vehicle speed, traffic volume, and occupancy rate. The traffic features, event features, and vehicle features are integrated and evaluated to determine the driving conditions at the intersection at this time, and a corresponding evaluation is given, which is used as an evaluation of the efficiency and quality of the current video recognition.

[0025] The detection range set during detection is generally based on the location where the video acquisition device is installed, which is 20-350m in the shooting direction corresponding to the starting point. The distances between the video acquisition devices relative to the road sections are also different according to the different roads, so as to identify the form of vehicles at each intersection.

[0026] The positioning integrity and trajectory integrity in step S2 are expressed as:

[0027] Positioning integrity: Evaluate whether the radar and vision data are consistent and check for sensor failures or abnormal data. For example, if the radar and vision detect the same vehicle at very different locations, there may be a sensor failure.

[0028] Trajectory integrity: Tracking the continuous position updates of the vehicle ensures the trajectory is consistent and reasonable. If there are sudden jumps or unreasonable changes in the trajectory, it may indicate data anomalies or sensor problems.

[0029] Positioning integrity includes obtaining the positioning accuracy deviation during positioning and the precise bit matching of the positioning data at this time to determine the integrity of the positioning data.

[0030] The positioning accuracy deviation is expressed as: obtaining the latitude and longitude reference value of the vehicle-mounted device in the vehicle and the latitude and longitude detection value of the vehicle in the radar and vision fusion data, and determining the number of moments corresponding to the latitude and longitude detection values ​​at this time to determine the positioning accuracy deviation.

[0031]

[0032] Among them, P ad is the positioning accuracy deviation, A1 k is the latitude and longitude reference value of the vehicle-mounted device at time k, A2 k is the latitude and longitude detection value of the vehicle in the radar-visual fusion data at time k, k = 1, 2, ..., K, K is the number of time points corresponding to the longitude and latitude detection values.

[0033] At the same time, determine the vehicle's positioning data matching accuracy and positioning data integrity. The positioning data matching accuracy is expressed as the number of matching errors between the output trajectory and the target vehicle in the radar vision system and the sample size; the positioning data integrity is expressed as the number of positioning data of the test vehicle in the radar vision system and the number of theoretical positioning data from the time of appearance to departure:

[0034]

[0035] Among them, P dm is the accurate position of the positioning data match, P di is the integrity of the positioning data; M is the number of positioning data of the test vehicle in the radar-vision fusion system, and N is the theoretical number of positioning data from the time it appears in the radar-vision system to the time it leaves.

[0036] The method of obtaining the track integrity is as follows: obtaining the detection range length of the radar vision equipment, the number of tracks and the length of the tracks, and determining the track integrity.

[0037]

[0038] Among them, TI is the track integrity, L is the detection range length of the radar equipment, A is the number of tracks, and I a is the length of the ath trajectory, a=1,2,3,......,A.

[0039] The integrity of the vehicle identification trajectory is obtained by calculating the data on different lanes.

[0040] If the positioning integrity and the trajectory integrity are less than a preset integrity threshold, the collected data is filled based on the corresponding data window during collection to determine the filled data.

[0041] like Figure 2 As shown, the processing method for performing the fusion processing in step S6 includes:

[0042] S61, based on the acquired vehicle characteristics, traffic characteristics, and event characteristics, construct a judgment matrix, extract each element in the vehicle characteristics, traffic characteristics, and event characteristics, calculate the relative weights between the elements of each vehicle characteristic, traffic characteristic, and event characteristic, and determine the value of each element in the judgment matrix.

[0043] S62, determining the priorities of the elements of vehicle characteristics, traffic characteristics, and event characteristics based on the relative weights among the elements of vehicle characteristics, traffic characteristics, and event characteristics.

[0044] S63, based on the overall priority of vehicle characteristics, traffic characteristics, and event characteristics, a consistency check is performed on the judgment matrix.

[0045] The values ​​of each element in the constructed judgment matrix are the relative weights between the elements of vehicle characteristics, traffic characteristics, and event characteristics, which are used to determine whether the currently set weights are accurate. At the same time, it is necessary to determine whether the relative weights of the elements of each vehicle characteristic, traffic characteristic, and event characteristic can meet the consistency test to find the most suitable judgment matrix.

[0046] The elements of event characteristics include parking event detection rate, parking event recognition accuracy, parking event recognition response time, pedestrian event detection rate, pedestrian event recognition accuracy, pedestrian event recognition response time, congestion event recognition detection rate, congestion event recognition accuracy, congestion event recognition response time, spilled object event detection rate, spilled object event recognition accuracy, and spilled object event recognition response time.

[0047] The elements of vehicle characteristics include longitude and latitude positioning deviation (outside the tunnel) / lane determination deviation (in the tunnel), positioning frequency, positioning data quality, license plate recognition accuracy, vehicle model recognition accuracy, and vehicle color recognition accuracy.

[0048] The elements of traffic characteristics include protection level, detection range, storage function, environmental adaptability, speed measurement accuracy deviation (in tunnel / outside tunnel), vehicle counting accuracy deviation, overall traffic operation status analysis capability, individual characteristic vehicle analysis capability, driving behavior analysis capability, traffic flow prediction accuracy, congestion event prediction accuracy, and average vehicle speed prediction accuracy.

[0049] The specific display methods of vehicle characteristics, event characteristics, and traffic characteristics are shown in Table 1, wherein the elements of vehicle characteristics, event characteristics, and traffic characteristics are numbered, and are divided into basic capabilities, event recognition capabilities, vehicle positioning and feature recognition capabilities, traffic data analysis and understanding capabilities, and traffic status prediction capabilities according to different feature acquisitions. The corresponding values ​​in Table 1 are obtained based on the preset values ​​of the current device, and the data of relevant accuracy are recognized in advance using an image recognition algorithm. The main problem solved by the present invention is how to determine the weights corresponding to the current different features or index values ​​after completing the data analysis and calculation, so as to adjust the system's evaluation effect on the radar data.

[0050] Table 1 Evaluation index system of radar and vision fusion system

[0051]

[0052]

[0053] For the set vehicle features, event features, and traffic features, the priority of the event features is higher than the priority of the vehicle features, which is higher than the priority of the traffic features.

[0054] And the priority of traffic features is half of that of vehicle features.

[0055] The relative importance of the elements in each vehicle feature, traffic feature, and event feature is determined based on the probability of each element co-occurring in the identified traffic events in the database.

[0056] The priorities of the elements of vehicle characteristics, event characteristics, and traffic characteristics in step S62 are:

[0057] The ratio of the product root of the relative importance of each factor to the product root is taken as the priority of each factor. Similarly, the overall priority of vehicle characteristics, event characteristics, and traffic characteristics is the sum of the priorities of each factor in the vehicle characteristics in the overall data.

[0058] According to the relative importance of each factor, an n×n judgment matrix is ​​constructed, and the relative importance of each factor is used as the index value of the judgment matrix. Based on the corresponding relative importance value in the judgment matrix, the priority of each factor is determined.

[0059]

[0060] Among them, W o is the nth root of the product of the oth element in the judgment matrix, n is the number of indicator values ​​in the judgment matrix, T ot is the relative importance value between the oth element and the tth element in the judgment matrix, o=1,2,......,n, t=1,2,3,......,n;

[0061]

[0062] Among them, ω o is the priority of the oth element among vehicle features, event features, and traffic features, It is the sum of the nth root of the product of vehicle characteristics, event characteristics, and traffic characteristics.

[0063] The method for obtaining the overall priority of vehicle characteristics, event characteristics, and traffic characteristics is the same as the method for obtaining the priority of each element in the vehicle characteristics, event characteristics, and traffic characteristics.

[0064] According to the specific contents of vehicle characteristics, event characteristics, and traffic characteristics, other priorities are set for the specific time of occurrence to quantify each specific event that occurs. At the same time, according to the priority of each element in the obtained vehicle characteristics, traffic characteristics, and event characteristics, the currently identified event and the priority of the identified event can be known, so that when the corresponding characteristics appear, timely processing and evaluation can be carried out.

[0065] As shown in Table 2-7, the priorities of corresponding vehicle characteristics, event characteristics, and traffic characteristics and the overall priorities of vehicle characteristics, event characteristics, and traffic characteristics are shown; including the five first-level indicators divided by vehicle characteristics, event characteristics, and traffic characteristics and the judgment matrix of the second-level indicators corresponding to the five first-level indicators, so as to determine the corresponding judgment matrix of vehicle characteristics, event characteristics, and traffic characteristics during calculation.

[0066] Table 2 Judgment matrix of the first-level indicator layer

[0067]

[0068]

[0069] Table 3 Secondary indicator judgment matrix of basic capability L1

[0070]

[0071] Table 4 Secondary indicator judgment matrix of event recognition capability L2

[0072]

[0073]

[0074] Table 5 Secondary index judgment matrix of vehicle positioning and feature recognition capability L3

[0075] L3 L31 L32 L33 L34 L35 L36 Priority L31 1 5 / 2 5 / 3 1 5 / 3 5 / 2 0.25 L32 2 / 5 1 2 / 3 2 / 5 2 / 3 1 0.1 L33 3 / 5 3 / 2 1 3 / 5 1 3 / 2 0.15 L34 1 5 / 2 5 / 3 1 5 / 3 5 / 2 0.25 L35 3 / 5 3 / 2 1 3 / 5 1 3 / 2 0.15 L36 2 / 5 1 2 / 3 2 / 5 2 / 3 1 0.1

[0076] Table 6 Secondary indicator judgment matrix of traffic data analysis and understanding ability L4

[0077]

[0078] Table 7 Secondary index judgment matrix of traffic state prediction capability L5

[0079]

[0080] After obtaining the priority, it is also necessary to verify the relative importance of each element in the vehicle characteristics, event characteristics, and traffic characteristics to determine whether the relative importance value is set correctly. The relative importance value set in the present invention is between 0 and 1, which is used to determine whether each element is slightly more important than other elements, so as to obtain the overall priority according to the relative importance of each element.

[0081] The method of performing consistency check on the judgment matrix in step S63 is:

[0082] Based on the overall priority of vehicle features, traffic features, and event features, the largest characteristic root in the judgment matrix is ​​extracted.

[0083] Among them, λ max is the maximum value of the characteristic root in the judgment matrix; n represents the number of values ​​in the judgment matrix; CI is the general consistency index of the judgment matrix. The larger the CI value, the greater the degree of deviation from complete consistency, and the smaller the CI value, the better the consistency.

[0084] When max =n, CI is 0, and the judgment matrix is ​​completely consistent. When the general consistency index CI of the judgment matrix M is larger, its complete consistency is worse. When CI is less than 0.1, the judgment matrix M meets the consistency requirements. If CI is greater than or equal to 0.1, it is necessary to re-compare the elements in the judgment matrix. Since the order of the judgment matrix will affect the consistent error of the acceptable judgment, when the order n of the judgment matrix is ​​larger, the consistency of the judgment matrix will be worse. At this time, the average random consistency index RI of the judgment matrix can be introduced to relax the consistency requirements for high-order judgment matrices.

[0085]

[0086] Among them, CI is the general consistency index of the judgment matrix; CR is the random consistency ratio of the judgment matrix; RI is the average random consistency index of the judgment matrix. The average random consistency index set in the judgment matrix is ​​obtained from the preset database.

[0087] For the judgment matrices of order 1 to 9, the average random consistency index RI is shown in Table 8 below.

[0088] Table 81-RI value of 9-order judgment matrix

[0089] Order n 1 2 3 4 5 6 7 8 9 RI 0.00 0.00 0.58 0.96 1.12 1.24 1.32 1.41 1.45

[0090] After consistency checking of the above judgment matrices, it was found that the above judgment matrices have good complete consistency and can be used to evaluate and analyze the capabilities of the radar and vision fusion system.

[0091] The comprehensive evaluation effect of the fused radar-vision fusion data determined in step S6 includes:

[0092] The index values ​​of the judgment matrix are standardized, the standardized evaluation index values ​​are determined, and a fuzzy evaluation matrix is ​​constructed according to the obtained evaluation index values; the judgment matrix of the fuzzy evaluation matrix is ​​determined, and the consistency check of the fuzzy evaluation matrix is ​​performed to confirm the optimal consistency judgment matrix corresponding to the fuzzy evaluation matrix.

[0093] Assume that there are n judgment matrix index values ​​that form the evaluation index for all m solutions; {x(i,j)|i=1,2,…n; j=1,2,…m}, and the index values ​​x(i,j) of the judgment matrix are all non-negative; at the same time, the index values ​​of the judgment matrix need to be normalized to determine the fuzzy evaluation matrix composed of the evaluation index, as shown below.

[0094] The standardization processing of the index values ​​of the judgment matrix includes: standardization processing of the larger the better type of index, standardization processing of the smaller the better type of index, and standardization processing of the medium the better type of index.

[0095] The standardized processing of the larger the better type of indicators is as follows:

[0096] r(i,j)=x(i,j) / [x max (i)+x min (i)];

[0097] The standardization of the smaller the better indicator is as follows:

[0098] r(i,j)=[x max (i)+x min (i)-x(i,j)] / [x max (i)+x min (i)];

[0099] The standardized processing of the Yuezhongyueyu type indicators is as follows:

[0100]

[0101] Among them, x min (i), x max (i), xmid (i) represents the minimum, maximum and middle value of the i-th index value, r(i,j) is the standardized evaluation index value, x(i,j) is the index value of the judgment matrix, and x mid (i,j) is the middle value between the i-th index value and the j-th index value, x max (i,j) is the maximum value of the i-th index value and the j-th index value, j=1,2,......,m, i=1,2,3,......,n.

[0102] The bigger the better type of indicators, which usually indicate some kind of positive benefit or increase in capacity. For traffic management, they may include:

[0103] Monitoring range: The larger the radar’s monitoring range, the wider the traffic area it can cover, which is more beneficial for traffic monitoring and accident prevention.

[0104] Detection distance: The longer the radar's detection distance, the stronger its ability to detect distant vehicles or obstacles in advance, which helps improve traffic safety.

[0105] Signal strength: When radar is monitoring traffic flow or speed, the stronger the signal strength, the clearer the reflected signal received by the radar and the higher the measurement accuracy.

[0106] The smaller the better. This type of indicator usually indicates a reduction in some negative effect or error. For traffic management, it may include:

[0107] Error rate: Whether it is speed measurement or distance measurement, the smaller the error rate, the more accurate the measurement result is, and the more beneficial it is for the analysis and application of traffic data.

[0108] False alarm rate: When radar is used for traffic monitoring, the lower the false alarm rate (i.e. the probability of misjudging a non-target object as a target object), the more stable and reliable the system is, reducing unnecessary interference and false alarms.

[0109] System loss: The smaller the loss of the radar system, the higher the energy conversion efficiency and the better the system performance, which helps to extend the service life of the equipment and reduce energy consumption.

[0110] The more in the middle, the better the index. This type of index usually means that a parameter reaches the best effect within a certain middle value or range. For traffic parameters, parameters that directly meet the characteristics of "the more in the middle, the better" may not be common, but they can be compared from related concepts:

[0111] Speed ​​distribution: In traffic flow management, although the speed of a single vehicle is not necessarily “the higher the speed, the better”, when the speed distribution of the entire traffic flow tends to be reasonable (such as close to the road design speed and with little fluctuation), traffic efficiency and safety are often higher; the “reasonable” speed distribution here can be regarded as an indirect concept of “the higher the speed, the better”.

[0112] Signal-to-noise ratio: Although the higher the signal-to-noise ratio, the better (indicating better signal quality), in some cases, when the signal-to-noise ratio reaches a certain level, further improving the signal-to-noise ratio may bring diminishing marginal benefits. However, this does not fully meet the definition of "the more the better", because in theory the higher the signal-to-noise ratio, the better; but from a practical application perspective, when the signal-to-noise ratio reaches a certain threshold, it may no longer be necessary to pay too much attention to its further improvement.

[0113] At this time, the traffic characteristics, event characteristics, and vehicle characteristics are divided into three types of indicators according to the current collection scene to determine what kind of standardization processing needs to be adopted at this time. For example, when the currently collected traffic characteristics require a larger scene analysis and the accuracy does not need to be considered, the larger the better indicator can be used. If the scene does not need to be considered carefully and only the accuracy is emphasized, the smaller the better indicator should be used. If both the scene and the accuracy need to be considered, the medium the better indicator should be used to standardize the current indicator.

[0114] The fuzzy evaluation matrix also includes the following when performing consistency check:

[0115] According to the fuzzy evaluation matrix R = {r(i,j)} n×m To construct the judgment matrix B of the fuzzy evaluation matrix = (b ij ) n×n , the sample standard deviation of each evaluation index can be used Reflect the influence of each evaluation index on the comprehensive evaluation, and use it to construct the judgment matrix B = (b ij ) n×n ;in, is the sample mean of each evaluation index.

[0116] The judgment matrix of the 1-9 level judgment scale of the fuzzy evaluation matrix is ​​shown below.

[0117]

[0118] Among them, s min 、s max is the minimum and maximum value of s(i), i=1,2,…,n; b m =min{9,int[s max / s min +0.5]} is the relative importance parameter value, b ijIt is the index value of the judgment matrix corresponding to the fuzzy evaluation matrix, s(j) represents the sample standard deviation of the evaluation index when it is the jth index, j = 1, 2, ..., m.

[0119] For the judgment matrix B = (b ij ) n×n Perform consistency check and calculate weight w i ,i=1,2,…n;According to the judgment matrix B=(b ij ) n×n The definition of is as follows.

[0120] b ij =w i / w j ,(i,j=1,2,…n);

[0121] Among them, the weight w i ,i=1,2,…n, and w i >0 and

[0122] And the above formula b ij =w i / w j ,(i,j=1,2,…n) has the following properties.

[0123] b ii =w i / w i =1;

[0124] b ji =w j / w i =1 / b ij ;

[0125] b ij b jk =(w i / w j )·(w j / w k )=w i / w k =b ik ;

[0126] Among them, the above formulas are respectively the identity of the judgment matrix, the reciprocal property of the judgment matrix, and the consistency of the judgment matrix.

[0127] If for the judgment matrix b ij =w i / w j ,(i,j=1,2,…n) has Then the judgment matrix B=(b ij ) n×nhas complete consistency. But in a complex real environment, B=(b ij ) n×n It does not necessarily have to meet the complete consistency condition, but it must have satisfactory consistency. If the judgment matrix B = (b ij ) n×n If there is no satisfactory consistency, the judgment matrix needs to be modified.

[0128] The optimal consistency judgment matrix is ​​expressed as:

[0129] Assume that the judgment matrix B = (b ij ) n×n The correction matrix is ​​C = (c ij ) n×n , the weight value of each element of the modified matrix C is still w i , i=1,2,…n, then the following equation is satisfied: C=(c ij ) n×n That is, the judgment matrix B = (b ij ) n×n The optimal consistency judgment matrix is ​​as follows:

[0130]

[0131] Among them, there are c ii =1,(i=1,2,…,n),1 / c ji =c ij ,c ij ∈[b ij -db ij ,b ij +db ij ],w i >0 and In the above formula, the objective function minCIC(n) is the consistency index coefficient (ConsistencyIndexCoefficient) of the judgment matrix, d is a non-negative parameter, and d∈[0,0.5].

[0132] If the judgment matrix B=(b ij ) n×n If the order n is different, the value of the consistency index coefficient CIC(n) will also be different. The average random consistency index coefficient RIC(n) of the judgment matrix can be introduced to measure whether the judgment matrix has satisfactory consistency. The value of the average random consistency index coefficient is shown in the following table.

[0133] Table 9 Average random consistency index coefficient RIC(n) values ​​of the judgment matrix

[0134] Order n 3 4 5 6 7 8 9 RIC(n) 0.578 0.487 0.451 0.377 0.321 0.308 0.277

[0135] The weight value w of each evaluation indexi Multiplying and accumulating with the evaluation index r(i,j) can obtain the index value of fuzzy comprehensive evaluation And decision analysis can be carried out according to the comprehensive evaluation index value.

[0136] According to the index value finally obtained by the current fuzzy evaluation matrix, it is possible to judge the current processing of different characteristic elements and determine whether the relative importance of the current settings conforms to the processing logic. The final index value can also reflect the accuracy of the current processing of radar-vision fusion data, thereby improving the efficiency and accuracy of system processing.

[0137] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention and they are still covered by the protection scope of the present invention.

Claims

1. A comprehensive quality measurement method for a radar-vision fusion system, characterized in that: include: S1, obtaining radar-visual fusion data, calibrating parameters of the radar-visual fusion data, and determining the position of the identified vehicle; S2, judging the position of the vehicle and determining the positioning integrity and trajectory integrity of the vehicle; S3, when the positioning integrity and trajectory integrity of the vehicle are greater than a preset integrity threshold, extracting vehicle features of the vehicle; S4, obtaining the traffic parameters corresponding to the radar-visual fusion data when the vehicle is moving, and extracting the traffic characteristics corresponding to the traffic parameters; S5, obtaining a traffic event corresponding to an intersection when the vehicle is moving, and obtaining event features; S6, based on the acquired vehicle characteristics, traffic characteristics, and event characteristics, the vehicle characteristics, traffic characteristics, and event characteristics are fused and processed to determine the comprehensive evaluation effect of the fused radar-visual fusion data; The processing methods for fusion processing include: S61, constructing a judgment matrix based on the acquired vehicle characteristics, traffic characteristics, and event characteristics, extracting each element in the vehicle characteristics, traffic characteristics, and event characteristics, calculating the relative weights between the elements of each vehicle characteristic, traffic characteristic, and event characteristic, and determining the value of each element in the judgment matrix; S62, determining the priorities of the elements of the vehicle characteristics, traffic characteristics, and event characteristics based on the relative weights between the elements of the vehicle characteristics, traffic characteristics, and event characteristics; S63, based on the overall priority of vehicle characteristics, traffic characteristics, and event characteristics, the judgment matrix is ​​checked for consistency; The priorities of the elements that determine vehicle characteristics, event characteristics, and traffic characteristics are: According to the relative importance of each factor, construct The judgment matrix is ​​constructed, and the relative importance of each factor is used as the index value of the judgment matrix. Based on the corresponding relative importance value in the judgment matrix, the priority of each factor is determined; ； in, is the nth root of the product of the oth element in the judgment matrix, is the number of indicator values ​​in the judgment matrix, is the relative importance value of the oth element and the tth element in the judgment matrix, , ; ； in, is the priority of the oth element among vehicle features, event features, and traffic features, It is the sum of the nth root of the product of vehicle characteristics, event characteristics, and traffic characteristics; The method of performing consistency check on the judgment matrix in step S63 is: Based on the overall priority of vehicle features, traffic features, and event features, the largest characteristic root in the judgment matrix is ​​extracted. ; in, is the maximum value of the characteristic root in the judgment matrix; n represents the number of indicator values ​​in the judgment matrix, and CI is the general consistency index of the judgment matrix; ； Among them, CI is the general consistency index of the judgment matrix; CR is the random consistency ratio of the judgment matrix; RI is the average random consistency index of the judgment matrix.

2. A comprehensive quality measurement method for a radar-visual fusion system according to claim 1, characterized in that: Positioning integrity includes obtaining the positioning accuracy deviation during positioning and the precise bit matching of the positioning data at this time to determine the integrity of the positioning data; The positioning accuracy deviation is expressed as follows: obtain the latitude and longitude reference value of the vehicle-mounted device in the vehicle and the latitude and longitude detection value of the vehicle in the radar-vision fusion data, and determine the number of moments corresponding to the latitude and longitude detection values ​​at this time to determine the positioning accuracy deviation: ； in, is the positioning accuracy deviation, is the latitude and longitude reference value of the on-board device in the vehicle at time k, is the latitude and longitude detection value of the vehicle in the radar-visual fusion data at time k, , It is the number of times corresponding to the longitude and latitude detection values; Determine the vehicle's positioning data matching accuracy and positioning data integrity. The positioning data matching accuracy is expressed as the number of matching errors between the output trajectory and the target vehicle in the radar vision system and the sample size; the positioning data integrity is expressed as the number of positioning data of the test vehicle in the radar vision system and the number of theoretical positioning data from the time of appearance to departure: ； ； in, Is the positioning data matching accurate position, is the integrity of the positioning data, B1 is the number of samples, B2 is the number of errors in matching the trajectory output by the radar vision system with the target vehicle; M is the number of positioning data of the test vehicle in the radar vision fusion system, and N is the number of theoretical positioning data from the time it appears in the radar vision system to the time it leaves.

3. The method for comprehensive determination of the quality of a radar-visual fusion system according to claim 1, characterized in that: The method of obtaining the track integrity is as follows: obtaining the detection range length of the radar vision equipment, the number of tracks and the length of the tracks, and determining the track integrity: ； in, is the track integrity, L is the detection range length of the radar equipment, A is the number of tracks, is the length of the ath trajectory, .

4. The method for comprehensive determination of the quality of a radar-visual fusion system according to claim 1, characterized in that: The comprehensive evaluation effect of the fused radar-vision fusion data determined in step S6 includes: The index values ​​of the judgment matrix are standardized, the standardized evaluation index values ​​are determined, and a fuzzy evaluation matrix is ​​constructed according to the obtained evaluation index values; the judgment matrix of the fuzzy evaluation matrix is ​​determined, and the consistency check of the fuzzy evaluation matrix is ​​performed to confirm the optimal consistency judgment matrix corresponding to the fuzzy evaluation matrix.

5. The method for comprehensive determination of the quality of a radar-visual fusion system according to claim 4, characterized in that: The standardization of the index values ​​of the judgment matrix includes: standardization of the larger the better index, standardization of the smaller the better index, and standardization of the medium the better index; The standardized processing of the larger the better type of indicators is as follows: ； The standardization of the smaller the better indicator is as follows: ； The standardized processing of the Yuezhongyueyu type index is shown in the following formula 4-7: ； in, , , Represents the minimum, maximum and middle value of the i-th indicator value, is the standardized evaluation index value, is the index value of the judgment matrix, is the middle value between the i-th index value and the j-th index value, is the maximum value of the i-th index value and the j-th index value, , .

6. A comprehensive quality measurement method for a radar-vision fusion system according to claim 4, characterized in that: The fuzzy evaluation matrix also includes the following when performing consistency check: According to the fuzzy evaluation matrix To construct the judgment matrix of the fuzzy evaluation matrix , using the sample standard deviation of each evaluation index Reflect the influence of each evaluation index on the comprehensive evaluation, and use it to construct a judgment matrix ;in, is the sample mean of each evaluation index; The judgment matrix for the 1-9 level judgment scale of the fuzzy evaluation matrix is ​​as follows; ； in, , for The minimum and maximum values ​​of is the relative importance parameter value, is the index value of the fuzzy evaluation matrix corresponding to the judgment matrix, The sample standard deviation of the evaluation index when it is expressed as the jth index, .

7. The method for comprehensive determination of the quality of a radar-visual fusion system according to claim 4, characterized in that: The optimal consistency judgment matrix is ​​expressed as: Assume the judgment matrix The correction matrix is , the weight values ​​of each element of the modified matrix C are , then the following is satisfied That is the judgment matrix The optimal consistency judgment matrix is ​​as follows: ； Among them, there are , , and ; In the above formula, the objective function is the consistency index coefficient of the judgment matrix, d is a non-negative parameter, .

Images