CN110163166B - Robust detection method for LED lighting lamp of highway tunnel - Google Patents
Robust detection method for LED lighting lamp of highway tunnel Download PDFInfo
- Publication number
- CN110163166B CN110163166B CN201910443626.2A CN201910443626A CN110163166B CN 110163166 B CN110163166 B CN 110163166B CN 201910443626 A CN201910443626 A CN 201910443626A CN 110163166 B CN110163166 B CN 110163166B
- Authority
- CN
- China
- Prior art keywords
- tunnel
- led
- area
- line segment
- point
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 claims abstract description 35
- 230000011218 segmentation Effects 0.000 claims abstract description 19
- 238000004458 analytical method Methods 0.000 claims abstract description 15
- 238000000605 extraction Methods 0.000 claims abstract description 4
- 238000005070 sampling Methods 0.000 claims abstract description 4
- 238000001914 filtration Methods 0.000 claims description 10
- 230000003044 adaptive effect Effects 0.000 claims description 8
- 238000009499 grossing Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 241001127637 Plantago Species 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 238000009434 installation Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 9
- 238000013461 design Methods 0.000 abstract description 5
- 230000008569 process Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000003292 glue Substances 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- DGAQECJNVWCQMB-PUAWFVPOSA-M Ilexoside XXIX Chemical compound C[C@@H]1CC[C@@]2(CC[C@@]3(C(=CC[C@H]4[C@]3(CC[C@@H]5[C@@]4(CC[C@@H](C5(C)C)OS(=O)(=O)[O-])C)C)[C@@H]2[C@]1(C)O)C)C(=O)O[C@H]6[C@@H]([C@H]([C@@H]([C@H](O6)CO)O)O)O.[Na+] DGAQECJNVWCQMB-PUAWFVPOSA-M 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- HODFCFXCOMKRCG-UHFFFAOYSA-N bitolterol mesylate Chemical compound CS([O-])(=O)=O.C1=CC(C)=CC=C1C(=O)OC1=CC=C(C(O)C[NH2+]C(C)(C)C)C=C1OC(=O)C1=CC=C(C)C=C1 HODFCFXCOMKRCG-UHFFFAOYSA-N 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000003063 flame retardant Substances 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 229910001507 metal halide Inorganic materials 0.000 description 1
- 150000005309 metal halides Chemical class 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003973 paint Substances 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 229910052708 sodium Inorganic materials 0.000 description 1
- 239000011734 sodium Substances 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- 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
Abstract
The invention discloses a robust detection method for a road tunnel LED lighting lamp. And taking the actual working real video sequence as input, and carrying out expansion analysis around the built-in design geometric structure of the road environment in the tunnel. Firstly, segment extraction based on an LSD operator and robust detection of the front blanking point are completed. And secondly, constructing a new coordinate system by taking the front blanking point as an original point, and accurately segmenting the dark area at the top of the tunnel. Based on the method, an optimal segmentation threshold value is reasonably selected by adopting a self-adaptive histogram analysis method, and the binarization robust segmentation of the LED lamp area is realized. And finally, based on the collinear assumption of the tunnel LED lamps, a random sampling consistency algorithm is adopted to robustly fit two straight lines formed by arranging the left and right LED lamps in the tunnel, so that the method not only plays a particularly prominent role in shielding highlight interference, but also fully obtains the arrangement sequence, the central point coordinates and the area size of the LED lamp regions. The method can robustly detect the missing LED lighting lamp under the complex condition in the tunnel of the highway, and has important significance for efficient acceptance check and daily maintenance of tunnel lighting.
Description
Technical Field
The invention belongs to the technical field of tunnel engineering safety protection, and provides a new basic technical means for efficient acceptance and daily maintenance of a tunnel lighting system by utilizing a computer video intelligent processing technology to automatically analyze vehicle-mounted video data of a road tunnel LED lighting lamp and realizing robust detection of the LED lamp aiming at complex scenes in a tunnel.
Background
The road tunnel lighting system is an important infrastructure for guaranteeing the driving safety and comfort in the tunnel. Compared with traditional Light source types such as metal halide lamps and high-pressure sodium lamps, Light-emitting Diode (LED) lamps have the characteristics of continuous adjustability, selectable color temperature, better color rendering property, energy conservation, environmental protection, long service life, small volume, low maintenance cost and the like, and can fully meet the requirements of creating a good Light environment in a tunnel and effectively operating and managing.
In the process of acceptance and operation and maintenance of the tunnel LED lighting system, the deficiency of the LED lighting lamps undoubtedly has important influence on constructing a good light environment in the tunnel. However, the tunnel lighting acceptance and maintenance technology is backward, so that the problems of high labor input cost and the like are inevitable. Current work acceptance is simply a one-time illumination measurement for a selected few road surface areas within the tunnel. In the daily maintenance process of the LED lamps in the highway tunnel, more modes of manual inspection and registration are adopted. The walking speed can be only determined by the inspector, and the work intensity of the inspector is huge. Real images of actual work of the tunnel LED lighting lamp in the large-scale highway area are rapidly acquired through the vehicle-mounted video system, and automatic analysis is performed by using a computer video intelligent processing technology, so that the efficiency of acceptance and operation maintenance of the tunnel LED lighting lamp can be effectively improved.
The accurate position and area size of the LED lamp area are obtained, so that the missing analysis and detection of the tunnel lighting lamp are completed, and the automatic segmentation of the LED lighting lamp area is realized by using a video image intelligent processing technology. According to the requirements of tunnel interior design and construction process, the interior full-section of the tunnel is sprayed with deep-color fire-retardant coating. Meanwhile, in order to ensure the reflectivity of the wall surface, milky wall paint is coated on the side wall of the tunnel within 3 m of height. The tunnel roof area occupying a sufficiently large area is all dark (but not fully black). The LED lighting lamps are arranged in the dark color area on the top of the tunnel according to a set rule, and each LED lamp normally emits light to form a high-brightness area. The method has the advantages that the gray difference between the high-brightness area of the LED lamp as the foreground and the dark background area at the top of the tunnel is large. However, the internal environment of the road tunnel is complex, and includes dynamic variable signs, electronic signs, LED strobe lights for capturing and supplementing light by a digital camera, and other tail lights of vehicles running around, which are used for monitoring and safety facilities with high brightness. Therefore, accurate detection and positioning of the LED lamp area are difficult to achieve only by using a simple image threshold segmentation method.
Disclosure of Invention
The invention aims to overcome various high-brightness interferences of complex scenes in a road tunnel, provides a vehicle-mounted video data intelligent analysis technology suitable for a road tunnel LED lighting lamp, and provides a robust method for quickly and accurately detecting the area of the LED lighting lamp around the geometric structure of the road environment in the tunnel, thereby practically providing a new basic technical means for efficient acceptance and daily maintenance of a tunnel lighting system.
In order to achieve the purpose, the invention specifically adopts the following technical scheme: the real video sequence of the actual work of the tunnel LED lighting lamp rapidly acquired by the vehicle-mounted video system is used as input, and the built-in design geometric structure surrounding the road environment in the tunnel is developed and analyzed. Segment extraction based on an LSD (line Segment detector) operator and robust detection of a front blanking point are completed firstly. After line segment characteristics reflecting geometric information of all composition structures in the tunnel are obtained by using an LSD operator, an effective parallel line set converged at a blanking point in front of a main vehicle is extracted through simple screening. Based on the detection result, robust detection of the front blanking point of the main vehicle is realized by adopting an M-estimator SAmple and consistency algorithm (MSAC). And secondly, constructing a new coordinate system by taking the front blanking point as an original point, detecting two key boundary lines of the tunnel top dark color region and the left and right white wall surface, and accurately segmenting the tunnel top dark color region. The invention adopts a self-adaptive histogram analysis method to reasonably select the optimal segmentation threshold value and realizes the binaryzation robust segmentation of the LED lamp area. And finally, analyzing the collinear assumption of the tunnel LED lamps on the basis of the front blanking point, and robustly fitting two straight lines formed by arranging the left and right LED lamps in the tunnel by adopting a RANdom SAmple Consensus (RANSAC) algorithm, thereby not only playing a particularly prominent role in shielding highlight interference, but also fully acquiring the arrangement sequence, the central point coordinate and the area size of the LED lamp area.
A robust detection method for a road tunnel LED lighting lamp is characterized by comprising the following steps:
step 1: robust detection of the plantago blanking point.
Step 1.1: and detecting line segments based on the LSD operator.
As an effective low-level feature of the image, the line segments can provide important information about the geometric content of the image. The highway tunnel is used as an artifact, and a regular geometric structure is presented on an internal environment main body, and the highway tunnel mainly comprises a front road surface, white side wall inner walls on two sides of a road, a top dark color area and the like. There are a number of line segment features in the scene, which are mainly reflected on the boundaries between the components, thereby providing the geometric structure information (the plane structure corresponding to each component) of the scene inside the tunnel.
Extracting a line segment set H representing the geometric structure information of the main body of the internal environment of the highway tunnel from the original tunnel LED illumination image by adopting an LSD algorithm, wherein s is a line segment in H and is defined by two end points a1And a2To define
s＝{a1,a2}＝{(x1,y1),(x2,y2)}
Wherein, a1＝(x1,y1) And a2＝(x2,y2) Respectively, the start and end points of s (in pixel coordinates).
Step 1.2: and screening the effective parallel line segment set.
There are multiple sets of parallel lines in the tunnel environment. The main vehicle front blanking point V ═ x (x) is formed by the dividing boundary line between the basic road in front of the main vehicle, the inner wall of the white side wall and the dark area on the top of the tunnel, and the dividing boundary line and the lane line on the road surface form a group of effective parallel lines, and the front blanking point V ═ x is formed on the image planev,yv). In order to avoid negative effects caused by line segment interference, the LSD detection line segment is simply screened to remove the front road space and the transverse parallel lines at the tail of other vehicles, so that corresponding effective parallel lines are extractedSet M of LSD segments (set of valid parallel segments). Let α be the inclination of the line segment s and β be the length of the line segment s, respectively calculated as follows:
the effective parallel line segments belonging to M need to satisfy the following two conditions at the same time. First, s needs to have a large tilt angle
αmin＜α＜αmax
Wherein alpha isminAnd alphamaxAnd a proper value range of the inclination angle of the effective parallel line segment is limited. Second, s needs to be of sufficient length
β＞βmin
Step 1.3: robust estimation of the plantago blanking point.
Let a1＝(x1,y1,1)T，a2＝(x2,y2,1)TAre respectively a point a1And a2The straight line defined by the line segment s is a1×a2＝(s1,s2,s3)TLet v be (x) the homogeneous coordinate of the front blanking point vv,yv,1)TThe straight line z ═ z1,z2,z3)TThe reference point r connecting the blanking point v and the line segment s, r is generally the midpoint of the line segment s, i.e. r ═ a1+a2) The distance d (s, v) from the front blanking point v to the line segment s can be defined as the absolute value of the sine of the angle θ between the straight line s and the straight line z:
constructing an optimized objective function
Adopting an MSAC method to carry out iterative solution, and when the following conditions are met, robustly estimating to obtain a front blanking point
Step 2: and (4) segmentation of a dark color area at the top of the tunnel and detection of a candidate area of the LED lamp.
Step 2.1: and (4) dividing the dark area at the top of the tunnel based on the front blanking point.
The key for realizing the accurate segmentation of the dark color area at the top of the tunnel is to robustly extract two key boundary lines b at the left and the right of the dark color area and the white side wall arearAnd bl. Based on the detection result of the front blanking point v, brAnd blThe structural relationship between them is highlighted: brAnd blAre respectively positioned at the left side and the right side of the front blanking point v and intersect with the front blanking point v.
For brAnd blThe detection process of (2) is divided into the following two steps. First, a coordinate system XOY is established in the image space with the anterior blanking point v as the origin. The active parallel line segments belonging to the set M can be converted to the coordinate system XOY according to:
{a1,a2and
respectively are two end points before and after the coordinate transformation of the effective parallel line segment s. In the new coordinate system XOY, each valid parallel line segment s falls into only one single quadrant: brWill fall into the fourth quadrant only, and blWill only fall into the third quadrant. Secondly, in effectIn the set M of parallel line segments, brAnd blA series of conditions as follows needs to be satisfied simultaneously:
1)bland brBoth end points of (2) are less than 0, blBoth end points of (b) are less than 0 on the abscissarThe opposite is true;
2)blthe dip angle falls within the range 0,90, and brThe inclination angle of (c) falls within the range {90,180 };
3)blor brShould be the LSD detection line segment with the greatest length in the respective quadrant.
Once the two critical boundaries b have been extractedrAnd blOne end (close to the front blanking point v) of the front panel can be directly extended to the front blanking point v, the other end (far from the front blanking point v) of the front panel can be extended outwards to intersect with the edge of the image space, and the intersection points are respectively set as v5,v2. Defining the upper left edge of an image space as v4And the upper right edge is v3And re-representing the front blanking point v as v1. From { v1,v2,v3,v4,v5And fifthly, connecting the points end to end in sequence to form a polygon R, which is the detection result of the dark color area at the top of the tunnel.
Step 2.2: and detecting candidate areas of the LED lamp.
Because the pixel scale distribution of the high-brightness area of the LED lamp as the foreground and the dark background area at the top of the tunnel is seriously uneven, the invention designs a method for self-adaptive threshold analysis according to the image gray histogram. The method considers a gray histogram h0Each peak of (t) corresponds to a different uniform region of the image, and there is a valley bottom between two adjacent peaks. The valley gray value is the optimal threshold for distinguishing different regions. For the original gray histogram h0(t) the problem of fluctuation, in order to find suitable peaks and valleys, the algorithm first uses a Gaussian kernel g (u) over h0(t) smoothing
Wherein the content of the first and second substances,
[·]indicating an integer truncation operation. W ≧ 3 is the window size of the Gaussian kernel function g (u), which determines the degree of smoothing. The larger the window of the Gaussian kernel function is, the larger the obtained smooth curve h1The flatter (t) is. Then, the gradation value is traversed from the highest gradation value 255 (certainly, peak value) to a direction in which the gradation value is smaller until h1(t) increase again
h1(T1)＜h1(T1-1)and h1(T1)＜h1(T1+1)
The valley bottom gradation value T thus obtained1The gray level threshold of the LED lamp is effectively detected.
Subjecting the grey value and T of each pixel in the deep color region R at the top of the tunnel1In comparison, R can be divided into two parts: high gray scale value region Rh and low gray scale value region RlRespectively defined as:
Rh＝{(x,y)∈M|I(x,y)≥T1}
Rl＝{(x,y)∈M|other}
wherein R ishThe detection result of the LED lighting lamp in the R is reflected. For RhThrough connected component analysis, the LED luminaire candidate region can be described as:
Ψ(1)＝{R1,R2,…,RL}
wherein R isjIs a high gray value region RhJ ═ 1,2, …, L.
Step 2.3: LED luminaire candidate area filtering
For any candidate region RjPerforming area filtering based on two conditions of size and aspect ratio so as to effectively shield much highlight interference of a complex tunnel scene;
the region filtering method specifically comprises the following steps:
first, culling smaller than a given area threshold σminAnd an LED lamp area ofUsing the maximum possible area threshold σmaxLimitation is made, namely the size of the candidate area of the LED lamp can be limited to
σmin＜σj＜σmax
Wherein σjIs the LED lamp region RjIs particularly represented by RjThe number of pixels contained;
secondly, calculating the candidate region R of the given LED lamp according to the minimum circumscribed rotating rectanglejWidth W ofjAnd height HjThereby increasing the aspect ratio Wj/HjIs limited to
ωmin＜Wj/Hj＜ωmax
Wherein, ω isminAnd ωmaxFor two parameters set in advance, and ωminωmax＝1；
And step 3: and (4) performing collineation hypothesis analysis on the tunnel LED lamps and fitting the LED lamp arrangement straight line.
Whether the LED lamps are collinear or not is an important condition which needs to be met by the LED lamp. According to the construction specifications of tunnel lighting lamps, the installation positions of the LED lamps need to meet certain requirements, and the LED lamps are generally arranged correspondingly according to the line type (straight line type or curve line type) of a road. The LED light fixture may take on a linear or curvilinear shape. Only those LED luminaires closest to the subject vehicle are considered, which not only have the highest resolution in the image, but also satisfy the assumption of a straight line type. Even for a curved road.
The tunnel LED lighting lamp is generally regularly arranged by dividing left and right sides to form a left straight line and a right straight line klAnd krRespectively next to two critical boundary lines blAnd br. Aiming at the self-adaptive detection and region filtering results of the LED illuminating lamp, the random sampling consistency algorithm (RANSAC) is adopted to robustly fit the two straight lines klAnd kr. Definition of κlAnd kapparAre respectively a straight line klAnd krThe slope of (a). When the RANSAC algorithm randomly selects the foreground points to estimate the initial fitting straight line, the invention limits the slope of the straight lineκlAnd kapparA certain range of values must be met to shield much of the highlight interference:
κ1min＜κl＜κ1max
κ2min＜κr＜κ2max
moreover, the RANSAC algorithm calculates the size of the number of support sets, which is favorable for the fitted straight line klAnd krPassing through the center point of the LED fixture area.
According to the basic graph scan conversion algorithm of computer graphics, along the left and right two fitting straight lines klAnd krSequentially determining a set of each pixel in an image space, and recording the passing LED lamp detection area, thereby obtaining the k-th position of each LED lamplAnd krThe installation and arrangement sequence of the components. Along klAnd krWhen pixel scan conversion is performed, it is necessary to perform the pixel scan conversion in a predetermined order. For klIn other words, the sequence from top left to bottom right needs to be selected. For krIn other words, the sequence from top right to bottom left needs to be selected. Thereby ensuring that all the LED lamp detection results are respectively fitted with straight lines (k)lOr kr) The above are arranged in order from near to far (relative to the subject vehicle), which can be specifically expressed as
Ψ(2)＝{Q1,Q2,…,QN}
Wherein Q isi＝{i,ci,σiIs a triplet, i 1,2, …, N. Triple QiThe elements are the arrangement sequence i of the LED lamp area and the central point coordinate ciAnd the size of the area σi。
Compared with the prior art, the invention has the following obvious advantages:
(1) the invention provides a robust method for rapidly and accurately detecting the LED lighting lamp area by adopting vehicle-mounted video data, and the robust method has important significance for efficient delivery acceptance and daily maintenance of tunnel lighting.
(2) The method adopts the robust detection method of the front blanking point based on the LSD line segment and the dark color area at the top of the tunnel, thereby not only avoiding the negative influence caused by the interference line segment, but also fully developing and analyzing the geometrical structure surrounding the road environment in the tunnel. The detection performance of the LED lamp area is improved aiming at various high-brightness interferences of complex scenes in the road tunnel.
(3) The invention adopts a self-adaptive histogram analysis method to reasonably select the optimal segmentation threshold value, realizes the binaryzation robust segmentation of the high-brightness region of the LED lamp, and solves the problems of accurate detection and positioning of the LED lamp region under the condition that the pixel number distribution of the foreground and background regions of the LED lamp is seriously uneven.
(4) The method analyzes the collinear assumption of the tunnel LED lamps on the basis of the front blanking point, robustly fits two straight lines formed by arranging the left and right LED lamps in the tunnel hole by adopting a random sampling consistency algorithm, plays a particularly prominent role in shielding highlight interference, and fully acquires the arrangement sequence, the central point coordinate and the area size of the LED lamp area.
Drawings
FIG. 1 is a general block diagram of a method to which the present invention relates;
fig. 2a-2f are the LED lighting vehicle video and the complex environment inside the tunnel in the gulf of Qingdao, glue: (a) an inlet reinforcement section; (b) an inlet section; (c) a transition section; (d) a middle section; (e) electronic indication boards for road passing through the holes; (f) the LED strobe light is used for snapshot supplementary lighting of the digital camera;
3a-3c front blanking point detection based on LSD line segments: (a) original images of the road in front of the vehicle-mounted video; (b) an LSD line segment; (c) effective parallel line segments and front blanking points;
FIG. 4 is based on a new coordinate system and a four quadrant distribution of the vehicle forward blanking points;
5a-5c robust detection of dark areas at the top of tunnels: (a) two key boundary lines between the dark color area and the white side wall area on the top of the tunnel; (b) the polygon range of the dark color area at the top of the tunnel; (c) a dark area at the top of the tunnel;
6a-6e LED luminaire candidate region detection based on adaptive threshold segmentation: (a) a dark area at the top of the tunnel; (b) based on the Otsu classical method; (c) the method provided by the invention; (d) an original gray level histogram; (e) primary Gaussian smoothing of the original histogram;
fig. 7a-7d RANSAC-based LED lamp alignment line fitting: (a) LED lamp candidate areas; (b) performing linear fitting based on RANSAC; (c) filtering results of the LED lamp area;
fig. 8 is a slope histogram of a line in which LED lamps are arranged, which is fitted by using RANSAC algorithm:
fig. 9 tunnel LED lighting lamp missing detection process and corresponding results: (a) a set of valid parallel line segments in front of the vehicle and a forward blanking point; (b) a dark area at the top of the tunnel; (c) LED luminaire candidate regions segmented based on adaptive thresholds; (d) post-processing and straight line fitting of the candidate area of the LED lamp; (e) and (5) dividing the LED lamp into results.
Detailed Description
The present invention will be further described with reference to the accompanying drawings and examples.
The embodiment of the invention is realized on a PC (personal computer) provided with VC2013 and OpenCV3.4, and the general block diagram is shown in figure 1. The tunnel LED lighting vehicle-mounted video used in the embodiment of the invention is shot in 7 months in 2017, and is shot by a photographer through a smart phone when the photographer passes through a sea-bottom tunnel in the gulf of Islands, and is shown in FIG. 2. The video records the complete driving process of the main vehicle in the tunnel, so that the road condition in the front of the vehicle at a distance can be observed, and the video also comprises LED illuminating lamps arranged on the left side and the right side of the top of the tunnel. The sea-bottom tunnel of the gulf of Qingdao glue is designed for 6 bidirectional lanes, and the daily average traffic is up to 5.3 thousands (2016 statistical data). The tunnel has a total length of 7800 meters, and the lighting in the tunnel is divided into a plurality of lighting sections, particularly comprising an entrance reinforcing section (see fig. 2a), an entrance section (see fig. 2b), a transition section (see fig. 2c) and a middle section (see fig. 2 d). The invention mainly aims at the robust detection of the LED lamp in the middle lighting section. Because the intermediate lighting section occupies the longest distance inside the tunnel, the most LED lighting fixtures are installed (about 3000 on a single side of the tunnel), and it takes a long time for the vehicle to pass through. Moreover, the LED lamps in the middle lighting section do not need to be arranged densely, and each LED lamp can be easily and effectively segmented on the image. The frame rate of the video is 25 frames per second, the image resolution is 720 x 404, and the total duration is 8 minutes and 48 seconds. It can be seen that the internal environment of the road tunnel is complex, including dynamic variable information signs (issuing warning information), electronic signs, etc., which present highlighted monitoring and safety facilities, as shown in fig. 2e and 2 f. In the traffic rush hour when taking a bus through the tunnel, the front and rear lights of other vehicles running around can also cause more serious influence.
The method provided by the embodiment of the invention specifically comprises the following steps:
step 1: robust detection of plantago blanking points
Fig. 3 shows a robust detection process of the pre-amble point v. Fig. 3a is a tunnel lighting vehicle-mounted video original image. Fig. 3b shows the detection result of the tunnel inner line segment based on the LSD operator (indicated by the red thin line). When an effective parallel line segment set M in front of a vehicle is screened, the value range of the inclination angle alpha is set to [ pi/60, pi/3 ], so that transverse parallel interference line segments are effectively eliminated. α is in units of radians (rad). The length β ranges from [30, ∞) in pixels. Figure 3c shows the set of screened valid parallel line segments (shown by the red bold lines). It is clearly seen that M is primarily present at the boundaries between the components of the tunnel interior design. Fig. 3c also shows the detection result of the vehicle front blanking point v, which is indicated by the black dot with red background at the middle position of each image.
Step 2: and (4) segmentation of a dark color area at the top of the tunnel and detection of a candidate area of the LED lamp.
Fig. 4 shows the establishment of a new coordinate system XOY in image space with the anterior blanking point v as the origin. Wherein the arrow direction indicates the positive direction of the coordinate axis. It is clear that in the new coordinate system XOY each valid parallel line segment s falls into only one single quadrant and in the two key left and right boundaries between the dark area at the top of the tunnel and the white sidewall wall, brWill fall into the fourth quadrant only, and blWill only fall into the third quadrant. In the new coordinate system XOY, two key boundary lines b can be accurately extracted from the effective parallel line segment set M through a series of simple conditionsrAnd blAs indicated by the thick blue line in fig. 5 a. FIG. 5b shows the key combination of two keysLine of demarcation brAnd blAnd obtaining the polygonal range of the dark area at the top of the tunnel. Accordingly, fig. 5c shows the detection result of the dark area on the top of the tunnel.
FIG. 6 illustrates a process for detecting LED luminaire candidate regions using an adaptive threshold segmentation algorithm. For the dark region R at the top of the tunnel shown in fig. 6a, the embodiment of the present invention first tries the Otsu classical segmentation method based on the optimal threshold extraction. However, the LED luminaire detection effect based on the Otsu segmentation method is not ideal (see fig. 6 b). This is because the pixel size distribution of the LED lamp area and the dark background area on the top of the tunnel is not uniform, see fig. 6d showing the gray histogram h of the dark area R on the top of the tunnel0(t)。h0(t) specifically indicates the number of pixels (ordinate) belonging to the gray value t (abscissa) in the original image R, where t is 0,1, …, 255. The high-brightness LED lamp area belongs to h0(t) is a region near the maximum gray value of 255, and a peak is formed at t 255. H is continuously reduced along with the gray value t0(t) decreases rapidly until h0(t) increases again, h over a longer segment of the gray scale range0(t) are relatively low. This reflects the large gray scale separation between the high light area of the LED fixture and the dark background area at the top of the tunnel. Fig. 6c shows the LED lighting fixture detection results obtained by the designed adaptive threshold analysis method, which can achieve accurate detection of LED fixture regions. In this adaptive threshold based image segmentation algorithm, the original gray histogram h is targeted0(t) problem of fluctuation, the algorithm uses Gaussian kernel function g (u) to h0(t) smoothing. The larger the window of the Gaussian kernel function is, the larger the obtained smooth histogram curve h1The flatter (t) is, as shown in fig. 6 e. The gaussian kernel function g (u) has a window size W of 3, and g (u) is specifically g (-1) ═ 0.2261, g (0) ═ 0.5478, and g (1) ═ 0.2261. T obtained based on adaptive threshold segmentation algorithm1The value range is [240,249 ] under constant change]. After the candidate regions of the LED lamp are selected, any candidate region R needs to be selectedjArea filtering based on two conditions of size and aspect ratio is carried out so as to effectively shield much highlight interference of a complex tunnel scene. This exampleIn, σjIs in the value range of [30,1000]Aspect ratio Wj/HjWith a limit of [1/3,3]。
And step 3: collineation hypothesis analysis of tunnel LED lamps and LED lamp arrangement straight line fitting
The collineation assumption of the LED lamps emphasized by the invention can be practically satisfied, and the collineation assumption plays a particularly prominent role in shielding highlight interference aiming at various complex conditions in a tunnel hole. Aiming at LED candidate areas (see figure 7a), in order to accurately extract LED lighting lamps in a highway tunnel, the RANSAC algorithm is used for fitting straight lines (k) formed by the LED lamp areas arranged on the left side and the right side in the tunnel holelAnd kr) As indicated by the two green segments in fig. 7 b. When the RANSAC algorithm randomly selects foreground points to estimate an initial fitting straight line, the method limits the slope kappa of the straight linelAnd kapparA certain value range must be satisfied to shield many high-brightness interferences of complex scenes inside the tunnel (see fig. 7 c).
The limit on the slope range for the line to be fitted cannot be determined simply empirically (observation by multiple experiments). Aiming at the candidate area of the LED lamp, a plurality of straight lines can be fitted by using RANSAC algorithm, the slope of the straight lines is always changed greatly, and the method is especially suitable for the condition that strong interference exists. The invention makes statistics of the distribution rule of all fitted straight line slopes obtained by RANSAC algorithm, such as a slope histogram shown in FIG. 8. Wherein, 100 containers are selected, and the slope range of the straight line to be fitted is [ -1,2.5 ]. It can be seen that the slopes of all the lines to be fitted are distributed in multiple peaks, and the two peaks (corresponding to the two ranges of [1.15,1.8) and [ -0.76,0.5) in the positive and negative slope ranges) are the true reflection of the lines formed by the sequential arrangement of the LED lamps along the left and right sides, while the interference caused by the horizontal parallel lines is near the slope of 0.
Fig. 9 shows a processing procedure of automatic detection of the tunnel LED lamp and a corresponding result. In the sequence of columns, each column represents a typical scene of an intermediate lighting segment inside the tunnel, respectively a general case (see the first column) and an actual complex case (see the other two columns). In line, fig. 9a shows a set M of valid parallel line segments in front of the vehicle (indicated by red line segments) and the detection result of the front blanking point v (indicated by black dots with red bottom with the respective images in the middle position) after simple screening. Each image also shows two key boundaries (indicated by blue line segments) between the dark areas at the top of the tunnel and the white sidewall areas. Fig. 9b shows a robust detection result of a dark color region at the top of the tunnel, which effectively limits the range of images for detecting the LED lamp and greatly reduces the complexity of subsequent processing. Fig. 9c shows LED luminaire candidate region detection results based on adaptive threshold segmentation. It can be seen that the high brightness interference caused by various electronic monitoring and safety devices inside the tunnel has a serious influence on the accurate detection of the LED lamp. Fig. 9d shows the results of the line fitting for the LED luminaire candidate regions. The two green straight lines represent the results of the straight line fitting for the foreground region of the LED luminaire. These straight lines preferably pass through the center point of the LED area. Through the judgment of collinearity or not, a lot of highlight interference of a complex tunnel scene can be thoroughly shielded. Only the LED lamp areas which are closer to the main vehicle and have higher resolution and are sequentially arranged at the left side and the right side in the tunnel hole can be finally reserved. Fig. 9e shows the LED fixture split result. Even the LED lamp with a small light-emitting area can still realize accurate detection by the method.
The invention uses a 64-bit computer with 8-core Intel to strong CPU (main frequency 3.30GHz) and 16GB memory. The average processing time per frame of image is about 21.49 milliseconds. The algorithm implementation used in each step is not specifically performance optimized. The most time consuming steps are mainly the detection of the anterior blanking points and the RANSAC-based line fitting.
Finally, it should be noted that: the above examples are only intended to illustrate the invention and do not limit the technical solutions described in the present invention; thus, while the present invention has been described in detail with reference to the foregoing examples, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted; all such modifications and variations are intended to be included herein within the scope of this disclosure and the appended claims.
Claims (1)
1. A robust detection method for a road tunnel LED lighting lamp is characterized by comprising the following steps:
step 1: robust detection of plantago blanking points
Step 1.1: line segment detection based on LSD operator
Extracting a line segment set H representing the geometric structure information of the main body of the internal environment of the highway tunnel from the original tunnel LED illumination image by adopting an LSD algorithm, wherein s is a line segment in H and is defined by two end points a1And a2To define the number of the channels to be used,
s＝{a1,a2}＝{(x1,y1),(x2,y2)}
wherein, a1＝(x1,y1) And a2＝(x2,y2) Respectively representing the starting point and the end point of s, expressed in pixel coordinates;
step 1.2: screening of sets of valid parallel line segments
The LSD detection line segment is simply screened, negative effects caused by interference line segments are eliminated, and an LSD line segment set M of effective parallel lines is extracted, wherein the specific extraction method comprises the following steps:
let α be the inclination of the line segment s and β be the length of the line segment s, respectively calculated as follows:
the effective parallel line segments belonging to M need to satisfy the following two conditions simultaneously: first, s needs to have a large tilt angle
αmin＜α＜αmax
Wherein alpha isminAnd alphamaxThe inclination angle of the effective parallel line segment is limited to have a proper value range, and secondly, s needs to have enough length
β＞βmin
Step 1.3: robust estimation of plantago blanking points
Let a1＝(x1,y1,1)T，a2＝(x2,y2,1)TAre respectively a point a1And a2The straight line defined by the line segment s is a1×a2＝(s1,s2,s3)TLet v be (x) the homogeneous coordinate of the front blanking point vv,yv,1)TThe straight line z ═ z1,z2,z3)TConnecting the blanking point v and a reference point r on the line segment s, r is taken as the middle point of the line segment s, i.e. r ═ a1+a2) And 2, the distance d (s, v) from the front blanking point v to the line segment s is defined as the absolute value of the sine value of an included angle theta between the straight line s and the straight line z:
constructing an optimized objective function
Adopting an MSAC method to carry out iterative solution, and when the following conditions are met, robustly estimating to obtain a front blanking point
Step 2: tunnel top dark color region segmentation and LED lamp candidate region detection
Step 2.1: tunnel top dark region segmentation based on front blanking points
The left and right boundary lines of the dark color area and the white side wall area at the top of the tunnel are respectively brAnd blAnd b isrAnd blThe front blanking points are respectively positioned at the left side and the right side of the front blanking point v and are intersected with the front blanking point v;
for brAnd blFirstly, taking a front blanking point v as an origin, establishing a coordinate system XOY in an image space, and converting effective parallel line segments belonging to a set M into the coordinate system XOY according to the following formula:
{a1,a2and
respectively two end points before and after the coordinate transformation of the effective parallel line segment s; in the new coordinate system XOY, each valid parallel line segment s falls into only one single quadrant: brWill fall into the fourth quadrant only, and blWill fall into the third quadrant only; second, in the set of valid parallel line segments M, brAnd blA series of conditions as follows needs to be satisfied simultaneously:
1)bland brBoth end points of (2) are less than 0, blBoth end points of (b) are less than 0 on the abscissarThe opposite is true;
2)blthe dip angle falls within the range 0,90, and brThe inclination angle of (c) falls within the range {90,180 };
3)blor brIs the LSD detection line segment with the maximum length in the respective quadrant;
once the two critical boundaries b have been extractedrAnd blRespectively extending one end close to the front blanking point v directly, and extending one end far away from the front blanking point outwards to intersect with the edge of the image space, and respectively setting the intersection point as v5,v2(ii) a Defining the upper left edge of an image space as v4And the upper right edge is v3And re-representing the front blanking point v as v1(ii) a From { v1,v2,v3,v4,v5Five points are sequentially connected end to form a polygon R which is the detection result of the deep color area at the top of the tunnel;
step 2.2: LED luminaire candidate area detection
Performing self-adaptive threshold analysis on a detection result R of a deep color region at the top of the tunnel by adopting an image gray histogram to extract an LED lamp candidate region, wherein the gray histogram h0Each peak value of (t) corresponds to different uniform areas in the detection result R, and a valley bottom exists between two adjacent peaks, and the gray value of the valley bottom is the optimal threshold value for distinguishing different areas;
the adaptive threshold analysis method specifically comprises the following steps:
firstly, using Gaussian kernel function g (u) to obtain original gray histogram h of detection result R0(t) smoothing to obtain a smoothed gray histogram h1(t), the formula is as follows:
wherein the content of the first and second substances,
[·]expressing integer truncation operation, W ≧ 3 is the window size of the Gaussian kernel function g (u), determines the degree of smoothing, the larger the window of the Gaussian kernel function is, the resulting smoothing curve h1(t) flatter;
then, h is paired in the direction in which the gradation value is small from the highest gradation value1(t) traversing the gray values until h1(t) is increased again by the amount of,
h1(T1)＜h1(T1-1)and h1(T1)＜h1(T1+1)
wherein, T1Representing valley bottom gray values, namely effectively detecting gray threshold values of the LED lamp;
finally, the gray value and T of each pixel in the deep color region R at the top of the tunnel1In comparison, R is divided into two parts: high gray scaleValue region RhAnd a low gray value region RlRespectively defined as:
Rh＝{(x,y)∈R|I(x,y)≥T1}
Rl＝{(x,y)∈R|other}
wherein R ishRepresenting the detection result of the LED lighting fixture in R, and I (x, y) represents the gray value of the pixel (x, y) for RhThrough connected component analysis, the LED luminaire candidate region can be described as:
Ψ(1)＝{R1,R2,…,RL}
wherein R isjIs a high gray value region RhJ ═ 1,2, …, L;
step 2.3: LED luminaire candidate area filtering
For any candidate region RjPerforming area filtering based on two conditions of size and aspect ratio so as to effectively shield much highlight interference of a complex tunnel scene;
the region filtering method specifically comprises the following steps:
first, culling smaller than a given area threshold σminAnd using the maximum possible area threshold σmaxLimitation is made, namely the size of the candidate area of the LED lamp can be limited to
σmin＜σj＜σmax
Wherein σjIs the LED lamp region RjIs particularly represented by RjThe number of pixels contained;
secondly, calculating a given LED lamp candidate region R according to the minimum circumscribed rotating rectanglejWidth W ofjAnd height HjThereby increasing the aspect ratio Wj/HjLimited to omegamin＜Wj/Hj＜ωmax
Wherein, ω isminAnd ωmaxFor two parameters set in advance, and ωmin·ωmax＝1；
And step 3: tunnel LED lamp arrangement straight line fitting
On the basis of meeting the collineation assumption of the tunnel LED lamps, linear fitting is carried out on the arrangement of the LED lamps based on a blanking point v in front of the vehicle, and the method specifically comprises the following steps:
the arrangement of the tunnel LED lighting lamps forms a left straight line k and a right straight line klAnd krRespectively next to two critical boundary lines blAnd brAiming at the self-adaptive detection and region filtering results of the LED lighting lamp, a random sampling consistency algorithm RANSAC is adopted to robustly fit the two straight lines klAnd krDefinition of κlAnd kapparAre respectively a straight line klAnd krWhen the RANSAC algorithm randomly selects foreground points to estimate an initial fitting straight line, the slope of the straight line is limited to be kappalAnd kapparThe following conditions must be satisfied:
κ1min＜κl＜κ1max
κ2min＜κr＜κ2max
according to the basic graph scan conversion algorithm of computer graphics, along the left and right two fitting straight lines klAnd krSequentially determining a set of each pixel in an image space, and recording the passing LED lamp detection area, thereby obtaining the k-th position of each LED lamplAnd krThe installation and arrangement sequence of the components; along klAnd krWhen pixel scan conversion is performed, it is necessary to perform the pixel scan conversion in a predetermined order, and k is a predetermined number of timeslFor k, it is necessary to choose the order from top left to bottom rightrIn other words, the sequence from top right to bottom left needs to be selected, so that all the detection results of the LED lamps are ensured to be respectively fitted with straight lines klOr krAre arranged in order from near to far relative to the subject vehicle, which is specifically shown as
Ψ(2)＝{Q1,Q2,…,QN}
Wherein Q isi＝{i,ci,σiIs a triplet, i 1,2, …, N, triplet QiThe elements are the arrangement sequence i of the LED lamp area and the central point coordinate ciAnd the size of the area σi。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910443626.2A CN110163166B (en) | 2019-05-27 | 2019-05-27 | Robust detection method for LED lighting lamp of highway tunnel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910443626.2A CN110163166B (en) | 2019-05-27 | 2019-05-27 | Robust detection method for LED lighting lamp of highway tunnel |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110163166A CN110163166A (en) | 2019-08-23 |
| CN110163166B true CN110163166B (en) | 2021-06-25 |
Family
ID=67632717
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910443626.2A Active CN110163166B (en) | 2019-05-27 | 2019-05-27 | Robust detection method for LED lighting lamp of highway tunnel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110163166B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852252B (en) * | 2019-11-07 | 2022-12-02 | 厦门市美亚柏科信息股份有限公司 | Vehicle weight-removing method and device based on minimum distance and maximum length-width ratio |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678212A (en) * | 2014-12-05 | 2016-06-15 | 现代摩比斯株式会社 | Method and apparatus for tunnel decision |
| WO2016194296A1 (en) * | 2015-06-04 | 2016-12-08 | Sony Corporation | In-vehicle camera system and image processing apparatus |
| CN106372619A (en) * | 2016-09-20 | 2017-02-01 | 北京工业大学 | Vehicle robustness detection and divided-lane arrival accumulative curve estimation method |
| CN106960179A (en) * | 2017-02-24 | 2017-07-18 | 北京交通大学 | Rail line Environmental security intelligent monitoring method and device |
| CN109086671A (en) * | 2018-07-04 | 2018-12-25 | 同济大学 | One kind being suitable for unpiloted night traffic lane line video detecting method |
| CN109299674A (en) * | 2018-09-05 | 2019-02-01 | 重庆大学 | A kind of lane change detection method violating the regulations of the tunnel based on car light |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101362324B1 (en) * | 2012-06-05 | 2014-02-24 | 현대모비스 주식회사 | System and Method for Lane Departure Warning |
| CN104655047B (en) * | 2015-02-16 | 2018-02-02 | 武汉武大卓越科技有限责任公司 | The rapid integrated measuring system in tunnel |
| CN104809433B (en) * | 2015-04-21 | 2017-12-15 | 电子科技大学 | A kind of zebra line detecting method based on maximum stable region and stochastical sampling |
-
2019
- 2019-05-27 CN CN201910443626.2A patent/CN110163166B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678212A (en) * | 2014-12-05 | 2016-06-15 | 现代摩比斯株式会社 | Method and apparatus for tunnel decision |
| WO2016194296A1 (en) * | 2015-06-04 | 2016-12-08 | Sony Corporation | In-vehicle camera system and image processing apparatus |
| CN106372619A (en) * | 2016-09-20 | 2017-02-01 | 北京工业大学 | Vehicle robustness detection and divided-lane arrival accumulative curve estimation method |
| CN106960179A (en) * | 2017-02-24 | 2017-07-18 | 北京交通大学 | Rail line Environmental security intelligent monitoring method and device |
| CN109086671A (en) * | 2018-07-04 | 2018-12-25 | 同济大学 | One kind being suitable for unpiloted night traffic lane line video detecting method |
| CN109299674A (en) * | 2018-09-05 | 2019-02-01 | 重庆大学 | A kind of lane change detection method violating the regulations of the tunnel based on car light |
Non-Patent Citations (4)
| Title |
|---|
| Detection and Tracking the Vanishing Point on the Horizon;Young-Woo Seo;《SEMANTIC SCHOLAR》;20140531;第1-13页 * |
| The comprehensive efficiency analysis of tunnel lighting based on visual performance;Xiaoqin Zhang等;《Advances in Mechanical Engineering》;20170404(第4期);第1-9页 * |
| 一种双正交消隐点的双目相机标定方法;赵亚凤等;《液晶与显示》;20161031;第31卷(第10期);第958-966页 * |
| 基于消隐点的结构化场景的三维重建;高翔;《中国优秀硕士学位论文全文数据库信息科技辑》;20080615(第6期);第I138-417页第2.3节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110163166A (en) | 2019-08-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109784344B (en) | Image non-target filtering method for ground plane identification recognition | |
| TWI607901B (en) | Image inpainting system area and method using the same | |
| US20140125801A1 (en) | On-line tunnel deformation monitoring system based on image analysis and its application | |
| CN110210451B (en) | Zebra crossing detection method | |
| CN101872546A (en) | Video-based method for rapidly detecting transit vehicles | |
| CN107895151A (en) | Method for detecting lane lines based on machine vision under a kind of high light conditions | |
| CN109410205B (en) | Crack extraction method under complex pavement background | |
| US20200151912A1 (en) | Vehicle surroundings recognition apparatus | |
| CN110163166B (en) | Robust detection method for LED lighting lamp of highway tunnel | |
| CN111354047A (en) | Camera module positioning method and system based on computer vision | |
| JP6413318B2 (en) | Vehicle detection device, system, and program | |
| CN110427979B (en) | Road water pit identification method based on K-Means clustering algorithm | |
| JP7206909B2 (en) | Attached matter detection device and attached matter detection method | |
| CN113989771A (en) | Traffic signal lamp identification method based on digital image processing | |
| CN109800641A (en) | Method for detecting lane lines based on threshold adaptive binaryzation and connected domain analysis | |
| Xin et al. | Robust detection of lighting LEDs by analyzing the geometric structure of the tunnel interior environment in vehicle-mounted video sequences | |
| KR101183211B1 (en) | Apparatus for segmentation processing on image of gauge module | |
| CN111695374B (en) | Segmentation method, system, medium and device for zebra stripes in monitoring view angles | |
| CN112150828B (en) | Method for preventing jitter interference and dynamically regulating traffic lights based on image recognition technology | |
| CN100371944C (en) | Greyscale image partition method based on light distribution character of reflection or penetrance | |
| JP2015207212A (en) | Vehicle detection device and system, and program | |
| JP3333468B2 (en) | Roadway image processing device for autonomous running of vehicles | |
| JP3232064B2 (en) | Target area extraction method for color images | |
| CN110321828B (en) | Front vehicle detection method based on binocular camera and vehicle bottom shadow | |
| Xin et al. | Automatic detection of missing-LEDs of road tunnels in vehicle-mounted video sequences |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240112 Address after: 8th Floor, Century Xingyuan Building, No. 20 Anyuan Road, Chaoyang District, Beijing, 100020 Patentee after: Huajie Engineering Consulting Co.,Ltd. Address before: 100124 No. 100 Chaoyang District Ping Tian Park, Beijing Patentee before: Beijing University of Technology |