TWI681372B - Detection system and detection method of traffic probe devices - Google Patents

Abstract

This invention discloses a detection system and a detection method of traffic probe devices, comprising obtaining a plurality of travel time samples of a plurality of road segments of the plurality of the probe devices, performing a deviation value filtering on the plurality of travel time samples by using various historical travel time probability distribution models, and inserting the filtered plurality of travel time samples into various historical travel time probability distribution models to determine the applicable model of each road segment, calculating the reliability of each probe device according to the filtered plurality of travel time samples and the applicable model of each road segment, integrating the reliability of each probe device to generate the detection reference value of each road segment, and detecting each detecting device according to the detection reference value of each road segment.

Description

Detection system and detection method of traffic detection equipment

This case relates to a detection technology for traffic detection equipment, in particular, to a system and method for detecting traffic detection equipment incorporating sources of heterogeneous detection.

For the intelligent transportation system, in order to provide real-time traffic information with a wider coverage, the quality of the traffic detection equipment itself and the quality of the data it can provide are relatively important, but the data quality of the traffic detection equipment may vary with time Gradually decreases.

The prior art proposes an error correction mechanism for road traffic sensors, where the road traffic sensors can be placed near the road or buried in the road, and continuously detect the speed of the passing vehicles to have historical speed data for a period of time, thereby, It can perform statistical processing or probability distribution processing on multiple historical speed data of road traffic sensors as reliable speed data, and then compare it with the currently detected data to reflect the possible failure of road traffic sensors.

However, the prior art only judges whether a road traffic sensor may fail based on historical data, and does not consider different road segments, different equipment, and different Different processing methods are designed for the same period or different vehicle conditions.

Therefore, how to continuously and automatically detect the output data of a variety of equipment to reduce the cost of detection, and thereby eliminate or update the equipment, and continue to maintain the accuracy of the data source.

In order to achieve the above and other purposes, the present case proposes a detection method for traffic detection equipment, which includes: obtaining a plurality of travel time samples for a plurality of road sections of a plurality of detection equipment; enclosing the plurality of travel time samples into various historical travel times Probability distribution model to determine the applicable model of each section; based on the plurality of travel time samples and the applicable model of each section, calculate the reliability of each detection equipment; fuse the reliability of each detection equipment to produce The detection reference value of each road section; and the detection equipment of each detection section according to the detection reference value of each road section.

In addition, calculating the reliability of each detection equipment based on the plurality of travel time samples and the applicable model of each road section includes: calculating the representative values of the plurality of travel time samples of each road section of each detection equipment, and The representative value of each section of the equipment is set into the applicable model of each section to calculate the reliability of each detection equipment. Furthermore, fusing the reliability of each detection and detection device to produce the detection reference value of each road segment includes: taking the reliability of each detection and detection device as a weight, and weighting the multiple travel time samples of each road segment , To produce the detection benchmark value of each road section.

Secondly, this case proposes a detection system for traffic detection and detection equipment, which includes: a travel time sample module to generate a plurality of travel time samples for a plurality of road sections of a plurality of detection and detection equipment; historical travel time probability distribution model The module is used to set the plurality of travel time samples into various historical travel time probability distribution models to determine the applicable model of each road segment; the reliability calculation module is used to determine the travel time samples and each The applicable model of the road section to calculate the reliability of each detection and detection equipment; the detection reference value calculation module to calculate the detection reference value of each road section based on the reliability of each detection and detection equipment; and the detection module, based on each road section The detection reference value is used to detect each detection equipment.

The historical travel time probability distribution model module is further used to insert the plurality of travel time samples into the various historical travel time probability distribution models, so as to filter the deviation values of the plurality of travel time samples, so as to cover the various historical travel The plurality of travel time samples of the time probability distribution model are the plurality of travel time samples filtered by the deviation value, thereby judging the applicable model of each road segment, and making the plurality of travel time based on which the reliability of each detection equipment is calculated The sample is a plurality of travel time samples filtered by the deviation value.

In addition, the detection system further includes a modeling module for establishing various historical travel time probability distribution models including congested historical travel time probability distribution models and smooth historical travel time probability distribution models based on historical data.

The plurality of detection equipments are classified as fixed-point detection traffic information sources and line segment detection traffic information sources. For the detection equipment classified as the fixed-point detection traffic information sources, the detection and detection equipment classified as the fixed-point detection traffic information sources use the road segments to which the devices belong or correspond and the speed of the detected points Obtain a travel time sample belonging to a fixed point, and for the detection equipment classified as the source of traffic information for the line segment, the travel time sample belonging to the line segment is obtained by using the path and road segment weight to which the device belongs or corresponds, And the travel time sample of the fixed point and the travel time sample of the line segment are the plurality of travel time samples, and the detection module is further used to classify the detection reference value of each road segment and the source of the traffic information of the fixed point detection The multiple travel time samples of each section of the detection equipment are compared to produce the qualification of the detection equipment classified as the source of traffic information for the fixed point detection, and the detection reference value of each section is combined and combined with the classification as the A plurality of travel time samples of the detection equipment of the line segment detection traffic information source are compared to produce the qualification of the detection equipment classified as the line segment detection traffic information source.

The fixed-point detection traffic information source is divided into static fixed-point detection traffic information source and dynamic fixed-point detection traffic information source, and the line segment detection traffic information source is classified into static line segment detection traffic information source and dynamic line segment detection traffic Information source. The static fixed-point detection traffic information source is a vehicle detector, the static line-section detection traffic information source is an etag, a Bluetooth detector, or a license plate image, and the dynamic fixed-point detection traffic information source or the dynamic line segment detects traffic The source of information is GPS of the vehicle, GPS of the mobile phone, or mobile signaling positioning.

Therefore, in this case, different calculation methods are designed according to the characteristics of various traffic detection equipments of heterogeneous detection sources, and the data are converted to the same time and space standards, and then based on the historical data of road segments, road types, weather, date and time, and other factors and the current received After comparative analysis and filtering of the data with severe deviations, the remaining data is dynamically adjusted to the current most suitable weight fusion output detection benchmark value, and this benchmark value is fed back to each traffic detection equipment for equipment output level measurement To detect low standards Retire the updated detection equipment.

2‧‧‧Detection system

21‧‧‧ Travel Time Sample Module

22‧‧‧Modeling module

23‧‧‧ Probability Distribution Model Module of Historical Travel Time

24‧‧‧Reliability calculation module

25‧‧‧Detection reference value calculation module

26‧‧‧Detection module

51‧‧‧Vehicle detector

52‧‧‧Bluetooth detector

53‧‧‧GPS detective vehicle

Section S001~S003‧‧‧

S201~S205‧‧‧Step

S301-1~S301-4, S302-1~S302-2, S303, S304, S305-1~S305-2, S306‧‧‧

S401~S405‧‧‧Step

Figure 1 is a block schematic diagram of an embodiment of a detection system for traffic detection and detection equipment in this case; Figure 2 is a schematic flow diagram of an embodiment of a detection method for traffic detection and detection equipment in this case; Figure 3 is a first example of a detection method for traffic detection and detection equipment in this case Schematic diagram of the specific embodiment; FIG. 4 is a schematic diagram of another embodiment of the detection method of the traffic detection equipment of the case; FIG. 5 is an illustration of an example of the detection system and method of the traffic detection equipment of the case; and FIG. 6 This is an example illustration of the travel time sample of the detection system and method of the traffic detection equipment in this case.

The following describes the implementation of this case through specific examples. Those skilled in the art can easily understand other advantages and effects of this case by the contents disclosed in this article. The structure, ratio, size, etc. shown in the drawings in this specification are only used to match the contents disclosed in the specification, for those familiar with this skill to understand and read, and are not used to limit the restrictive conditions that can be implemented in this case. Therefore, any modification, alteration or adjustment should still fall within the scope of the technical content disclosed in this case, without affecting the efficacy and purpose achieved in this case.

Please refer to Figure 1, which is the detection system of the traffic detection equipment in this case Block diagram of the embodiment. The detection system 2 includes a travel time sample module 21, a modeling module 22, a historical travel time probability distribution model module 23, a reliability calculation module 24, a detection reference value calculation module 25, and a detection module 26.

The travel time sample module 21 can generate a plurality of travel time samples of a plurality of sections of a plurality of detection equipment. A plurality of detection equipment can be classified as a fixed-point detection traffic information source and a line segment detection traffic information source. This is a heterogeneous detection source. For the detection equipment classified as a fixed-point detection traffic information source, the road segment and the corresponding part of the device and the detected Obtain the travel time samples belonging to the fixed point by measuring the speed, and for the detection equipment classified as the source of traffic information for the line segment detection, use the path and the weight of the device to which the device belongs or correspond to obtain the travel time samples belonging to the line segment, then the fixed point travel time sample and The travel time sample of the line segment is a plurality of travel time samples.

The modeling module 22 can establish various historical travel time probability distribution models including congested historical travel time probability distribution models and smooth historical travel time probability distribution models based on historical data. Modeling can be done in advance or after obtaining travel time samples.

The historical travel time probability distribution model module 23 can set a plurality of travel time samples into various historical travel time probability distribution models to determine the applicable model of each road segment. The historical travel time probability distribution model module 23 can further include a plurality of travel time samples into various historical travel time probability distribution models to filter the deviation values of the plurality of travel time samples, so as to fit into various historical travel time probability distribution models. The plurality of travel time samples are the plurality of travel time samples filtered by the deviation value, thereby determining the applicable model of each road segment, and enabling the calculation of the reliability of each detection equipment based on the plurality of travel The row time sample is a plurality of travel time samples filtered by the deviation value.

The reliability calculation module 24 can calculate the reliability of each detection equipment based on a plurality of travel time samples and applicable models of each road segment. In other words, the reliability calculation module 24 can calculate the representative values of the plurality of travel time samples of each section of each detection equipment, and insert the representative values of each section of each detection equipment into the applicable model of each section to calculate the value of each detection equipment Reliability.

The detection reference value calculation module 25 can calculate the detection reference value of each road segment according to the reliability of each detection equipment. In other words, the detection reference value calculation module 25 may weight the reliability of each detection equipment as a weight, and average the plurality of travel time samples of each road segment to produce a detection reference value of each road segment.

The detection module 26 can detect each detection equipment according to the detection reference value of each road section. In other words, the detection module 26 can compare the detection reference value of each road section with a plurality of travel time samples of each road section of the detection equipment classified as the fixed-point detection traffic information source, and use the output classification as the fixed-point detection traffic information source The qualification of the detection equipment of the road and the detection reference value of each road segment are combined and compared with the multiple travel time samples of the detection equipment classified as the source of traffic information for line segment detection, and the output is classified as the source of traffic information for line segment detection. The qualification of the detection equipment.

Please refer to FIG. 2, which is a schematic flowchart of an embodiment of the detection method of the traffic detection equipment in this case.

In step S201, a plurality of travel time samples of a plurality of road sections of a plurality of detection equipment are obtained. For the detection equipment classified as a fixed-point detection traffic information source, the travel time samples belonging to the fixed point are obtained by using the road segment to which the device belongs or the corresponding and the speed of the detected point of the device, and for the line segment The detection equipment that detects the source of traffic information obtains the travel time samples belonging to the line segment by using the path of the device or the corresponding path and the weight of the device road segment. The fixed-point travel time sample and the travel time sample of the line segment are a plurality of travel time samples.

In step S202, a plurality of travel time samples are set into various historical travel time probability distribution models to determine the applicable model of each road segment. It should be noted that various historical travel time probability distribution models can be used to filter the deviation values of multiple travel time samples to filter out samples that are too large or too small, and then the filtered travel time samples can be set into various historical travel time probabilities. The distribution model is used to obtain the probability of what kind of model each road section may be. The model with the highest probability is the applicable model. In addition, various historical travel time probability distribution models can be modeled before, simultaneously, or after obtaining a plurality of travel time samples.

In step S203, the reliability of each detection equipment is calculated according to the plurality of travel time samples and the applicable models of each road segment, that is, the representative values of the plurality of travel time samples of each road segment of each detection equipment are calculated, and then the representative values are set Applicable models into each section to calculate the reliability of each detection equipment.

In step S204, the reliability of each detection equipment is fused to produce the detection reference value of each road section, that is, the weighted average of the plurality of travel time samples of each road section is weighted with the reliability of each detection equipment as the weight, Output the detection benchmark value of each road section.

In step S205, each detection and detection equipment is detected according to the detection reference value of each road section, that is, the detection reference value of each road section and the plurality of travel times of each road section of the detection equipment classified as a fixed-point detection traffic information source The samples are compared to determine the eligibility of the detection equipment classified as the fixed-point detection traffic information source, and to combine the detection benchmark values of each road segment with the multiple travel time of the detection equipment classified as the line segment detection traffic information source The samples are compared to determine the eligibility of the detection equipment classified as the source of traffic information for line segment detection. According to this, based on the qualification, the detection equipment can be maintained or decommissioned.

Accordingly, the detection methods of the traffic detection equipment in this case can be roughly summarized as follows.

(1) According to the principle of detection and detection, the detection equipment is classified and the road section is corresponding: the detection equipment is divided into four types according to the detection position is dynamic or static and the detection principle is fixed point or line segment; static detection can be obtained by querying the pre-established correspondence table Corresponding road segments, dynamic detection can obtain corresponding road segments through real-time map correspondence; the number of corresponding road segments for fixed-point detection is one, and line segment detection can be one or more road segments. (2) Calculate the travel time sample of each equipment detection: fixed-point detection is converted from the length of the road segment by the speed of the point detected by the equipment to the travel time of the road segment; the line detection detects the path travel time of the equipment detected by the weight of each road segment Convert to road travel time. (3) Summarize travel time samples: mark and summarize travel time samples according to source type, equipment ID, and road section ID, so as to calculate the reliability of each equipment based on road sections and carry out fusion output detection benchmark value. (4) Establish a probability distribution model of historical travel time of congestion and smoothness according to the situation classification: select the probability distribution model, classify the historical data of the road segment according to the characteristics of the road segment, weather, date time, congestion or not, and train each situation classification separately Parameters of the model to establish the probability distribution model of historical travel time. (5) According to the model Bias value filtering: Given a probability threshold, filter out the outliers with extremely low probability in the historical travel time. (6) Decide whether the current road condition is congested and select the model: Set the travel time samples calculated from all current sources of the road segment to the congestion or smooth historical travel time probability distribution model, and select the one with the highest probability as the current road condition and the next The model used. (7) Calculate the reliability of the equipment based on the model: take a representative value of the travel time sample statistics of each equipment, and then fuse the representative value of the equipment. The current reliability calculation method is that the representative value of the detection equipment is embedded in the history of the current road conditions. The probability calculated by the travel time probability distribution model; because the reliability of the equipment often changes with time, the reliability of multiple periods can be considered. The calculation method is the product of the reliability of the current and past periods. (8) The benchmark value of fusion output detection: the current fusion travel time is weighted on the basis of reliability and the current fusion travel time is used as the benchmark value. (9) Comparison of reference value and detection sample: according to the combination of detection range of each device, the detection reference value is compared. The detection range of the fixed-point detection source is a single road segment, which can be directly compared with the sample produced by the device; line segment detection The source combines the reference values of all road sections in the detection range and then compares the combined reference value with the equipment output sample. (10) Output test results of each device: use the absolute error ratio as an indicator, and consider the sample with the absolute error ratio less than the error threshold as the qualified sample, so that the qualified ratio of the output sample of each device can be calculated.

Please refer to FIG. 3, which illustrates a schematic flow chart of a specific embodiment of the detection method of the traffic detection equipment in this case.

First, the information provided by various detection equipment can be used as a source of information, and the detection position of various traffic detection equipment is dynamic or static, and The detection principle is fixed point or line segment, and the information sources can be divided into static fixed point detection, dynamic fixed point detection, static line segment detection, and traffic information source for dynamic line segment detection. The detector installed under the road or on the side of the road is static detection, and can be further divided into single-point erection such as vehicle detector VD (vehicle detector) to detect static fixed-point detection of the single-point speed of the road Traffic information sources, such as etag, Bluetooth, license plate images, etc. need to match the detection data of more than two detectors to detect the road line speed and travel time of static line detection traffic information source. In addition, the detection and detection vehicle technologies such as vehicle GPS, mobile phone GPS, and mobile signaling positioning are classified as dynamic detection. Since GPS data can simultaneously have a single point speed, it can be further subdivided into dynamic fixed-point detection including point speed The source of traffic information and the source of detected traffic information for dynamic line segments that do not include point speed. This requires matching the same detection vehicle at two points in time and comparing the paths to obtain the average speed and travel time of the road segment.

As shown in Figure 3, the four traffic information sources can be subjected to different detection processes. Among them, the detection process of static fixed-point traffic information sources can be as follows: steps S301-1, S302-1, S303, S304, S305-1 and S306, the detection flow of the traffic information source for dynamic fixed-point detection can be as steps S301-2, S302-1, S303, S304, S305-1, and S306, and the detection flow of the traffic information source for static line segment detection can be as in step S301- 3. S302-2, S303, S304, S305-2 and S306, the detection process of the traffic information source for dynamic line detection can be as steps S301-4, S302-2, S303, S304, S305-2 and S306.

As shown in step S301-1, the source of traffic information for static fixed-point detection Because the location of the device does not change with time, the corresponding road segment can be queried through the correspondence table of the pre-established detection equipment ID and the corresponding road segment; as shown in step S301-3, the source of traffic information for static line segment detection is also because the device location does not Over time, the route can also be queried through pre-established pairing of two device IDs. This route may be composed of one or more road segments; as shown in step S301-2, the source of traffic information for dynamic fixed-point detection is due to the location of the device It will change with time, and the current latitude and longitude position needs to be mapped to the closest road segment through the real-time map correspondence algorithm; as shown in step S301-4, the traffic information source for dynamic line detection also needs to be calculated through the real-time map correspondence algorithm The method matches the latitude and longitude of the two paired devices to the path. Similarly, this path can be composed of one or more road segments.

，其中，路徑R由l條路段S_i所 組成，i

{1,2,…,I}，T_R為路徑旅行時間，

為路段i之旅行時間，

為路段i之權重，權重可為該路段的歷史旅行時間或順暢旅行時間，歷史旅行時間可為在該路段在歷史同一日期種類、天氣等分類下之平均旅行時間，順暢旅行時間則可為路段長度除以道路速限。接著如步驟S303和S304所示，各探偵設備轉換後取得的路段旅行時間樣本會依照進行來源種類、設備ID、路段ID標示並歸納，以利後續以路段為基準計算各探偵設備的可靠度並進行融合產出檢測基準值。所歸納出之旅行時間樣本可參考第6圖。 Then, after each detection equipment is mapped to a road segment, the point velocity or path travel time data detected by the equipment needs to be converted into road segment travel time to obtain a road segment travel time sample contributed by each detection equipment to its detection road segment. For the spot detection equipment, as shown in step S302-1, the vehicle speed sample can be converted into a travel time sample by dividing the length of the road segment by the point speed. Since the range of the spot detection only corresponds to one road segment, this can be Minimize the error in converting point speed to travel time. For the detection equipment of line segment detection, as shown in step S302-2, the time difference detected by the same vehicle through its paired device is the detected travel time sample, but the path range of the line segment detection often includes multiple road segments , According to the weight of each road segment, the detected travel time of the route can be converted to the travel time of each road segment within the range of the route. The conversion method can be expressed as

, Where the path R consists of l road segments S _i , i

{1,2,…,I}, _TR is the route travel time,

Is the travel time of section i,

It is the weight of the road segment i. The weight can be the historical travel time or smooth travel time of the road segment. The historical travel time can be the average travel time of the road segment under the same historical date category and weather classification. The smooth travel time can be the road segment Length divided by road speed limit. Next, as shown in steps S303 and S304, the travel time samples of the road sections obtained after the conversion of each detection equipment will be marked and summarized according to the type of source, equipment ID, and road ID, so that the reliability of each detection equipment can be calculated based on the road section. The benchmark value of the fusion output test. Refer to Figure 6 for the sample travel time.

In addition, for the detection equipment of fixed-point detection, as shown in step S305-1, the detection reference value is compared with the detection road sample. For the detection equipment for line segment detection, as shown in step S305-2, the reference values are combined and the detection path samples are compared. Finally, as shown in step S306, the qualification ratio of each device is calculated to produce the detection result of each device. The more detailed contents of the output detection benchmark value and the subsequent calculation of the pass ratio are shown in Figure 4.

Please refer to FIG. 4, as shown in step S401, a historical travel time probability distribution model under different situation classifications is established, which is used as a basis for deviation value filtering and reliability calculation. In detail, the situation classification can categorize the historical data of road segments according to the characteristics of road segments, weather, date and time, congestion or smoothness, because the average travel time required for different road segments to pass under different weather and date conditions may be different. Smoothness is also an important feature. Congestion will have greater variation and longer average travel time than smooth travel time. Failure to consider this feature will seriously affect the accuracy of the historical travel time probability distribution model. Then choose the appropriate probability distribution model, For example, uniform distribution or normal distribution, etc., and use the classified historical data to train the model parameters under each situation classification, so as to establish the probability distribution of historical travel time that can be described mathematically. The pre-established historical travel time probability distribution model of each road segment can be stored to provide quick query and utilization during detection.

ρ

1，給定機率門檻值ρ的數值可以依據壅塞與順暢的模型計算出上界B_U與下界B_L，並依此進行過濾；例如P_C(X>B_U)=ρ,P_F(X<B_L)=ρ，P_C(x)為壅塞時的旅行時間機率分布函數、P_F(x)為順暢時的旅行時間機率分布函數，若ρ=0.01，則表示壅塞時旅行時間大於B_U或順暢時旅行時間小於B_L之機率為百分之一，如此便可過濾掉在歷史旅行時間中出現機率極低的異常值。 Next, as shown in step S402, the travel time samples of the links of each source can be filtered by the deviation value. In detail, the historical travel time probability distribution model of congestion and smoothness can be created or queried through the current road segment, weather, date and time, and the filtering method can use the historical travel time probability distribution model to assume the area of normal travel time distribution when the segment upper bound lower bound B _U B _L, when the travel time is greater than the upper bound congestion zone B _U or smooth zone is less than the travel time of probability lower bound threshold value B _L ρ, 0

ρ

1. The value of the given probability threshold ρ can be calculated according to the congestion and smooth model of the upper bound B _U and the lower bound B _L , and filtered accordingly; for example, P _C (X>B _U )=ρ,P _F (X <B _L )=ρ, P _C (x) is the travel time probability distribution function during congestion, and P _F (x) is the travel time probability distribution function during smooth congestion, if ρ=0.01, it means that the travel time during congestion is greater than B The probability that the travel time is less than B _L when _U or smooth travel is 1%, so that outliers with extremely low probability in historical travel time can be filtered out.

T)，或順暢的機率P_F(X

T)，T為該路段所有旅行時間 樣本的集合，並選取機率大者為當前的路況與接下來使用的模型。 Then, as shown in step S403, it is determined whether the current road condition is congested or smooth, and then the probability distribution model of historical travel time to use congestion or smoothness for the next reliability calculation can be selected. In detail, the travel time samples calculated from all the current sources of the road segment are respectively included in the congestion or smooth historical travel time probability distribution model to obtain the probability of traffic congestion under the sample P _C (X

T), or smooth probability P _F (X

T), T is a collection of all travel time samples of the road segment, and the one with the highest probability is selected as the current road condition and the model to be used next.

,

， 其中，

當前時段t之下探偵設備D_i之可靠度，P_t(x)為當前時段t之下路況的旅行時間機率分布函數，

為當前時段t之下探偵設備D_i產出之所有旅行時間樣本的集合，

為探偵設備D_i第j個樣本的權重，其中權重的計算方式可為探偵設備D_i第j個樣本套入當前路況之歷史旅行時間機率分布模型所計算出的機率。如此便可計算出所有探偵設備在當前時段之可靠度，接著如步驟S405所示，融合時再依照各可靠度為比例對樣本進行加權平均，便可產出當前之融合旅行時間作為檢測基準值。 After judging that the current road condition is completed, as shown in step S404, the current travel time sample of the road segment can be classified according to the source and equipment to calculate the reliability respectively. In detail, because the number of samples produced by each device varies, you must first take a representative value from the sample statistics of the device, where the representative value can be the simple average of the sample, the weighted average according to the probability distribution, and the median And other statistical data, and then fuse the representative value of each device. The current reliability calculation method is the probability calculated by the representative value of the detection equipment into the historical travel time probability distribution model of the current road conditions, which can be expressed as

,

, among them,

The reliability of the detection equipment D _i under the current period t, P _t (x) is the travel time probability distribution function of the road conditions under the current period t,

Is a collection of all travel time samples produced by the detection equipment D _{i under the} current time period t,

It is the weight of the jth sample of the detection equipment D _i . The weight can be calculated as the probability calculated by the probability distribution model of the historical travel time of the jth sample of the detection equipment D _i into the current road conditions. In this way, the reliability of all detection equipment in the current time period can be calculated. Then, as shown in step S405, the samples are weighted and averaged in proportion to each reliability during fusion to produce the current fusion travel time as the detection reference value .

,τ

t,t-1,...,k，即 當前與過去k個時段之可靠度乘積，在考慮探偵設備可能在某些時段有缺漏的狀況下，k的選取範圍應為該探偵設備無缺漏的時段，並統一選取的時段個數以公平衡量各探偵設備的表現。最後同理可依照多時段的可靠度為比例進行加權平均產出當前融合旅行時間作為檢測基準值。 In addition, because the reliability of the detection equipment often changes with time, if only the reliability of the current time period is referenced, it is easy to cause deviations. The result of the fusion may be biased to the historical travel time in the same situation, causing detection deviations. To avoid this situation, The reliability of more time periods can be considered together, and the current fusion weight is determined according to the reliability performance of the detection equipment during these time periods, so that the performance of the detection equipment can be measured more effectively. Specifically, the reliability calculation method considering multiple time periods can be expressed as

,τ

t,t-1,...,k, which is the reliability product of the current and the past k time periods, considering that the detection equipment may be missing in certain periods, the range of k should be that the detection equipment is not missing Time period, and the number of time periods selected uniformly to fairly measure the performance of each detection equipment. Finally, in the same way, the weighted average of the current fusion travel time can be used as the detection reference value in proportion to the reliability of multiple periods.

， 其中

為當前時段t之下探偵設備D_i產出之所有旅行時間樣本的集合，s_t為當前時段t之基準值，n(

)為樣本數，

，ε為誤差門檻值。使用合格比例可以 衡量各探偵設備之樣本，當合格比例持續低於平均水準便可以判斷該設備出現異常，可進行設備維護或更換以排除 異常。 After the detection reference value is generated, the output level of each detection equipment can be detected according to this standard. Returning to Figure 3, for the fixed-point detection source, as shown in step S305-1, since the detection range is a single road segment, no additional combination of detection reference values is needed, and it can be directly compared with the travel time samples of the detection equipment; for line segments For the detection source, as shown in step S305-2, the detection reference values of all the road sections in the detection range are added and combined, and then the combined detection reference values are compared with the travel time samples of the detection equipment. For the comparison method, the absolute error ratio can be used as an indicator, and the sample with the absolute error ratio less than the error threshold can be regarded as a qualified sample, so that the qualified ratio of the travel time sample of the device can be calculated, which can be expressed as

, among them

Is the set of all travel time samples produced by the detection equipment D _{i under the} current period t, s _t is the reference value of the current period t, n(

) Is the number of samples,

, Ε is the error threshold. The qualified ratio can be used to measure the samples of each detection equipment. When the qualified ratio continues to be below the average level, it can be judged that the equipment is abnormal, and the equipment can be maintained or replaced to eliminate the abnormality.

Please refer to Figures 5 and 6, which illustrate the application of the detection system and method for the traffic detection equipment in this case. As shown in Fig. 5, the route has a section S001, a section S002, and a section S003, in which the speed limit of the section is 60 kilometers per hour, the length of the section S001 is 840 meters, and the length of the sections S002 and S003 is 450 meters. A static fixed-point detection source (such as the vehicle detector 51 for detecting the road segment S002) and a static line detection source (a pair of Bluetooth detectors 52 for detecting the road segment S001~ S003). On a certain day from 08:00 to 08:05, there were five vehicles passing through this path, and one of them was a GPS detection vehicle 53. Therefore, two dynamic fixed-point detection sources (GPS detection vehicle 53 are used to detect Measured sections S001 and S003) and a dynamic line detection source (GPS detection vehicle 53 for detecting sections S001~S003). During this period, the vehicle speed (km/h) detected by the vehicle detector 51 is 95, 62, 60, 58, 32 in sequence, and the vehicle travel time (seconds) detected by the Bluetooth detector 52 respectively For 100, 107, 107, 112, 200, the GPS detection vehicle 53 detects the fixed vehicle speed (km/h) is 67 in the section S001, 60 in the section S003, and the travel time from the section S003 to the section S001 is 89 second. Through the method as shown in Figure 3, dividing the length of the road segment by the speed can calculate the travel time sample of the fixed-point detection from the contribution of the road segment, and using the smooth travel time as the weight, the line detection can be calculated from the contribution of the passing road segment Travel time sample, so you can get all the travel time samples contained in the three sections of this period, as shown in Figure 6.

Next, establish a historical travel time probability distribution model for each road segment, search for historical data according to the road segment, weather (whether it rains), date type (day of the week), time period (08: 00 ~ 08: 05), and find out The historical data is divided into two groups through congestion or smoothness, and the normal distribution is selected as the probability distribution model to train and model the two groups of data. The model trained by normal distribution can be expressed as N(μ,σ ² ), μ is the expected value of normal distribution, σ ² is the variance of normal distribution, and the smooth models established by historical data for road sections S001, S002, and S003 can be: N(50,152.62), N(29,29.71), N(28,28.22); the congestion model is N(160,1840.24), N(67,463.97), N(68,261.04) in order.

After that, the deviation value is filtered and the current road conditions are determined. 2% is used as the probability threshold for filtering, that is, the lowest 2% in the smooth model and the highest 2% in the congestion model are filtered as outliers, and the road segment S001, The upper bounds of S002 and S003 are approximately 246, 110, and 100, respectively, and the lower bounds are approximately 25, 18, and 17, respectively. In this way, it can be known that the first sample detected by the vehicle detector 51 of the road segment S002 is a deviation value, which needs to be filtered. After filtering, the remaining samples are respectively inserted into the congestion and smooth model to compare the probability. It can be known that the road conditions of the three road sections are smooth.

，μ=29，σ ²=29.71，將各樣本套入函數可 求得各樣本的權重，計算出的代表值約為

，藍芽偵測器52計算出的 代表值約為27.81，GPS探偵車53僅有一個樣本23，可直接用做代表值。利用各設備的代表值套入模型可分別計算出此時段之可靠度，以單時段之可靠度計算為例，其中車輛偵測器51的可靠度約為0.068、藍芽偵測器52的可靠度約為0.071、GPS探偵車53的可靠度約為0.039，依可靠度為權重融合得到路段S002此時段的檢測基準值約為

。同理可以計算路段S001之 檢測基準值約為47.08、路段S003之檢測基準值約為26.67。 Next, calculate the reliability and fuse the detection reference value of the output road section. Take the road section S002 as an example, the sample is set into the weight of the probability calculated by the model N(29,29.71). The weighted average is calculated to calculate the representative value, where The travel time sample detected by the vehicle detector 51 has been excluded from filtering out the sample of 17, and the probability density function of N (29, 29.71) is assigned according to the normal state

, Μ =29, σ ² =29.71, the weight of each sample can be obtained by inserting each sample into the function, and the calculated representative value is about

The representative value calculated by the Bluetooth detector 52 is about 27.81, and the GPS detection vehicle 53 has only one sample 23, which can be directly used as the representative value. The reliability of this period can be calculated separately by using the representative value of each device into the model. Taking the reliability calculation of a single period as an example, the reliability of the vehicle detector 51 is about 0.068, and the reliability of the Bluetooth detector 52 Degree is about 0.071, and the reliability of the GPS reconnaissance vehicle 53 is about 0.039. According to the reliability, the fusion is obtained as the weight and the detection reference value of the road segment S002 during this period is about

. Similarly, it can be calculated that the detection reference value of the road section S001 is about 47.08, and the detection reference value of the road section S003 is about 26.67.

Finally, the travel time samples of each detection and detection equipment are detected, and the detection reference values of the road section S001, the road section S002, and the road section S003 are combined to obtain the detection base value of the path section S001~S003 as 26.45+47.08+26.67=100.2, and the error threshold The value is set to 20%. It can be calculated that there are 3 qualified samples of the vehicle detector 51, 5 total samples, and 60% of the qualified samples, and 4 qualified samples of the Bluetooth detector 52, the total sample There are 5 pens and the qualified ratio is 80%. There are 3 qualified samples of GPS detection vehicle 53 (2 dynamic fixed points and 1 dynamic line segment). There are 3 total samples and the qualified ratio is 100%. In this way, the detection for each time period can continuously measure the output level of all detection equipment.

To sum up, in this case, different detection treatments can be adopted for the detection equipment according to the heterogeneous detection sources, and the source data can be respectively mapped to the road segments according to the detection range, and the detection benchmark value can be fused based on the road segment, and then based on each The detection range of the detection equipment reconstructs the detection reference value of each road segment and compares it with the travel time samples of the detection equipment to produce detection result. Therefore, this case can detect all information sources in a fully automatic and highly scalable manner, without the need to install additional detection equipment or manual detection. When the source of detection equipment is increased, the detection results can be directly output, and there is no need to design the detection method again. . In addition, in this case, historical data can be classified according to the characteristics of road segments, weather, date and time, congestion or smoothness, and the probability distribution model of travel time under such classification can be calculated statistically as a filter deviation value to determine whether the current road condition is congested And the basis for reliability calculation. Then, it can accumulate real-time data for a period of time, and calculate the current weight distribution of each data source for fusion in real time, that is, considering the reliability of multiple periods to produce a more accurate detection benchmark value, thereby improving the accuracy of detection.

The above-mentioned embodiments are only illustrative of the effect of this case, and are not used to limit this case. Anyone who is familiar with this skill can modify and change the above-mentioned embodiments without departing from the spirit and scope of this case. Therefore, the scope of protection of rights in this case should be as listed in the scope of patent application mentioned later.

S201~S205‧‧‧Step

Claims (14)

A detection method for traffic detection equipment, the detection method includes: obtaining a plurality of travel time samples of a plurality of road sections of a plurality of detection equipment; the plurality of travel time samples are included in a probability distribution model including congestion history travel time and smooth history The historical travel time probability distribution model of the travel time probability distribution model to determine the model applicable to each road segment; based on the plurality of travel time samples and the models applicable to each road segment, calculate the reliability of each detection equipment; The reliability of the detection equipment is merged to produce the detection reference value of each road section; and each detection equipment is detected according to the detection reference value of each road section. The detection method as described in item 1 of the patent application scope further includes: inserting the plurality of travel time samples into the historical travel time probability distribution model to filter the deviation values of the plurality of travel time samples so as to fit into the The plurality of travel time samples of the historical travel time probability distribution model are the plurality of travel time samples filtered by the deviation value, thereby judging the model applicable to each road segment, and allowing the calculation of the reliability of each detection equipment based on the complex number The travel time samples are a plurality of travel time samples filtered by the deviation value. The detection method as described in item 1 of the patent application scope further includes: searching historical data according to road sections, weather, dates or time periods, to divide the historical data found into congestion data and smooth data, and then using uniform distribution or normal distribution Training and modeling the congestion data and the smooth data, thereby establishing the probability distribution model of the historical travel time probability and the smooth historical travel time probability distribution model, where the model trained with the normal distribution is expressed as N(μ,σ ² ), μ is the expected value of normal distribution, and σ ² is the number of variations of normal distribution. The detection method as described in item 1 of the patent application scope, wherein calculating the reliability of each detection equipment based on the plurality of travel time samples and the model applicable to each section includes: calculating each section of the detection equipment The representative values of the plurality of travel time samples are used to set the representative values of each section of the detection equipment to the applicable model of each section to calculate the reliability of the detection equipment. The detection method as described in item 1 of the patent application scope, wherein calculating the reliability of each detection equipment based on the plurality of travel time samples and the model applicable to each road segment includes: based on the plurality of travel time samples and each The model applicable to the road section calculates the reliability of the detection equipment in multiple periods. The detection method as described in item 1 of the patent application scope, in which the reliability of each detection and detection device is fused to produce a detection reference value for each of the sections including: taking the reliability of each detection and detection device as the weight, for The multiple travel time samples of each road segment are weighted average to produce the detection reference value of each road segment. The detection method as described in item 1 of the patent application scope, wherein the multiple detection equipments are classified into fixed-point detection traffic information sources and line detection Measure traffic information sources. For the detection equipment classified as the fixed-point detection traffic information source, the travel time samples belonging to the fixed point are obtained by using the road segment to which the device belongs or the corresponding and the speed of the detected point of the device. The detection equipment of the traffic information source obtains the travel time samples belonging to the line segment by using the path to which the device belongs or corresponds and the weight of the device road segment, and the fixed point travel time sample and the travel time sample of the line segment are the plurality of travel time samples. The detection method as described in item 7 of the patent application scope, wherein the detection of each detection equipment according to the detection reference value of each road section includes: classifying the detection reference value and classification of each road section as the fixed-point detection traffic information The multiple travel time samples of each section of the source detection equipment are compared to produce the qualification of the detection equipment classified as the fixed-point detection traffic information source, and the detection benchmark values of each section are combined and combined with the classification Multiple travel time samples of the detection equipment for detecting traffic information sources for the line segment are compared, and the eligibility of the detection equipment for detecting traffic information sources for the line segment is output. A detection system for traffic detection equipment, the detection system includes: a travel time sample module for generating a plurality of travel time samples for a plurality of sections of a plurality of detection equipment; a historical travel time probability distribution model module for Put the plurality of travel time samples into the historical travel time probability distribution model including the congested historical travel time probability distribution model and the smooth historical travel time probability distribution model to determine the applicable model for each road segment; the reliability calculation module uses According to the plurality of travel time samples The model applicable to each road section and each road section calculates the reliability of the detection equipment; the detection reference value calculation module is used to calculate the detection reference value of each road section based on the reliability of each detection equipment; and the detection module , Used to detect each detection equipment based on the detection reference value of each road section. The detection system as described in item 9 of the patent application scope, wherein the historical travel time probability distribution model module is further used to insert the plurality of travel time samples into the historical travel time probability distribution model to control the plurality of travels The time samples are filtered by the deviation value, so that the plurality of travel time samples nested in the historical travel time probability distribution model are the plurality of travel time samples filtered by the deviation value, thereby judging the applicable model for each road segment, and making the calculation The plurality of travel time samples based on the reliability of each detection equipment are the plurality of travel time samples filtered by the deviation value. The detection system as described in item 9 of the patent application scope also includes a modeling module for searching historical data according to road segments, weather, dates or time periods, to divide the historical data found into congestion data and smooth data, and then Use uniform distribution or normal distribution to train and model the congestion data and the smooth data, thereby establishing the probability distribution model of the historical travel time of congestion and the probability distribution model of the smooth historical travel time, where the model is expressed by the normal distribution Is N(μ,σ ² ), μ is the expected value of normal distribution, and σ ² is the variance of normal distribution. The detection system as described in item 9 of the patent application scope, in which the plural Each detection equipment is classified as a fixed-point detection traffic information source and a line segment detection traffic information source. For the detection equipment classified as the fixed-point detection traffic information source, it uses the road segment to which the device belongs or corresponds and the speed of the detected point to obtain the A fixed-point travel time sample, and for the detection equipment classified as the source of traffic information for the line segment, the travel time sample belonging to the line segment is obtained by using the device's or corresponding path and road segment weight, and the fixed point travel time sample and the line segment The travel time sample of is the plurality of travel time samples, and the detection module is further used to divide the detection reference value of each road section and the plurality of travel of each road section of the detection equipment classified as the fixed-point detection traffic information source Time samples are compared to determine the eligibility of the detection equipment classified as the fixed-point detection traffic information source, and the detection reference value of each road segment is combined and combined with the detection equipment classified as the line segment detection traffic information source. A plurality of travel time samples are compared to produce the qualification of the detection equipment classified as the source of traffic information for the line segment. The detection system as described in item 12 of the patent application scope, wherein the fixed-point detection traffic information source is divided into static fixed-point detection traffic information source and dynamic fixed-point detection traffic information source, and the line segment detection traffic information source is It is classified as static line segment detection traffic information source and dynamic line segment detection traffic information source. The detection system as described in item 13 of the patent application scope, wherein the static fixed-point detection traffic information source is a vehicle detector, and the static line segment detection traffic information source is an etag, a Bluetooth detector, or a license plate image, and The dynamic fixed-point detection traffic information source or the dynamic line segment detection traffic information The source of the information is the GPS of the vehicle, the GPS of the mobile phone or the positioning of the mobile signaling.

Images