vehicle detection method

A method for checking, distributing and controlling counter-plate vehicles  
CN 104103173 A
Summary
A method for arresting, investigating, and deploying vehicles with duplicate license plates, comprising the following steps: 1) data extraction; 2) lower bound calculation of travel time; 3) identification of vehicles using a fast hash algorithm of license plate numbers; And the data with the same license plate number is used to determine whether it is a suspicious vehicle with a duplicate license plate. The steps are as follows: time suspiciousness calculation, summary suspiciousness calculation, 5) Selection of the best control point: sort the suspicious degree of the vehicle license plate to obtain a list of suspicious license plate vehicles , when the degree of suspicion is greater than the set threshold, it is determined as a suspicious dummy vehicle; the trajectory of the suspicious dummy vehicle is analyzed, and the vehicle driving trajectory table is established according to the bayonet number and elapsed time, and the bayonet number that has passed the most times in the same time slot is as the best control point. The invention provides a method for detecting and deploying a license plate vehicle with fast calculation speed, high recognition accuracy and good reliability, and also has the function of deploying, controlling and intercepting.
Claim (3)
1. A method for arresting, investigating and deploying a vehicle with a license plate, characterized in that: the method comprises the following steps: 1) Data extraction and input data are: license plate number, the time when the vehicle passes through the bayonet, vehicle type, body color, bayonet number, Extract required data and carry out preprocessing; 2) Travel time lower bound calculation The two bayonet directly take the shortest distance according to a straight line, and the calculation formula of travel time lower bound T is: T=H/V; Wherein, H is the shortest straight line of the two bayonet Distance; V is the maximum speed; 3) Use the license plate number fast hash algorithm to identify the vehicle. After a piece of data is read, the Hash code is obtained through the Hash algorithm, and the Hash code is used as the key, and the license plate number, body color, vehicle type, and bayonet number are used as value; first determine whether the Hash table corresponding to the calculated Hash code has a value, if it does not indicate that it is a new vehicle record, then add the vehicle information of the license plate number, body color, vehicle type, and bayonet number to the Hash table; if the Hash code already exists , then query whether the license plate number has a collision or the same license plate number does not exist as a new vehicle, and add the vehicle information of the license plate number, body color, vehicle type, bayonet number to the Hash table; 4) The Hash code is generated when the suspected degree of the duplicated license plate is calculated. The same data with the same license plate number is used to determine whether it is a suspected vehicle with a duplicate license plate. The steps are as follows: 4. 1. Time suspicious degree calculation: Calculate the difference between the elapsed time of the car passing through the two bayonet ports as t, and when t < T, as Temporal suspicious set of license plates; The calculation formula of suspicious degree is: Pt = BU-Pi where, Pl is the time confidence degree, and the time confidence degree represents the accuracy of the elapsed time data measured by the sensor; 4. 2. Calculation of the aggregated suspicious degree: Steps 4.1 Obtain the time suspicious degree Pi, the vehicle type suspicious degree P2 and the body color suspicious degree P3 respectively represent the accuracy of the sensor to measure the vehicle type and body color, the vehicle type suspicious degree P2, and the body color suspicious degree P3 according to the measurement accuracy of the detector Obtained, is the ratio of the total number of correctly measured vehicles to the total number of vehicles; the calculation formula of the aggregated suspicious degree P is as follows: p = la-PiMi-PsMi-pj 5) The best deployment control point is selected to sort the suspicious degree of vehicle decks to obtain suspicious sets If the suspicious degree is greater than the set threshold, it will be judged as a suspicious dummy vehicle; conduct trajectory analysis on the suspicious dummy vehicle, establish a vehicle trajectory table according to the bayonet number and the elapsed time, and count the most passing times in the same time slot. The bayonet number is used as the best deployment point.
2. The method for arresting, investigating and deploying a license plate vehicle as claimed in claim 1, wherein: in the step 4), the calculation of the degree of suspicion of the license plate vehicle further comprises: 4.3) The calculation of the degree of suspicion combined with historical data By recording the suspicious degree of historical deck vehicles, and updating the vehicle suspicious degree P_, the formula of the suspicious degree P combined with the historical data: P = .
3. A method for arresting, checking and dispatching a vehicle with a license plate as claimed in claim 1 or 2, characterized in that: in the step 1), the preprocessing includes the following process: 1. 1) For the license plate number, there is a missing part in the elapsed time. 1.2) Divide the license plate number into two parts, the first part is the first two digits of the number, representing the region to which the vehicle belongs; the second part is the last five digits of the number; 1. 3) For the data of vehicle type and body color that are empty Completion, unified as NA unknown.
illustrate
A method for arresting, investigating, and deploying vehicles with fake license plates

technical field

The invention belongs to the field of intelligent traffic, relate to a kind of vehicle detection method of dummy card, especially a kind of method of checking and dispatching of dummy vehicle.

Background technique

[0002] The dummy car is commonly known as a cloned car, which refers to the vehicle that travels on the road by forging or illegally obtaining procedures such as other vehicle license plates and driving licenses. Most of the license plate cars are unknown and have no legal procedures, such as illegal zousi cars, stolen cars, scrapped cars, etc. It is impossible for these vehicles to apply for legal license plates, and they have to apply license plates in order to drive on the road. At the same time, if you use other license plates, you can drive illegally at will. Even if you are caught by the electronic police, you will not find yourself on your head, so you can get away with impunity, which is very harmful to the society. The cost of the decks is low, and by using fake decks, large trucks can evade taxes and fees as high as tens of thousands of yuan each year, disrupting the economic order. At the same time, the legitimate rights and interests of real car owners are damaged. Therefore, it is very necessary to detect and carry out investigation and control of vehicles with fake plates.

[0003] In the published patents and the patents under examination, the design of the detection method for the fake license plate vehicle is too ideal, and an effective method of investigation and control cannot be given. These methods include:

Method one (patent number: CN201110280822.6): based on the time in the bayonet database and the geographical position to identify whether to set the card, if the theoretical shortest travel time of two points is greater than the difference of the elapsed time that has been measured then think that this car is Set car.

[0005] 方法二(专利号:CN201110300956.X):基于内存的卡口数据,通过设置各个卡口 的时间阈值,当该时间阈值大于两卡口的经过时间则判断该车为套牌车。

[0006] 方法三(专利号:CN201310034242. 8):通过将城市道路划分为网格,对车辆行驶 轨迹进行分析,如果行驶轨迹不连续则为套牌车。

[0007] 然而在实际应用中,上述方法的效果并不理想。主要原因是由于:(1)这些方法未 充分考虑前端传感器在分析车辆特征时的误差,例如,在实际应用中,车牌识别精度一般为 80%左右,车辆颜色、车型等均可能因为外部环境影响而产生一定的错误率。(2)这些方法 只考虑到检测套牌车,未考虑如何布控拦截套牌车。

发明内容

[0008] 为了克服已有套牌车辆检测方法的计算速度较慢、识别精度较低、可靠性较差、无 布控拦截功能的不足,本发明提供了一种计算速度较快、识别精度较高、可靠性良好,兼有 布控拦截功能的套牌车辆缉查布控方法。

[0009] 本发明解决其技术问题所采用的技术方案是:

[0010] 一种套牌车辆缉查布控方法,所述方法包括以下步骤:

[0011] 1)数据提取

[0012] 输入数据为:车牌号码、车辆经过卡口的时间、车辆类型、车身颜色、卡口号码,提 取所需数据并进行预处理;

[0013] 2)行程时间下界计算

[0014] 将两卡口直接按照直线取最短距离,行程时间下界T的计算公式为:T = H/V ;其 中,Η为两卡口最短直线距离;V为最大速度;

[0015] 3)采用车牌号码快速hash算法识别车辆

[0016] 当一条数据读入后,经过Hash算法得到Hash编码,将Hash编码作为键,车牌号 码、车身颜色、车辆类型、卡口号码作为值;首先判断计算得到的Hash编码对应的Hash表是 否有值,没有说明是全新车辆记录,则将车牌号码、车身颜色、车辆类型、卡口号码车辆信息 加入Hash表中;如果Hash编码已经存在,则查询是否车牌号码存在碰撞不存在相同车牌号 码则作为新增车辆,将车牌号码、车身颜色、车辆类型、卡口号码车辆信息加入Hash表;

[0017] 4)套牌车嫌疑度计算

[0018] 当产生Hash编码相同且车牌号相同的数据,判断是否为套牌嫌疑车,步骤如下:

[0019] 4· 1.时间可疑度计算:

[0020] 计算该车通过两卡口的经过时间之差为t,当t < T时,为时间可疑套牌车;可疑 度的计算公式为:Pi = 1 - * Pi

[0021] 其中,Pl为时间置信度,时间置信度表示传感器测量的经过时间数据的准确度;

[0022] 4. 2.汇总可疑度计算:

[0023] 步骤4. 1得到时间可疑度Pi,步骤4. 1得到时间可疑度Pi,车辆类型可疑度P2、车 身颜色可疑度P3分别代表传感器测量车辆类型、车身颜色的准确度,车辆类型可疑度P 2、车 身颜色可疑度P3根据检测器的测量精度得到,是正确测量车辆总数与总车辆数的比值;

[0024] 汇总可疑度P的计算公式如下:

[0025] P = 1- (1-P) * (1-P2) * (1-P3)

[0026] 5)最佳布控点选择

[0027] 对车辆套牌可疑度进行排序得到可疑套牌车辆表,当可疑度大于设定阈值,判定 为可疑套牌车辆;

[0028] 对可疑套牌车辆进行轨迹分析,根据卡口号码以及经过时间,建立车辆行驶轨迹 表,将同一时间槽内经过次数最多的卡口号作为最佳布控点。

[0029] 进一步,所述步骤4)中,套牌车嫌疑度计算还包括:

[0030] 4. 3)与历史数据结合的可疑度计算

[0031] 通过记录历史套牌车辆可疑度,并更新车辆可疑度PMW,与历史数据结合的可 疑度 P 的公式:P = 1-(l-pj * (l-pnew)。

[0032] 再进一步,所述步骤1)中,预处理包括如下过程:

[0033] 1. 1)对车牌号码,经过时间有缺失的数据进行过滤;

[0034] 1. 2)将车牌号码分为两部分。第一部分为号码前两位,代表车辆所属地区;第二 部分为号码后五位;

[0035] 1. 3)对于车辆类型、车身颜色为空的数据进行补全,统一为NA未知。

[0036] 本发明的有益效果主要表现在:1、计算速度快:本发明通过Hash算法计算出对应 车牌号码的Hash编码。可通过Hash编码快速进行车牌号码的查找、去除与统计新增车辆 信息。

[0037] 2、量化嫌疑度:该方法解决了传感器在收集传播过程中可能是数据丢失或者产生 错误,通过对嫌疑度的定量分析,找到最可疑的套牌车辆。

[0038] 3、自学习:该算法在计算套牌车辆嫌疑度时,充分利用了历史数据,通过对套牌嫌 疑度历史数据的更新,帮助交通管理部门有针对性的处理高嫌疑度的套牌车。

[0039] 4、数据融合:该算法不再仅对经过时间进行判断,而是加入了车辆类型,车身颜色 等数据,该算法并不局限于这些数据,如果传感器能够采集更具体的数据,也可以很方便的 进行嫌疑度计算。

[0040] 5、最佳布控点:该算法通过对套牌嫌疑度较高的车辆,进行轨迹分析。找到在特定 时段该车频繁出现的卡口号码,将该点作为最佳稽查布控点,具有较好的现实意义。

附图说明

[0041] 图1是套牌车辆缉查布控方法中套牌车检测流程图。

[0042] 图2是套牌嫌疑度的流程图。

[0043] 图3是最佳布控点的流程图。

[0044] 图4是布控的示意图。

具体实施方式

[0045] 下面结合附图对本发明作进一步描述。

[0046] 参照图1〜图4,一种套牌车辆缉查布控方法,所述方法包括以下步骤:

[0047] 1)数据提取

[0048] 提取所需数据并进行简单预处理;数据的来源、形式有很高的通用性,输入数据 为:车牌号码、车辆经过卡口的时间、车辆类型、车身颜色、卡口号码等数据。预处理包括如 下:

[0049] 对车牌号码,经过时间有缺失的数据进行过滤。

[0050] 根据国家普通车牌号码规范,将车牌号码分为两部分。第一部分为号码前两位,代 表车辆所属地区;第二部分为号码后五位,由两个英文字母和三个数字组成。数据提取也按 这种规则划分车牌号码。

[0051] 对于车辆类型、车身颜色为空的数据进行补全,统一为NA未知。

[0052] 2)行程时间下界计算

[0053] 通过卡口信息数据表可以得到各个卡口之间的最短路径,为计算方便,将两卡口 直接按照直线取最短距离。本发明假设车辆最大速度为1〇〇千米/小时,行程时间下界T 的计算公式为:T = H/V ;其中,Η为两卡口最短直线距离;V为最大速度。

[0054] 3)采用车牌号码快速hash算法识别车辆

[0055] 当一条数据读入后,经过Hash算法,将Hash码作为键,车牌号码、车身颜色、车辆 类型、卡口号码等作为值。首先判断计算得到的Hash编码对应的Hash表是否有值,没有说 明是全新车辆记录,则将车牌号码、车身颜色、车辆类型、卡口号码等车辆信息加入Hash表 中;如果Hash编码已经存在,则查询是否车牌号码存在碰撞不存在相同车牌号码则作为新 增车辆,将车牌号码、车身颜色、车辆类型、卡口号码等车辆信息加入Hash表;

[0056] 车牌号码快速hash算法:Hash就是把任意长度的输入通过Hash算法,变换成固 定长度的输出,该输出就是Hash值。这种转换是一种映射,也就是,输入值必定有唯一的 Hash值与之对应。不同的输入值可能会生成相同的Hash码,而不可能从Hash码来唯一的 确定输入值。通过查找Hash编码能快速查找数据,对大数据查询具有较高的速度。数学描 述如下:

[0057] 输入值:keyl ;key2 ;

[0058] Hash 码:F (keyl) ;F (key2);

[0059] 其中,F为Hash函数。

[0060] 当keyl尹key2,然而F(keyl) = F(key2);这种现象称为碰撞。本文通过二次比 较消除碰撞。

[0061] Hash函数F表示方法如下:

[0062] hash=0; Λ 0: for(inl j = 0; j < 5; j++) \ hash = (hash << 4> 十 carmmilxxcharAiQ^ if((x = hash & OxFOOOOOOOL) != 0) t \ hash Λ= (π » 24); } hash&= ~x; }

[0063] 其中,carnumber为经过数据提取的车牌号码后5位;carnumber. charAt (j)方法 将第j位对应的字符转化为对应ASCII码。经过迭代运算、位移运算、位运算得出Hash码。

[0064] 4)套牌车嫌疑度计算

[0065] 经过上面步骤,当产生Hash码相同且车牌号相同的数据,判断是否为套牌嫌疑 车,具体步骤如下:

[0066] 4. 1.时间可疑度计算:

[0067] 计算该车通过两卡口的经过时间之差为t,该卡口理论最小时间通过步骤2)

[0068] 步骤可得为T,当t < T时,为时间可疑套牌车。

[0069] 可疑度的计算公式为A = [l-(t/T)~2]*Pl

[0070] 其中,Pl为时间置信度,时间置信度表示传感器测量的经过时间数据的准确度。

[0071] 4. 2.汇总可疑度计算:

[0072] 对一条完整数据包括车牌号码、车身颜色、车辆类型、卡口号码、经过时间等;进行 上述步骤可以得到时间可疑度Pi,

[0073] 车辆类型可疑度P2、车身颜色可疑度P3分别代表传感器测量车辆类型、车身颜色 的准确度,车辆类型可疑度P 2、车身颜色可疑度P3根据检测器的测量精度得到,是正确测量 车辆总数与总车辆数的比值;

[0074] 汇总可疑度P的计算公式如下:

[0075] P = 1- (1-P) * (1-P2) * (1-P3)

[0076] 4.3.与历史数据结合的可疑度计算

[0077] 本发明通过记录历史套牌车辆可疑度,并更新当前车辆可疑度Ρη",当某辆可 疑套牌车辆反复出现可疑记录,通过与历史数据的结合可以提高该车的可疑度,最后对可 疑度较高的车辆进行轨迹分析得到最佳布控点。

[0078] 与历史数据结合的可疑度P的公式:P = Ι-α-Ρ^Μΐ-Ρ^)

[0079] 6)最佳布控点选择

[0080] 对车辆套牌可疑度进行排序得到可疑套牌车辆表,当可疑度大于设定阈值,判定 为可疑套牌车辆;

[0081] 对套牌车辆进行轨迹分析,根据卡口号码以及经过时间,可以建立车辆行驶轨迹 表。将同一时间槽内经过次数最多的卡口号作为最佳布控点。

[0082] 本实施例采用杭州市821个卡口传感器采集的数据。卡口位置信息格式如下:

[0083]

 

Figure CN104103173AD00081

[0084] 抽取了杭州市一周所有卡口记录的数据,总共47483502条。格式如下:

[0085]

 

Figure CN104103173AD00082

[0086]

[0087] 行程时间下界计算模块,GPS计算卡口之间最短距离方法如下:

[0088] EARTH_RADIUS = 6378137 ;

[0089] 经纬度转弧度公式:rad = d* π /180. 0 ;

[0090] 通过两卡口的经纬度求最短距离方法如下:

[0091] radLatl = rad(yl);

[0092] radLat2 = rad(y2);

[0093] a = radLatl-radLat2 ;

[0094] b = rad (xl)-rad (x2);

[0095] s = 2*Math. asin (Math, sqrt (Math, pow (Math, sin (a/2),2)+Math. cos (radLatl)*Math. cos(radLat2)*Math. pow(Math, sin(b/2),2)));

[0096] s = s*EARTH_RADIUS ;

[0097] 其中A卡口经纬度为xl,yl ;B卡口经纬度为x2,y2。s为AB卡口之间的最短距 离。

[0098] Hash码计算:根据上述车牌号码快速hash算法经过五次迭代,可以得到对应Hash 码如下:

[0099]

[0100]胃套牌车嫌疑度计算:某辆车经过A,B

 

Figure CN104103173AD00091

两卡口用时8分钟;该卡口最短行程时间为 10分钟;该车车辆类型、车身颜色均发生改变。该车历史套牌可疑度为〇. 5 ;则该车的套牌 嫌疑度计算如下:

[0101] 时间可疑度计算:

[0102] 根据可疑度的计算公式A = [l-(t/T)~2]*Pl

[0103] 可得:P! =

 

Figure CN104103173AD00092

* ρ! = 0.18,其中卩丨=0· 5

[0104] 汇总可疑度计算:

[0105] 根据汇总可疑度计算公式:P = ι-α-ρχι-ρχι-ρρ

[0106] 根据实际数据与专家经验经分析,令传感器对车辆类型、车身颜色的识别准确度 分别为Ρ2 = Ρ3 = 〇· 3

[0107] 可得:Ρ = 1-(l-Pj *(1-Ρ2)*(1-Ρ3) =0· 5982

[0108] 与历史数据结合的可疑度计算

[0109] 根据与历史数据结合的公式:ρ = ι-α-ρ^Μΐ-ρ^)

[0110] 可得:Ρ = 1-(1-Pild)*(l-Pnew) = 0· 7991

[0111] 通过汇总可疑度计算以及与历史可疑度相结合,更新得到该车套牌可疑度为 0. 7991,当该车反复出现套牌可疑情况,该可疑度还会继续升高。对可疑度较高的车辆,交 通管理部门可以有针对性的调查,及时发现套牌车辆。

[0112] 最佳布控点选择方法:通过对一周47483502条数据进行套牌可疑度分析,找到了 可疑度最高的一车辆信息如下:

[0113]

 

Figure CN104103173AD00093

[0114] 将时间划分为一小时时间槽,统计该车辆各个开卡在该时间槽的通过次数:

[0115]

 

Figure CN104103173AD00101

[0116] 通过卡口号码-时间槽信息表找到在5点到6点间该车有4次经过2148801卡口, 如下:

[0117]

[0118]

 

Figure CN104103173AD00102

[0119] 通过该表可以选择在5点20到6点间在2148801卡口对该车进行稽查以确认该 车是否为套牌车辆。

[0120] 浙A3B108车行驶轨迹如图4,标记点(黑点)为2148801卡口地理位置。

专利引用
引用的专利 申请日期 公开日 申请人 专利名
CN101261771A * 2008年3月14日 2008年9月10日 康华武 一种道路车辆身份的自动稽查方法
CN101540105A * 2009年4月15日 2009年9月23日 四川川大智胜软件股份有限公司 一种基于车辆牌照识别和网格化监控的套牌车检测方法
CN101587643A * 2009年6月8日 2009年11月25日 宁波大学 一种套牌车的识别方法
CN103730007A * 2013年12月20日 2014年4月16日 南威软件股份有限公司 一种卡口过车实时比对分析报警方法及其报警系统
JP2014002534A *       没有名称
US20140074567 * 2013年9月12日 2014年3月13日 Accenture Global Services Limited Electronic Toll Management
分类
   
国际分类号 G08G1/017
法律事件
日期 代码 事件 说明
2014年10月15日 C06 Publication  
2014年11月12日 C10 Entry into substantive examination  
2016年8月24日 C14 Grant of patent or utility model

Guess you like

Origin http://43.154.161.224:23101/article/api/json?id=326453316&siteId=291194637