Operational evaluation method for signalized intersections based on radar vision fusion detection

Abstract

To provide an operational evaluation method for signalized intersections based on radar vision fusion detection.SOLUTION: A radar vision fusion all-in-one machine is installed at a signalized intersection for real-time detection of the intersection. A detection area is divided into a grid pattern based on a flow path. The number of vehicles in queue, the number of vehicles crossing a stop line, and the number of vehicles passing through the stop line within each time interval are tabulated, and a delay time, a secondary stop rate, a queue length, and a grid occupancy rate are calculated. Hourly and daily operational evaluation indices using a Topsis evaluation model based on an entropy-weighting method is calculated to evaluate congestion conditions at five different levels. Continuous and accurate vehicle dynamic spatial data is provided through the radar vision fusion all-in-one machine. In addition, the Topsis evaluation model based on the entropy-weighting method is used to quantify the evaluation indices and improve the objectivity and accuracy of the evaluation.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent traffic detection, and particularly to a method for evaluating the operation of signal intersections based on radar vision fusion detection.

Background Art

[0002] Radar vision fusion detection technology is widely used in aspects such as traffic information perception and traffic target detection. Currently, the radar vision fusion all-in-one machine integrates emerging technologies such as active forward scanning and millimeter-wave 3DHD detection, and realizes wide-range tracking of targets, high-precision ranging, constant speed measurement, and multi-dimensional direction detection in traffic scenes. In addition, radar vision fusion maximally exerts the characteristics of a composite sensor, and by converting single-point perception into multi-point joint control, it effectively solves the technical problem of large detection deviations caused by single radar or video detection due to shielding or interference in harsh environments. In view of this point, radar vision fusion technology has obvious advantages in the evaluation of signal intersection operation, and by using the joint calibration method, various basic data of vehicles collected by different devices within a certain range of the intersection are corrected to ensure the accuracy and effectiveness of the signal intersection.

[0003] At present, the data types for evaluating the operation of signal intersections are generally classified into trajectory data and speed data. Trajectory data is mainly divided into the GPS trajectories of running vehicles and the GPS trajectories of floating cars, and speed data is further divided into vehicle speed and road section speed. From the spatio-temporal dimensional analysis obtained from the above data, it was found that the current signal intersection evaluation data has the following defects. (1) In the time dimension, periodic data collection is insufficient, the time continuity is low, and it depends on past statistical data, so the real-time performance is not high. (2) In the space dimension, a static drift phenomenon occurs in the target trajectory, that is, there is a large deviation between GPS positioning and the actual trajectory, and it is necessary to preprocess the data in advance according to the drift points, so the time cost of evaluation has increased.

[0004] Mathematical methods such as the analytic hierarchy process, fuzzy mathematics method, gray correlation method, and entropy weighting method are used to establish evaluation indicators for intersection operation, and then the operation status of the intersection is divided into different levels to distinguish between smooth and congested states. However, many of the above methods have problems such as subjective analysis, ambiguous definitions, and qualitative evaluations, so there are doubts about the reliability of the evaluation indicators.

Summary of the Invention

Problems to be Solved by the Invention

[0005] The object of the present invention is to overcome the disadvantages of the prior art and provide an operation evaluation method for signal intersections based on radar vision fusion detection. This method selects four main evaluation parameters, namely total intersection delay, secondary stop rate, queue length, and grid occupancy, through the Topsis evaluation model, calculates the time-based operation evaluation indicator and daily operation evaluation indicator of the intersection, and avoids the influence of a single indicator and subjective factors on the evaluation accuracy. In addition, the evaluation method of the present invention dynamically selects the main evaluation indicators by the primary and secondary factor analysis method and assigns specific values to each main indicator, so it is more persuasive than qualitative evaluation.

Means for Solving the Problems

[0006] An operation evaluation method for signal intersections based on radar vision fusion detection, comprising: Step S1 of installing a mature radar vision fusion all-in-one machine at the intersection signal, setting the rated operation parameters, and detecting the static information of the intersection and the dynamic information of the vehicle in real time; Step S2 of constructing a digital twin scene of the intersection, defining a rectangular detection area from the stop line of each inflow road to the maximum range of radar vision fusion detection, dividing small grids with the same vehicle length as the unit, and giving the coordinates of the intersections of the small grids; Based on the stop line, set the detection time interval, aggregate the number of vehicles that stop and start across the stop line of the waiting queue within each time interval, the number of vehicles that stop and start without crossing the stop line, calculate the total intersection delay, secondary stop rate, waiting queue length, and grid occupancy in step S3, and Using the Topsis evaluation model, with every 15 minutes per hour or every hour per day as the evaluation period Ai, take the total intersection delay, secondary stop rate, waiting queue length, and grid occupancy as the scheme indicators a ij of each evaluation period, calculate the operation evaluation indicators by time and the operation evaluation indicators by day, and evaluate the congestion status of the signal intersection by dividing it into five levels in step S4 The above method including

[0007] Preferably, step S1 specifically includes the following steps: Step S11 of defining the rated operation parameters of the radar vision fusion all-in-one machine, the maximum scanning distance D of the radar vision fusion all-in-one machine, the scanning frequency f, and the vehicle resolution ε; Define the static information of the intersection as the number of lanes t m of the m-th intersection inflow road, and the width W m,n of the n-th lane of the m-th intersection inflow road in step S12; Define the dynamic information of the vehicle as the two-dimensional coordinates of the k-th detected vehicle, vehicle type classification (label small passenger cars as vehicle, medium-sized cars as midvehicle, and large cars as bigvehicle), the real-time speed v k of the k-th vehicle, and the real-time azimuth angle φ k of the k-th vehicle in step S13, and S14 of jointly calibrating the two-dimensional coordinates of the detected vehicle using coordinate radar vision fusion.

[0008]

Number

[0009] Preferably, the construction of the digital twin scene of the intersection in step S2 is Specifically: Define the coordinates of the intersections of the small grids, use the stop line as the baseline of the rectangular area, and use the joint calibration method in step 1 to determine the coordinates (u 0,m , v 0,m ) of the intersection of the stop line and each incoming lane, where u 0,m is the abscissa of the intersection of the m-th lane and the stop line, v 0,m is the ordinate of the intersection of the m-th lane and the stop line, and define the coordinates of the remaining intersections as (u r,m , v r,m ), r is the r-th small grid up to the maximum detection range direction with respect to the stop line, and step S21 of defining the lane close to the U-axis as the first lane (m = 1) is included.

[0010] Preferably, step S3 specifically includes the following steps.

[0011] S31: The step of calculating the intersection delay.

[0012]

Equation

[0013] The method for detecting the number of stopped vehicles S i and the number of non-stopped vehicles nS i is as follows: Taking the stop line coordinates as the detection reference, within one signal cycle, for a certain vehicle coordinate (uk , v k ) crosses the stop line coordinates or the speed changes from 0 to v k When this happens, S i is accumulated once; for a vehicle coordinate (u k , v k ), when it crosses and the speed changes from v k to v k1 , nS i is accumulated once. S32: Step of calculating the secondary stop rate.

[0014]

Number

[0015] S33: Step of calculating the queue length.

[0016]

Number

[0017] S34: Step of calculating the grid occupancy rate.

[0018]

Number

[0019] Preferably, the calculation steps of the time-based operation evaluation index and the daily operation evaluation index Topsis in step S4 are as follows.

[0020] S41: Standardization step The forward index r ij is represented by the following formula.

[0021]

Equation

[0022] S42: Step of determining the attribute weight W j (1) Calculate the entropy value EH of the j-th index by the entropy weight method. j

[0023]

Equation

[0024] (2) Calculate the time difference coefficient FH by the entropy weight method. j

[0025] (Equation 8) FH j = 1 - EH j , j = 1, 2, 3, 4

[0026] (3) Determine the weight WH of the time-based operation evaluation by the entropy weight method. j

[0027]

Equation

[0028] Similarly, determine the weight WD of the daily operation evaluation. j

[0029]

Equation

[0030] S43: Step of calculating the weighted norm matrix A.

[0031]

Number

[0032] S44: Step of determining the positive ideal solution and the negative ideal solution of matrix A.

[0033]

Number

[0034] S45: Step of calculating the Euclidean distance between the positive ideal solution and the negative ideal solution in each evaluation period.

[0035]

Number

[0036] S46: Step of calculating the relative proximity of Ai to the ideal solution in each evaluation period.

[0037]

Number

[0038] ch * i According to [ ], each evaluation period is normalized from large to small and sorted, and the evaluation periods ranked higher are more excellent.

[0039]

Number

[0040]

Number

[0041] Also, the classification criteria for the evaluation index, that is, AH k ∈[0, 20%] means very smooth within 1 hour, AH k ∈[20%, 40%] means slightly smooth within 1 hour, AH k ∈[40%, 60%] means smooth within 1 hour, AH k ∈[60%, 80%] means slightly congested within 1 hour, AH k ∈[80%, 100%] is defined as very congested within 1 hour.

Advantages of the Invention

[0042] The advantages and technical effects of the present invention are as follows.

[0043] The present invention better solves the problem that data in intersection operation evaluation exists in two dimensions of time and space. With the radar vision fusion detection technology, the data has strong temporal continuity and real-time performance. By abandoning the conventional GPS coordinate system and establishing a unique coordinate system in the form of grid division, it avoids the processing of static drift points and reduces the computational complexity of cloud processing.

Brief Description of the Drawings

[0044]

Figure 1

Figure 2

Embodiments for Carrying Out the Invention

[0045] To further understand the content, features, and effects of the present invention, the following examples will be given and described in detail with reference to the accompanying drawings. It should be noted that these examples are merely illustrative and do not limit the present invention, nor are they intended to limit the protection scope of the present invention.

[0046] Taking the intersection of a certain city as an example, as shown in Figure 1, the operation evaluation method of the signal intersection based on radar vision fusion detection according to the present invention includes the following steps.

[0047] (1) First, complete the acquisition of basic data and detect the static information of the intersection and the dynamic information of vehicles in real time.

[0048] Suppose the operation status of the intersection evaluated every 15 minutes within 1 hour is Y1, Y2, Y3, Y4, and the four attributes of the total intersection delay, secondary stop rate, queue length, and occupancy are X1, X2, X3, X4. The results for every 15 minutes are obtained by calculation respectively, and a ij (i = 1, 2, 3, 4, j = 1, 2, 3, 4) represents the possible value of scheme Xi for attribute Yi (also called the original weight).

[0049] [Table 1]

[0050] The static information of the intersection is the number of lanes t of the m-th intersection inflow road m , and the width W of the n-th lane of the m-th intersection inflow road m,n is

[0051] The dynamic information of the vehicle is the two-dimensional coordinates (u k , v k ) of the k-th detected vehicle, vehicle type classification (labeling small passenger cars as vehicle, medium-sized cars as midvehicle, and large cars as bigvehicle), the real-time speed v of the k-th vehicle k , and the real-time azimuth angle φ of the k-th vehicle k is

[0052] The joint calibration of the radar vision fusion of the two-dimensional coordinates of the vehicle is represented by the following formula.

[0053]

Equation

[0054] (2) Defining a rectangular detection area from the stop line of each inflow road to the maximum range of radar vision fusion detection, dividing it into small grids with the same vehicle length as the unit, and giving the coordinates of the intersections of the small grids.

[0055] Definition of the coordinates of the intersections of the small grids: Taking the stop line as the baseline of the rectangular area, determining the coordinates (u 0,m , v 0,m ) of the intersections of the stop line and each lane by the joint calibration method in step 1, where u 0,m is the abscissa of the intersection of the m-th lane and the stop line, v 0,m is the ordinate of the intersection of the m-th lane and the stop line, and defining the coordinates of the remaining intersections as (u r,m , v r,m ), and r is the r-th small grid in the direction of the maximum detection range with the stop line as the reference.

[0056] (3) Calculating the total intersection delay, secondary stop rate, queue length, and grid occupancy rate.

[0057] Specifically, the intersection delay is represented by the following formula.

[0058]

Equation

[0059] The number of stopped vehicles $s$ i and the number of non-stopped vehicles $nS$ i detection method: Taking the stop line coordinates as the detection reference, within one signal cycle, when the coordinates of a vehicle $(u$ k , $v$ k ) exceed the stop line coordinates or the speed changes from 0 to $v$ k , $S$ i is incremented by 1; when the coordinates of a vehicle $(u$ k , $v$ k ) are exceeded and the speed changes from $v$ k to $v$ k1 , $nS$ i is incremented by 1, Specifically, the following stop rate is expressed by the following formula.

[0060]

Equation

[0061] The waiting queue length in step 3 is expressed by the following formula.

[0062]

Equation

[0063] The grid occupancy rate in Step 3 is expressed by the following formula.

[0064] [Equation] [In the formula, δ is the grid occupancy rate, O is the area occupied by the waiting queue vehicles in the grid, Q is the total area of the grid, M is the total number of intersection inflow roads, I is the total number of lanes of the intersection inflow roads, and O m,i is the grid area occupied by the waiting queue vehicles in the i-th lane of the m-th intersection inflow road.

[0065] (4) Step of calculating the time-based operation evaluation index and the daily operation evaluation index using the TOPSIS evaluation model and dividing them into 5 levels to evaluate the congestion status of the signalized intersection.

[0066] Step 1: Standardization: The forward index r ij is expressed by the following formula.

[0067] [Equation]

[0068] Step 2: Determination of the attribute weight W j : (1) Calculate the entropy value EH of the j-th index by the entropy weight method. j

[0069] [Equation]

[0070] (2) Calculate the time difference coefficient FH j by the entropy weight method.

[0071] (Equation 24) FH j = 1 - EHj , where j = 1, 2, 3, 4

[0072] (3) Determine the weight WH of the time - based operation evaluation by the entropy weighting method j .

[0073]

Number

[0074] Similarly, determine the weight WD of the daily operation evaluation j too.

[0075]

Number

[0076] Step 3: Calculate the weighted norm matrix A

[0077]

Number

[0078] Step 4: Determine the positive ideal solution and negative ideal solution of matrix A

[0079]

Number

[0080] Step 5: Calculate the Euclidean distance between the positive ideal solution and the negative ideal solution for each evaluation period

[0081]

Number

[0082] Step 6: Calculate the relative closeness of Ai to the ideal solution for each evaluation period

[0083]

Number

[0084] Step 7: ch * i According to this, normalize and sort the evaluation periods from large to small respectively. The evaluation periods ranked higher are more excellent.

[0085]

Number

[0086]

Number

[0087] Also, the classification criteria of the evaluation index, that is, AH k ∈[0, 20%] means very smooth within 1 hour, AH k ∈[20%, 40%] means slightly smooth within 1 hour, AH k ∈[40%, 60%] means smooth within 1 hour, AH k ∈[60%, 80%] means slightly congested within 1 hour, AH k ∈[80%, 100%] is defined as very congested within 1 hour.

[0088] Finally, for the parts not described in the present invention, mature products and technical means in the prior art are adopted.

[0089] Those skilled in the art will be able to make various improvements or changes based on the above description, and it will be understood that such improvements and changes are also included in the protection scope of the appended patent claims of the present invention.

Claims

1. A method for evaluating the operation of a signal intersection based on radar vision fusion detection, comprising: Step S1 of installing a mature radar vision fusion all-in-one machine at the intersection signal, setting the rated operation parameters, and detecting the static information of the intersection and the dynamic information of the vehicle in real time; Step S2 of constructing a digital twin scene of the intersection, defining a rectangular detection area from the stop line of each inflow road to the maximum detection range of radar vision fusion detection, dividing it into small grids with the same vehicle length as the unit, and giving the coordinates of the intersections of the small grids; Based on the stop line, setting the detection time interval, aggregating the number of vehicles in the waiting queue, the number of vehicles stopping and starting across the stop line, and the number of vehicles stopping and starting without crossing the stop line within each time interval, and calculating the total intersection delay, secondary stop rate, waiting queue length, and grid occupancy rate; Step S3, and Using the TOPSIS evaluation model, with the evaluation period Ai being every 15 minutes per hour or every hour per day, the total intersection delay, secondary stop rate, queue length, and grid occupancy are used as the scheme indicators a of each evaluation period ｉｊ to calculate the time-based operation evaluation indicators and daily operation evaluation indicators, and including step S4 of evaluating the congestion status of the signal intersection by dividing it into five levels The said step S3 specifically includes the following steps S31: The step of calculating the intersection delay; 【Number 2】 [where N is the total number of stopped and starting vehicles, n ｉ，ｊ is the number of vehicles that stopped at different times on the intersection approach road within the j-th Δt in the i-th minute, n is the detection time (minutes), m is the number of intervals divided into 1 minute, S is the total number of stopped vehicles, S ｉ is the number of stopped vehicles on the intersection inflow road within i minutes, α is the total intersection delay, Δt is the detection time interval within 1 minute, β is the average delay of each stopped and starting vehicle, t ｍａｘ is the maximum detection time.] Number of stopped vehicles S ｉ and number of non-stopped vehicles nS ｉ Detection method: Taking the stop line coordinates as the detection criterion, within one signal cycle, when a vehicle coordinate (u ｋ , v ｋ ) crosses the stop line coordinates or the speed changes from 0 to v ｋ , S ｉ is cumulated once; when a vehicle coordinate (u ｋ , v ｋ ) is crossed and the speed changes from v ｋ to v ｋ１ , nS ｉ is cumulated once. S32: The step of calculating the secondary stop rate; 【Number 3】 [where r is the secondary stop rate of the entire intersection, and ds is the cumulative number of secondary stops when a certain vehicle coordinate (u ｋ , v ｋ ) does not cross the stop line coordinate, the speed state changes to 0 or v ｋ１ , and further decelerates to 0] S33: The step of calculating the waiting queue length; 【Number 4】 [where L is the total length of the waiting queue at the intersection, M is the total number of incoming roads at the intersection, I is the total number of lanes of the incoming roads at the intersection, and l ｍ，ｉ is the total length of the waiting vehicle queue in the i-th lane of the m-th incoming road at the intersection.] S34: The step of calculating the grid occupancy rate; 【Number 5】 [where δ is the grid occupancy rate, O is the area occupied by the queuing vehicles in the grid, Q is the total area of the grid, M is the total number of intersection inflow roads, I is the total number of lanes of the intersection inflow roads, and O ｍ，ｉ is the grid area occupied by the queuing vehicles in the i-th lane of the m-th intersection inflow road.] The calculation steps of the operation evaluation index by time and the Topsis of the daily operation evaluation index in the said step S4 are as follows: S41: Standardization step; Forward pointer r ｉｊ is represented by the following equation, 【Number 6】 S42: Determining the attribute weight w ｊ step, (1) Calculate the entropy value $EH$ of the $j$-th index by the entropy weighting method ｊ and 【Number 7】 (2) Calculate the time difference coefficient FH ｊ using the entropy weighting method. (Equation 8) FH ｊ = 1 - EH ｊ , j = 1, 2, 3, 4 (3) Determine the weight WH of the operation evaluation by time using the entropy weighting method ｊ and 【Number 9】 Similarly, the weight WD of the daily operation evaluation is also determined. ｊ ​ 【Number 10】 S43: The step of calculating the weighted norm matrix A; 【Number 11】 S44: The step of determining the positive ideal solution and the negative ideal solution of the matrix A; 【Number 12】 S45: The step of calculating the Euclidean distance between the positive ideal solution and the negative ideal solution in each evaluation period; 【Number 13】 S46: The step of calculating the relative proximity of Ai to the ideal solution in each evaluation period; 【Number 14】 ch ＊ ｉ Accordingly, each evaluation period is normalized and sorted from largest to smallest, and the evaluation periods arranged at the top are more excellent 【Number 15】 Similarly, 【Number 16】 [where AH ｋ is the operation evaluation index for the i-th evaluation period within 1 hour, AD ｋ is the operation evaluation index for the i-th evaluation period within 1 day, a ｉ，ｊ is the value that X ｉ can take with respect to attribute Y ｉ .] Classification criteria for evaluation indicators, i.e., AH ｋ ∈ [0, 20%] means very smooth within 1 hour, AH ｋ ∈ [20%, 40%] means slightly smooth within 1 hour, AH ｋ ∈ [40%, 60%] means smooth within 1 hour, AH ｋ ∈ [60%, 80%] means slightly congested within 1 hour, AH ｋ ∈ [80%, 100%] is defined as very congested within 1 hour A method for evaluating the operation of a signal intersection based on radar vision fusion detection, characterized in that.

2. The said step S1 includes Step S11 of defining the rated operation parameters of the radar vision fusion all-in-one machine, the maximum scanning distance D of the radar vision fusion all-in-one machine, the scanning frequency f, and the vehicle resolution ε; The static information of the intersection is the number of lanes t of the m-th intersection inflow road ｍ , the width W of the n-th lane of the m-th intersection inflow road ｍ，ｎ and step S12 for defining them The dynamic information of the vehicle, the two-dimensional coordinates of the k-th detection vehicle, vehicle type classification (labeling a small passenger car as vehicle, a medium-sized vehicle as midvehicle, and a large vehicle as bigvehicle), the real-time speed v of the k-th vehicle ｋ , the real-time azimuth angle φ of the k-th vehicle ｋ Step S13 defined as such, and S14 of jointly calibrating the two-dimensional coordinates of the detected vehicle using coordinate radar vision fusion The method for evaluating the operation of a signal intersection based on radar vision fusion detection according to claim 1, characterized in that it includes the above. 【Number 1】 [where (u ｋ , v ｋ ) are the two-dimensional pixel coordinates of the vehicle after co-calibration by the Zhang calibration method, and (x ｋ , y ｋ ) are the coordinates after distortion correction of the camera. ]

3. The construction of the digital twin scene of the intersection in the said step S2 is Specifically, define the coordinates of the intersections of the small grids, use the stop line as the baseline of the rectangular area, and determine the coordinates (u ０，ｍ , v ｏ，ｍ ) of the intersection of the stop line and each inflow lane by the joint calibration method in Step 1. Here, u ０，ｍ is the abscissa of the intersection of the m-th lane and the stop line, v ｏ，ｍ is the ordinate of the intersection of the m-th lane and the stop line. Define the coordinates of the remaining intersections as (u ｒ，ｍ , v ｒ，ｍ ), where r is the r-th small grid up to the maximum detection range direction with respect to the stop line, and define the lane closest to the u-axis as the first lane (m = 1). The method for evaluating the operation of a signal intersection based on radar vision fusion detection according to claim 2 is characterized by including Step S21.

Images