WO2020034722A1 - 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法 - Google Patents

一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法 Download PDF

Info

Publication number
WO2020034722A1
WO2020034722A1 PCT/CN2019/089077 CN2019089077W WO2020034722A1 WO 2020034722 A1 WO2020034722 A1 WO 2020034722A1 CN 2019089077 W CN2019089077 W CN 2019089077W WO 2020034722 A1 WO2020034722 A1 WO 2020034722A1
Authority
WO
WIPO (PCT)
Prior art keywords
corner
license plate
class
corner points
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/089077
Other languages
English (en)
French (fr)
Inventor
李旭
曹继文
倪培州
韦坤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Original Assignee
Southeast University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southeast University filed Critical Southeast University
Priority to US17/057,915 priority Critical patent/US11367215B2/en
Publication of WO2020034722A1 publication Critical patent/WO2020034722A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B62—LAND VEHICLES FOR TRAVELLING OTHERWISE THAN ON RAILS
    • B62D—MOTOR VEHICLES; TRAILERS
    • B62D15/00—Steering not otherwise provided for
    • B62D15/02—Steering position indicators ; Steering position determination; Steering aids
    • B62D15/029—Steering assistants using warnings or proposing actions to the driver without influencing the steering system
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60P—VEHICLES ADAPTED FOR LOAD TRANSPORTATION OR TO TRANSPORT, TO CARRY, OR TO COMPRISE SPECIAL LOADS OR OBJECTS
    • B60P3/00—Vehicles adapted to transport, to carry or to comprise special loads or objects
    • B60P3/12—Vehicles adapted to transport, to carry or to comprise special loads or objects for salvaging damaged vehicles
    • B60P3/125—Vehicles adapted to transport, to carry or to comprise special loads or objects for salvaging damaged vehicles by supporting only part of the vehicle, e.g. front- or rear-axle
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
    • B60R11/00—Arrangements for holding or mounting articles, not otherwise provided for
    • B60R11/04—Mounting of cameras operative during drive; Arrangement of controls thereof relative to the vehicle
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/24—Aligning, centring, orientation detection or correction of the image
    • G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G06V20/625—License plates
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
    • B60R11/00—Arrangements for holding or mounting articles, not otherwise provided for
    • B60R2011/0042—Arrangements for holding or mounting articles, not otherwise provided for characterised by mounting means
    • B60R2011/008—Adjustable or movable supports
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20112—Image segmentation details
    • G06T2207/20164—Salient point detection; Corner detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30248—Vehicle exterior or interior
    • G06T2207/30252—Vehicle exterior; Vicinity of vehicle
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition

Definitions

  • the invention belongs to the field of intelligent operation of road rescue equipment, and particularly relates to a method for inducing positive azimuth dragging of road rescue equipment based on the characteristics of the license plate corners.
  • the rescue environment in a positive direction is relatively complicated. Most of them face a narrow and congested environment, and the working space is severely restricted.
  • the trailer is sandwiched between the left and right vehicles and arranged in a "three" shape. Implement drag and rescue.
  • the towing device for road rescue equipment is mainly composed of a folding arm, a telescopic arm, a swing arm, and two side support arms.
  • the road rescue equipment is located in front of the front of the trailer.
  • the road rescue equipment passes When reversing, align the supporting arms on both sides with the two front wheels of the trailer, fix the tires on the two front wheels of the trailer, and finally pull the crane by the folding arm to pull the trailer away.
  • towing operation in the positive direction mainly relying on the driver's experience, by repeatedly reversing the position to correct the position, the alignment of the support arm and the front wheel of the trailer is completed, and the towing operation is completed.
  • Depends on the operator's operating experience it takes too long, and the rescue efficiency is low.
  • the present invention proposes a positive azimuth tow induction method based on the characteristics of the license plate corners to assist the driver in carrying out tow operations, thereby achieving the purpose of improving the rescue efficiency of road rescue equipment. Due to the complexity and variability of the tow environment, the positive-direction drag induction method proposed by the present invention should not only have good real-time performance, but also have the ability to adapt to various vehicle models and environmental changes, and the ability to resist various noises and interferences. This is also a prerequisite guarantee for the reliability of the method.
  • the present invention proposes a method for inducing positive azimuth dragging of road rescue equipment that has both good real-time performance and strong environmental adaptability and anti-interference ability.
  • the technical solution adopted by the present invention is: a method for inducing positive azimuth dragging of road rescue equipment based on the corner feature of a license plate.
  • the image acquisition and pre-processing method is as follows: install a vehicle camera on the rear arm of the road rescue equipment, collect the image of the working area behind the trailer, and first convert the color acquisition image into a grayscale image, reduce Small image processing data volume, smooth filtering for grayscale images.
  • step (2) the corner detection is performed on the smoothed gray image by using a Harris corner detection algorithm to obtain the corner points in the image, and all the obtained corner points are stored in the corner point set O;
  • the corners obtained by Harris corner detection are optimized to obtain a high-quality corner set A, as follows:
  • each corner a n n 1,2,3 ..., N in the high-quality corner set A as a class, and each class has one and only one corner.
  • the class spacing D rg between two, D rg represents the class spacing between the r- th class and the g- th class, and the class spacing D rg is defined as each corner point in the r- th class and each corner point in the g-th class.
  • the average value of the Euclidean distance between each class currently has only one corner point, that is, only the corner point a r in the r- th class, and only the corner point a g in the g- th class, so the class distance D rg is calculated at this time. for:
  • d rg represents the Euclidean distance between the corner points a r and a g , ie among them, with The horizontal and vertical coordinates of the corner point a r , with The horizontal and vertical coordinates of the corner point a g , respectively;
  • M g represents the number of corner points in the g-th class.
  • d jk represents the Euclidean distance between the corner points p j and q k , that is, among them, with The horizontal and vertical coordinates of the corner point p j , with The horizontal and vertical coordinates of the corner point q k respectively;
  • the class spacing matrix H is updated to obtain a new V ⁇ V class spacing matrix. And return to the above sub-step (3.2).
  • step (5) the method of dragging induction is as follows: according to the position of the license plate center (x 0 , y 0 ) determined in the license plate positioning in step (4) in the image, a direction prompt is given in real time to induce The driver performs the reverse operation: if the trailer license plate center is on the left side of the image center in the image, the driver is prompted to fall to the right; if the trailer license plate center is on the right side of the image center in the image, the driver is prompted to fall left, In this way, the alignment of the support arms on both sides with the two front wheels of the trailer is completed, and then the trailer is fixed and held away by pulling.
  • the induction method proposed by the present invention fully considers and utilizes the character corner characteristics of different vehicles on the license plate, and can adapt to various different vehicle models.
  • the processing speed is fast and the real-time performance is good.
  • the induction method proposed by the present invention uses gray images for corner detection, and performs corner optimization before corner clustering, which effectively improves the operation speed of the induction method and ensures Real-time.
  • FIG. 1 is a flowchart of a forward azimuth tow induction method for road rescue equipment according to the present invention
  • FIG. 2 is a schematic perspective view of the road rescue equipment towing operation in a positive direction
  • FIG. 3 is a top view of a road rescue equipment in a positive azimuth towing operation
  • FIG. 4 is a schematic diagram of a structure of a towing device of a road rescue equipment and a camera installation position;
  • FIG. 5 is a frame image acquired
  • FIG. 6 is a pre-processing result of an acquired image
  • FIG. 7 is a diagram of a corner detection result of a captured image
  • FIG. 8 is a diagram of the optimized result of the corner points of the acquired image
  • FIG. 9 is a graph of the clustering results of the collected image corners
  • FIG. 10 is a result of filtering effective corner points of the license plate characters of the collected image
  • FIG. 11 is a map of license plate positioning results.
  • the towing device of road rescue equipment using a pickup truck as an example is mainly composed of a folding arm, a telescopic arm, a swing arm, and two side supporting arms.
  • the specific form is shown in FIG. 4.
  • the road rescue equipment is located in front of the trailer's head.
  • the road rescue equipment is aligned with the two front wheels of the trailer to be reversed through the reverse operation.
  • the two front wheels are fixed with a fixed wheel.
  • they are pulled and lifted by folding arms, and will be towed away by the trailer.
  • the operation of aligning the front wheel of the trailer with the support arm of the road rescue equipment is mainly dependent on human factors such as the technical experience of the driver, and usually requires reversing the vehicle several times. Corrected that the operation took too long and the rescue efficiency was low.
  • the present invention proposes a positive directional dragging induction method for road rescue equipment that not only has good real-time performance, but also has strong environmental adaptability and anti-interference ability.
  • the method mainly includes the following steps:
  • the induction method proposed in the present invention is mainly for drag-and-rescue rescue of blue-card small cars.
  • the vehicle license is a common feature of most social vehicles and is universal.
  • the license plate characters are printed characters, and the character strokes contain relatively rich corner information. Therefore, corner features can be extracted through the corner detection.
  • Typical corner detection methods include Moravec corner detection, Susan corner detection, Fast corner detection, Harris corner detection and other algorithms.
  • the Moravec corner detection algorithm is simple, but the amount of calculation is large, and the false detection rate is high.
  • the point detection algorithm is not sensitive to noise, but the positioning accuracy is poor.
  • Harris corner detection algorithm is simple, the calculation amount is not large, the corner points are extracted uniformly, the adaptability is strong, and the stability is good.
  • a Harris corner detection algorithm is used to detect the smoothed grayscale image to obtain the corner points in the image. See FIG. 7 of the description, and all the obtained corner points are stored in the corner point set O.
  • the location of Harris corner detection is relatively accurate, and the corner extraction is relatively uniform, but most of the obtained corner locations appear in batches in the form of neighborhoods. See FIG. 7 of the description.
  • the amount of calculation in the class process improves the real-time performance of the induction method, and the corners obtained by the Harris corner detection can be optimized according to the corner strength to obtain a more representative corner position.
  • the effective working distance of the positive direction dragging operation is generally 1 to 5 meters.
  • the focal length of the camera is selected to be 4 to 8 mm.
  • the size of the image collected by the camera in the present invention is 960 ⁇ 540.
  • the minimum allowable distance between the corners can be set to 5 pixels.
  • the corner points obtained by Harris corner detection are optimized to obtain a high-quality corner point set A. See FIG. 8 of the description.
  • the specific sub-steps are:
  • the class spacing D rg between two, D rg represents the class spacing between the r- th class and the g-th class.
  • the class spacing D rg is defined as each corner point in the r- th class and each of the g-th class. The average value of the Euclidean distance between the corner points. Because there is only one corner point in each class, that is, only the corner point a r in the r- th class and the corner point a g in the g- th class, the class distance D The calculation formula of rg is:
  • d rg represents the Euclidean distance between the corner points a r and a g , ie among them, with The horizontal and vertical coordinates of the corner point a r , with The horizontal and vertical coordinates of the corner point a g are respectively expressed.
  • the class spacing matrix H is updated to obtain a new V ⁇ V class spacing matrix. And return to the above sub-step (3.2).
  • the domestic blue license plate has national standards, and its printed characters are seven digits, each of which contains at least one corner point, and the length and width of the seven printed characters The ratio is about 4; 2
  • the camera is installed at the middle fixed position of the folding arm of the road rescue equipment and horizontally faces the rear of the road rescue equipment. In the process of towing operation in a positive direction, the trailer is located in the rear area of the road rescue equipment. The distance is generally 1 to 5 meters, and the focal length of the camera can be selected to be 4 to 8 mm.
  • the size of the image collected by the camera in the present invention is 960 ⁇ 640. Within the effective working distance of 1 to 5 meters, the area of the license plate character area in the collected image is The pixel size varies between 300 and 5000.
  • the road rescue equipment is generally located in front of the front of the trailer, and the two are arranged in a "one" shape.
  • the on-board camera is installed on the folding arm of the road rescue equipment.
  • the central axis of the camera collection area is consistent with the central axis of the road rescue equipment and its towing device, and the front license plate of the trailer is generally located in the middle of its front.
  • the left-right positional relationship between the center of the output picture and the center of the license plate of the trailer in the picture determines the relative positional relationship of the road rescue equipment and the trailer.
  • the driver faces the front of the road rescue equipment, so according to the position of the license plate center (x 0 , y 0 ) determined in the license plate positioning in step (4) in the image, see the instruction manual FIG. 11 gives a direction prompt in real time to induce the driver to perform a reverse operation: if the trailer license plate center is on the left side of the image center in the image, the driver is prompted to fall to the right; if the trailer license plate center is in the image center On the right side, the driver is prompted to fall to the left, thereby completing the alignment of the support arms on both sides with the two front wheels of the trailer, and then fixing the trailer by holding the tire and pulling it away.
  • the method provided by the present invention can realize real-time induction during the positive directional dragging operation of rescue equipment, and effectively improve the rescue efficiency of road rescue equipment.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Mechanical Engineering (AREA)
  • Transportation (AREA)
  • Combustion & Propulsion (AREA)
  • Chemical & Material Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，该方法结合道路救援装备的结构及正方位拖牵作业的特点，首先通过安装摄像头采集作业区域图像，并进行灰度处理及高斯平滑滤波；接着对平滑后的灰度图像进行角点检测并根据角点强度实现优选；随后对优选得到的角点进行层次聚类；然后筛选出车牌字符有效角点集合，实现车牌定位；进而根据车牌定位的结果实施拖牵诱导，从而达到提高道路救援装备救援效率的目的。该诱导方法具有良好的实时性、环境适应力和抗干扰能力，有效提高了道路救援装备的救援效率。

Description

一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法 技术领域
本发明属于道路救援装备智能化操作领域，尤其涉及一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法。
背景技术
随着社会经济的不断增长，我国道路交通基础设施建设及汽车工业得到飞速发展，机动车保有量在大幅增加，汽车在行驶途中由于驾驶不当或车辆自身故障等导致的车辆无法正常行驶等特殊情况与日俱增，作为汽车后市场的道路清障救援行业逐渐被社会和民众重视，道路清障救援行业对保障道路运输安全高效的重要性日益凸显。但是由于交通事故的发生具有无法彻底预防的客观必然性，因此在很多汽车交通事故后，常有因清障救援不及时或救援效率不高而导致城市交通拥堵或二次连锁事故发生，比如道路救援装备无法快速准确从正方位把事故车辆托离现场而导致交通堵塞就是其中之一。一般正方位救援环境比较复杂，大多面对狭窄及拥堵环境，作业空间严重受限，被拖车夹在左右两车中间，呈“三”字形排列，此时道路救援装备只能从被拖车的前方实施拖牵救援。
道路救援装备的拖牵装置主要由折臂、伸缩臂、摆臂及两侧托臂组成，在正方位拖牵作业实施的过程中，道路救援装备位于被拖车车头的前方，首先道路救援装备通过倒车作业，将其两侧托臂分别与被拖车两前轮进行对准，然后对被拖车两前轮进行抱胎固定，最后通过折臂牵拉起吊，将被拖车拖离。但是传统情况下在进行正方位拖牵作业时，主要凭借驾驶员经验，通过多次反复倒车进行位置修正，实现托臂与被拖车前轮的对准，进而完成拖牵作业，这种现状过于依赖作业人员操作经验，且费时过长，救援效率低下。
为了解决上述问题，本发明提出一种基于车牌角点特征的正方位拖牵诱导方法辅助驾驶员实施拖牵作业，从而达到提高道路救援装备救援效率的目的。由于拖牵环境的复杂多变性，本发明所提出的正方位拖牵诱导方法不仅应具有良好的实时性，还应具有适应各种车型及环境变化的能力以及抗各种噪声和干扰的能力，这也是该方法可靠性的前提保证。
发明内容
发明目的：针对以上问题，本发明提出一种既具有良好实时性，又具有较强环境自适应能力和抗干扰能力的道路救援装备正方位拖牵诱导方法。
技术方案：为实现本发明的目的，本发明所采用的技术方案是：一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，该方法包括如下步骤：
(1)采集拖车后方作业区域图像并对图像进行预处理；
(2)对预处理后的图像进行角点检测及优选；
(3)对优选的角点进行聚类；
(4)根据优选的角点对车牌定位并得到车牌中心；
(5)根据车牌中心实施拖牵诱导。
其中，在步骤(1)中，图像采集及预处理方法如下：在道路救援装备尾部折臂上安装车载摄像头，对拖车后方作业区域图像采集，并且先将彩色采集图像转成灰度图像，减小图像处理数据量，对灰度图像平滑滤波。
其中，在步骤(2)中，采用哈里斯角点检测算法对平滑后的灰度图像进行角点检测，获得图像中的角点，并将得到的所有角点存储到角点集合O中；对哈里斯角点检测获得的角点进行优选，得到优质角点集合A，方法如下：
(2.1)按照角点横、纵坐标依次递增的顺序，遍历角点集合O中的所有角点，并在以每个角点为中心，半径为5像素的圆形区域内，将该区域角点强度最大的角点保存到优质角点集合A中；
(2.2)上述遍历结束后，对优质角点集合A中重复保存的具有相同坐标及角点强度的角点，只保留其中一个，并将多余的相同角点从优质角点集合A中删除，最后可得到N个不同的优质角点。
其中，在步骤(3)中，角点聚类方法如下：根据角点之间的距离对优选得到的角点集合A进行凝聚层次聚类，得到I个车牌字符预选角点集合B i，i＝1,2,3…,I，其中，i表示预选角点集合的序号，方法如下：
(3.1)把优质角点集合A中的每个角点a nn＝1,2,3…,N，各视为一个类，每个类中有且只有一个角点，计算当前所有类两两之间的类间距D rg，D rg表示第r类和第g类之间的类间距，定义类间距D rg为第r类中的每个角点与第g类中的每个角点之间欧式距离的平均值，由于当前每个类中有且只有一个角点，即第r类中只有角点a r，第g类中只有角点a g，故此时类间距D rg计算公式为：
D rg＝d rg
式中，d rg表示角点a r和a g之间的欧式距离，即
Figure PCTCN2019089077-appb-000001
Figure PCTCN2019089077-appb-000002
其中，
Figure PCTCN2019089077-appb-000003
和
Figure PCTCN2019089077-appb-000004
分别表示角点a r的横、纵坐标，
Figure PCTCN2019089077-appb-000005
和
Figure PCTCN2019089077-appb-000006
分别表示角点a g的横、纵坐标；
根据类间距D rg的物理意义，则有D rg＝D gr，且当r＝g时，D rg＝0，将上述计算得到的所有类间距用矩阵形式表示，可得到初始的N×N类间距矩阵
Figure PCTCN2019089077-appb-000007
Figure PCTCN2019089077-appb-000008
矩阵H中的r行g列元素即为D rg；
(3.2)遍历当前的欧式距离矩阵H，找到矩阵H中的非对角线最小元素，即当前类间距最小值，设为D st，且s≠t，表示第s类和第t类为当前距离最近的两个类，若D st＜D th，则将第t类中的角点归并到第s类中，进而将第s类和第t类合并为一个新类，并将合并后的角点总类数记为V，进入子步骤(3.3)；否则结束聚类计算，可得到I个车牌字符预选角点集合B i，i＝1,2,3…,I，其中，i表示预选角点集合的序号；I表示车牌字符预选角点集合的总数，且I＝V，该子步骤的判别条件中，D st为当前矩阵H中的非对角线最小元素；D th为类间距最小值阈值；
(3.3)重新计算当前剩余的类与合并得到的新类两两之间的类间距D rg，此时类间距D rg的计算公式为：
Figure PCTCN2019089077-appb-000009
式中，M r表示第r类中的角点个数，该类中的角点表示为p j，j＝1,2,3…,M r；M g表示第g类中的角点个数，该类中的角点表示为q k，k＝1,2,3…,M g；d jk表示角点p j和q k之间的欧式距离，即
Figure PCTCN2019089077-appb-000010
其中，
Figure PCTCN2019089077-appb-000011
和
Figure PCTCN2019089077-appb-000012
分别表示角点p j的横、纵坐标，
Figure PCTCN2019089077-appb-000013
和
Figure PCTCN2019089077-appb-000014
分别表示角点q k的横、纵坐标；
用上述重新计算得到的类间距，根据类间距D rg及类间距矩阵H的物理意义， 对类间距矩阵H进行更新，可得到新的V×V类间距矩阵
Figure PCTCN2019089077-appb-000015
并重新返回上述子步骤(3.2)。
其中，在步骤(4)中，车牌定位方法如下：对于步骤(3)中聚类得到的车牌字符预选角点集合B i，i＝1,2,3…,I，从车牌字符预选角点集合B i中筛选出车牌字符有效角点集合C，进而确定车牌位置，方法如下：
(4.1)初始化i＝1；
(4.2)若车牌字符预选角点集合Bi满足
Figure PCTCN2019089077-appb-000016
则进入子步骤(4.3)，否则进入子步骤(4.5)，该子步骤的判别条件中，
Figure PCTCN2019089077-appb-000017
为车牌字符角点集合Bi中的角点个数；num th为车牌字符角点集合的角点个数阈值；
(4.3)遍历车牌字符预选角点集合Bi，确定该集合中角点的最大横坐标
Figure PCTCN2019089077-appb-000018
最小横坐标
Figure PCTCN2019089077-appb-000019
最大纵坐标
Figure PCTCN2019089077-appb-000020
最小纵坐标
Figure PCTCN2019089077-appb-000021
并定义该角点集合的最小包围矩形的中心
Figure PCTCN2019089077-appb-000022
长
Figure PCTCN2019089077-appb-000023
及宽
Figure PCTCN2019089077-appb-000024
即
Figure PCTCN2019089077-appb-000025
Figure PCTCN2019089077-appb-000026
进而计算车牌字符预选角点集合Bi的最小包围矩形的长宽比
Figure PCTCN2019089077-appb-000027
及面积
Figure PCTCN2019089077-appb-000028
即
Figure PCTCN2019089077-appb-000029
若满足
Figure PCTCN2019089077-appb-000030
且
Figure PCTCN2019089077-appb-000031
Figure PCTCN2019089077-appb-000032
则进入子步骤(4.4)，否则进入子步骤(4.5)。该子步骤的判别条件中，
Figure PCTCN2019089077-appb-000033
表示车牌字符预选角点集合B i的最小包围矩形的长宽比；β min表示车牌字符预选角点集合B i的最小包围矩形的长宽比低阈值；β max表示车牌字符预选角点集合B i的最小包围矩形的长宽比高阈值；
(4.4)若车牌字符预选角点集合B i满足
Figure PCTCN2019089077-appb-000034
且
Figure PCTCN2019089077-appb-000035
则判定该角点集合B i为车牌字符有效角点集合C，并确定车牌中心(x 0,y 0)、长l 0及宽w 0，即
Figure PCTCN2019089077-appb-000036
结束车牌定位过程，进入(5)实施拖牵诱导；否则进入子步骤(4.5)；该子步骤的判别条件中，
Figure PCTCN2019089077-appb-000037
表示车牌字符预选角点集合B i的最小包围矩形的面积；γ min表示车牌字符预选角点集合B i的最小包围矩形的面积低阈值；γ max表示车牌字符预选角点集合B i的最小包围矩形的面积 高阈值；
(4.5)若i＜I，则将i的值增1，重新返回上述子步骤(4.2)；否则结束车牌定位过程，返回步骤(1)图像采集及预处理。
其中，在步骤(5)中，实施拖牵诱导方法如下：根据步骤(4)中车牌定位中已确定的车牌中心(x 0,y 0)在图像中的位置，实时给出方向提示，诱导驾驶员进行倒车作业：若图像中被拖车车牌中心在图像中心的左侧，则提示驾驶员向右倒；若图像中被拖车车牌中心在图像中心的右侧，则提示驾驶员向左倒，从而完成两侧托臂与被拖车两前轮的对准，进而对被拖车进行抱胎固定，将其牵拉拖离。
有益效果：与现有技术相比，本发明的技术方案具有以下有益技术效果：
1、针对道路救援装备的结构及正方位拖牵作业工作特点而提出，能有效加快正方位拖牵车轮对准过程，对拖牵作业形成诱导辅助，提高道路救援装备救援效率。
2、诱导可靠性高，抗干扰能力强，具有良好的环境适应能力。本发明所提出的诱导方法充分考虑并利用了不同车辆在车牌上共同具有的字符角点特征，能够适应各种不同的车型。
3、处理速度快，实时性好，本发明所提出的诱导方法采用灰度图像进行角点检测，并在角点聚类前先进行角点优选，有效提高了该诱导方法的运行速度，保证了实时性。
附图说明
图1是本发明提出的道路救援装备正方位拖牵诱导方法的流程图；
图2是道路救援装备正方位拖牵作业的立体示意图；
图3是道路救援装备正方位拖牵作业的俯视图；
图4是道路救援装备拖牵装置结构及摄像头安装位置示意图；
图5是采集的某帧图像；
图6是采集图像预处理结果图；
图7是采集图像角点检测结果图；
图8是采集图像角点优选结果图；
图9是采集图像角点聚类结果图；
图10是采集图像车牌字符有效角点筛选结果图；
图11是车牌定位结果图。
具体实施方式
下面结合附图和实施例对本发明的技术方案作进一步的说明。
随着社会经济的不断增长，我国道路交通基础设施建设及汽车工业得到飞速发展，机动车保有量在大幅增加，汽车在行驶途中由于驾驶不当或车辆自身故障等导致的车辆无法正常行驶等特殊情况与日俱增，作为汽车后市场的道路清障救援行业逐渐被社会和民众重视，道路清障救援行业对保障道路运输安全高效的重要性日益凸显。但是由于交通事故的发生具有无法彻底预防的客观必然性，在很多交通事故发生后，常有因清障救援不及时或救援效率太低而导致的城市交通拥堵或二次连锁事故，比如道路救援装备无法快速准确从正方位将事故车拖离现场而导致交通堵塞就是其中之一。究其原因，一方面正方位救援环境比较复杂，大多面对狭窄及拥堵环境，作业空间严重受限，如说明书附图2和说明书附图3所示，被拖车夹在左右两车中间，呈“三”字形排列，此时道路救援装备只能从被拖车的前方实施拖牵救援；另一方面道路救援装备智能化水平低，没有利用现有科学技术手段，对拖牵作业进行诱导辅助，也是导致救援效率低的一个重要因素，但目前对于道路救援装备正方位拖牵诱导相关技术尚未见有研究。
以皮卡救援车为例的道路救援装备的拖牵装置主要由折臂、伸缩臂、摆臂及两侧托臂组成，具体形式如附图4所示。在正方位拖牵作业实施的过程中，道路救援装备位于被拖车车头的前方，首先道路救援装备通过倒车作业，将其两侧托臂分别与被拖车两前轮进行对准，然后对被拖车两前轮进行抱胎固定，最后通过折臂牵拉起吊，将被拖车拖离。在传统的正方位拖牵作业过程中，将被拖车前轮与道路救援装备托臂进行对准这一操作，其效率主要依赖于驾驶员技术经验等人为因素，通常需要多次反复倒车进行位置修正，作业耗时过长，救援效率低下。
针对道路救援装备正方位拖牵作业的特点，本发明提出一种既具有良好实时性，又具有较强环境自适应能力和抗干扰能力的道路救援装备正方位拖牵诱导方法，总流程参见说明书附图1，该方法主要包括下面几个步骤：
(1)图像采集及预处理
在道路救援装备尾部折臂上离地面40～60厘米处的中间位置上安装车载摄像头，摄像头水平朝向救援装备的后方，其采集范围为包含救援装备摆臂、托臂 及被拖车车头等关键部位的后方作业区域，具体形式如说明书附图4所示。对采集的后方作业区域图像，参见说明书附图5，进行预处理，参见说明书附图6：先将彩色采集图像转成灰度图像，减小图像处理数据量，提高该方法的实时性，然后对灰度图像进行3×3邻域高斯平滑滤波，消除一些不必要的孤立点和凸起。
(2)角点检测及优选
应当指出，目前国内大部分社会车辆为蓝牌小型汽车，本发明所提出的诱导方法主要面向蓝牌小型汽车的拖牵救援，而车辆牌照是大部分社会车辆的共同特征，具有普遍性，且车牌字符属于印刷字符，字符笔画中蕴含比较丰富的角点信息，因此可通过角点检测提取车牌区域角点特征。
典型的角点检测方法包括Moravec角点检测、Susan角点检测、Fast角点检测、Harris角点检测等算法，其中Moravec角点检测算法简单，但计算量大，误检率很高，Susan角点检测算法对噪声不敏感，但定位精度较差，Harris角点检测算法简单，计算量不大，角点提取较为均匀，适应性强，具有较好的稳定性。本发明采用哈里斯(Harris)角点检测算法对平滑后的灰度图像进行检测，获得图像中的角点，参见说明书附图7，并将得到的所有角点存储到角点集合O中。
Harris角点检测的位置相对准确，角点提取也相对较为均匀，但所得到的角点位置大多还是以邻域的形式成批出现的，参见说明附图7，为了进一步减小后续角点聚类过程中的计算量，提高该诱导方法的实时性，可根据角点强度对上述Harris角点检测获得的角点进行优选，获得更具代表性的角点位置。正方位拖牵作业的有效作业距离一般为1～5米，摄像头的焦距选择为4～8毫米，本发明中摄像头采集的图像尺寸为960×540，可设置角点间最小允许距离为5像素，对Harris角点检测获得的角点进行优选，得到优质角点集合A，参见说明书附图8，具体子步骤为：
(2.1)按照角点横、纵坐标依次递增的顺序，遍历角点集合O中的所有角点，并在以每个角点为中心，半径为5像素的圆形区域内，将该区域角点强度最大的角点保存到优质角点集合A中；
(2.2)上述遍历结束后，对优质角点集合A中重复保存的具有相同坐标及角点强度的角点，只保留其中一个，并将多余的相同角点从优质角点集合A中删除，最后可得到N个不同的优质角点。
(3)角点聚类
车牌区域字符角点分布密集且有规则，可根据角点之间的距离对优选得到的角点集合A进行凝聚层次聚类，参见说明书附图9，得到I个车牌字符预选角点集合B i，i＝1,2,3…,I，其中，i表示预选角点集合的序号，具体子步骤为：
(3.1)把优质角点集合A中的每个角点a nn＝1,2,3…,N，各视为一个类，每个类中有且只有一个角点，计算当前所有类两两之间的类间距D rg，D rg表示第r类和第g类之间的类间距，本发明中定义类间距D rg为第r类中的每个角点与第g类中的每个角点之间欧式距离的平均值，由于当前每个类中有且只有一个角点，即第r类中只有角点a r，第g类中只有角点a g，故此时类间距D rg计算公式为：
D rg＝d rg
式中，d rg表示角点a r和a g之间的欧式距离，即
Figure PCTCN2019089077-appb-000038
Figure PCTCN2019089077-appb-000039
其中，
Figure PCTCN2019089077-appb-000040
和
Figure PCTCN2019089077-appb-000041
分别表示角点a r的横、纵坐标，
Figure PCTCN2019089077-appb-000042
和
Figure PCTCN2019089077-appb-000043
分别表示角点a g的横、纵坐标。
根据类间距D rg的物理意义，则有D rg＝D gr，且当r＝g时，D rg＝0，将上述计算得到的所有类间距用矩阵形式表示，可得到初始的N×N类间距矩阵
Figure PCTCN2019089077-appb-000044
Figure PCTCN2019089077-appb-000045
矩阵H中的r行g列元素即为D rg；
(3.2)遍历当前的欧式距离矩阵H，找到矩阵H中的非对角线最小元素，即当前类间距最小值，设为D st，且s≠t，表示第s类和第t类为当前距离最近的两个类。若D st＜D th，则将第t类中的角点归并到第s类中，进而将第s类和第t类合并为一个新类，并将合并后的角点总类数记为V，进入子步骤(3.3)；否则结束聚类计算，可得到I个车牌字符预选角点集合B i，i＝1,2,3…,I，其中，i表示预选角点集合的序号；I表示车牌字符预选角点集合的总数，且I＝V。该子步骤的判别条件中，D st为当前矩阵H中的非对角线最小元素；D th为类间距最小值阈值，该阈值在10～15内取值；
(3.3)重新计算当前剩余的类与合并得到的新类两两之间的类间距D rg，此时类间距D rg的计算公式为：
Figure PCTCN2019089077-appb-000046
式中，M r表示第r类中的角点个数，该类中的角点表示为p j，j＝1,2,3…,M r；M g表示第g类中的角点个数，该类中的角点表示为q k，k＝1,2,3…,M g；d jk表示角点p j和q k之间的欧式距离，即
Figure PCTCN2019089077-appb-000047
其中，
Figure PCTCN2019089077-appb-000048
和
Figure PCTCN2019089077-appb-000049
分别表示角点p j的横、纵坐标，
Figure PCTCN2019089077-appb-000050
和
Figure PCTCN2019089077-appb-000051
分别表示角点q k的横、纵坐标。
用上述重新计算得到的类间距，根据类间距D rg及类间距矩阵H的物理意义，对类间距矩阵H进行更新，可得到新的V×V类间距矩阵
Figure PCTCN2019089077-appb-000052
并重新返回上述子步骤(3.2)。
(4)车牌定位
对于步骤(3)中聚类得到的车牌字符预选角点集合B i，i＝1,2,3…,I，根据车牌尺寸的先验性知识，从车牌字符预选角点集合B i中筛选出车牌字符有效角点集合C，参见说明书附图10，进而确定车牌位置，参见说明书附图11，具体子步骤为：
(4.1)初始化i＝1；
(4.2)若车牌字符预选角点集合B i满足
Figure PCTCN2019089077-appb-000053
则进入子步骤(4.3)，否则进入子步骤(4.5)。该子步骤的判别条件中，
Figure PCTCN2019089077-appb-000054
为车牌字符角点集合B i中的角点个数；num th为车牌字符角点集合的角点个数阈值，该阈值在6～8内取值；
(4.3)遍历车牌字符预选角点集合B i，确定该集合中角点的最大横坐标
Figure PCTCN2019089077-appb-000055
最小横坐标
Figure PCTCN2019089077-appb-000056
最大纵坐标
Figure PCTCN2019089077-appb-000057
最小纵坐标
Figure PCTCN2019089077-appb-000058
并定义该角点集合的最小包围矩形的中心
Figure PCTCN2019089077-appb-000059
长
Figure PCTCN2019089077-appb-000060
及宽
Figure PCTCN2019089077-appb-000061
即
Figure PCTCN2019089077-appb-000062
Figure PCTCN2019089077-appb-000063
进而计算车牌字符预选角点集合B i的最小包围矩形的长宽比
Figure PCTCN2019089077-appb-000064
及面积
Figure PCTCN2019089077-appb-000065
即
Figure PCTCN2019089077-appb-000066
若满足
Figure PCTCN2019089077-appb-000067
且
Figure PCTCN2019089077-appb-000068
Figure PCTCN2019089077-appb-000069
则进入子步骤(4.4)，否则进入子步骤(4.5)。该子步骤的判别条件中，
Figure PCTCN2019089077-appb-000070
表示车牌字符预选角点集合B i的最小包围矩形的长宽比；β min表示车牌字符预选角点集合B i的最小包围矩形的长宽比低阈值，该阈值在3～3.5内取值；β max表示车牌字符预选角点集合B i的最小包围矩形的长宽比高阈值，该阈值在4.5～5内取值。
(4.4)若车牌字符预选角点集合B i满足
Figure PCTCN2019089077-appb-000071
且
Figure PCTCN2019089077-appb-000072
则判定该角点集合B i为车牌字符有效角点集合C，并确定车牌中心(x 0，y 0)、长l 0及宽w 0，即
Figure PCTCN2019089077-appb-000073
结束车牌定位过程，进入(5)实施拖牵诱导；否则进入子步骤(4.5)。该子步骤的判别条件中，
Figure PCTCN2019089077-appb-000074
表示车牌字符预选角点集合B i的最小包围矩形的面积；γ min表示车牌字符预选角点集合B i的最小包围矩形的面积低阈值，该阈值在200～300内取值；γ max表示车牌字符预选角点集合B i的最小包围矩形的面积高阈值，该阈值在5000～5500内取值；
(4.5)若i＜I，则将i的值增1，重新返回上述子步骤(4.2)；否则结束车牌定位过程，返回步骤(1)图像采集及预处理。
上述车牌字符有效角点集合筛选算法的两点补充说明：①国内的蓝色牌照具有国家标准，其印刷字符均为七位，其中每个字符至少含有一个角点，七位印刷字符的长宽比约为4；②摄像头安装在道路救援装备折臂的中间固定位置处，且水平朝向道路救援装备后方，在进行正方位拖牵作业的过程中，被拖车位于道路救援装备的后方区域，作业距离一般为1～5米，摄像头的焦距可选择为4～8毫米，本发明中摄像头采集的图像尺寸为960×640，在有效作业距离1～5米范围内，采集图像中车牌字符区域的像素大小在300～5000范围内变化。
(5)实施拖牵诱导
如说明书附图4所示，在正方位拖牵作业的实施过程中，道路救援装备一般位于被拖车车头的前方，二者近似呈“一”字形排列，车载摄像头安装于道路救援装备折臂上中间固定位置处，且水平朝向道路救援装备后方，摄像头采集区域的 中轴线与道路救援装备及其拖牵装置的中轴线相一致，而被拖车前车牌一般位于其车头中间位置，因而可根据摄像头输出画面的中心与画面中被拖车车牌中心的左右位置关系判断道路救援装备与被拖车的相对位置关系。
在实施正方位拖牵作业的过程中，驾驶员面向道路救援装备前方，故可根据步骤(4)中车牌定位中已确定的车牌中心(x 0,y 0)在图像中的位置，参见说明书附图11，实时给出方向提示，诱导驾驶员进行倒车作业：若图像中被拖车车牌中心在图像中心的左侧，则提示驾驶员向右倒；若图像中被拖车车牌中心在图像中心的右侧，则提示驾驶员向左倒，从而完成两侧托臂与被拖车两前轮的对准，进而对被拖车进行抱胎固定，将其牵拉拖离。
经过上述步骤，本发明提出的方法能够实现救援装备正方位拖牵作业过程中的实时诱导，有效提高了道路救援装备的救援效率。

Claims (6)

  1. 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，其特征在于，该方法包括如下步骤：
    (1)采集拖车后方作业区域图像并对图像进行预处理；
    (2)对预处理后的图像进行角点检测及优选；
    (3)对优选的角点进行聚类；
    (4)根据优选的角点对车牌定位并得到车牌中心；
    (5)根据车牌中心实施拖牵诱导。
  2. 根据权利要求1所述的一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，其特征在于，在步骤(1)中，图像采集及预处理方法如下：在道路救援装备尾部折臂上安装车载摄像头，对拖车后方作业区域图像采集，并先将彩色采集图像转成灰度图像，然后对灰度图像进行平滑滤波处理。
  3. 根据权利要求2所述的一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，其特征在于，在步骤(2)中，采用哈里斯角点检测算法对平滑后的灰度图像进行检测，获得图像中的角点，并将得到的所有角点存储到角点集合O中；对哈里斯角点检测算法获得的角点进行优选，得到优质角点集合A，方法如下：
    (2.1)按照角点横、纵坐标依次递增的顺序，遍历角点集合O中的所有角点，并在以每个角点为中心，半径为5像素的圆形区域内，将该区域角点强度最大的角点保存到优质角点集合A中；
    (2.2)上述遍历结束后，对优质角点集合A中重复保存的具有相同坐标及角点强度的角点，只保留其中一个，并将多余的相同角点从优质角点集合A中删除，最后可得到N个不同的优质角点。
  4. 根据权利要求1所述的一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，其特征在于，在步骤(3)中，角点聚类方法如下：根据角点之间的距离对优选得到的角点集合A进行凝聚层次聚类，得到I个车牌字符预选角点集合B i，i＝1,2,3…,I，其中，i表示预选角点集合的序号，方法如下：
    (3.1)把优质角点集合A中的每个角点a nn＝1,2,3…,N，各视为一个类，每个类中有且只有一个角点，计算当前所有类两两之间的类间距D rg，D rg表示第r类和第g类之间的类间距，定义类间距D rg为第r类中的每个角点与第g类中的每个角点之间欧式距离的平均值，由于当前每个类中有且只有一个角点，即第r类中只有角点a r，第g类中 只有角点a g，故此时类间距D rg计算公式为：
    D rg＝d rg
    式中，d rg表示角点a r和a g之间的欧式距离，即
    Figure PCTCN2019089077-appb-100001
    其中，
    Figure PCTCN2019089077-appb-100002
    和
    Figure PCTCN2019089077-appb-100003
    分别表示角点a r的横、纵坐标，
    Figure PCTCN2019089077-appb-100004
    和
    Figure PCTCN2019089077-appb-100005
    分别表示角点a g的横、纵坐标；
    根据类间距D rg的物理意义，则有D rg＝D gr，且当r＝g时，D rg＝0，将上述计算得到的所有类间距用矩阵形式表示，可得到初始的N×N类间距矩阵
    Figure PCTCN2019089077-appb-100006
    Figure PCTCN2019089077-appb-100007
    矩阵H中的r行g列元素即为D rg；
    (3.2)遍历当前的欧式距离矩阵H，找到矩阵H中的非对角线最小元素，即当前类间距最小值，设为D st，且s≠t，表示第s类和第t类为当前距离最近的两个类，若D st＜D th，则将第t类中的角点归并到第s类中，进而将第s类和第t类合并为一个新类，并将合并后的角点总类数记为V，进入子步骤(3.3)；否则结束聚类计算，可得到I个车牌字符预选角点集合B i，i＝1,2,3…,I，其中，i表示预选角点集合的序号；I表示车牌字符预选角点集合的总数，且I＝V，该子步骤的判别条件中，D st为当前矩阵H中的非对角线最小元素；D th为类间距最小值阈值；
    (3.3)重新计算当前剩余的类与合并得到的新类两两之间的类间距D rg，此时类间距D rg的计算公式为：
    Figure PCTCN2019089077-appb-100008
    式中，M r表示第r类中的角点个数，该类中的角点表示为p j，j＝1,2,3…,M r；M g表示第g类中的角点个数，该类中的角点表示为q k，k＝1,2,3…,M g；d jk表示角点p j和q k之间的欧式距离，即
    Figure PCTCN2019089077-appb-100009
    其中，
    Figure PCTCN2019089077-appb-100010
    和
    Figure PCTCN2019089077-appb-100011
    分别表示角点p j的横、纵坐标，
    Figure PCTCN2019089077-appb-100012
    和
    Figure PCTCN2019089077-appb-100013
    分别表示角点q k的横、纵坐标；
    用上述重新计算得到的类间距，根据类间距D rg及类间距矩阵H的物理意义，对类间距矩阵H进行更新，可得到新的V×V类间距矩阵
    Figure PCTCN2019089077-appb-100014
    并重新返回 上述子步骤(3.2)。
  5. 根据权利要求4所述的一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，其特征在于，在步骤(4)中，车牌定位方法如下：对于步骤(3)中聚类得到的车牌字符预选角点集合B i，i＝1,2,3…,I，从车牌字符预选角点集合B i中筛选出车牌字符有效角点集合C，进而确定车牌位置，方法如下：
    (4.1)初始化i＝1；
    (4.2)若车牌字符预选角点集合B i满足
    Figure PCTCN2019089077-appb-100015
    则进入子步骤(4.3)，否则进入子步骤(4.5)，该子步骤的判别条件中，
    Figure PCTCN2019089077-appb-100016
    为车牌字符角点集合B i中的角点个数；num th为车牌字符角点集合的角点个数阈值；
    (4.3)遍历车牌字符预选角点集合B i，确定该集合中角点的最大横坐标
    Figure PCTCN2019089077-appb-100017
    最小横坐标
    Figure PCTCN2019089077-appb-100018
    最大纵坐标
    Figure PCTCN2019089077-appb-100019
    最小纵坐标
    Figure PCTCN2019089077-appb-100020
    并定义该角点集合的最小包围矩形的中心
    Figure PCTCN2019089077-appb-100021
    长
    Figure PCTCN2019089077-appb-100022
    及宽
    Figure PCTCN2019089077-appb-100023
    即
    Figure PCTCN2019089077-appb-100024
    Figure PCTCN2019089077-appb-100025
    进而计算车牌字符预选角点集合B i的最小包围矩形的长宽比
    Figure PCTCN2019089077-appb-100026
    及面积
    Figure PCTCN2019089077-appb-100027
    即
    Figure PCTCN2019089077-appb-100028
    若满足
    Figure PCTCN2019089077-appb-100029
    且
    Figure PCTCN2019089077-appb-100030
    则进入子步骤(4.4)，否则进入子步骤(4.5)，该子步骤的判别条件中，
    Figure PCTCN2019089077-appb-100031
    表示车牌字符预选角点集合B i的最小包围矩形的长宽比；β min表示车牌字符预选角点集合B i的最小包围矩形的长宽比低阈值；β max表示车牌字符预选角点集合B i的最小包围矩形的长宽比高阈值；
    (4.4)若车牌字符预选角点集合B i满足
    Figure PCTCN2019089077-appb-100032
    且
    Figure PCTCN2019089077-appb-100033
    则判定该角点集合B i为车牌字符有效角点集合C，并确定车牌中心(x 0,y 0)、长l 0及宽w 0，即
    Figure PCTCN2019089077-appb-100034
    Figure PCTCN2019089077-appb-100035
    结束车牌定位过程，进入(5)实施拖牵诱导；否则进入子步骤(4.5)；该子步骤的判别条件中，
    Figure PCTCN2019089077-appb-100036
    表示车牌字符预选角点集合B i的最小包围矩形的面积；γ min表示车牌字符预选角点集合B i的最小包围矩形的面积低阈值；γ max表示车牌字符预选角点集合B i的最小包围矩形的面积高阈值；
    (4.5)若i＜I，则将i的值增1，重新返回上述子步骤(4.2)；否则结束车牌定位过程，返回步骤(1)图像采集及预处理。
  6. 根据权利要求5所述的一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法，其特征在于，在步骤(5)中，实施拖牵诱导方法如下：根据步骤(4)中车牌定位中已确定的车牌中心(x 0,y 0)在图像中的位置，实时给出方向提示，诱导驾驶员进行倒车作业：若图像中被拖车车牌中心在图像中心的左侧，则提示驾驶员向右倒；若图像中被拖车车牌中心在图像中心的右侧，则提示驾驶员向左倒，从而完成两侧托臂与被拖车两前轮的对准，进而对被拖车进行抱胎固定，将其牵拉拖离。
PCT/CN2019/089077 2018-08-13 2019-05-29 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法 Ceased WO2020034722A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/057,915 US11367215B2 (en) 2018-08-13 2019-05-29 Positive azimuth towing guidance method for road rescue equipment based on license plate corner features

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810920734.X 2018-08-13
CN201810920734.XA CN109376734B (zh) 2018-08-13 2018-08-13 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法

Publications (1)

Publication Number Publication Date
WO2020034722A1 true WO2020034722A1 (zh) 2020-02-20

Family

ID=65404377

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/089077 Ceased WO2020034722A1 (zh) 2018-08-13 2019-05-29 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法

Country Status (3)

Country Link
US (1) US11367215B2 (zh)
CN (1) CN109376734B (zh)
WO (1) WO2020034722A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120235860A (zh) * 2025-05-29 2025-07-01 西安嘉和华亨热系统有限公司 一种散热器振动试验中的表面裂纹智能识别方法

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109376734B (zh) * 2018-08-13 2020-07-31 东南大学 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法
US12183028B2 (en) * 2019-12-04 2024-12-31 Beijing Smarter Eye Technology Co. Ltd. Object-based short range measurement method, device and system, and storage medium
CN113077513B (zh) * 2021-06-03 2021-10-29 深圳市优必选科技股份有限公司 视觉定位方法、装置和计算机设备
CN119942520A (zh) * 2024-12-15 2025-05-06 北京工业大学 一种基于车牌信息的车辆编队3d目标检测方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103632550A (zh) * 2013-11-05 2014-03-12 广东工业大学 一种基于固定摄像头的道路车辆识别方法和装置
US20160200359A1 (en) * 2013-08-29 2016-07-14 Robert Bosch Gmbh Method for steering a vehicle
CN106027997A (zh) * 2016-07-13 2016-10-12 孙福盛 一种无线视频辅助集卡定位装置及方法
CN107861510A (zh) * 2017-11-01 2018-03-30 龚土婷 一种智能车辆驾驶系统
CN109376734A (zh) * 2018-08-13 2019-02-22 东南大学 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9238483B2 (en) * 2011-04-19 2016-01-19 Ford Global Technologies, Llc Trailer backup assist system with trajectory planner for multiple waypoints
WO2014152470A2 (en) * 2013-03-15 2014-09-25 Tk Holdings, Inc. Path sensing using structured lighting
US9464886B2 (en) * 2013-11-21 2016-10-11 Ford Global Technologies, Llc Luminescent hitch angle detection component
US9696723B2 (en) * 2015-06-23 2017-07-04 GM Global Technology Operations LLC Smart trailer hitch control using HMI assisted visual servoing
US20160375831A1 (en) * 2015-06-23 2016-12-29 GM Global Technology Operations LLC Hitching assist with pan/zoom and virtual top-view
CN112805207B (zh) * 2018-05-08 2023-03-21 大陆汽车系统公司 用于自动车辆倒车的用户可调轨迹
US11603100B2 (en) * 2018-08-03 2023-03-14 Continental Autonomous Mobility US, LLC Automated reversing by following user-selected trajectories and estimating vehicle motion

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160200359A1 (en) * 2013-08-29 2016-07-14 Robert Bosch Gmbh Method for steering a vehicle
CN103632550A (zh) * 2013-11-05 2014-03-12 广东工业大学 一种基于固定摄像头的道路车辆识别方法和装置
CN106027997A (zh) * 2016-07-13 2016-10-12 孙福盛 一种无线视频辅助集卡定位装置及方法
CN107861510A (zh) * 2017-11-01 2018-03-30 龚土婷 一种智能车辆驾驶系统
CN109376734A (zh) * 2018-08-13 2019-02-22 东南大学 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120235860A (zh) * 2025-05-29 2025-07-01 西安嘉和华亨热系统有限公司 一种散热器振动试验中的表面裂纹智能识别方法

Also Published As

Publication number Publication date
CN109376734A (zh) 2019-02-22
US11367215B2 (en) 2022-06-21
CN109376734B (zh) 2020-07-31
US20210312653A1 (en) 2021-10-07

Similar Documents

Publication Publication Date Title
WO2020034722A1 (zh) 一种基于车牌角点特征的道路救援装备正方位拖牵诱导方法
CN109435852B (zh) 一种针对大型卡车的全景式辅助驾驶系统及方法
CN105825696B (zh) 基于信号灯信息提示的驾驶辅助系统
CN110443225A (zh) 一种基于特征像素统计的虚实车道线识别方法及其装置
CN110069986A (zh) 一种基于混合模型的交通信号灯识别方法及系统
CN110210363A (zh) 一种基于车载图像的目标车辆压线检测方法
CN109299674B (zh) 一种基于车灯的隧道违章变道检测方法
CN110733416B (zh) 一种基于逆透视变换的车道偏离预警方法
CN201825037U (zh) 高速公路车辆的车道偏离报警装置
CN103902985B (zh) 一种基于roi的强鲁棒性实时车道侦测算法
CN100545858C (zh) 基于小波变换的复杂背景中的车牌提取方法
CN106128115A (zh) 一种基于双摄像机检测道路交通信息的融合方法
CN108470142B (zh) 基于逆透视投影和车道距离约束的车道定位方法
CN106887004A (zh) 一种基于块匹配的车道线检测方法
CN105654073A (zh) 一种基于视觉检测的速度自动控制方法
CN109285169B (zh) 一种基于车轮识别的道路救援装备侧方位拖牵诱导方法
CN105966314B (zh) 基于双低成本摄像头的车道偏离预警方法
CN109949593A (zh) 一种基于路口先验知识的交通信号灯识别方法及系统
CN113635905B (zh) 一种面向驾驶行为保险的风险实时提示系统及装置
CN103481842B (zh) 一种车辆检测与跟踪模式的切换方法
CN110569739B (zh) 一种基于深度学习和胎噪声谱图的汽车辅助系统及方法
CN108563976B (zh) 一种基于车窗位置的多方向车辆颜色识别方法
CN107220632B (zh) 一种基于法向特征的路面图像分割方法
CN115303263A (zh) 一种基于神经网络的自动泊车方法及系统
CN110349415B (zh) 一种基于多尺度变换的行车速度测量方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19850320

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19850320

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19850320

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 06/09/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19850320

Country of ref document: EP

Kind code of ref document: A1