CN102136150B - Method for detecting speed limit sign based on image processing technology - Google Patents

Abstract

The invention discloses a method for detecting a speed limit sign based on an image processing technology. A circle scanning method for scanning from the inside to the outside is provided according to the unique corresponding relation between the color and shape pair of the speed limit sign and an inner core pattern, so that a circle can be successfully detected, the presence of a red pixel in an inner core can be judged, and the speed limit sign can be quickly distinguished from most prohibition signs (with a red pixel or a non-circular frame in the inner core). In combination with judgment on the high consistency of each black region in the inner core, a non-speed-limit sign can be quickly eliminated, and successful detection of the speed limit sign is realized.

Description

A kind of speed(-)limit sign detection method based on image processing techniques

Technical field

The present invention relates to a kind of recognition methods of road signs, especially relate to a kind of speed(-)limit sign detection method based on image processing techniques.

Background technology

Road signs identification (Traffic Sign Recognition; TSR) system is as intelligent transportation system (Intelligent Transportation System; ITS) important component part of research field can be used for DAS (Driver Assistant System) and automated driving system.As the important means of transportation information service systems, the main task of TSR system is an analyzing and processing traffic scene image, from the traffic scene figure of complicacy, detects traffic sign, and it is carried out correct differentiation ^[1]At present, " hypervelocity " is the one of the main reasons that causes Urban Road Traffic Accidents.For guaranteeing the unimpeded and traffic safety of road traffic, in conjunction with situation such as road alignment, traffic, facilities along the line, various countries are provided with corresponding speed(-)limit sign one after another with the control speed of a motor vehicle, have reduced accident rate to a certain extent.Speed(-)limit sign is as the important means of control vehicle running speed, and effect is outstanding in traffic control.Therefore, in the TSR systematic study, realize that the correct identification of speed(-)limit sign just seems particularly important.

China directly traffic sign relevant with traffic safety is divided into prohibitory sign, Warning Mark, warning notice 3 big classes, has 118 kinds (not comprising the sign that can derive from).Wherein, the quantity that prohibitory sign is set on urban road and the highway has 44 kinds at most, and speed(-)limit sign then belongs to a kind of of prohibitory sign.

Summary of the invention

Technical matters to be solved by this invention provides a kind of speed(-)limit sign detection method based on image processing techniques that can fast and effeciently speed(-)limit sign be distinguished from 118 kinds of road signs.

The present invention solves the problems of the technologies described above the technical scheme that is adopted: a kind of speed(-)limit sign detection method based on image processing techniques, and this detection method may further comprise the steps:

1. to a traffic sign, outwards make the scanning circle from centre of figure, record is through the least radius r of red pixel point _MinWith maximum radius r _Max, the ratio that defines both is m, i.e. m=r _Max/ r _Min, when m ∈ (1,1.3] time, judge that this sign is made up of the circular frame of redness, get into next step, otherwise be non-speed(-)limit sign;

2. judge and remove least radius r _MinWith maximum radius r _MaxBetween red pixel outside, less than least radius r _MinThe zone in have or not red pixel, have then to be non-speed(-)limit sign, otherwise get into next step;

3. judge less than least radius r _MinThe zone in have or not black picture element, have then to get into next step, otherwise be non-speed(-)limit sign;

4. less than least radius r _MinEach black region of regional interscan, the record black pixel point maximum height h _MaxWith minimum constructive height h _MinThe ratio that defines both is n, i.e. n=h _Max/ h _Min, when n ∈ [1,1.2], judge that this is masked as speed(-)limit sign, otherwise be non-speed(-)limit sign;

5. judging whether to accomplish the scanning of all traffic signs, is detection of end then, otherwise to next traffic sign, repeating step 1.～step 4..

Step concrete steps 1. are:

(1) obtaining the circumscribed rectangular shaped rim of red area-of-interest and setting area, serves as the round heart O of scanning with the frame center, and r ∈ (0, L) be sweep radius, 1 pixel of step-length is done scanning from inside to outside and is justified, and wherein, r is a sweep radius, and L is the distance that the frame summit is arrived at the center;

(2) scanning area-of-interest, detect circular:

A. when scanning circle on through the zone during the most inboard point, it is r that record scans radius of circle this moment _Min

B. when scanning circle through regional outermost, it is r that record scans radius of circle at this moment _Max

C. computational discrimination formula m=r _Max/ r _Min, if satisfy m ∈ (1,1.3], judge that then this area-of-interest is for circular; Then removing should the zone if do not satisfy.

Step concrete steps 4. are:

1) extracts the interior black kernel of annulus, and be partitioned into kernel area;

2) the region height consistance is judged, detects the pure digi-tal kernel:

A. scan each black kernel area, the maximum height of posting field black pixel point is h _Max, the minimum constructive height of regional black pixel point is h _Min

B. calculate n=h _Max/ h _Min, if satisfy n ∈ [1,1.2], be the pure digi-tal kernel then, confirm that this is masked as speed(-)limit sign, then do not remove this kernel and circle ring area if do not satisfy.

Compared with prior art; Advantage of the present invention is following: according to the speed(-)limit sign color shape pair unique corresponding relation with the kernel pattern, propose a kind of circular scanning from inside to outside, can successfully detect circle; Can also whether there be red pixel by the judgement symbol kernel; Can be fast speed(-)limit sign be distinguished from most of prohibitory sign (red pixel is arranged the kernel, or non-circular frame), in conjunction with to the conforming judgement of each black region height in the kernel; Can remove non-speed(-)limit sign fast, realize the successful detection of speed(-)limit sign.The present invention with compare based on the circular detection method of Hough conversion usually, overcome it and on three dimensions, carried out evidence and add up and need to consume very large time and space, be unfavorable for the shortcoming of the real-time detection of traffic sign.

Description of drawings

Fig. 1 is the prohibitory sign in China's traffic sign;

Fig. 2 a is a speed(-)limit sign common in the prohibitory sign;

Fig. 2 b is the structural representation of cylindrical in the speed(-)limit sign;

Fig. 3 is the basic framework synoptic diagram of detection method of the present invention.

Embodiment

Embodiment describes in further detail the present invention below in conjunction with accompanying drawing.

As shown in the figure, the speed(-)limit sign in China's traffic sign is its basic border color with redness, and white is its background color, and the kernel pattern is highly consistent ater numeral, and frame all is circular.According to national national standard, there is certain relation in the road speed in the size of speed(-)limit sign setting and this highway section speed limit scope, and is as shown in table 1.

The relation of table 1 speed(-)limit sign size and road speed

Analysis to the color geometric attribute and the kernel pattern of speed(-)limit sign; Can obtain an important conclusion: exist unique corresponding relationship between the color-geometric attribute of speed(-)limit sign and the kernel pattern; Promptly satisfy frame on the red circular basis, each black region height is consistent in its kernel redfree pixel and the kernel.

If C in arbitrary road signs _RThe color attribute of indicator sign is red, S _CThe expression shape attribute is circular, and the color shape that then satisfies red circle marker to subclass does

VCSP＝{(C _R，S _C)} (1)

If satisfying the kernel attribute of red circle marker is K _L| L=1,2 ..., 4, wherein, K ₁, K ₂, K ₃And K ₄Implication respectively as follows:

1. K ₁The sign of pixel redly in the expression kernel is like the mark P closed to motor vehicles 38 among Fig. 1;

2. K ₂Indicator sign kernel redfree pixel and do not have black region is like the traffic prohibited sign P41 among Fig. 1;

3. K ₃Indicator sign kernel redfree pixel and each black region height are inconsistent, like the weight limit sign P7 among Fig. 1;

4. K ₄Indicator sign kernel redfree pixel and each black region height are consistent, like the speed(-)limit sign P5 among Fig. 1.

So speed(-)limit sign color shape pair with the corresponding relation of kernel pattern does

SLS＝(C _R，S _C，K ₄) (2)

Speed(-)limit sign logical relation (2) is the speed(-)limit sign detection model.It shows, in 3 big types of 118 kinds of traffic signs of China, as long as what satisfy the SLS relation condition is exactly speed(-)limit sign, ungratifiedly can be used as the redundant area filtering.

Based on above consideration, a kind of speed(-)limit sign detection method based on image processing techniques of the present invention may further comprise the steps:

1. obtaining the circumscribed rectangular shaped rim of red area-of-interest and setting area, serves as the round heart O of scanning with the frame center, and r ∈ (0, L) be sweep radius, 1 pixel of step-length is made the scanning circle from inside to outside, and wherein, r is a sweep radius, and L is the distance that the frame summit is arrived at the center;

(2) scanning area-of-interest, detect circular:

A. when scanning circle on through the zone during the most inboard point, it is r that record this moment is scanned radius of circle through the minimum of red pixel point _Min

B. when scanning circle through regional outermost, record is r through the maximum scan radius of circle of red pixel point at this moment _Max

C. the ratio that defines both is m, i.e. m=r _Max/ r _Min, computational discrimination formula m=r _Max/ r _Min, if satisfy m ∈ (1,1.3], judge that then this area-of-interest gets into next step for circular, then be non-speed(-)limit sign if do not satisfy, removing should the zone;

2. judge and remove least radius r _MinWith maximum radius r _MaxBetween red pixel outside, less than least radius r _MinThe zone in have or not red pixel, have then to be non-speed(-)limit sign, otherwise get into next step;

3. judge less than least radius r _MinThe zone in have or not black picture element, the black kernel that then extracts in the annulus is arranged, and is partitioned into kernel area, get into next step, otherwise be non-speed(-)limit sign;

4. less than least radius r _MinEach black region of regional interscan, carry out the region height consistance and judge, detect the pure digi-tal kernel:

A. scan each black kernel area, the maximum height of posting field black pixel point is h _MaxWith minimum constructive height be h _Min

B. the ratio that defines both is n, i.e. n=h _Max/ h _Min, calculate n=h _Max/ h _Min, if satisfy n ∈ [1,1.2], be the pure digi-tal kernel then, confirm that this is masked as speed(-)limit sign, be non-speed(-)limit sign then if do not satisfy, remove this kernel and circle ring area;

5. judging whether to accomplish the scanning of all traffic signs, is detection of end then, otherwise to next traffic sign, repeating step 1.～step 4..

Claims (2)

1. speed(-)limit sign detection method based on image processing techniques is characterized in that this detection method may further comprise the steps:

1. to a traffic sign, outwards make the scanning circle from centre of figure, record is through the least radius r of red pixel point _MinWith maximum radius r _Max, the ratio that defines both is m, i.e. m=r _Max/ r _Min, when m ∈ (1,1.3] time, judge that this sign is made up of the circular frame of redness, get into next step, otherwise be non-speed(-)limit sign, concrete steps are:

(1) obtaining the circumscribed rectangular shaped rim of red area-of-interest and setting area, serves as the round heart O of scanning with the frame center, and r ∈ (0, L) be sweep radius, 1 pixel of step-length is done scanning from inside to outside and is justified, and wherein, r is a sweep radius, and L is the distance that the frame summit is arrived at the center;

(2) scanning area-of-interest, detect circular:

A. when scanning circle on through the zone during the most inboard point, it is r that record scans radius of circle this moment _Min

B. when scanning circle through regional outermost, it is r that record scans radius of circle at this moment _Max

C. computational discrimination formula m=r _Max/ r _Min, if satisfy m ∈ (1,1.3], judge that then this area-of-interest is for circular; Then remove should the zone if do not satisfy;

2. judge and remove least radius r _MinWith maximum radius r _MaxBetween red pixel outside, less than least radius r _MinThe zone in have or not red pixel, have then to be non-speed(-)limit sign, otherwise get into next step;

3. judge less than least radius r _MinThe zone in have or not black picture element, have then to get into next step, otherwise be non-speed(-)limit sign;

4. less than least radius r _MinEach black region of regional interscan, the record black pixel point maximum height h _MaxWith minimum constructive height h _MinThe ratio that defines both is n, i.e. n=h _Max/ h _Min, when n ∈ [1,1.2], judge that this is masked as speed(-)limit sign, otherwise be non-speed(-)limit sign;

5. judging whether to accomplish the scanning of all traffic signs, is detection of end then, otherwise to next traffic sign, repeating step 1. ~ step 4..

2. a kind of speed(-)limit sign detection method based on image processing techniques as claimed in claim 1 is characterized in that step concrete steps 4. are:

1) extracts the interior black kernel of annulus, and be partitioned into kernel area;

2) the region height consistance is judged, detects the pure digi-tal kernel:

A. scan each black kernel area, the maximum height of posting field black pixel point is h _Max, the minimum constructive height of regional black pixel point is h _Min

B. calculate n=h _Max/ h _Min, if satisfy n ∈ [1,1.2], be the pure digi-tal kernel then, confirm that this is masked as speed(-)limit sign, then do not remove this kernel and circle ring area if do not satisfy.

Images