CN102354457B - General Hough transformation-based method for detecting position of traffic signal lamp - Google Patents

General Hough transformation-based method for detecting position of traffic signal lamp Download PDF

Info

Publication number
CN102354457B
CN102354457B CN 201110325073 CN201110325073A CN102354457B CN 102354457 B CN102354457 B CN 102354457B CN 201110325073 CN201110325073 CN 201110325073 CN 201110325073 A CN201110325073 A CN 201110325073A CN 102354457 B CN102354457 B CN 102354457B
Authority
CN
China
Prior art keywords
image
point
traffic lights
space
gradient
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.)
Active
Application number
CN 201110325073
Other languages
Chinese (zh)
Other versions
CN102354457A (en
Inventor
鲁帅
冯瑞
金城
薛向阳
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.)
Fudan University
Original Assignee
Fudan 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 Fudan University filed Critical Fudan University
Priority to CN 201110325073 priority Critical patent/CN102354457B/en
Publication of CN102354457A publication Critical patent/CN102354457A/en
Application granted granted Critical
Publication of CN102354457B publication Critical patent/CN102354457B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention belongs to the technical field of digital image processing and pattern recognition and discloses a general Hough transformation-based method for detecting the position of a traffic signal lamp, which comprises the following steps: by taking an image sequence as input, calculating image gradient information by adopting a first-order differential operator on an image by using image gradation; establishing a traffic lamp shape describing function; defining mapping from an image point to the space of an accumulator; and searching the extreme value of the space of the accumulator to acquire the coordinates of the traffic lamp to provide positional information for traffic lamp state identification. The general Hough transformation-based method for detecting the position of the traffic signal lamp can cope with the significant change of illumination, is insensitive to the influence caused by image sampling color cast and can cope with the traffic lamp with various common shapes.

Description

Traffic lights method for detecting position based on generalised Hough transform
Technical field
The invention belongs to Digital Image Processing and mode identification technology, be specifically related to traffic lights method for detecting position in the video image.
Technical background
Development along with economy, urban population and motor vehicle number rapid growth, the incidence of traffic hazard and annual because toll on traffic is very surprising, one of major reason wherein are exactly automobile driver, and beyond the invisible the situation of irregular driving is very general the traffic police.In order to regulate the traffic more widely, the traffic police is provided non-at-scene Law enforcement capability, electronic police system arises at the historic moment as intelligent transportation system (Intelligent Transportation System) important component.It has been realized round-the-clock, the large-scale management expectancy of urban road, simultaneously, for traffic police provides clearly video image evidence, management level have been improved, reduce the casualties and the property loss that cause because of traffic hazard, indirectly created huge economic worth.
Early stage electronic police system mainly is the violation information [1] of obtaining vehicle by bury inductive coil underground below road foundation, but inductive coil is difficult for installing and is subject to the maintenance on road surface and destroys.The electronic police system that new non-inductive coil detects so people begin one's study.Along with the development of technology, grow up gradually based on the method that video detects, can meet the tendency in the daily management of road monitoring.
The traffic lights signal is identified requisite constituent in the electronic police system that is based on video.Electronic police system all needs the information of traffic lights signal in the vehicles peccancy testing processes such as red light running, red light violation left/right rotation or green light follow up in violation of rules and regulations, the accuracy of simultaneously traffic lights signal differentiation also directly seriously affects the subsequent treatment of electronic police system.The main fixedly monocular cam that adopts in the at present Intelligent road monitoring, the position of traffic lights in the visual field is relatively stable.But because the impact of physical environment outside the venue, as strong wind make that camera rocks, the dealing oversize vehicle causes frame to stand in the camera shake etc. on road surface, so that the traffic lights position probing becomes very necessary work in the image.
People have proposed many methods for the traffic lights test problems in recent years, can roughly be divided into two classes: a kind of image color information and another kind of utilizing only utilizes image pixel intensity to process.As far back as Chung et al.[2 in 2002] to the coloured image that fixing camera is taken, get access to the coordinate of traffic lights in image in conjunction with background modeling, Fuzzy Processing and morphological operation.They are transformed into the HSV space to the RGB image first, utilize background model to estimate the ambient lighting parameter, respectively the image of two passages of color harmony intensity carried out obfuscation, carry out morphological operations by a similar round template, extract all possible candidate region in the image, then add the final coordinate of time sequence information screening of traffic lights.In addition, Shioyama et al.[3] also be that coloured image is set up traffic lights in Farnsowrth color space template, calculate three kinds of different colours lamps two-dimensional histogram separately, obtain the probability that single pixel in the searching image belongs to lamp by Histogram backprojection, then clustering cleaning reaches the purpose that traffic lights detects.Afterwards, Shen et al.[4] changed the colour model of the dimensional Gaussian distribution indication lamp in HSI space into.Yet these methods all have same defective, all can't thoroughly solve the impact that ambient lighting changes.Although [3] utilize Gaussian distribution to attempt each color change of match traffic lights, the change color tolerance is promoted to some extent, still lamp line secular variation is not outside the venue had a robustness.Traffic lights because light the colour cast phenomenon occurs easily in the decay of communication process, the impact that becomes pixel that strength information is lost, has limited the usability of algorithm by the imageing sensor imaging greatly simultaneously.
Abandon the traffic lights color information, only utilize pixel intensity, Omachi et al.[5] for the architectural feature of traffic lights circle, first with gray level image according to pixels intensity carry out cluster, take the image gradient edge as feature, carry out the Hough transformation of standard under the circular equation model of the traffic lights of setting up, the extreme value of search projecting space is determined the coordinate of traffic lights.This method has also detected other circular object, and the result has been caused very large interference.So the author has improved oneself method 2010 [6], utilizes the planform of three traffic lights to set up curve side's group, adopts equally the standard Hough transformation to find the solution the position of traffic lights.These methods have all solved the color factors of instability very first-class, yet because adopted the standard two-dimentional geometric figure to carry out the traffic lights model description, the traffic lights imaging results differs greatly in this and the real scene, has finally lacked practicality.
Traffic lights detects impacts such as being subject to illumination variation, color deflection and shape difference, and present certain methods all can not reach gratifying accuracy rate in actual applications.For the problems referred to above, the present invention is directed to the broad applicability of color instability and traffic lights model, the traffic lights detection algorithm based on generalised Hough transform has been proposed.
Summary of the invention
The object of the present invention is to provide a kind of traffic lights detection algorithm based on generalised Hough transform, obtain the traffic lights coordinate in the monitoring image, for the signal lamp state recognition provides positional information, and can prepare for the deeper application in back.
Technical scheme of the present invention is: take image sequence as input, use gradation of image, image is adopted first order differential operator computed image gradient information.By the method that document [7] is mentioned, set up the traffic lights function of describing the shape, define a kind of mapping from the picture point to the Accumulator space, search Accumulator space extreme value is obtained the coordinate of signal lamp, for further traffic light status identification provides positional information.
Concrete steps of the present invention are successively:
Step 1. is gray space with the image of input from color space conversion.
Generally input picture is the coloured image with R, G, three passages of B, and each passage is 8 bit-depths in R, G, B color model, and the discrete range of intensity is 0 to 255.So the formula of conversion can be expressed as:
(10)。
Wherein , , The channel strength value of each pixel of respectively expression correspondence, Intensity level corresponding to gray space after the conversion.
Step 2. is utilized first order differential operator computed image gradient magnitude and direction.
Image array is a discrete two-dimensional function, and functional gradient calculates and can obtain by consecutive point being carried out difference processing.Consecutive point on the horizontal direction are carried out the brightness of difference processing on can the detection of vertical direction change, the brightness on the horizontal direction is changed the consecutive point difference that can adopt on the vertical direction obtain.So image can be expressed as in the gradient magnitude of certain point:
(11)。
Gradient direction can be expressed as:
(12)。
Wherein , Respectively to image , The differential of direction.Because integral image brightness grow or die down and just each pixel is added or deduct a constant, and this constant will not affect the calculating of image gradient.If the change of picture contrast, that also is that each pixel be multiply by a constant, and this constant equally also can be divided out and do not affected the calculating of image gradient direction when the compute gradient direction.
Step 3. is set up the mapping function of traffic lights shape.
If Be in the arbitrary shaped region reference point (such as the barycenter in zone, Fig. 2), It is the traffic lights shape border Upper any point, note With Between difference vector be:
(13)。
Arrive The axle clamp angle is , the point To the border Distance be The definition frontier point The border orientation angles at place is , by formula (12) compute gradient direction, as can be known:
(14)。
Can find With It is orientation Function, will The possible span in angle is divided into discrete Plant possible state , be denoted as:
(15)。
Wherein Angle step, definition Direction parameter:
(16)。
All boundary point is set up corresponding relation by fixed reference point function, Here it is traffic lights shape function model.
Step 4. utilizes the traffic lights shape function to do spatial mappings.
After having calculated the gradient magnitude and direction of image sequence, gradient is mapped to the totalizer parameter space greater than 0 pixel.Specific practice is: first with the accumulator array zero clearing, to all frontier points , according to the gradient direction that step 3 is calculated, utilize the traffic lights function of setting up to shine upon:
(17)
(18)
To respective point Accumulator value add 1.
The maximal value of step 5. search Accumulator space.
Utilize formula (17), (18) that all boundary pixels are mapped to parameter space, the pixel that is arranged in traffic lights in that image sequence can be added in the mode of " ballot " the parameter space same point.Along with the carrying out of " ballot ", a maximum point can appear in parameter space.This extreme point is exactly the parameter of the reference point in the original image space.
Step 6. is calculated the traffic lights position.
Utilize each traffic lights and reference point relative coordinate in the former map space, can inverting calculate the final position of traffic lights.
Principal feature of the present invention has:
(1) can answer the marked change of illumination.Be described because ambient lighting changes to be similar to linear model, the present invention is not subjected to the PARAMETERS IN THE LINEAR MODEL variable effect.
(2) impact that the image sampling colour cast is caused is insensitive.Owing to do not adopt color information, so be indifferent to the luminous change color of traffic lights.As long as intensity still exists during the traffic lights imaging, still can locate exactly traffic lights.
(3) can tackle the traffic lights of multiple Common Shape.The form of traffic lights in image changes with the structure of self, has and can change such as physical dimension variation between mounting means (transverse and longitudinal), lamp and the lamp etc.Owing to utilize the traffic lights image modeling, can change easily the traffic lights image, so that algorithm can be tackled the traffic lights under the various scenes.
Description of drawings
Fig. 1 is process flow diagram of the present invention.
Fig. 2 is modeling synoptic diagram of the present invention.
Fig. 3 is image gradient size and Orientation presentation graphs.
Fig. 4 is coordinate projection accumulation result figure.
Embodiment
In order to make purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, is not intended to limit the present invention.
Fig. 1 has provided the process flow diagram based on generalised Hough transform traffic lights detection algorithm, utilizes traffic lights topography as template, and take image sequence to be detected as input, concrete steps are:
1, at first image to be detected and template image are converted to gray level image, utilize formula (10).
2, utilizing canny operator [8] to obtain gray level image exists With The derivative of direction calculates gradient magnitude and the direction of gray level image, again such as Fig. 3 according to formula (11) and formula (12).
3, set up traffic lights shape function model.To having the pixel of gradient in the traffic lights image, set up it to the discrete function of reference point according to gradient direction, adopt formula (16).
4, inspection image mapped to be measured is arrived Accumulator space.According to the mapping relations of having set up, utilize formula (17), (18), the pixel that has gradient in the image to be detected is projected to Accumulator space.
5, the maximum value of search Accumulator space.The maximum value correspondence of Accumulator space the reference point of traffic lights in the image to be detected, such as Fig. 4.
6, calculate traffic lights coordinate in the image to be detected.
With reference to selected works:
[1] Andrew H. S. Lai and Nelson H. C. Yung, "Vehicle-Type Identification Through Automated Virtual Loop Assignment and Block-Based Direction-Biased Motion Estimation," in Proc. IEEE Conf. Intelligent Transportation Systems, 2001, 1(2), 86-97.
[2] Y. Chung, J. Wang, and S. Chen, "A vision-based traffic light detection system at intersections," Journal of National Taiwan Normal University: Mathematics, Science & Technology, 2002, 47(1), 67-86.
[3] T. Shioyama, H. Wu, N. Nakamura, and S. Kitawaki, "Measurement of the length of pedestrian crossing and detection of traffic lights from image data," Measurement Science and Technology, 2002, 13, 1450-1457.
[4] Y. Shen, U. Ozguner, K. Redmill, and J. Liu, "A Robust Video based Traffic Light Detection Algorithm for Intelligent Vehicles," Intelligent Vehicles Symposium, 2009, 3, 521-526.
[5] M. Omachi and S. Omachi, "Detection of Traffic Light Using Structural Information," ICSP2010 Proceedings.
[6] M. Omachi and S. Omachi, "Traffic Light Detection with Color and Edge Information," Computer Science and Information Technology, 2009, 8, 284-287.
[7] D. H. Ballard, "Generalizing the Hough Transform to detect arbitrary shapes," Pattern Recognition, 1981, 13(2), 111-122.
[8] Canny, J., "A Computational Approach to Edge Detection," IEEE Trans. PAMI, 1986, 11(1), 549-553.。

Claims (2)

1. the traffic lights method for detecting position based on generalised Hough transform is characterized in that take image sequence as input, uses gradation of image, and image is adopted first order differential operator computed image gradient information; Set up the traffic lights function of describing the shape, the mapping of definition from the picture point to the Accumulator space, search Accumulator space extreme value is obtained the coordinate of signal lamp, for traffic light status identification provides positional information; Concrete steps are:
Step 1. is gray space with the image of input from color space conversion
The image of described input has the coloured image of R, G, three passages of B, and each passage is 8 bit-depths in R, G, B color model, and the discrete range of intensity is 0 to 255; The formula of described conversion is:
(1)
Wherein , , The channel strength value of each pixel of respectively expression correspondence, Intensity level corresponding to gray space after the conversion;
Step 2. is utilized first order differential operator computed image gradient magnitude and direction
Obtain functional gradient calculating by consecutive point being carried out difference processing: the consecutive point on the horizontal direction are carried out difference processing, the brightness that obtains on the vertical direction changes, adopt the consecutive point difference processing on the vertical direction, the brightness that obtains on the horizontal direction changes, and image is expressed as in the gradient magnitude of certain point:
(2)
Gradient direction is expressed as:
(3)
Wherein , It is respectively the differential to image x, y direction;
Step 3. pair traffic lights shape is set up mapping function
If The reference point in the arbitrary shaped region, It is the traffic lights shape border Upper any point, note With Between difference vector be:
(4)
Arrive The axle clamp angle is , the point To the border Distance be , the definition frontier point The border orientation angles at place is , by gradient direction formula in the formula (3):
(5)
Definition With It is orientation Function, will The possible span in angle is divided into discrete Plant possible state, be denoted as:
,k=1,2,3,...,m (6)
Wherein Angle step, definition Be direction parameter:
(7)
All boundary point by fixed reference point Formula corresponding relation, is namely got traffic lights shape function model;
Step 4. utilizes the traffic lights shape function to carry out spatial mappings
After calculating the gradient magnitude and direction of image sequence, gradient is mapped to the totalizer parameter space greater than 0 pixel, specific practice is: first with the accumulator array zero clearing, to all frontier points , according to the gradient direction that step 3 is calculated, utilize the traffic lights shape function of setting up to shine upon:
(8)
(9)
To respective point Accumulator value add 1;
The maximal value of step 5. search Accumulator space
Utilize formula (8), (9) that all boundary pixels are mapped to parameter space, the pixel that is arranged in traffic lights in the image sequence is added to the parameter space same point in the mode of " ballot "; Along with the carrying out of " ballot ", parameter space can search a maximum point, and this extreme point is exactly the parameter of the reference point in the original image space;
Step 6. is calculated actual traffic lamp position coordinates
Utilize each traffic lights and reference point relative coordinate in the former map space, the final position of traffic lights is calculated in inverting.
2. the traffic lights method for detecting position based on generalised Hough transform according to claim 1 is characterized in that orientation in the step 3 The radix of discrete set is 10; In order to strengthen traffic lights appearance model tolerance, adopt the dispersion degree of 10-bin.
CN 201110325073 2011-10-24 2011-10-24 General Hough transformation-based method for detecting position of traffic signal lamp Active CN102354457B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201110325073 CN102354457B (en) 2011-10-24 2011-10-24 General Hough transformation-based method for detecting position of traffic signal lamp

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201110325073 CN102354457B (en) 2011-10-24 2011-10-24 General Hough transformation-based method for detecting position of traffic signal lamp

Publications (2)

Publication Number Publication Date
CN102354457A CN102354457A (en) 2012-02-15
CN102354457B true CN102354457B (en) 2013-10-16

Family

ID=45578016

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201110325073 Active CN102354457B (en) 2011-10-24 2011-10-24 General Hough transformation-based method for detecting position of traffic signal lamp

Country Status (1)

Country Link
CN (1) CN102354457B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102663345B (en) * 2012-03-07 2014-04-16 中盟智能科技(苏州)有限公司 Method and apparatus for automatic identification of traffic lights
WO2014078979A1 (en) * 2012-11-20 2014-05-30 Harman International Industries, Incorporated Method and system for detecting traffic lights
CN103063166B (en) * 2013-01-05 2015-03-18 山西省电力公司大同供电分公司 Detection method and device for wind deflection angle of suspension type composite insulator chain
EP3036730B1 (en) 2013-08-20 2019-12-18 Harman International Industries, Incorporated Traffic light detection
CN103679733B (en) * 2013-12-18 2018-06-05 浙江宇视科技有限公司 A kind of signal lamp image processing method and its device
CN104866823B (en) * 2015-05-11 2018-12-28 重庆邮电大学 A kind of moving vehicles detection and tracking method based on monocular vision
CN105141912B (en) * 2015-08-18 2018-12-07 浙江宇视科技有限公司 A kind of method and apparatus of signal lamp reorientation
CN107886033B (en) * 2016-09-30 2021-04-20 比亚迪股份有限公司 Method and device for identifying circular traffic light and vehicle
CN107992788B (en) * 2016-10-27 2020-09-15 比亚迪股份有限公司 Method and device for identifying traffic light and vehicle
CN109816724B (en) * 2018-12-04 2021-07-23 中国科学院自动化研究所 Three-dimensional feature extraction method and device based on machine vision
CN112036392A (en) * 2020-07-22 2020-12-04 四川长宁天然气开发有限责任公司 Automatic identification method for states of UPS (uninterrupted Power supply) indicator lamps of production station of shale gas field
CN112908006A (en) * 2021-04-12 2021-06-04 吉林大学 Method for identifying state of road traffic signal lamp and counting down time of display

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0762326A2 (en) * 1995-09-12 1997-03-12 Matsushita Electric Industrial Co., Ltd. Object detecting apparatus in which the position of a planar object is estimated by using hough transform
US5762292A (en) * 1988-09-08 1998-06-09 Daimler-Benz Aerospace Ag Apparatus for identification and tracking of objects
CN101968813A (en) * 2010-10-25 2011-02-09 华北电力大学 Method for detecting counterfeit webpage

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5762292A (en) * 1988-09-08 1998-06-09 Daimler-Benz Aerospace Ag Apparatus for identification and tracking of objects
EP0762326A2 (en) * 1995-09-12 1997-03-12 Matsushita Electric Industrial Co., Ltd. Object detecting apparatus in which the position of a planar object is estimated by using hough transform
CN101968813A (en) * 2010-10-25 2011-02-09 华北电力大学 Method for detecting counterfeit webpage

Non-Patent Citations (8)

* Cited by examiner, † Cited by third party
Title
《基于霍夫变换和几何特征的图像识别方法》;顾筠;《技术纵横》;20070831;第86-88页 *
《基于霍夫变换理论的图形识别》;杨治明等;《重庆工业高等专科学校学报》;20021130;第17卷(第4期);第16-18页 *
叶富东.基于霍夫变换的图形检测算法.《湖北生态工程职业技术学院学报》.2011,第9卷(第3期),第46-49页.
基于广义霍夫变换的芯片检测;张小军;《计算机工程》;20091231;第35卷(第23期);第252-254页 *
基于霍夫变换的图形检测算法;叶富东;《湖北生态工程职业技术学院学报》;20110331;第9卷(第3期);第46-49页 *
张小军.基于广义霍夫变换的芯片检测.《计算机工程》.2009,第35卷(第23期),第252-254页.
杨治明等.《基于霍夫变换理论的图形识别》.《重庆工业高等专科学校学报》.2002,第17卷(第4期),第16-18页.
顾筠.《基于霍夫变换和几何特征的图像识别方法》.《技术纵横》.2007,第86-88页.

Also Published As

Publication number Publication date
CN102354457A (en) 2012-02-15

Similar Documents

Publication Publication Date Title
CN102354457B (en) General Hough transformation-based method for detecting position of traffic signal lamp
CN105260699B (en) A kind of processing method and processing device of lane line data
Son et al. Real-time illumination invariant lane detection for lane departure warning system
US9704060B2 (en) Method for detecting traffic violation
CN105426864B (en) One kind being based on the matched Multi-lane Lines Detection method of equidistant marginal point
Lan et al. Vehicle speed measurement based on gray constraint optical flow algorithm
CN110379168B (en) Traffic vehicle information acquisition method based on Mask R-CNN
WO2013129361A1 (en) Three-dimensional object detection device
CN103206957B (en) The lane detection and tracking method of vehicular autonomous navigation
CN102938057B (en) A kind of method for eliminating vehicle shadow and device
KR20110131678A (en) Road and lane detection system for intelligent transportation system and method therefor
Qingsong et al. A detection and recognition method for prohibition traffic signs
CN107315095B (en) More vehicle automatic speed-measuring methods with illumination adaptability based on video processing
CN105005989A (en) Vehicle target segmentation method under weak contrast
CN202134079U (en) Unmanned vehicle lane marker line identification and alarm device
CN103544489A (en) Device and method for locating automobile logo
CN105913000B (en) A kind of wireless vehicle tracking based on video analysis
CN106802144A (en) A kind of vehicle distance measurement method based on monocular vision and car plate
CN108520528B (en) Mobile vehicle tracking method based on improved difference threshold and displacement matching model
CN104766344A (en) Vehicle detecting method based on moving edge extractor
CN103680145A (en) Automatic pedestrian and vehicle recognition method based on local image characteristics
CN106203302B (en) The pedestrian detection and statistical method that view-based access control model and wireless aware combine
CN103680148B (en) A kind of taxi recognition methods
Xuan et al. Robust lane-mark extraction for autonomous driving under complex real conditions
CN104240268B (en) A kind of pedestrian tracting method based on manifold learning and rarefaction representation

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
C14 Grant of patent or utility model