CN109145797B - Square induction method for road rescue equipment based on vehicle bottom shadow positioning license plate - Google Patents
Square induction method for road rescue equipment based on vehicle bottom shadow positioning license plate Download PDFInfo
- Publication number
- CN109145797B CN109145797B CN201810915174.9A CN201810915174A CN109145797B CN 109145797 B CN109145797 B CN 109145797B CN 201810915174 A CN201810915174 A CN 201810915174A CN 109145797 B CN109145797 B CN 109145797B
- Authority
- CN
- China
- Prior art keywords
- image
- point
- license plate
- vehicle bottom
- candidate
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
Abstract
A square induction method of road rescue equipment for positioning a license plate based on vehicle bottom shadow is characterized by firstly setting a vehicle bottom shadow region of interest; secondly, performing vehicle bottom shadow segmentation operation on the gray level image of the area; then, extracting the shadow of the vehicle bottom; further, determining a license plate region of interest according to the position of the vehicle bottom shadow, converting the region image into a gray image, and performing Gaussian smooth filtering, vertical Privitt edge detection and binarization processing; then, screening edge points according to the proposed screening algorithm of the candidate point set and the effective point set; then, connecting the effective points by adopting a morphological method, and screening according to the area and length characteristics of the license plate region to obtain the license plate region and the license plate center position; and finally, carrying out dragging induction according to the position difference of the license plate center and the image center in the transverse direction. The positive induction method provided by the invention has strong environmental adaptability and anti-interference capability.
Description
Technical Field
The invention belongs to the field of intelligent operation of road rescue equipment, and particularly relates to a vehicle bottom shadow positioning license plate-based road rescue equipment square induction method.
Background
With the rapid development of the transportation and automobile industry, the mileage and the automobile holding capacity of the expressway in the world are rapidly increased, and the road traffic accidents occupy the largest proportion in the sudden public safety incidents, and in recent years, the proportion of the road traffic accidents to the total number of all kinds of safety accidents is high, even exceeds 70%, and the number of deaths accounts for 83% of the total number of deaths of all kinds of safety accidents, so that the expressway with the greatest importance of public safety issues and government departments is a public safety issue. After a road traffic accident occurs, if rapid and efficient obstacle clearing rescue and emergency treatment cannot be carried out on an accident vehicle in time, for example, on a two-way road, light road rescue equipment cannot rapidly and accurately drag the accident vehicle away from a lane from a square position, so that traffic jam is caused, the safety smoothness of the road is influenced, and even a secondary accident is caused. At present, the main reason of low square towing induction efficiency of light road rescue equipment is that the intelligent level of the rescue equipment is low, and the existing scientific technology is not utilized to carry out induction assistance on towing operation.
The light road rescue equipment is towing equipment with relatively single structure and function in the road rescue equipment, is used for towing some small and medium accident cars, and is positioned at the tail part of the rescue car and mainly comprises a folding arm, a telescopic arm, a swing arm and supporting arms at two sides. According to the invention, the vehicle-mounted camera is mounted on a folding arm at the tail part of the road rescue equipment, the height from the ground is 40-60 cm, the pointed direction is parallel to the longitudinal axis of the vehicle body of the rescue vehicle, and the horizontal direction points to the rear. In the process of implementing the square towing operation, the rescue vehicle is positioned in front of the accident vehicle, the road rescue equipment aligns the support arms at two sides with two front wheels of the accident vehicle respectively through the reversing operation, then fixes the front wheels of the accident vehicle through the support arms at two sides, and finally folds the arms to draw and lift, so as to tow the accident vehicle away from the site. In the process of square towing induction in the traditional situation, the operation of aligning the bracket arm of the road rescue equipment with the front wheel of the accident vehicle depends on the experience of a driver, and therefore, the time consumption is long and the efficiency is low.
Aiming at the problems, the technical difficulty to be solved by the invention is how to judge the relative position relation between the rescue vehicle and the accident vehicle by using the knowledge in the aspect of digital images according to the images acquired by the camera, thereby achieving the purpose of improving the rescue efficiency of the light road rescue equipment.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problems in the prior art, the invention provides a vehicle bottom shadow positioning license plate-based road rescue equipment square induction method, which gives an auxiliary prompt in the towing operation process and realizes accurate and efficient towing induction, thereby achieving the purpose of improving the rescue efficiency of light road rescue equipment.
The technical scheme is as follows: in order to realize the purpose of the invention, the technical scheme adopted by the invention is as follows: a vehicle bottom shadow positioning license plate-based road rescue equipment square induction method comprises the following steps:
(1) determining a vehicle bottom shadow region of interest;
(2) dividing the shadow of the vehicle bottom;
(3) extracting the shadow of the vehicle bottom;
(4) determining a license plate region of interest and preprocessing an image;
(5) determining a candidate point set and an effective point set;
(6) positioning a license plate;
(7) and (5) carrying out dragging induction.
In the step (1), firstly, the collected rear working area color image of the accident vehicle is set as an image I0And for image I0Is copied to obtain a color image I'0(ii) a Then intercepting the image I0The area 1/2 below the vehicle bottom shadow area to obtain the vehicle bottom shadow area-of-interest image, and converting the vehicle bottom shadow area-of-interest image into a gray image I1(ii) a In addition, the overall image coordinate system OXY is defined as: its coordinate origin and original image I0The top left corners of the image are coincident, the OX axis is horizontally to the right along the image, and the OY axis is vertically downward along the image;
in the step (2), a two-time adaptive threshold segmentation method is adopted to obtain the gray level image I in the step (1)1Carrying out vehicle bottom shadow segmentation to obtain a segmented image I2The method comprises the following specific steps:
(2.1) calculating a first adaptive threshold th based on the first adaptive threshold calculation formula1(ii) a And according to the threshold th1In the image I1Middle screened gray value less than th1The image points of (1) are set as a point set consisting of the image points; the first adaptive threshold calculation formula is as follows:
th1＝μ1-σ1/μ1
wherein, mu1、σ1As an image I1Mean and variance of the gray levels of th1Is a first time adaptive threshold;
(2.2) calculating a second adaptive threshold th based on the second adaptive threshold calculation formula2(ii) a And according to the threshold th2For image I1Carrying out binarization processing on the image: gray value less than th2Has a gray value of 255, whichThe gray value of other image points is set to zero, thus obtaining the image I2(ii) a The second adaptive threshold calculation formula is as follows:
th2＝μ2-σ2/μ2
wherein, mu2、σ2Mean and variance of the gray levels, th, of the set of points alpha2Is a second adaptive threshold;
wherein, in the step (3), the image I obtained in the step (2) is subjected to2Carrying out vehicle bottom shadow extraction operation to obtain a vehicle bottom shadow effective area beta and position information thereof, and specifically comprising the following steps:
(3.1) selecting rectangular structural elements of 5 × 5 size for image I2Performing an opening operation to obtain M connected regions AmWhere M is 1,2,3 …, and M represents a connected region amM is a connected region AmThe total number of (c);
(3.2) according to the area characteristics of the vehicle bottom shadow, carrying out treatment on M communication areas AmScreening, wherein M is 1,2,3 …, M, and N vehicle bottom shadow candidate areas B are obtainednN is 1,2,3 …, and N represents the candidate underbody shadow area BnThe number N is the total number of the candidate areas of the vehicle bottom shadow; then according to the length characteristics of the vehicle bottom shadow, N vehicle bottom shadow candidate areas B are processednScreening is carried out, so that the effective area beta of the underbody shadow and the position information thereof are obtained, wherein N is 1,2,3 …, and N, and the specific sub-steps are as follows:
(3.2.1) initializing m ═ 1, n ═ 0;
(3.2.2) if the region A is connectedmThe conditions are satisfied:
then substep (3.2.3) is entered, else substep (3.2.4) is entered, wherein,
denotes a connected region AmArea of (S)thIndicates the area AmAn area threshold of (d);
(3.2.3) increasing the value of n by 1 and AmJudging as a vehicle bottom shadow candidate area BnThen, the region B is obtainednIs long
Go to substep (3.2.4), in which
Are respectively the candidate regions B of the vehicle bottom shadownThe maximum and minimum of the median image point with respect to the abscissa of the overall image coordinate system OXY;
(3.2.4) if M < M, increasing M by 1 and restarting substep (3.2.2); otherwise, let N be N, go to substep (3.2.5);
(3.2.5) if N ≠ 0, in N vehicle bottom shadow candidate areas BnOf the lengths of
N is 1,2,3 …, N, corresponding candidate area B of the underbody shadownAs the effective area beta of the vehicle bottom shadow, obtaining the minimum value x of the abscissa of the effective area of the vehicle bottom shadow relative to the coordinate system OXY of the whole imageminMaximum value x of abscissamaxOrdinate minimum value yminAnd maximum value y of ordinatemax(ii) a If N is 0, continuing to execute the step (4);
in the step (4), determining a license plate region of interest according to whether the vehicle bottom shadow is successfully extracted in the step (3), and then preprocessing an image of the license plate region of interest to obtain a binarized image I4And T edge points CtT is 1,2,3 …, T, where T denotes the edge point CtThe sequence number of (1) and T is the total number of the edge points, and the specific steps are as follows:
(4.1) if N is 0, namely the vehicle bottom shadow is not successfully extracted, the color image I 'in the step (1) is processed'0Color image I as license plate region of interest3(ii) a Otherwise, according to the coordinate information of the effective area beta of the shadow under the vehicle, carrying out color image I'0To proceed to the corresponding positionObtaining a license plate region of interest color image I3(ii) a The interception range is: relative to the overall image coordinate system OXY, x0Is (x)min-xth,xmax+xth)，y0Is of value y0∈(0,ymax)，x0、y0Are each picture I'0Abscissa, ordinate, x, of the middle image pointthIs the range expansion threshold on the abscissa;
(4.2) license plate region of interest image I3Carrying out pretreatment operation: first, for image I3Making duplication to obtain image I3'; next, image I is imaged3Converting the image into a gray image, and performing Gaussian smooth filtering on the gray image to remove noise interference; then, performing edge detection in the vertical direction on the image subjected to Gaussian filtering by using a Privitt operator to obtain an image subjected to edge detection; then, the image after the edge detection is subjected to binarization processing, namely in the image after the edge detection, the gradient amplitude along the horizontal direction of the image is larger than a threshold value PthThe gray value of the image point is 255, and the gray values of other image points are 0, so that the image I after binarization is obtained4And T edge points CtT is 1,2,3 …, T, wherein PthA threshold value for binarization processing;
in the step (5), aiming at the color characteristics and the densely distributed characteristics of the edge points of the license plate region characters, a screening algorithm of a license plate region character edge candidate point set and an effective point set is provided, and T edge points C in the step (4) are subjected to screening algorithmtScreening is carried out, T is 1,2,3 …, T, the screening algorithm comprises two steps, a candidate point set is determined firstly, the candidate point set belongs to rough selection, then an effective point set is determined aiming at the candidate point set, namely, fine selection is carried out, and the screening algorithm comprises the following specific steps:
(5.1) determining a rough selection process of the candidate point set: according to the color characteristics of the edge points of the characters in the license plate area with the blue background and white characters, the image I after binarization is subjected to image binarization4Upper T edge points CtScreening to obtain roughly selected image I5And K candidate edge points EkK is 1,2,3 …, K, where K represents an edgeEdge point EkK is the total number of candidate edge points, and the specific sub-steps are as follows:
(5.1.1) initializing t ═ 1, k ═ 0;
(5.1.2) Using edge points CtCoordinate values in the color image I 'relative to the overall image coordinate system OXY'3Find the corresponding image point (x) with the same coordinate valuet,yt)，xt、ytThe abscissa and ordinate of the point with respect to the overall image coordinate system OXY are respectively set as the image point (x)t,yt) Is a point (x)t-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1), point (x)t+1,yt+1)；
(5.1.3) respectively solving the points (x) according to the conversion formula from the red, green and blue color space to the HSI color spacet-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1) and point (x)t+1,ytA hue component, a saturation component, and a brightness component of + 1);
(5.1.4) pairs of image points (x)t,yt) Six surrounding image points, i.e. points (x)t-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1) and point (x)t+1,yt+1), color discrimination is performed one by one: if the image point satisfies the condition (H)min＜H＜Hmax) And (S > S)min) Judging the color of the image point as blue; if the image point satisfies the condition (S < S)max) And (I > I)min) Judging the color of the image point as white; wherein H, S, I represents the hue component, saturation component and brightness component of the image point, respectively, Hmax、HminHigh and low threshold values, S, respectively, for hue componentsmax、SminHigh and low thresholds, I, respectively, for the saturation componentminIs a low threshold for the luminance component;
(5.1.5) if the image point (x)t-1,yt) Is blue, image point (x)t-1,yt-1) or image points (x)t-1,yt+1) is blue, while the image point (x)t+1,yt) Is white, the image point (x)t+1,yt-1) or image points (x)t+1,yt+1) is white, the k value is added by 1 to the edge point CtIs judged as a candidate point EkAnd will point EkIs set to be 255, the substep (5.1.7) is carried out, otherwise, the substep (5.1.6) is carried out;
(5.1.6) if the image point (x)t-1,yt) Is white, the image point (x)t-1,yt-1) or image points (x)t-1,yt+1) is white, while the image point (x)t+1,yt) Is blue, image point (x)t+1,yt-1) or image points (x)t+1,yt+1) is blue, the k value is added by 1 to the edge point CtIs judged as a candidate point EkAnd will point EkIs set to 255, otherwise point E is setkSetting the gray value of (4) to 0, and entering the substep (5.1.7);
(5.1.7) if t<T, increasing T by 1, and re-entering the substep (5.1.2); otherwise, ending the rough selection process, and making K equal to K, thereby obtaining K candidate edge points EkAnd roughly selected image I5，k＝1,2,3…,K；
(5.2) a selection process of determining the effective point set: according to the characteristic that the character edge points in the license plate area are dense, K candidate edge points E are subjected tokScreening to obtain selected image I6And an effective point set, K is 1,2,3 …, and K, and the specific steps are as follows:
(5.2.1) initializing k ═ 1;
(5.2.2) at the edge point EkScanning in a 7 × 7 area around the center, and if other candidate edge points are detected, retaining the edge point EkOtherwise, the point E is pointed outkSetting the gray value to zero;
(5.2.3) if K is less than K, increasing the value of K by 1, and returning to the substep (5.2.2); otherwise, finishing the fine selection process to obtainTo the refined image I6And a set of valid points; if the effective point set is an empty set, returning to the step (1); otherwise, executing the step (6);
in the step (6), the effective point sets obtained in the step (5) are connected by adopting a morphological method to obtain connected regions with different sizes and shapes, and are screened according to the area and length characteristics of the license plate region to obtain the license plate effective region and position information, and the specific steps are as follows:
(6.1) for the image I selected in the step (5)6Processing by morphological method to obtain J connected regions F with different sizes and shapesjJ is 1,2,3 …, and J denotes a connected region FjJ is the total number of connected regions, and the specific sub-steps are as follows:
(6.1.1) selecting rectangular structural elements with the size of 3 multiplied by 5 for the image I6Performing four times of expansion operation, connecting the effective point sets, communicating a license plate region, and then performing four times of corrosion operation on the expanded image by using rectangular structural elements with the same size to keep the size of the license plate region unchanged;
(6.1.2) selecting rectangular structural elements with the size of 3 multiplied by 3, carrying out open operation on the image after corrosion operation, further eliminating small noise areas in the non-license plate area, and obtaining J communicated areas F with different sizes and shapesj，j＝1,2,3…,J；
(6.2) according to the area characteristics of the license plate region, carrying out alignment on J connected regions FjScreening J to 1,2,3 … and J to obtain W license plate candidate regions DwW is 1,2,3 …, and W represents the license plate candidate region DwW is a license plate candidate region DwThe total number of (c); then, according to the length characteristics of the license plate regions, the W license plate candidate regions D are subjected towScreening, wherein W is 1,2,3 …, and W, obtaining the license plate effective area gamma and the abscissa x of the license plate center relative to the overall image coordinate system OXYmidThe specific substeps are as follows:
(6.2.1) initializing j ═ 1, w ═ 0;
(6.2.2) if the region F is connectedjThe conditions are satisfied:
then go to substep (6.2.3); otherwise, go to substep (6.2.4), in which
Indicates a connected region FjArea of (S)thFIndicates the region FjAn area threshold of (d);
(6.2.3) increasing the value of w by 1 and FjJudged as a license plate candidate region DwThen, the candidate region D of the license plate is obtainedwIs long
Go to substep (6.2.4) in which
xmaxw、xminwRespectively are license plate candidate regions DwThe maximum and minimum of the median image point with respect to the abscissa of the overall image coordinate system OXY;
(6.2.4) if J < J, increasing J by 1 and restarting substep (6.2.2); otherwise, let W equal W, go to substep (6.2.5);
(6.2.5) if W is 0, indicating that no license plate candidate area is obtained, and returning to the step (1); otherwise, in W license plate candidate areas DwIs selected to be the maximum value
W is 1,2,3 …, W, corresponding connected region DwAs the license plate effective area gamma, and obtaining the minimum value x of the abscissa of the license plate effective area relative to the overall image coordinate system OXYDminMaximum value of abscissa is xDmaxThe abscissa x of the license plate center relative to the overall image coordinate system OXYmidIs (x)Dmin+xDmax)/2；
In the step (7), the abscissa x of the center of the license plate of the accident vehicle obtained in the step (6) is usedmidAnd image I0Comparing the horizontal coordinates of the centers, giving a direction prompt in real time and inducing a driver to performBacking operation: relative to the overall image coordinate system OXY, if the horizontal coordinate x of the center of the license plate of the accident vehiclemidSmaller than image I0The abscissa of the center, the center of the license plate in the image is on the left side of the center of the image, and then the driver is prompted to back up to the right; otherwise, the driver is prompted to back left, so that the support arms on the two sides are aligned with the two front wheels of the accident vehicle, the accident vehicle is further clamped and fixed, and the accident vehicle is pulled and pulled away.
Has the advantages that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects:
1. aiming at the structure of light road rescue equipment and the working characteristics of square towing operation, the invention provides a square induction method of the road rescue equipment based on vehicle bottom shadow positioning license plate, which can effectively accelerate the process of aligning the supporting arms at two sides of the light road rescue equipment with the front wheel of an accident vehicle, thereby forming induction on towing operation and improving the rescue efficiency of the light road rescue equipment.
2. The induction reliability is high, the anti-interference capability is high, the induction method fully considers and utilizes similar characteristics on license plates of different vehicles, and the robustness is strong.
3. The vehicle bottom shadow segmentation method has good environment adaptability, adopts a twice self-adaptive threshold segmentation method in the induction method to segment the vehicle bottom shadow, and can adapt to vehicle models of various sizes.
Drawings
FIG. 1 is a flow chart of a method for inducing positive drag according to the present invention;
FIG. 2 is a schematic diagram of a light road rescue equipment for square towing operation on a two-way road;
FIG. 3 is a front view of a light road rescue equipment square tow induction;
FIG. 4 is a top view of a light road rescue equipment square tow induction;
FIG. 5 is a global image coordinate system OXY;
FIG. 6 is an exemplary image of a frame captured by a camera;
FIG. 7 is a gray scale image of a vehicle bottom shadow region of interest;
FIG. 8 is an image after segmentation of the underbody shadow;
FIG. 9 is a sectional image of the vehicle bottom shadow after the opening operation;
FIG. 10 is a graph of the results of underbody shading extraction;
FIG. 11 is a license plate region-of-interest image captured according to vehicle bottom shadow position information;
FIG. 12 is an image after the detection of the Prewitt operator edge in the region of interest where the license plate is located;
FIG. 13 is a license plate region-of-interest binarized image;
FIG. 14 is a image after the coarse selection;
FIG. 15 is a refined image;
FIG. 16 is an image of a license plate after a region of interest corrosion operation;
FIG. 17 is a result image of a license plate region of interest opening operation;
FIG. 18 is a result image after license plate location.
Detailed Description
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
With the rapid development of the transportation and automobile industry, the mileage of expressways and the automobile holding capacity are rapidly increased all over the world, and the road traffic accidents occupy the largest proportion in the sudden public safety incidents. In recent years, road traffic accidents account for a high proportion of all kinds of safety accidents, even more than 70%, and death accounts for 83% of all kinds of safety accidents, and become the most concerned public safety problem for the masses and government departments. After a road traffic accident occurs, if rapid and efficient obstacle clearing rescue and emergency treatment cannot be timely performed on an accident vehicle, for example, on a two-way road, see the attached drawing 2 in the specification, the light road rescue equipment cannot rapidly and accurately lift the accident vehicle from a lane from a square position, so that traffic jam is caused, the safety smoothness of the road is influenced, and even a secondary accident is caused. At present, the main reason of low square towing induction efficiency of light road rescue equipment is that the intelligent level of the rescue equipment is low, and the existing scientific technology is not utilized to carry out induction assistance on towing operation.
The towing device of light road rescue equipment (such as a pickup type rescue vehicle) is positioned at the tail part of the rescue vehicle and mainly comprises a folding arm, a telescopic arm, a swing arm and supporting arms at two sides, and the drawing 3 and the drawing 4 in the specification are shown. In the process of implementing the square towing operation, the rescue vehicle is positioned in front of the accident vehicle, referring to the attached figures 3 and 4 in the specification, the road rescue equipment aligns the support arms on two sides with two front wheels of the accident vehicle respectively through the reversing operation, then fixes the front wheels of the accident vehicle through the support arms on the two sides, and finally folds the arms to tow and lift the accident vehicle away from the site. In the traditional situation, in the process of carrying out square towing operation, the operation of aligning the bracket arm of the road rescue equipment with the front wheel of the accident vehicle depends on the experience of a driver, and therefore, the time consumption is long, and the efficiency is low. In the invention, the vehicle-mounted camera is arranged on the folding arm, the pointed direction is parallel to the longitudinal axis of the vehicle body of the rescue vehicle, the horizontal direction is backward, and the folding arm is positioned in the middle of the vehicle width, so the central axis of an image area acquired by the camera is consistent with the central axes of road rescue equipment and a towing device thereof, and the front license plate of an accident vehicle is generally positioned in the middle of the vehicle head of the accident vehicle. Aiming at the characteristic, the invention positions the license plate in the collected original image, then compares the abscissa of the center of the original image with the abscissa of the center of the license plate of the accident vehicle, thereby determining the relative position relationship between the rescue vehicle and the accident vehicle in the current state, then gives a direction prompt, forms an induction for the towing operation, and further improves the efficiency of the square towing induction.
Aiming at the working characteristics of square towing operation of light road rescue equipment, as shown in the attached figure 1 of the specification, the invention provides a square induction method of road rescue equipment based on vehicle bottom shadow positioning license plate, which comprises seven steps as follows:
(1) determining the region of interest of the vehicle bottom shadow
The vehicle-mounted camera is installed on a folding arm at the tail of the road rescue equipment, the height from the ground is 40-60 cm according to the attached drawing 3 of the specification, the pointed direction is parallel to the longitudinal axis of the rescue vehicle body, and the horizontal direction points to the rear. Firstly, the camera is arranged behind the accident vehicleCollecting the square operation area to obtain a color image I of the rear operation area0See figure 6 of the specification; secondly, for image I0Is copied to obtain a color image I'0For use in the following step (4); then, the shadow of the bottom of the vehicle is positioned in the image I according to the accident0Feature of the lower part, capturing image I0The area 1/2 below the vehicle bottom shadow area to obtain the vehicle bottom shadow area-of-interest image, and converting the vehicle bottom shadow area-of-interest image into a gray image I1See figure 7 of the specification.
In addition, in the present invention, the overall image coordinate system OXY is defined as: coordinate origin and original image I0The top left corner of the image coincides with the OX axis horizontally to the right of the image and the OY axis vertically down the image, see figure 5 of the specification.
(2) Vehicle bottom shadow segmentation
The shadow area at the bottom of the car is darker in the whole image, and is a characteristic with stronger robustness. Therefore, the method is an effective detection method for firstly segmenting the vehicle bottom shadow from the image, then narrowing the vehicle license plate detection range according to the position characteristics of the vehicle license plate above the vehicle bottom shadow, and then positioning the vehicle license plate.
The vehicle bottom shadow detection method comprises a model-based method and a feature-based method: the model-based approach requires strict assumptions and does not adapt well to the changes in various environments; the feature-based method is sensitive to noise, but can quickly detect the underbody shadow. There are many vehicle bottom shadow detection methods based on characteristics, and two-time adaptive threshold segmentation methods are one of the vehicle bottom shadow detection methods. Under the outdoor condition, the environment is complex and changeable, the image collected by the camera can be interfered by light intensity and large-area shadow, and the interference can not be effectively inhibited by adopting one-time threshold segmentation, so that the method adopts a two-time self-adaptive threshold segmentation method with strong robustness to the image I1Carrying out vehicle bottom shadow segmentation to obtain a segmented image I2See figure 8 of the specification. The method comprises the following specific steps:
(2.1) calculating a first adaptive threshold th based on the first adaptive threshold calculation formula1(ii) a And according to the threshold th1In the image I1Middle screened gray value less than th1The image points of (1) are set as a point set consisting of the image points; the first adaptive threshold calculation formula is as follows:
th1＝μ1-σ1/μ1
wherein, mu1、σ1As an image I1Mean and variance of the gray levels of th1Is a first time adaptive threshold;
(2.2) calculating a second adaptive threshold th based on the second adaptive threshold calculation formula2(ii) a And according to the threshold th2And carrying out binarization processing on the image: gray value less than th2The gray value of the image point is set to be 255, and the gray values of other image points are set to be zero, thereby obtaining the image I2(ii) a The second adaptive threshold calculation formula is as follows:
th2＝μ2-σ2/μ2
wherein, mu2、σ2Mean and variance of the gray levels, th, of the set of points alpha2Is a second adaptive threshold;
(3) vehicle bottom shadow extraction
Image I after segmentation of vehicle bottom shadow2There are many small block noise interferences. Therefore, image I is first morphologically processed2And processing to obtain a connected region set containing the vehicle bottom shadow region, and then screening a plurality of connected regions according to certain conditions to obtain the vehicle bottom shadow effective region. The method comprises the following specific steps:
(3.1) selecting rectangular structural elements of 5 × 5 size for image I2Opening operation is carried out, most small noise is removed while the vehicle bottom shadow area is reserved, and M connected areas A are obtainedmAnd M is 1,2,3 …, M, see the description and the attached figure 9. Wherein m represents a connected region AmM is a connected region AmThe total number of (c).
(3.2) in M connected regions AmAnd M is 1,2,3 …, and the area of the vehicle bottom shadow area is large and the length is long. So, first according to the shadow of the vehicle bottomArea characteristics, according to certain conditions, for M connected regions AmScreening to obtain N vehicle bottom shadow candidate areas BnN is 1,2,3 …, and N represents the candidate underbody shadow area BnN is the candidate area B of the vehicle bottom shadownThe total number of (c); then according to the length characteristics of the vehicle bottom shadow, N vehicle bottom shadow candidate areas B are processednAnd screening to extract the effective area beta of the vehicle bottom shadow and obtain the position information of the effective area beta. It should be noted that the effective working distance of the square towing is generally 1-5 m, the focal length of the camera can be selected to be 4-8 mm, the size of the image collected by the camera is fixed to be 640 x 360, and the size of the pixels in the shadow area of the vehicle bottom in the image can be changed within 1000-9000. The method comprises the following specific substeps:
(3.2.1) initializing m ═ 1, n ═ 0;
(3.2.2) if the region A is connectedmThe conditions are satisfied:
then go to substep (3.2.3); otherwise, go to substep (3.2.4), wherein,
denotes a connected region AmArea of (S)thDenotes a connected region AmThe area threshold value of (2) can be within 1000-1500;
(3.2.3) increasing the value of n by 1 and AmJudging as a vehicle bottom shadow candidate area BnThen, the candidate area B of the vehicle bottom shadow is obtainednIs long
Go to substep (3.2.4), in which
Are respectively the candidate regions B of the vehicle bottom shadownThe abscissa of the middle image point with respect to the overall image coordinate system OXYTarget maximum and minimum values;
(3.2.4) if M < M, increasing M by 1 and restarting substep (3.2.2); otherwise, let N be N, go to substep (3.2.5);
(3.2.5) if N ≠ 0, in N vehicle bottom shadow candidate areas BnOf the lengths of
N is 1,2,3 …, N, corresponding candidate area B of the underbody shadownAs the effective area beta of the vehicle bottom shadow, the effective area beta of the vehicle bottom shadow is marked on the image I1Referring to the attached figure 10 of the specification, the minimum value x of the abscissa of the effective area of the vehicle bottom shadow relative to the coordinate system OXY of the whole image is obtainedminMaximum value x of abscissamaxOrdinate minimum value yminAnd maximum value y of ordinatemax(ii) a If N is 0, the extraction operation of the vehicle bottom shadow is failed, and the step (4) is continuously executed.
(4) Determining region of interest of license plate and image preprocessing
If the extraction of the vehicle bottom shadow in the step (3) fails, the color image I 'in the step (1) is processed'0Color image I as license plate region of interest3(ii) a Otherwise, according to the coordinate information of the effective area beta of the vehicle bottom shadow, carrying out comparison on the image I'0Intercepting the corresponding position to obtain a license plate region-of-interest color image I3See, figure 11 of the specification; the interception range is: relative to the overall image coordinate system OXY, x0Is (x)min-xth,xmax+xth)，y0Is of value y0∈(0,ymax)。x0、y0Are each picture I'0Abscissa, ordinate, x, of the middle image pointthIs a range expansion threshold on the abscissa, threshold xthThe value range of (1) is 90-100. Wherein, set xthThe reason for this is that the light is complicated in the open air, the direction of light irradiation changes, the phenomenon that the detected vehicle bottom shadow deviates to one side wheel exists, and x is setthThe robustness of the algorithm can be improved.
Because the license plate is interested in the regional picture I3There is much noise, so it is necessary to apply to the image I3Carrying out pretreatment operation: first, for image I3Making duplication to obtain image I3', to be used in the step (5) described later; next, image I is imaged3Converting the image into a gray image, and performing Gaussian smoothing filtering on the gray image to remove the interference of noise; then, according to the characteristic that the number plate region has more vertical edges of characters, a Prewitt operator is adopted to perform edge detection on the image after gaussian filtering in the vertical direction, so as to obtain an image after edge detection, and refer to the attached figure 12 in the specification.
The edge detection algorithms are various, such as Prewitt algorithm, Sobel algorithm, Laplacian algorithm and the like, the difference of the algorithms is that the used gradient operators are different, and in order to ensure the real-time performance of the induction method, the Prewitt algorithm with smaller calculation amount is adopted. Wherein the Prewitt operator has the form of image horizontal direction
In a form perpendicular to the image
Because the number plate region characters have more edge points in the vertical direction, the invention only adopts the Prewitt operator in the horizontal direction of the image to carry out edge detection. Accordingly, the gradient magnitude calculation formula in the image horizontal direction at the image point (x, y) is:
Gx(x,y)＝g(x+1,y-1)+g(x+1,y)+g(x+1,y+1)-g(x-1,y-1)-g(x-1,y)-g(x-1,y+1)
wherein G isx(x, y) is the gradient amplitude of the image point (x, y) along the image horizontal direction, and g (x, y) is a function of the image gray value.
In order to reduce the calculation amount in the following rough selection process, it is further necessary to perform binarization processing on the image after edge detection according to the magnitude of the gradient amplitude of the image point after edge detection along the image horizontal direction, that is, in the image after edge detection, the gradient amplitude along the image horizontal direction is greater than the threshold value PthThe gray value of the image point is 255, and the gray values of other image points are 0, thereby obtaining T edge points CtT1, 2,3 …, T, and the binarized image I4See figure 13 of the specification. Wherein t represents an edge point CtAnd T is the total number of edge points. In order to reserve the character edge points of the license plate area to the maximum extent and remove more noise points at the same time, the threshold value P of the binarization processingthTaking values within 80-120.
(5) Determination of candidate point set and effective point set
Binary image I4There are also many interference points, so it is necessary to match the T edge points CtScreening was performed with T ═ 1,2,3 …, T. The screening basis is the basic characteristics of the character edge points of the license plate area, including color characteristics, dense characteristics of the edge points of the license plate area and the like.
The edge point screening comprises two steps: firstly, determining a candidate point set, belonging to rough selection, and then determining an effective point set aiming at the candidate point set, namely fine selection. The rough selection process is based on the color characteristics of the edge points of characters in the license plate area, and it should be noted that the license plates of most of domestic social vehicles at present are in blue-bottom white characters, so the induction method provided by the invention is suitable for small automobiles with the license plates in blue-bottom white characters under the condition of sufficient light in the daytime; the selection process is based on the characteristic that the character edge points in the license plate area are dense. Through the two-step screening, the influence of a plurality of interferences can be effectively removed.
Wherein, the processing object of the rough selection process is the edge point CtT is 1,2,3 …, T, and edge point CtThe image I of the place4Is a binary image, if the rough selection is carried out according to the color characteristics, the binary image I is needed to be utilized4Corresponding color image I3' of the color information. For binary images I4Each edge point C oftBased on the coordinate values of the color image I relative to the overall image coordinate system OXY3Finding the corresponding points with the same coordinate values, wherein the coordinate values of the corresponding points are relative to the whole image coordinate system OXY; then using the color information of the adjacent point of the corresponding point to the edge point C according to a certain conditiontIt is effective to carry out the screeningThe screening method of (1). The specific steps of edge point screening are as follows:
(5.1) determining a rough selection process of the candidate point set: for the binarized image I4Upper T edge points CtScreening to obtain roughly selected image I5And K candidate edge points EkK is 1,2,3 …, K, where K denotes the edge point EkK is the total number of candidate edge points, and the specific sub-steps are as follows:
(5.1.1) initializing t ═ 1, k ═ 0;
(5.1.2) Using edge points CtIn the obtained color image I 'relative to the coordinate values of the overall image coordinate system OXY'3Find the corresponding image point (x) with the same coordinate valuet,yt)，xt、ytThe abscissa and ordinate of the point with respect to the overall image coordinate system OXY are respectively set as the image point (x)t,yt) Is a point (x)t-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1), point (x)t+1,yt+1)；
(5.1.3) respectively solving the points (x) according to the conversion formula from the red, green and blue color space to the HSI color spacet-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1) and point (x)t+1,yt+1) a hue (H) component, a saturation (S) component and a brightness (I) component;
(5.1.4) pairs of image points (x)t,yt) Six surrounding image points, i.e. points (x)t-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1) and point (x)t+1,yt+1), color discrimination is performed one by one: if the image point satisfies the condition Hmin＜H＜HmaxAnd S > SminJudging the color of the image point as blue; if the image point satisfies the condition S < SmaxAnd I > IminThen the color of the image point is determined to be white. Wherein H, S, I represents the hue component, saturation component and brightness component of the image point, respectively, Hmax、HminHigh and low thresholds, respectively, for hue components, threshold HmaxTaking a value within 240-245 and taking a threshold value HminTaking a value within 130-140; smax、SminHigh and low thresholds, respectively, of the saturation component, threshold SmaxThe value is within 0.25-0.35, and the threshold value S isminTaking a value within 0.2-0.25; i isminIs a low threshold, threshold I, of the luminance componentminTaking values within 80-85;
(5.1.5) if the image point (x)t-1,yt) Is blue, image point (x)t-1,yt-1) or image points (x)t-1,yt+1) is blue, while the image point (x)t+1,yt) Is white, the image point (x)t+1,yt-1) or image points (x)t+1,yt+1) is white, the k value is added by 1 to the edge point CtIs judged as a candidate point EkAnd will point EkIs set to be 255, the substep (5.1.7) is carried out, otherwise, the substep (5.1.6) is carried out;
(5.1.6) if the image point (x)t-1,yt) Is white, the image point (x)t-1,yt-1) or image points (x)t-1,yt+1) is white, while the image point (x)t+1,yt) Is blue, image point (x)t+1,yt-1) or image points (x)t+1,yt+1) is blue, the k value is added by 1 to the edge point CtIs judged as a candidate point EkAnd will point EkThe gray value of is 255; otherwise point E will be pointed outkSetting the gray value of (4) to 0, and entering the substep (5.1.7);
(5.1.7) if t<T, increasing T by 1, and re-entering the substep (5.1.2); otherwise, ending the rough selection process, and making K equal to K to obtain K candidate edge points EkAnd roughly selected image I5K1, 2,3 …, K, image I5See description figure 14.
Supplementary description of the above roughing algorithm: classic color space has redThe red, green and blue color space can be well matched with the characteristic that human eyes have strong sensibility to red, green and blue, such as the green and blue color space, the HSI color space and the like, so that the color image is generally represented in the red, green and blue color space, such as the color image I 'in the invention'3. The three color components of red, green and blue are easily affected by illumination, and the three components of hue, saturation and brightness of the HSI color space are relatively stable. Therefore, in the above-mentioned rough selection step, the hue component, the saturation component and the brightness component of the image point are obtained without directly using the three color components of red, green and blue of the image point.
(5.2) a selection process of determining the effective point set: according to the characteristic that the character edge points in the license plate area are dense, K candidate edge points E are subjected tokScreening to obtain selected image I6And the effective point set, K is 1,2,3 …, K, and the specific sub-steps are as follows:
(5.2.1) initializing k ═ 1;
(5.2.2) at the edge point EkScanning in a 7 × 7 area around the center, and if other candidate edge points are detected, retaining the edge point EkOtherwise, the point E is pointed outkSetting the gray value to zero;
(5.2.3) if K is less than K, increasing the value of K by 1, and returning to the substep (5.2.2); otherwise, ending the selection process to obtain the selected image I6And set of significant points, image I6See specification figure 15. If the effective point set is an empty set, returning to the step (1); otherwise, continuing to execute the step (6).
(6) License plate location
For the license plate region-of-interest image, after the effective point set is determined, the effective point set can be connected, and the rectangular region of the license plate is restored. Meanwhile, a small amount of interference of connected areas of non-license plates exists on the image. Therefore, according to the characteristics of larger area and longer length of the license plate region in the image, the connected region is screened, and the license plate effective region gamma and the abscissa x of the license plate center relative to the whole image coordinate system OXY can be obtainedmidThereby completing the license plate positioning. Among them, the method for connecting the effective point sets isThe method adopts a morphological method for processing, effectively reduces the rectangular region of the license plate, and can filter the interference of partial noise points. The license plate positioning method specifically comprises the following steps:
(6.1) on the refined image I6Processing by morphological method to obtain J connected regions F with different sizes and shapesjJ is 1,2,3 …, J. Wherein j represents a connected region FjJ is the total number of connected regions, and the specific sub-steps are as follows:
(6.1.1) selecting rectangular structural elements with the size of 3 multiplied by 5 for the image I6Performing expansion operation for four times, connecting the effective point sets to communicate with a license plate region, and then performing corrosion operation for four times on the expanded image by using rectangular structural elements with the same size to keep the size of the license plate region unchanged, wherein the result image refers to a point set formed by carefully selected edge points, and the effective edge points refer to a point set formed by the carefully selected edge points in the specification, and are shown in the attached figure 16 of the specification;
(6.1.2) selecting rectangular structural elements with the size of 3 multiplied by 3, carrying out open operation on the image after corrosion operation, further eliminating small noise areas of non-license plate areas, and obtaining J communicated areas F with different sizes and shapesjJ is 1,2,3 …, J, and the image after the calculation is shown in the figure 17 of the specification;
(6.2) according to the area characteristics of the license plate region, carrying out alignment on J connected regions FjScreening J to 1,2,3 … and J to obtain W license plate candidate regions DwW is 1,2,3 …, and W represents the license plate candidate region DwW is a license plate candidate region DwThe total number of (c); then, according to the length characteristics of the license plate regions, the W license plate candidate regions D are subjected towScreening is carried out, W is 1,2,3 …, W, thereby obtaining the license plate effective area gamma and the abscissa x of the license plate center relative to the overall image coordinate system OXYmidThe specific substeps are as follows:
(6.2.1) initializing j ═ 1, w ═ 0;
(6.2.2) if the region F is connectedjThe conditions are satisfied:
substep (6) is entered.2.3); otherwise, go to substep (6.2.4). Wherein the content of the first and second substances,
indicates a connected region FjArea of (S)thFIndicates a connected region FjThe area threshold value of (1) should be pointed out, the effective working distance of the square towing is generally 1-5 meters, the focal length of the camera can be selected to be 4-8 millimeters, the size of the image collected by the camera is fixed to be 640 multiplied by 360, the pixel size of the license plate area in the image can be changed within 400-5000, and the threshold value S is based on the conditionsthFThe value can be taken within 100-400;
(6.2.3) increasing the value of w by 1 and FjJudged as a license plate candidate region DwThen, the candidate region D of the license plate is obtainedwIs long
Go to substep (6.2.4). Wherein
xmaxw、xminwRespectively are license plate candidate regions DwMaximum and minimum values of the abscissa of the middle image point relative to the overall image coordinate system OXY;
(6.2.4) if J < J, increasing J by 1 and restarting substep (6.2.2); otherwise, let W equal W, go to substep (6.2.5);
(6.2.5) if W is 0, indicating that no license plate candidate area is obtained, and returning to the step (1); otherwise, in W license plate candidate areas DwIs selected to be the maximum value
W is 1,2,3 …, W, corresponding connected region DwAs the license plate effective area gamma, and obtaining the minimum value x of the abscissa of the license plate effective area relative to the overall image coordinate system OXYDminMaximum value of abscissa is xDmaxThe horizontal coordinate x of the center of the license platemidIs (x)Dmin+xDmax)/2，xmidRelative to the overall image coordinate system OXY. Wherein the content of the first and second substances,the license plate effective area gamma is marked on the color image I0See figure 18 of the specification.
(7) Carrying out towing induction
In the process of light road rescue, the abscissa x of the center of the license plate of the accident vehicle obtained in the step (6) is usedmidAnd image I0Comparing the horizontal coordinates of the centers, giving a direction prompt in real time, and inducing a driver to perform backing operation: relative to the overall image coordinate system OXY, if the horizontal coordinate x of the center of the license plate of the accident vehiclemidSmaller than image I0The abscissa of the center, the center of the license plate in the image is on the left side of the center of the image, and then the driver is prompted to back up to the right; otherwise, the driver is prompted to back left, so that the support arms on the two sides are aligned with the two front wheels of the accident vehicle, the accident vehicle is further clamped and fixed, and the accident vehicle is pulled and pulled away.
Claims (5)
1. A square induction method for positioning license plates of road rescue equipment based on vehicle bottom shadow is characterized by comprising the following steps:
(1) determining a vehicle bottom shadow region of interest; firstly, setting a collected rear operation area color image of an accident vehicle as an image I0And for image I0Is copied to obtain a color image I'0(ii) a Then intercepting the image I0The area 1/2 below the vehicle bottom shadow area to obtain the vehicle bottom shadow area-of-interest image, and converting the vehicle bottom shadow area-of-interest image into a gray image I1(ii) a In addition, the overall image coordinate system OXY is defined as: its coordinate origin and original image I0The top left corners of the image are coincident, the OX axis is horizontally to the right along the image, and the OY axis is vertically downward along the image;
(2) dividing the shadow of the vehicle bottom; adopting a twice self-adaptive threshold segmentation method to obtain a gray level image I in the step (1)1Carrying out vehicle bottom shadow segmentation to obtain a segmented image I2The method comprises the following specific steps:
(2.1) calculating a first adaptive threshold th based on the first adaptive threshold calculation formula1(ii) a And according to the threshold th1In the image I1Middle screened gray value less than th1Is shown inSetting a point set composed of the image points as alpha; the first adaptive threshold calculation formula is as follows:
th1＝μ1-σ1/μ1
wherein, mu1、σ1As an image I1Mean and variance of the gray levels of th1Is a first time adaptive threshold;
(2.2) calculating a second adaptive threshold th based on the second adaptive threshold calculation formula2(ii) a And according to the threshold th2For image I1Carrying out binarization treatment: gray value less than th2The gray value of the image point is set to be 255, and the gray values of other image points are set to be zero, thereby obtaining the image I2(ii) a The second adaptive threshold calculation formula is as follows:
th2＝μ2-σ2/μ2
wherein, mu2、σ2Mean and variance of the gray levels, th, of the set of points alpha2Is a second adaptive threshold;
(3) extracting the shadow of the vehicle bottom; for the image I obtained in the step (2)2Carrying out vehicle bottom shadow extraction operation to obtain a vehicle bottom shadow effective area beta and position information thereof, and specifically comprising the following steps:
(3.1) selecting rectangular structural elements of 5 × 5 size for image I2Performing an opening operation to obtain M connected regions AmWhere M is 1,2,3 …, and M represents a connected region amM is a connected region AmThe total number of (c);
(3.2) according to the area characteristics of the vehicle bottom shadow, carrying out treatment on M communication areas AmScreening, wherein M is 1,2,3 …, M, and N vehicle bottom shadow candidate areas B are obtainednN is 1,2,3 …, and N represents the candidate underbody shadow area BnThe number N is the total number of the candidate areas of the vehicle bottom shadow; then according to the length characteristics of the vehicle bottom shadow, N vehicle bottom shadow candidate areas B are processednScreening is carried out, so that the effective area beta of the underbody shadow and the position information thereof are obtained, wherein N is 1,2,3 …, and N, and the specific sub-steps are as follows:
(3.2.1) initializing m ═ 1, n ═ 0;
(3.2.2) if the region A is connectedmThe conditions are satisfied:
then substep (3.2.3) is entered, else substep (3.2.4) is entered, wherein,
denotes a connected region AmArea of (S)thIndicates the area AmAn area threshold of (d);
(3.2.3) increasing the value of n by 1 and AmJudging as a vehicle bottom shadow candidate area BnThen, the region B is obtainednIs long
Go to substep (3.2.4), in which
Are respectively the candidate regions B of the vehicle bottom shadownThe maximum and minimum of the median image point with respect to the abscissa of the overall image coordinate system OXY;
(3.2.4) if M < M, increasing M by 1 and restarting substep (3.2.2); otherwise, let N be N, go to substep (3.2.5);
(3.2.5) if N ≠ 0, in N vehicle bottom shadow candidate areas BnOf the lengths of
Nmax is more than or equal to 1 and less than or equal to N, and corresponding vehicle bottom shadow candidate area B is divided intonmaxAs the effective area beta of the vehicle bottom shadow, obtaining the minimum value x of the abscissa of the effective area of the vehicle bottom shadow relative to the coordinate system OXY of the whole imageminMaximum value x of abscissamaxOrdinate minimum value yminAnd maximum value y of ordinatemax(ii) a If N is 0, continue to execute step (4)；
(4) Determining a license plate region of interest and preprocessing an image;
(5) determining a candidate point set and an effective point set;
(6) positioning a license plate;
(7) and (5) carrying out dragging induction.
2. The square induction method for the road rescue equipment based on the underbody shadow positioning of the license plate as claimed in claim 1, which is characterized in that: in the step (4), determining a license plate region of interest according to whether the vehicle bottom shadow is successfully extracted in the step (3), and then preprocessing the image of the license plate region of interest to obtain a binarized image I4And T edge points CtT is 1,2,3 …, T, where T denotes the edge point CtThe sequence number of (1) and T is the total number of the edge points, and the specific steps are as follows:
(4.1) if N is 0, namely the vehicle bottom shadow is not successfully extracted, the color image I 'in the step (1) is processed'0Color image I as license plate region of interest3(ii) a Otherwise, according to the coordinate information of the effective area beta of the shadow under the vehicle, carrying out color image I'0Intercepting the corresponding position to obtain a license plate region-of-interest color image I3(ii) a The interception range is: relative to the overall image coordinate system OXY, x0Is (x)min-xth,xmax+xth)，y0Is of value y0∈(0,ymax)，x0、y0Are each picture I'0Abscissa, ordinate, x, of the middle image pointthIs the range expansion threshold on the abscissa;
(4.2) license plate region of interest color image I3Carrying out pretreatment operation: first, for image I3Was copied to give image I'3(ii) a Next, image I'3Converting the image into a gray image, and performing Gaussian smoothing filtering on the gray image to remove the interference of noise; then, performing edge detection in the vertical direction on the image subjected to Gaussian smooth filtering by using a Privitt operator to obtain an image subjected to edge detection; then, the image after edge detection is processedLine binarization processing, i.e. in the image after edge detection, the gradient amplitude in the horizontal direction of the image is larger than a threshold value PthThe gray value of the image point is 255, and the gray values of other image points are 0, so that the image I after binarization is obtained4And T edge points CtT is 1,2,3 …, T, wherein PthIs a threshold value of the binarization processing.
3. The vehicle bottom shadow positioning license plate-based road rescue equipment square induction method according to claim 2, characterized in that: in step (5), the T edge points C in step (4) are processedtScreening is carried out, T is 1,2,3 …, T, the screening algorithm comprises two steps, a candidate point set is determined firstly, the candidate point set belongs to rough selection, then an effective point set is determined aiming at the candidate point set, namely, fine selection is carried out, and the screening algorithm comprises the following specific steps:
(5.1) determining a rough selection process of the candidate point set: according to the color characteristics of the edge points of the characters in the license plate area with the blue background and white characters, the image I after binarization is subjected to image binarization4T edge points C oftScreening to obtain roughly selected image I5And K candidate edge points EkK is 1,2,3 …, K, where K denotes the edge point EkK is the total number of candidate edge points, and the specific sub-steps are as follows:
(5.1.1) initializing t ═ 1, k ═ 0;
(5.1.2) Using edge points CtCoordinate values in the color image I 'relative to the overall image coordinate system OXY'3Find the corresponding image point (x) with the same coordinate valuet,yt)，xt、ytThe abscissa and ordinate of the point with respect to the overall image coordinate system OXY are respectively set as the image point (x)t,yt) Is a point (x)t-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1), point (x)t+1,yt+1)；
(5.1.3) respectively solving the points (x) according to the conversion formula from the red, green and blue color space to the HSI color spacet-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1) and point (x)t+1,ytA hue component, a saturation component, and a brightness component of + 1);
(5.1.4) pairs of image points (x)t,yt) Six surrounding image points, i.e. points (x)t-1,yt-1), point (x)t-1,yt) Point (x)t-1,yt+1), point (x)t+1,yt) Point (x)t+1,yt-1) and point (x)t+1,yt+1), color discrimination is performed one by one: if the image points simultaneously satisfy the condition Hmin＜H＜HmaxAnd S > SminJudging the color of the image point as blue; if the image points simultaneously satisfy the condition S < SmaxAnd I > IminJudging the color of the image point as white; wherein H, S, I represents the hue component, saturation component and brightness component of the image point, respectively, Hmax、HminHigh and low threshold values, S, respectively, for hue componentsmax、SminHigh and low thresholds, I, respectively, for the saturation componentminIs a low threshold for the luminance component;
(5.1.5) if the image point (x)t-1,yt) Is blue, image point (x)t-1,yt-1) or image points (x)t-1,yt+1) is blue, while the image point (x)t+1,yt) Is white, the image point (x)t+1,yt-1) or image points (x)t+1,yt+1) is white, the k value is added by 1 to the edge point CtIs judged as a candidate edge point EkAnd will point EkIs set to be 255, the substep (5.1.7) is carried out, otherwise, the substep (5.1.6) is carried out;
(5.1.6) if the image point (x)t-1,yt) Is white, the image point (x)t-1,yt-1) or image points (x)t-1,yt+1) is white, while the image point (x)t+1,yt) Is blue, image point (x)t+1,yt-1) or image points (x)t+1,yt+1) is a blue color, and,the k value is added by 1 to obtain the edge point CtIs judged as a candidate edge point EkAnd will point EkIs set to 255, otherwise point E is setkSetting the gray value of (4) to 0, and entering the substep (5.1.7);
(5.1.7) if t<T, increasing T by 1, and re-entering the substep (5.1.2); otherwise, ending the rough selection process, and making K equal to K, thereby obtaining K candidate edge points EkAnd roughly selected image I5，k＝1,2,3…,K；
(5.2) a selection process of determining the effective point set: according to the characteristic that the character edge points in the license plate area are dense, K candidate edge points E are subjected tokScreening to obtain selected image I6And the effective point set, K is 1,2,3 …, K, and the specific sub-steps are as follows:
(5.2.1) initializing k ═ 1;
(5.2.2) at the edge point EkScanning in a 7 × 7 area around the center, and if other candidate edge points are detected, retaining the edge point EkOtherwise, the point E is pointed outkSetting the gray value to zero;
(5.2.3) if K is less than K, increasing the value of K by 1, and returning to the substep (5.2.2); otherwise, ending the selection process to obtain the selected image I6And a set of valid points; if the effective point set is an empty set, returning to the step (1); otherwise, step (6) is executed.
4. The vehicle bottom shadow positioning license plate-based road rescue equipment square induction method according to claim 3, characterized in that: in the step (6), the effective point sets obtained in the step (5) are connected by adopting a morphological method to obtain connected regions with different sizes and shapes, and screening is carried out according to the area and length characteristics of the license plate region to obtain the license plate effective region and position information, and the specific steps are as follows:
(6.1) for the image I selected in the step (5)6Processing by morphological method to obtain J connected regions F with different sizes and shapesjJ is 1,2,3 …, and J denotes a connected region FjJ is the total number of connected regions, and the specific sub-steps are as follows:
(6.1.1) selecting rectangular structural elements with the size of 3 multiplied by 5 for the image I6Performing four times of expansion operation, connecting the effective point sets, communicating a license plate region, and then performing four times of corrosion operation on the expanded image by using rectangular structural elements with the same size to keep the size of the license plate region unchanged;
(6.1.2) selecting rectangular structural elements with the size of 3 multiplied by 3, carrying out open operation on the image after corrosion operation, further eliminating small noise areas in the non-license plate area, and obtaining J communicated areas F with different sizes and shapesj，j＝1,2,3…,J；
(6.2) according to the area characteristics of the license plate region, carrying out alignment on J connected regions FjScreening J to 1,2,3 … and J to obtain W license plate candidate regions DwW is 1,2,3 …, and W represents the license plate candidate region DwW is a license plate candidate region DwThe total number of (c); then, according to the length characteristics of the license plate regions, the W license plate candidate regions D are subjected towScreening, wherein W is 1,2,3 …, and W, obtaining the license plate effective area gamma and the abscissa x of the license plate center relative to the overall image coordinate system OXYmidThe specific substeps are as follows:
(6.2.1) initializing j ═ 1, w ═ 0;
(6.2.2) if the region F is connectedjThe conditions are satisfied:
then go to substep (6.2.3); otherwise, go to substep (6.2.4), in which
Indicates a connected region FjArea of (S)thFIndicates the region FjAn area threshold of (d);
(6.2.3) increasing the value of w by 1 and FjJudged as a license plate candidate region DwThen, the candidate region D of the license plate is obtainedwIs long
Go to substep (6.2.4) in which
xmaxw、xminwRespectively are license plate candidate regions DwThe maximum and minimum of the median image point with respect to the abscissa of the overall image coordinate system OXY;
(6.2.4) if J < J, increasing J by 1 and restarting substep (6.2.2); otherwise, let W equal W, go to substep (6.2.5);
(6.2.5) if W is 0, indicating that no license plate candidate area is obtained, and returning to the step (1); otherwise, in W license plate candidate areas DwIs selected to be the maximum value
Corresponding connected region DwmaxAs the license plate effective area gamma, and obtaining the minimum value x of the abscissa of the license plate effective area relative to the overall image coordinate system OXYDminMaximum value of abscissa is xDmaxThe abscissa x of the license plate center relative to the overall image coordinate system OXYmidIs (x)Dmin+xDmax)/2。
5. The vehicle bottom shadow positioning license plate-based road rescue equipment square induction method according to claim 4, characterized in that: in the step (7), the abscissa x of the center of the license plate of the accident vehicle obtained in the step (6) is usedmid and image I0Comparing the horizontal coordinates of the centers, giving a direction prompt in real time, and inducing a driver to perform backing operation: relative to the overall image coordinate system OXY, if the horizontal coordinate x of the center of the license plate of the accident vehiclemidSmaller than image I0The abscissa of the center, the center of the license plate in the image is on the left side of the center of the image, and then the driver is prompted to back up to the right; otherwise, the driver is prompted to back left, so that the support arms on the two sides are aligned with the two front wheels of the accident vehicle, the accident vehicle is further clamped and fixed, and the accident vehicle is pulled and pulled away.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810915174.9A CN109145797B (en) | 2018-08-13 | 2018-08-13 | Square induction method for road rescue equipment based on vehicle bottom shadow positioning license plate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810915174.9A CN109145797B (en) | 2018-08-13 | 2018-08-13 | Square induction method for road rescue equipment based on vehicle bottom shadow positioning license plate |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109145797A CN109145797A (en) | 2019-01-04 |
| CN109145797B true CN109145797B (en) | 2021-06-25 |
Family
ID=64792636
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810915174.9A Active CN109145797B (en) | 2018-08-13 | 2018-08-13 | Square induction method for road rescue equipment based on vehicle bottom shadow positioning license plate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109145797B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009095B (en) * | 2019-03-04 | 2022-07-29 | 东南大学 | Road driving area efficient segmentation method based on depth feature compressed convolutional network |
| CN110298247B (en) * | 2019-05-23 | 2022-04-12 | 东南大学 | Road rescue equipment square towing induction method based on deep learning |
| CN111145184B (en) * | 2019-06-12 | 2022-07-29 | 东南大学 | Dragging process safety early warning method based on multi-scale feature fusion |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104590372A (en) * | 2013-10-31 | 2015-05-06 | 福特全球技术公司 | Method and system for monitoring placement of a target on a trailer |
| CN106293385A (en) * | 2015-06-23 | 2017-01-04 | 通用汽车环球科技运作有限责任公司 | There is the connection auxiliary of pan/zoom and virtual top view |
| CN107133594A (en) * | 2017-05-11 | 2017-09-05 | 南宁市正祥科技有限公司 | The detection method of front side moving vehicle |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130226390A1 (en) * | 2012-02-29 | 2013-08-29 | Robert Bosch Gmbh | Hitch alignment assistance |
| CN103279755A (en) * | 2013-03-25 | 2013-09-04 | 东莞中国科学院云计算产业技术创新与育成中心 | Vehicle bottom shadow characteristics-based rapid license plate positioning method |
| US9696723B2 (en) * | 2015-06-23 | 2017-07-04 | GM Global Technology Operations LLC | Smart trailer hitch control using HMI assisted visual servoing |
| CN105825589A (en) * | 2016-05-20 | 2016-08-03 | 深圳大学 | Electric vehicle wireless charging alignment system and method based on vehicle license plate recognition |
| CN108229248A (en) * | 2016-12-14 | 2018-06-29 | 贵港市瑞成科技有限公司 | Vehicle checking method based on underbody shade |
-
2018
- 2018-08-13 CN CN201810915174.9A patent/CN109145797B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104590372A (en) * | 2013-10-31 | 2015-05-06 | 福特全球技术公司 | Method and system for monitoring placement of a target on a trailer |
| CN106293385A (en) * | 2015-06-23 | 2017-01-04 | 通用汽车环球科技运作有限责任公司 | There is the connection auxiliary of pan/zoom and virtual top view |
| CN107133594A (en) * | 2017-05-11 | 2017-09-05 | 南宁市正祥科技有限公司 | The detection method of front side moving vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109145797A (en) | 2019-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109784344B (en) | Image non-target filtering method for ground plane identification recognition | |
| CN102509098B (en) | Fisheye image vehicle identification method | |
| CN110414355A (en) | The right bit sky parking stall of view-based access control model and parking stall line detecting method during parking | |
| CN103077384A (en) | Method and system for positioning and recognizing vehicle logo | |
| CN109145797B (en) | Square induction method for road rescue equipment based on vehicle bottom shadow positioning license plate | |
| CN102663760B (en) | Location and segmentation method for windshield area of vehicle in images | |
| CN104036262B (en) | A kind of method and system of LPR car plates screening identification | |
| Li et al. | Nighttime lane markings recognition based on Canny detection and Hough transform | |
| CN105488454A (en) | Monocular vision based front vehicle detection and ranging method | |
| CN104050450A (en) | Vehicle license plate recognition method based on video | |
| CN102938057B (en) | A kind of method for eliminating vehicle shadow and device | |
| CN103927548B (en) | Novel vehicle collision avoiding brake behavior detection method | |
| CN102419820A (en) | Method for rapidly detecting car logo in videos and images | |
| CN109285169B (en) | Road rescue equipment side direction towing induction method based on wheel identification | |
| CN202134079U (en) | Unmanned vehicle lane marker line identification and alarm device | |
| CN106887004A (en) | A kind of method for detecting lane lines based on Block- matching | |
| CN108256521A (en) | For the effective coverage localization method of body color identification | |
| CN107256633B (en) | Vehicle type classification method based on monocular camera three-dimensional estimation | |
| CN103324958B (en) | Based on the license plate locating method of sciagraphy and SVM under a kind of complex background | |
| CN109635737A (en) | Automobile navigation localization method is assisted based on pavement marker line visual identity | |
| CN109376733B (en) | Square towing induction method for road rescue equipment based on license plate positioning | |
| CN106022268A (en) | Identification method and device of speed limiting sign | |
| CN106407951A (en) | Monocular vision-based nighttime front vehicle detection method | |
| Wang et al. | License plate location algorithm based on edge detection and morphology | |
| CN106203273A (en) | The lane detection system of multiple features fusion, method and senior drive assist system |
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 |