CN114241438A - Traffic signal lamp rapid and accurate identification method based on prior information - Google Patents
Traffic signal lamp rapid and accurate identification method based on prior information Download PDFInfo
- Publication number
- CN114241438A CN114241438A CN202111419474.6A CN202111419474A CN114241438A CN 114241438 A CN114241438 A CN 114241438A CN 202111419474 A CN202111419474 A CN 202111419474A CN 114241438 A CN114241438 A CN 114241438A
- Authority
- CN
- China
- Prior art keywords
- traffic signal
- signal lamp
- image
- sum
- identification method
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 82
- 238000001514 detection method Methods 0.000 claims abstract description 17
- 238000012937 correction Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 13
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000003708 edge detection Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 abstract description 6
- 238000012549 training Methods 0.000 abstract description 4
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 8
- 230000000052 comparative effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 230000002411 adverse Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- QXAITBQSYVNQDR-UHFFFAOYSA-N amitraz Chemical compound C=1C=C(C)C=C(C)C=1N=CN(C)C=NC1=CC=C(C)C=C1C QXAITBQSYVNQDR-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000000746 body region Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Images
Landscapes
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention aims to provide a traffic signal lamp rapid and accurate identification method based on prior information, and belongs to the technical field of intelligent traffic information detection. Aiming at the problem that the existing HSV color space identification method is difficult to select and verify a component range so that an identification result is accurate, the method adds a step of detecting the main body area of the traffic signal lamp, and then sums the gray values of the main body area so as to judge the type of the traffic signal lamp and finish identification. The method can still keep high identification accuracy rate under the condition of serious color distortion in the image, and improves the adaptability under different environments and weathers; meanwhile, because the invention does not adopt a machine learning or deep learning method, the requirements on equipment are greatly reduced, long-time model selection and training are avoided, and the advantages of rapidness and accuracy in recognition are taken into consideration.
Description
Technical Field
The invention belongs to the technical field of intelligent traffic information detection, and particularly relates to a traffic signal lamp rapid and accurate identification method based on traffic signal lamp structural characteristic prior information.
Background
An Intelligent Transportation System (ITS) is a comprehensive System integrating many disciplines such as artificial intelligence, computer vision, automatic control principle, communication and the like. The visual scene information is a key ring in ITS scene information, and includes important road Traffic safety information such as Traffic lights, lane lines, Traffic signs, and the like, so how to effectively and accurately identify Traffic Lights (TLRs) is an inevitable important topic in ITS.
Cameras, as a relatively reliable and low-cost sensor, are widely used in visual scene recognition in ITS. Due to interference of factors such as weather, environment, camera characteristics and the like, images acquired by the camera may be interfered by adverse factors such as color distortion, target area occlusion, distortion and the like, so that the identification accuracy of the traffic signal lamp is reduced. With the continuous development of computer vision, pattern recognition, artificial intelligence, deep learning and other disciplines, numerous methods for detecting and recognizing traffic signals appear, which are generally classified into four types: a color-based traffic signal light recognition method, a shape-based traffic signal light recognition method, a traffic signal light recognition method based on multi-feature (color, shape, map information, etc.) fusion, and a deep learning-based traffic signal light recognition method. The shape-based traffic signal lamp identification method generally uses a Hough circle detection method to identify a traffic signal lamp, so that the traffic signal lamp is easily interfered by factors such as environmental object shielding, automobile tail lamps, image distortion, color 'blooming effect' (blooming effect) and the like; most traffic signal lamp recognition methods based on multi-feature fusion introduce feature information such as a Histogram of Oriented Gradient (HOG), a high-precision map, a traffic signal lamp structure and the like on the basis of color and shape features, and use a Support Vector Machine (SVM) to train and classify, so that a better result than other algorithms can be obtained on a small sample training set, but large-scale training samples are difficult to implement, and the problem of multi-classification is difficult to solve; the traffic signal lamp identification method based on deep learning adopts algorithms such as Convolutional Neural Networks (CNN), Region-CNN (RCNN), fast-RCNN and the like, and has the advantages of strong learning capacity, but has the defects of large calculated amount, poor portability, high hardware requirement, complex model design, low interpretability and the like.
The color characteristic is the most obvious characteristic of the traffic signal lamp, the identification method is simple to realize, and the requirement on equipment hardware is not high, so that the traffic signal lamp identification method based on the color is widely used. The traffic Signal lamp identification method based on color type usually adopts a color space judgment method, and people of Hassan N et al (Hassan N, Kong W, Wah C K.A Comparative S tudy on HSV-based and Deep left-based Object Detection Algorithms for Peer Trafic Light Signal registration [ C ]. The 3rd International Conference on Intelligent autonomous systems.2020.) propose a Pedestrian traffic Signal lamp Detection method based on HSV color space and Deep Learning, wherein The identification accuracy is 93.25% under The condition of adopting only HSV color space, but The method depends on factors such as ambient Light, weather and The like to a great extent, and each component of HSV is difficult to select a proper range, so that The accuracy is not high enough in The actual general identification process.
Therefore, how to further research the color space method to improve the recognition accuracy becomes a hot point of research.
Disclosure of Invention
Aiming at the problems in the background art, the invention aims to provide a traffic signal lamp rapid and accurate identification method based on prior information. Aiming at the problem that the existing HSV color space identification method is difficult to select and verify a component range so that an identification result is accurate, the method adds a step of detecting the main body area of the traffic signal lamp, and then sums the gray values of the main body area so as to judge the type of the traffic signal lamp and finish identification.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a traffic signal lamp fast and accurate identification method based on prior information comprises the following steps:
step 1, preprocessing the traffic signal lamp image, including image inclination correction and image size standardization processing;
step 2, judging the type of the traffic signal lamp by adopting an HSV color space method for the image obtained after the preprocessing in the step 1, and if the type of the traffic signal lamp can be judged, directly outputting a judgment result; otherwise, entering step 3;
and 4, trisecting the main body area in the step 3 along the vertical direction, summing the gray pixel values of each part obtained by the trisection, wherein the summation is to sum the gray pixel values in the horizontal direction along the vertical direction of the image, and then judging the type of the traffic signal lamp according to the gray value results of the trisections.
Further, the prior information is: the traffic signal lamp is in a vertical shape, and has a red light at the upper part, a yellow light at the middle part and a green light at the lower part.
Further, the traffic signal lamp image in the step 1 is a local area where a single traffic signal lamp is located.
Further, the image tilt correction in step 1 includes four steps of RGB color image graying processing, Canny edge detection, Hough line detection, and line tilt angle compensation rotation.
Further, the specific process of the straight line tilt angle compensation in the image tilt correction is as follows: and selecting the inclination angle of the straight line with the longest length in all the detected straight lines as a rotation compensation angle to realize image inclination correction.
Further, the image size normalization in step 1 may adopt a nearest neighbor interpolation method, a bilinear interpolation method, or a double-triple interpolation method.
Further, in step 3, if the angle θ range of the straight line satisfies θ ∈ [ -10,10] ≦ 80,110], the straight line is regarded as a horizontal straight line or a vertical straight line.
The mechanism of the invention is as follows: the traffic signal lamp type identification method has the advantages that the structural characteristics of the traffic signal lamp are used as prior information, under the condition that color information is not relied, the summation interval of the gray level image pixel value summation method is determined by searching the main body area where the traffic signal lamp is located, the type of the traffic signal lamp is judged according to the distribution condition of the gray level image pixel value summation result, and the problem of low identification accuracy caused by color distortion is effectively solved.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
the traffic signal lamp rapid and accurate identification method based on the traffic signal lamp structure characteristic prior information provided by the invention provides a traffic signal lamp identification method based on the gray pixel value summation of the main body area of the traffic signal lamp on the basis of the traditional HSV color space identification traffic signal lamp type, can still keep high identification accuracy rate under the condition of serious color distortion in an image, and improves the adaptability under different environments and weathers; meanwhile, because the invention does not adopt a machine learning or deep learning method, the requirements on equipment are greatly reduced, long-time model selection and training are avoided, and the advantages of rapidness and accuracy in recognition are taken into consideration. The intelligent traffic system built based on the identification method has low cost and high accuracy, and can be applied to visual traffic scene identification in unmanned vehicles or auxiliary driving, thereby improving the road traffic safety.
Drawings
Fig. 1 is a block diagram illustrating a process of identifying a traffic signal lamp by using HSV color space in the prior art.
Fig. 2 is a schematic diagram of prior information of a signal lamp according to the present invention.
FIG. 3 is a block diagram of a process for identifying traffic lights according to the method of the present invention.
Fig. 4 is a schematic diagram of the image tilt correction process.
Fig. 5 is a schematic diagram of image size normalization.
FIG. 6 is a schematic diagram of detection of a body region of a traffic signal.
FIG. 7 is a schematic diagram of summation of gray pixel values in a traffic signal main body area.
Fig. 8 is a graph of the result of summing gray pixel values to identify the type of traffic signal in embodiment 1 of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the following embodiments and accompanying drawings.
Fig. 1 is a flow chart of identifying a traffic signal lamp by using an HSV color space method in the prior art, and it can be seen from the flow chart that the method includes an image preprocessing method and an HSV color space determination method, all pictures applicable to the method are pictures divided in a local area where the traffic signal lamp is located, the traffic signal lamp is vertical, a red light is above, a yellow light is in the middle, a green light is below, and the red light, the yellow light and the green light are used as prior information, and the specific process of identification is as follows:
step 1, preprocessing the traffic signal lamp image, including image inclination correction and image size standardization processing, specifically including,
step 1.1. the image is subjected to tilt correction processing, comprising the following steps:
step 1.1.1, converting the RGB color image of the traffic signal lamp into a gray image, wherein the conversion formula is as follows:
Gray=0.299·R+0.587·G+0.114·B (1)
step 1.1.2, Canny edge detection is carried out, firstly, the gray level image obtained in the step 1.1 is subjected to Gaussian filter smoothing treatment according to the following formula,
wherein (x, y) is the coordinate of any point in the mask, (u)x,uy) The coordinates of the center point of the mask are taken; σ is the standard deviation;
substituting the coordinates of each position into a Gaussian filter function, wherein the obtained value (f (x, y)) is the coefficient of the mask;
if the mask size is (2k +1) × (2k +1), the calculation formula of each element value in the mask is as follows:
wherein ,H(x,y)The coefficient of any point (x, y) in the mask is obtained;
then to H(x,y)And normalizing the calculated result, and if the result is in an integer form, performing normalization according to the following formula:
wherein ,is the inverse of the sum of the mask coefficients; h(-k,k)The coefficient value of the upper left corner of the mask;
if the result is in the form of a decimal number, multiplication is not requiredThe rest are kept consistent;
after the normalized mask matrix is obtained, performing convolution operation of the mask and the image pixel point regions with the same size:
where I is the original pixel matrix, HnormalThe method comprises the following steps of performing convolution operation on a Gaussian convolution kernel, and performing convolution operation on a pixel matrix of a new image obtained after the convolution operation; (7) the formula is that the gaussian smoothing filtering operation is completed, and fig. 4(a) is a gray image and an image after the gaussian smoothing filtering processing;
then calculating the gradient amplitude and direction of the smoothed new image, and convolving the input image with Sobel horizontal operator and vertical operator to obtain horizontal gradient component dxAnd a vertical gradient component dy:
dx=f(x,y)*Sobelx(x,y)
dy=f(x,y)*Sobely(x,y) (8)
The magnitude of the image gradient M (x, y) is calculated as:
M(x,y)=|dx(x,y)|+|dy(x,y)| (9)
the azimuth angle is the included angle between the image gradient direction and the x axis:
finally, non-maximum suppression of the amplitude along the gradient direction:
horizontal edge, i.e. gradient direction is vertical:
αM∈[0,22.5)∪(-22.5,0]∪(157.5,180]∪[-180,-157.5) (11)
135 ° edge, i.e. gradient direction 45 °:
αM∈[22.5,67.5)∪[-157.5,-112.5) (12)
vertical edges, i.e. gradient direction is horizontal:
αM∈[67.5,112.5]∪[-112.5,-67.5] (13)
45 ° edge, i.e. gradient direction 135 °:
αM∈(112.5,157.5]∪[-67.5,-22.5] (14)
at each point, comparing the domain center (x, y) (the actual position of the mask center on the image) with two pixels along the corresponding gradient direction, if the center pixel is the maximum value, retaining, otherwise, centering 0, and fig. 4(b) is the image after Canny edge extraction;
step 1.1.3, carrying out Hough line detection to determine a rotation compensation angle: assuming that the distance between the straight line and the origin is s and the polar angle is theta, each point on the straight line satisfies the following conditions:
s=x*cosθ+y*sinθ (15)
therefore, the length of any one straight line segment can be calculatedGet LiMaximum value of (1): l isMAX=max{L1,...,Li,.., fig. 4(c) is a graph in which the longest straight line segment in the image is detected;
mixing L withi=LMAXTheta corresponding to timeiTaking a negative value as an angle of rotation compensation of the original image, the image tilt correction is thereby completed, as shown in fig. 4 (d);
step 1.2, standardizing the size of the image, unifying the size of the image after rotation compensation to 30 multiplied by 30, and facilitating subsequent unified processing; there are three main methods for changing the size of an image: nearest neighbor interpolation, bilinear interpolation, bicubic interpolation;
taking a bilinear interpolation method as an example, the method has small calculated amount and simultaneously has higher output image quality:
setting 5 pixel points in the original image: q00(coordinates: (h)0,w0))、Q01(coordinates: (h)0,w1))、Q10(coordinates: (h)1,w0)) and Q11(coordinates: (h)1,w1) Point P (coordinates: (h, w)) is the projection of the pixel points of the target graph on the original graph;
for each pixel point of the target graph, finding four points which are most relevant to the pixel point on the original graph (namely, the points with the shortest distance satisfy w)1-w0=1,h1-h01) and obtains its pixel value by interpolation calculation, which is specifically calculated as follows:
f(P)≈(1-u)×(1-v)×f(Q00)+(1-u)×v×f(Q01) +u×(1-v)×f(Q10)+u×v×f(Q11) (16)
wherein u, v and Q (Q)00And P) coordinate-dependent (u-h)0,v=w-w0) FIG. 5 is a diagram of an image after the tilt correction and the image size normalization;
step 2, judging the type of the traffic signal lamp by adopting an HSV color space method for the image obtained after the preprocessing in the step 1, wherein the specific process is as follows: the conversion formula for converting an RGB color image into an HSV color space is as follows,
S=(max(R,G,B)-min(R,G,B))/max(R,G,B) (17)
V=max(R,G,B)
wherein H is hue, S is saturation, and V is brightness; h parameter ranges for red, yellow, green: 180 is more than or equal to Hred≥150and0≥Hred≥10,60≥Hyellow≥10,100≥HgreenThe parameter ranges of not less than 70 and S, V are all satisfied: 255 is more than or equal to Slow,255≥V≥140,Slow=int(average(Scurrent_image)×1.3);
Counting the SUM number of pixel points respectively satisfying the HSV color range of red, yellow and greenred、SUMyellow、 SUMgreen。
The existing method is commonly used: setting threshold T10, if SUMred、SUMyellow、SUMgreenIf only one value of the color is equal to or greater than T, the corresponding color is the judgment result of the traffic signal lamp; if at least one SUM is equal to or more than T, comparing the maximum values of the three, and the corresponding color is the judgment result of the traffic signal lamp.
However, in real life, the selection of H, S and V thresholds is different, so that the recognition result is inaccurate, and SUM also existsred、SUMyellow、SUMgreenThe value of (A) does not satisfy the possibility that SUM is more than or equal to T (namely, the type of the traffic signal lamp cannot be identified); meanwhile, the colors of some traffic lights in the actually photographed traffic light image may significantly change, and a judgment error may occur if the judgment is made based on the colors only.
The invention provides a traffic signal lamp fast and accurate identification method based on prior information, which comprises the following steps:
step 1, preprocessing the traffic signal lamp image, including image inclination correction and image size standardization processing;
step 2, judging the type of the traffic signal lamp by adopting an HSV color space method for the image obtained after the pretreatment in the step 1;
retaining the horizontal straight line and the vertical straight line obtained by the Hough straight line detection in the step 1.4, wherein the polar angle theta of the straight line satisfies theta ∈ [ -10,10 ]. U [80,110], as shown in FIG. 6 (a);
defining the upper left corner of the image as the origin, the horizontal right side as the x-axis, and the vertical down side as the y-axis, assuming that the horizontal central line y of the image is 15 and the vertical central line x is 15, then y istop∈[0,14],ybottom∈[15,29](ii) a Is provided with a horizontal straight line which corresponds to a vertical coordinate yiAt ytopWithin the coordinate range of (1), takeIf yiAt ybottomWithin the coordinate range of (1), takeAndnamely, the vertical coordinates corresponding to the upper and lower boundaries of the main area of the traffic signal lamp, and the horizontal coordinates corresponding to the left and right boundaries of the main area of the traffic signal lamp can be obtained in the same wayAnda rectangular area where the red and green main body is located can be obtained, as shown in fig. 6 (b);
and 4, trisecting the main body area in the step 3, summing the gray pixel values of each part obtained by the trisection, wherein the summation is to sum the gray pixel values in the horizontal direction along the vertical direction of the image, and then judging the type of the traffic signal lamp according to the gray value results of the trisections, and the specific process is as follows:
because the structural characteristics of the traffic signal lamp are known, the gray pixel value summation of the traffic light main body area can be judged; according to the step 3, the coordinates of the upper left corner of the rectangular area where the traffic signal lamp main body is located are as follows:lower left corner coordinates:coordinates of upper right cornerCoordinates of lower right cornerAs shown in fig. 7; defining the summation as summing the gray image pixel values in the x-axis direction along the y-axis direction, then y is the ordinate in the imageiFor summing the pixels in the x-axis direction, there are:
wherein ,is the gray scale image midpoint (x)i,yi) The summation range of the gray pixel values in the x-axis direction is as follows:the summation range of the gray pixel values in the y-axis direction is as follows:
according to the prior information that the green light is on the upper part, the yellow light is in the middle part and the green light is on the lower partTrisecting:
the summation process is as follows:
respectively obtaining the sum of gray value pixels of the three summation intervalsIf it isNamely, the lamp is judged to be red; if it isNamely, the lamp is judged to be yellow; if it isI.e. a green light.
Example 1
1484 traffic light data images which are collected are processed by the method, wherein 904 red traffic light images, 536 green traffic light images and 44 yellow traffic light images are obtained.
The data result chart for identifying the traffic signal light type in the present embodiment is shown in fig. 8. Wherein, the summation result of the red light gray pixel values is intensively distributed in the longitudinal axis interval yred(4,11) as shown in FIG. 8 (a); the summation result of the yellow light gray pixel values is centrally distributed in the vertical axis interval yyellow(12,17) as shown in FIG. 8 (b); the summation result of the green light gray pixel values is centrally distributed in the vertical axis interval ygreen(18,25) as shown in FIG. 8 (c). In 1484 traffic light data images in total, the correct number of traffic light identification is 1458, the identification accuracy is 98.248%, and the average time for processing a single picture is about 16.33 ms;
the results show that the identification method has high identification precision and high running speed, and can be used for identifying the traffic signal lamp in real time.
Comparative example 1
Only the identification method of HSV color space is used.
The same 1484 traffic light data images collected are processed, 904 red traffic light pictures, 536 green traffic light pictures, and 44 yellow traffic light pictures.
The correct number of identifications for this comparative example was 1396, the accuracy was 94.070%, which is a drop of 4.178% compared to approximately 13.87ms for an average single picture.
Comparative example 2
The method is adopted to identify the traffic signal lamp, and only the detection of the signal lamp main body area in the step 3 is not carried out. Namely, the gray scale image pixel value summation area is defined as the whole image, and the gray scale pixel value summation range in the x-axis direction is as follows: [0,29], the summation range of gray pixel values in the y-axis direction is: [0,29], calculation and judgment were performed according to the equations (19) and (20) in step 4, and the number of correct identifications was 1231, the accuracy was 82.951%, which was lower than 15.287%, and the average processing time for a single picture was about 17.95 ms (without region detection, the image processing area was increased, and the processing time was increased).
This is because the background portion is gray-white and the traffic light is gray-white after the image is grayed, and the gray pixel values of both are close to each other. Therefore, the traffic signal lamp main body area detection in the step 3 effectively ensures the high accuracy of identifying the traffic signal lamp by the gray pixel value summation method.
The traffic signal lamp identification method based on the HSV color space is adopted to primarily screen the types of the traffic signal lamps, and meanwhile, a method for summing gray pixel values of a main body area of the traffic signal lamp is further provided aiming at adverse factors such as color distortion and the like caused by weather, camera characteristics and the like in partial images, so that the types of the traffic signal lamps are effectively identified, the time cost and the equipment cost are greatly reduced, and meanwhile, the high accuracy of the traffic signal lamp identification can be ensured.
While the invention has been described with reference to specific embodiments, any feature disclosed in this specification may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise; all of the disclosed features, or all of the method or process steps, may be combined in any combination, except combinations where mutually exclusive features and/or steps are present.
Claims (8)
1. A traffic signal lamp fast and accurate identification method based on prior information is characterized by comprising the following steps:
step 1, preprocessing the traffic signal lamp image, including image inclination correction and image size standardization processing;
step 2, judging the type of the traffic signal lamp by adopting an HSV color space method for the image obtained after the preprocessing in the step 1, and if the type of the traffic signal lamp can be judged, directly outputting a judgment result; otherwise, entering step 3;
step 3, detecting the main body area of the traffic signal lamp in the image obtained after the preprocessing in the step 1, wherein the detection comprises Canny edge detection, Hough straight line detection and traffic signal lamp boundary determination; the specific process of determining the boundary is as follows: screening all detected straight lines, setting an angle range which needs to be met by the straight lines, and reserving horizontal and vertical straight lines; aiming at the horizontal straight lines, taking two horizontal straight lines closest to the horizontal central line of the signal lamp image as the upper and lower boundaries of the traffic signal lamp main body area; regarding the vertical straight lines, taking two vertical straight lines closest to the vertical central line as left and right boundaries of the traffic signal lamp main body area;
and 4, trisecting the main body area in the step 3 along the vertical direction, summing the gray pixel values of each part obtained by the trisection, wherein the summation is to sum the gray pixel values in the horizontal direction along the vertical direction of the image, and then judging the type of the traffic signal lamp according to the gray value results of the trisections.
2. The identification method of claim 1, wherein the a priori information is: the traffic signal lamp is in a vertical shape, and has a red light at the upper part, a yellow light at the middle part and a green light at the lower part.
3. The identification method according to claim 1, wherein the traffic signal image in step 1 is a local area where a single traffic signal is located.
4. The recognition method according to claim 1, wherein the image tilt correction in step 1 comprises four steps of RGB color image graying processing, Canny edge detection, Hough line detection, and line tilt angle compensation rotation.
5. The identification method according to claim 4, wherein the straight line tilt angle compensation in the image tilt correction comprises: and selecting the inclination angle of the straight line with the longest length in all the detected straight lines as a rotation compensation angle to realize image inclination correction.
6. The method according to claim 1, wherein the image size normalization in step 1 is performed by nearest neighbor interpolation, bilinear interpolation or bicubic interpolation.
7. The identification method according to claim 1, wherein in step 3, if the angle θ range of the straight line satisfies θ ∈ [ -10,10] uegou [80,110], the straight line is regarded as a horizontal straight line or a vertical straight line.
8. The identification method according to claim 1, wherein the specific process of judging the type of the traffic signal lamp by adopting the HSV color space method in the step 2 is as follows:
the conversion formula for converting an RGB color image into an HSV color space is as follows,
S=(max(R,G,B)-min(R,G,B))/max(R,G,B)
V=max(R,G,B)
wherein H is hue, S is saturation, and V is brightness; h parameter ranges for red, yellow, green: 180 is more than or equal to Hred≥150and0≥Hred≥10,60≥Hyellow≥10,100≥HgreenThe parameter ranges of not less than 70 and S, V are all satisfied: 255 is more than or equal to Slow,255≥V≥140,Slow=int(average(Scurrent_image)×1.3);
Counting the SUM number of pixel points respectively satisfying the HSV color range of red, yellow and greenred、SUMyellow、SUMgreen(ii) a Setting a threshold T, if SUMred、SUMyellow、SUMgreenIf only one of the values of (1) meets the condition that SUM is more than or equal to T, the corresponding color is the judgment result of the traffic signal lamp; if at least one SUM is equal to or more than T, comparing the maximum values of the three, and the corresponding color is the judgment result of the traffic signal lamp.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111419474.6A CN114241438B (en) | 2021-11-26 | 2021-11-26 | Traffic signal lamp rapid and accurate identification method based on priori information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111419474.6A CN114241438B (en) | 2021-11-26 | 2021-11-26 | Traffic signal lamp rapid and accurate identification method based on priori information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114241438A true CN114241438A (en) | 2022-03-25 |
CN114241438B CN114241438B (en) | 2023-05-16 |
Family
ID=80751277
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111419474.6A Active CN114241438B (en) | 2021-11-26 | 2021-11-26 | Traffic signal lamp rapid and accurate identification method based on priori information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114241438B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116152785A (en) * | 2023-04-24 | 2023-05-23 | 之江实验室 | Traffic signal lamp detection method and system based on CBAM combined with HSB mode |
CN117292360A (en) * | 2023-11-24 | 2023-12-26 | 福思(杭州)智能科技有限公司 | Traffic signal lamp identification method and device, storage medium and electronic equipment |
Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104021378A (en) * | 2014-06-07 | 2014-09-03 | 北京联合大学 | Real-time traffic light recognition method based on space-time correlation and priori knowledge |
CN104050827A (en) * | 2014-06-06 | 2014-09-17 | 北京航空航天大学 | Traffic signal lamp automatic detection and recognition method based on visual sense |
CN105160924A (en) * | 2015-08-25 | 2015-12-16 | 公安部第三研究所 | Video processing-based intelligent signal lamp state detection method and detection system |
CN105913041A (en) * | 2016-04-27 | 2016-08-31 | 浙江工业大学 | Pre-marked signal lights based identification method |
CN106295605A (en) * | 2016-08-18 | 2017-01-04 | 宁波傲视智绘光电科技有限公司 | Traffic lights detection and recognition methods |
CN106650641A (en) * | 2016-12-05 | 2017-05-10 | 北京文安智能技术股份有限公司 | Traffic light positioning and identification method, device and system |
CN106682597A (en) * | 2016-12-14 | 2017-05-17 | 四川长虹电器股份有限公司 | Method for identifying traffic lights |
CN106909937A (en) * | 2017-02-09 | 2017-06-30 | 北京汽车集团有限公司 | Traffic lights recognition methods, control method for vehicle, device and vehicle |
CN107507142A (en) * | 2017-08-11 | 2017-12-22 | 江苏航天大为科技股份有限公司 | Signal lamp image processing method |
CN108108761A (en) * | 2017-12-21 | 2018-06-01 | 西北工业大学 | A kind of rapid transit signal lamp detection method based on depth characteristic study |
CN110210316A (en) * | 2019-05-07 | 2019-09-06 | 南京理工大学 | Traffic lights digit recognition method based on gray level image |
KR20200101509A (en) * | 2019-01-30 | 2020-08-28 | (주)컨트롤웍스 | Method and apparatus for detecting traffic light |
CN112183427A (en) * | 2020-10-10 | 2021-01-05 | 厦门理工学院 | Rapid extraction method for arrow-shaped traffic signal lamp candidate image area |
US20210350152A1 (en) * | 2020-05-11 | 2021-11-11 | Toyota Research Institute, Inc. | Structural object detector for hierarchical ontology for traffic light handling |
-
2021
- 2021-11-26 CN CN202111419474.6A patent/CN114241438B/en active Active
Patent Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104050827A (en) * | 2014-06-06 | 2014-09-17 | 北京航空航天大学 | Traffic signal lamp automatic detection and recognition method based on visual sense |
CN104021378A (en) * | 2014-06-07 | 2014-09-03 | 北京联合大学 | Real-time traffic light recognition method based on space-time correlation and priori knowledge |
CN105160924A (en) * | 2015-08-25 | 2015-12-16 | 公安部第三研究所 | Video processing-based intelligent signal lamp state detection method and detection system |
CN105913041A (en) * | 2016-04-27 | 2016-08-31 | 浙江工业大学 | Pre-marked signal lights based identification method |
CN106295605A (en) * | 2016-08-18 | 2017-01-04 | 宁波傲视智绘光电科技有限公司 | Traffic lights detection and recognition methods |
CN106650641A (en) * | 2016-12-05 | 2017-05-10 | 北京文安智能技术股份有限公司 | Traffic light positioning and identification method, device and system |
CN106682597A (en) * | 2016-12-14 | 2017-05-17 | 四川长虹电器股份有限公司 | Method for identifying traffic lights |
CN106909937A (en) * | 2017-02-09 | 2017-06-30 | 北京汽车集团有限公司 | Traffic lights recognition methods, control method for vehicle, device and vehicle |
CN107507142A (en) * | 2017-08-11 | 2017-12-22 | 江苏航天大为科技股份有限公司 | Signal lamp image processing method |
CN108108761A (en) * | 2017-12-21 | 2018-06-01 | 西北工业大学 | A kind of rapid transit signal lamp detection method based on depth characteristic study |
KR20200101509A (en) * | 2019-01-30 | 2020-08-28 | (주)컨트롤웍스 | Method and apparatus for detecting traffic light |
CN110210316A (en) * | 2019-05-07 | 2019-09-06 | 南京理工大学 | Traffic lights digit recognition method based on gray level image |
US20210350152A1 (en) * | 2020-05-11 | 2021-11-11 | Toyota Research Institute, Inc. | Structural object detector for hierarchical ontology for traffic light handling |
CN112183427A (en) * | 2020-10-10 | 2021-01-05 | 厦门理工学院 | Rapid extraction method for arrow-shaped traffic signal lamp candidate image area |
Non-Patent Citations (3)
Title |
---|
CUIZHU BAO等: "Safe driving at traffic lights: An image recognition based approach" * |
VIJAY JOHN等: "Saliency map generation by the convolutional neural network for real-time traffic light detection using template matching" * |
郭碧筱: "基于图像处理的智能交通信号控制系统研究" * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116152785A (en) * | 2023-04-24 | 2023-05-23 | 之江实验室 | Traffic signal lamp detection method and system based on CBAM combined with HSB mode |
CN117292360A (en) * | 2023-11-24 | 2023-12-26 | 福思(杭州)智能科技有限公司 | Traffic signal lamp identification method and device, storage medium and electronic equipment |
CN117292360B (en) * | 2023-11-24 | 2024-02-09 | 福思(杭州)智能科技有限公司 | Traffic signal lamp identification method and device, storage medium and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN114241438B (en) | 2023-05-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107679520B (en) | Lane line visual detection method suitable for complex conditions | |
CN109657632B (en) | Lane line detection and identification method | |
CN109886896B (en) | Blue license plate segmentation and correction method | |
CN109785291B (en) | Lane line self-adaptive detection method | |
Kluge | Extracting road curvature and orientation from image edge points without perceptual grouping into features | |
CN106599792B (en) | Method for detecting hand driving violation behavior | |
CN107844750A (en) | A kind of water surface panoramic picture target detection recognition methods | |
CN107895375B (en) | Complex road route extraction method based on visual multi-features | |
CN114241438B (en) | Traffic signal lamp rapid and accurate identification method based on priori information | |
WO2023273375A1 (en) | Lane line detection method combined with image enhancement and deep convolutional neural network | |
CN112819094A (en) | Target detection and identification method based on structural similarity measurement | |
CN107563330B (en) | Horizontal inclined license plate correction method in surveillance video | |
CN111126306A (en) | Lane line detection method based on edge features and sliding window | |
Youjin et al. | A robust lane detection method based on vanishing point estimation | |
CN111832388B (en) | Method and system for detecting and identifying traffic sign in vehicle running | |
CN114331986A (en) | Dam crack identification and measurement method based on unmanned aerial vehicle vision | |
Ma et al. | An all-weather lane detection system based on simulation interaction platform | |
CN111860509A (en) | Coarse-to-fine two-stage non-constrained license plate region accurate extraction method | |
CN113066050A (en) | Method for resolving course attitude of airdrop cargo bed based on vision | |
CN115841633A (en) | Power tower and power line associated correction power tower and power line detection method | |
Devane et al. | Lane detection techniques using image processing | |
CN111723805B (en) | Method and related device for identifying foreground region of signal lamp | |
Stević et al. | Vision-based extrapolation of road lane lines in controlled conditions | |
CN115100615A (en) | End-to-end lane line detection method based on deep learning | |
CN113989771A (en) | Traffic signal lamp identification method based on digital image processing |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |