CN103679121A - Method and system for detecting roadside using visual difference image - Google Patents
Method and system for detecting roadside using visual difference image Download PDFInfo
- Publication number
- CN103679121A CN103679121A CN201210342374.2A CN201210342374A CN103679121A CN 103679121 A CN103679121 A CN 103679121A CN 201210342374 A CN201210342374 A CN 201210342374A CN 103679121 A CN103679121 A CN 103679121A
- Authority
- CN
- China
- Prior art keywords
- roadside
- region
- anaglyph
- module
- road
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
The invention provides a method for detecting a roadside using a visual difference image. The method comprises the following steps: estimating a roadside region on a visual difference image and removing noise, determining a roadside region by the relationship of road true width and roadside inclination angle on the visual difference image, and detecting the type and position of the roadside by judging the visual difference between related pixel pots on the visual difference image.
Description
Technical field
The present invention relates to a kind of new method of utilizing anaglyph fast detecting road roadside type.
Background technology
Along with popularizing of automobile, daily communications and transportation is more and more busier, and traffic safety becomes the focus that people pay close attention to day by day.Therefore, people wish that the vehicle that participates in traffic can possess more Intelligent Characteristics, thereby substitute a part of safe matters of human pilot judgement.Aspect of traffic safety relates to whether safety of the composed component of vehicle own, and another aspect relates to whether safety of environment that vehicle advances.For the latter, the situation of mainly understanding road surface by driver judges.Yet, driver in tired or note less than in the situation that, tend to cause danger.Therefore, people need a kind of method and system can in the situation that driver fails to judge, replace driver to carry out the judgement of some road safety situations, to evade danger.
US Patent No. 6813370B1 utilizes the monochrome information on gray-scale map and the range information that calculated by anaglyph detects the white line on road surface.This patent result that white line detects in previous frame obtains the white line surveyed area of this frame, and utilizes monochrome information and range information to detect white line.
US Patent No. 7346190B2 detects the traffic lines on road by a kind of method of projection mapping.The method is carried out projection by the point in the traffic lines region in anaglyph according to the direction vertical with road, obtains a histogram after projection, then detects the traffic lines on road according to the information on histogram.Although it has been used three-dimensional information and has detected, the projection histogram that has mainly utilized road vertical direction to obtain.When the height in road roadside is different (tilted road surface), this method is just not handy.
The method that above patent is described can not provide the method in a kind of fast detecting different kinds of roads roadside effectively.Utilize traditional road road boundary detection method of anaglyph need to use U disparity map and V disparity map.It first obtains position and each roadside to be selected straight line of end point on U disparity map, and then on V disparity map, calculates the roadside type under each roadside to be selected straight line.This method can obtain road boundary detection result very accurately, but it is very consuming time.Each pixel in anaglyph has been used to more than one time.The number of the roadside to be selected straight line staying on U figure, depends on threshold value.The judgement of every roadside straight line, all needs to calculate for 2 times on U disparity map and V disparity map.
Summary of the invention
Therefore, people need accurately a kind of and detect rapidly the method in roadside.This patent provides a kind of method, can fast detecting goes out type and the position in road roadside.The method can be used in the multiple application of vehicle-mounted vidicon.
The invention provides a kind of method that adopts anaglyph to detect roadside, comprising: in anaglyph, estimate region, road surface denoising; The corresponding relation in anaglyph by road actual width and angle of inclination, roadside, determines region, roadside; And by the parallax difference between related pixel point in judgement anaglyph, detect type and the position in roadside.
According to employing anaglyph of the present invention, detect the method in roadside, describedly in anaglyph, estimate that region, road surface denoising comprise: utilize boundary operator, on the basis of former disparity map, generate edge image; Road surface equation by estimating on V disparity map is partitioned into region, road surface in edge disparity map; And denoising module is utilized the monotonicity that on roadside, pixel is worth on edge disparity map, removal noise.
According to employing anaglyph of the present invention, detect the method in roadside, wherein said by road actual width and angle of inclination, roadside the corresponding relation in anaglyph determine that region, roadside comprises: by past n frame, road actual width and angle of inclination, the roadside corresponding relation in anaglyph, sets up region, roadside estimation model; Region, roadside estimation model and Common Prediction Method based on setting up, dope the road boundary detection region of present frame; And update module is by the road boundary detection result of present frame and the trusted area accumulated result of past m frame, upgrades region, roadside estimation model.
The method that detects roadside according to employing anaglyph of the present invention, wherein said region, the roadside estimation model of setting up comprises: select the area-of-interest on image to reduce computing time; Adopt certain computing method, by the location positioning of above-mentioned zone cathetus out; The corresponding relation in anaglyph by road actual width and roadside straight line, determines the region, roadside in area-of-interest; And utilize roadside extensibility to come region, growth source roadside.
The method that detects roadside according to employing anaglyph of the present invention, wherein said type and the position that detects roadside comprises: scan region, roadside one time, utilize the difference of parallax value between pixel, only leave the pixel on useful roadside; Adopt certain computing method, by the location positioning of above-mentioned zone cathetus out; And according to the roadside type of pixel on straight line, thereby the roadside type of definite whole piece straight line
According to another aspect of the present invention, provide a kind of system that adopts anaglyph to detect roadside, having comprised: anaglyph pretreatment module, in anaglyph, estimate region, road surface denoising; Region, roadside estimation module, the corresponding relation in anaglyph by road actual width and angle of inclination, roadside, determines region, roadside; And road boundary detection module, by judging the parallax difference between related pixel point in anaglyph, thereby detect type and the position in roadside.
The system that detects roadside according to employing anaglyph of the present invention, described anaglyph pretreatment module comprises:
Edge disparity map generation module, utilizes boundary operator, on the basis of former disparity map, generates edge image;
Road surface Region Segmentation module, the road surface equation by estimating on V disparity map is partitioned into region, road surface in edge disparity map; And
Denoising module, utilizes the monotonicity that on roadside, pixel is worth on edge disparity map, removes noise.
According to employing anaglyph of the present invention, detect the system in roadside, region, described roadside estimation module comprises: region, roadside MBM, by in past n frame, road actual width and angle of inclination, the roadside corresponding relation in anaglyph, set up region, roadside estimation model; Roadside regional prediction module, region, roadside estimation model and Common Prediction Method based on setting up, dope the road boundary detection region of present frame; Region, roadside estimation model update module, by the road boundary detection result of present frame and the trusted area accumulated result of past m frame, upgrades region, roadside estimation model.
The system that detects roadside according to employing anaglyph of the present invention, region, described roadside MBM comprises: area-of-interest is selected module, selects the area-of-interest on image to reduce computing time; Straight line generation module, adopts certain computing method, by the location positioning of above-mentioned zone cathetus out; Angle restriction area judging module, the corresponding relation in anaglyph by road actual width and roadside straight line, determines the region, roadside in area-of-interest; And roadside region growing module, utilize roadside extensibility to come region, growth source roadside.
The system that detects roadside according to employing anaglyph of the present invention, described road boundary detection module comprises: pixel generation module to be selected, scan region, roadside one time, utilize the difference of parallax value between pixel, only leave the pixel on useful roadside; Straight line generation module, adopts certain computing method, by the location positioning of above-mentioned zone cathetus out; And roadside type judging module, according to the roadside type of pixel on straight line, thus the roadside type of definite whole piece straight line.
Generally, this patent has proposed a kind of in anaglyph, the method based on disparity map character direct-detection roadside.Estimate region, road surface and utilizing roadside parallax value monotonicity to carry out after denoising, the method for quick in roadside comprises following 2 parts: the road boundary detection region that is generated present frame by region, roadside estimation model, by the parallax difference between related pixel point in anaglyph, in road boundary detection region, detect type and the position in roadside again.By in past n frame, road actual width and angle of inclination, the roadside corresponding relation in anaglyph, set up region, roadside estimation model.By the road boundary detection result of present frame and the trusted area accumulated result of past m frame, upgrade region, roadside estimation model again.Scan region, current roadside one time, only leave the pixel on useful roadside, by judging the parallax difference between these related pixel points, determine the roadside type of each pixel, thereby determine the actual type in whole piece roadside.
Visible, the present invention is the corresponding relation in anaglyph by road actual width and angle of inclination, roadside, determines road boundary detection region.Wherein utilize the difference of parallax value between pixel to come the disposable various roadsides that detect on road surface, comprise white line, curb stone etc.
Accompanying drawing explanation
By reading the detailed description of following the preferred embodiments of the present invention of considering by reference to the accompanying drawings, will understand better above and other target of the present invention, feature, advantage and technology and industrial significance.
Utilization shown in Fig. 1 the schematic diagram of image processing system of the method according to this invention and system;
Shown in Fig. 2 is the block scheme of image processing module;
Shown in Fig. 3 is to adopt according to the module map of road boundary detection system of the present invention;
Shown in Fig. 4 is to adopt according to the data flow schematic diagram of road boundary detection method of the present invention;
Shown in Fig. 5 is according to the main flow chart of road boundary detection method of the present invention.
Shown in Fig. 6 A is according to the process flow diagram of the anaglyph preprocessing process of road boundary detection system of the present invention;
The window schematic diagram that means the point on roadside shown in Fig. 6 B;
Shown in Fig. 7 is according to the process flow diagram of region, the roadside estimation procedure of road boundary detection system of the present invention;
Shown in Fig. 8 is according to the process flow diagram of region, the roadside modeling process of road boundary detection system of the present invention;
Shown in Fig. 9 is according to the process flow diagram of the road boundary detection process of road boundary detection system of the present invention;
Shown in Figure 10 A and 10B is according to the process flow diagram of the pixel generative process to be selected of road boundary detection system of the present invention;
Shown in Figure 11 is edge disparity map and road surface area schematic;
Shown in Figure 12 A-12C is region, the roadside estimation model modeling process schematic diagram according to road boundary detection method of the present invention;
Shown in Figure 13 A and 13B is the road boundary detection process schematic diagram according to road boundary detection method of the present invention;
Shown in Figure 14 A and 14B is principle of stereoscopic vision figure;
Shown in Figure 15 A-15C is the schematic diagram of the angular relationship between two roadsides and Rwidth.
Embodiment
Below in conjunction with accompanying drawing, the embodiment of the present invention is described.
Utilization shown in Fig. 1 the schematic diagram of image processing system of the method according to this invention and system.Image processing system is arranged on vehicle 100.This system comprises binocular camera shooting head 101 or two independent cameras, and image processing module 102.Camera 101 is arranged on the position near vehicle mirrors, for catching the scene of vehicle front.The image of the vehicle front scene of catching is using the input as image processing module 102.Image processing module 102 is analyzed the picture signal of input, calculates their scene classification, and based on classification, give different weights and merge two width images, and the institute's fused images that obtains detection road surface white line.Vehicle control module 103 receives the signal of being exported by image processing module 102, and the current white line obtaining according to detection generates travel direction and travel speed that control signal is controlled vehicle 100.
Shown in Fig. 2 is the structural representation of the image processing system of the method according to this invention and system.Camera 101 comprises imageing sensor 201 and camera digital signal processor (digital signal processing, DSP) 202.Imageing sensor 201 is converted to electronic signal by light signal, the image when vehicle in front 100 the place aheads of catching is converted to analog picture signal, then imports result into camera DSP202.If needed, camera 101 can further include camera lens, filter, etc.Although only provided in an embodiment of the present invention two cameras, embodiments of the present invention can comprise a plurality of cameras 101, after these camera registrations, can catch multiple image simultaneously.
Camera DSP202 is converted to data image signal by analog picture signal, and sends to image processing module 102.Image input interface 203 time interval is in accordance with regulations obtained image.Disparity map image-forming module 204 utilizes principle of stereoscopic vision, and a pair of digital picture of input is converted to disparity map.Then anaglyph is written into internal memory 206, by program 207, is analyzed and is processed.Image is herein processed and is comprised multiple operation, such as denoising, sets up region, roadside estimation model, detects roadside type etc.Program 207 in ROM is carried out a series of operation and is detected final roadside type.In this process, CPU 205 is responsible for control operation and arithmetic operation, for example, by interface, obtain data, carries out image processing etc.
Shown in Fig. 3 is to adopt according to the module map of road boundary detection system of the present invention.For a width anaglyph to be detected, anaglyph pretreatment unit F1 estimates region, road surface by road surface equation in anaglyph, and the monotonicity denoising based on roadside parallax value.Region, road surface determines by every some position of road surface, and no matter it is level road or jolts road surface, goes up a slope or descending.If region, roadside estimation model is not also set up, so by roadside regional model modeling unit F2, according in past n frame, road actual width and angle of inclination, the roadside corresponding relation in anaglyph, sets up model.If region, roadside estimation model is set up, so just by roadside regional prediction unit F 3 according to this model and Common Prediction Method, dope the road boundary detection region of present frame.Then road boundary detection unit F 4 is by judging the parallax difference between the related pixel point staying in these regions, thereby detects type and the position in roadside.This method only need scan region, roadside one time, only leaves the pixel on useful roadside, then passes through the roadside type of these pixels, thereby detects the actual type in whole piece roadside.Roadside regional model updating block F5, by the road boundary detection result of present frame and the trusted area accumulated result of past m frame, upgrades region, roadside estimation model.
Shown in Fig. 4 is to adopt according to the data flow schematic diagram of road boundary detection method of the present invention.First by stereoscopic camera, obtain anaglyph S1 and S2.Then through region, road surface choose with the operation such as denoising after, obtain pretreated anaglyph S3.Region, road surface estimation model S4 is by past n frame, and road actual width and angle of inclination, the roadside corresponding relation in anaglyph is set up.Then model S4 thus, in conjunction with Common Prediction Method, obtains region, the roadside S5 of the prediction on anaglyph S2 to be detected.In the S5 of region, roadside, scan one time, only leave the pixel S6 on useful roadside, each pixel has a roadside type.Through over-fitting pixel S6, obtain the position in roadside; The roadside type that on roadside, pixel number is maximum, is the type in whole piece roadside.Obtain final road boundary detection result S7.
Shown in Fig. 5 is according to the main flow chart of road boundary detection method of the present invention.First, anaglyph pretreatment module 21 is estimated region, road surface denoising on edge disparity map.Then, region, roadside estimation module 22 by road actual width and angle of inclination, roadside the corresponding relation in anaglyph predict region, roadside.Finally, road boundary detection module 23, in the region, roadside of prediction, is utilized the parallax difference between related pixel point, detects type and the position in roadside.All these will be discussed in the back in detail.
Shown in Fig. 6 A is according to the process flow diagram of the anaglyph preprocessing process of road boundary detection system of the present invention.In dense disparity map picture, be difficult to detect roadside, so the first step, need to first generate edge disparity map, the work that edge disparity map generation module 211 is done.Can to former dense disparity map, look like to process by boundary operators such as Sobel operators, obtain edge disparity map.Then by estimating to obtain road surface equation, be partitioned into region, road surface, the work that road surface Region Segmentation module 212 is done.Region, road surface comprises the various objects on road surface and road surface.We carry out approximate evaluation road surface equation with elevation information.Have a lot of methods to estimate for road surface herein, a kind of implementation method is to adopt the method based on V disparity map.When we obtain after the equation of road surface, according to elevation information, get suitable threshold value, remove the point in the edge anaglyph outside region, road surface, only leave the region, road surface on edge disparity map.Figure 11 is the schematic diagram after edge anaglyph and region, road surface are chosen.
d(above_pt)<d(p)<d(below_pt)
Here, d (p) is the parallax value at p point place, and above_pt is the p point point of position above, and below_pt is the point of p point lower position, and particular location is as shown below.All these points all in a denoising window, the window of 3*3 or 5*5 size for example.Shown in Fig. 6 B is the schematic diagram of this window.
All roadsides meet at the place of this point of end point a long way off, and roadside is all below end point, therefore, when denoising window is in end point lower left time, the upper right side that above_pt should be ordered at p, the lower left that below_pt should be ordered at p; When denoising window is when end point is bottom-right, the upper left side that above_pt should be ordered at p, the lower right that below_pt should be ordered at p; When denoising window is under end point, above_pt should order at p directly over, below_pt should order at p under.
Shown in Fig. 7 is according to the process flow diagram of region, the roadside estimation procedure of road boundary detection system of the present invention.For an input picture to be detected, if region, roadside estimation model is not also set up, so first by region, roadside MBM 221, set up this model.Region, roadside MBM 221 is according in past n frame, and road actual width and angle of inclination, the roadside corresponding relation in anaglyph, sets up model.Remove noisy line, only leave kerb line.This module will describe in detail in the back.If region, roadside estimation model has been set up, roadside regional prediction module 222, according to this model and Common Prediction Method, dopes the road boundary detection region of present frame.This can reduce the scope of road boundary detection afterwards, reduces the time of road boundary detection.In a kind of implementation method, Kalman filtering is selected as the method for prediction.Target of prediction comprises four parameters (x direction position coordinates (x-position), y direction position coordinates (y-position), width (width), highly (height)), these four parametric descriptions the information in region, roadside.
State vector is expressed as x (t)=(x, y, dx, dy, w, h) T, x wherein, y is the coordinate of target's center's point, dx, dy be central point in the change of horizontal and vertical direction, w, h is width and the height of target of prediction.State-transition matrix is expressed as:
Predictor formula: x (t+ Δ)=A x (t).
Region, roadside estimation model update module 223, by the road boundary detection result of present frame and the trusted area accumulated result of past m frame, is upgraded region, roadside estimation model.The testing result of present frame is for region, real-time update roadside estimation model.Road boundary detection will be carried out at set intervals in region, non-roadside in anaglyph, if believable roadside detected, is so just equivalent to occur new region, roadside, and the confidence level in region, new roadside adds up along with the increase of frame number.If the confidence level in region, new roadside surpasses certain threshold value, this region, roadside is just added in the middle of the estimation model of region, roadside so.The confidence level in region, a roadside depends on the confidence level in the roadside detecting in region.The confidence level in roadside is definite by the credible length in roadside on the one hand, and also the change in location in successive frame is relevant with it on the other hand.
Shown in Fig. 8 is according to the process flow diagram of region, the roadside modeling process of road boundary detection system of the present invention.Area-of-interest (ROI) selects module 2211 to select the area-of-interest on road surface to reduce operation time.Because the parallax value in region is nearby relatively accurate, in a kind of implementation method, in anaglyph, the region of general lower 1/3 height is selected as area-of-interest, as shown in rectangle in Figure 12 A.Straight line generation module 2212, by certain generation method of straight line, obtains the straight line in area-of-interest, as the method for Hough transformation, and RANSAC algorithm, or least-square fitting approach etc.In a kind of implementation method, as shown in Figure 12 A, straight line shows with blue line segment, and has put on sequence number.They are not collinear on road surface, wherein, the 1st, buildings bottom margin line, 4 and 5 is white line edge lines, and 6 and 7 is arrow edge lines, and 2 and 3 also have 8 to be respectively curb stone edge line.These straight lines based on area-of-interest, angle restriction area judging module 2213, by road actual width and the corresponding relation of roadside straight line in anaglyph, is determined the region, roadside in area-of-interest.Can get rid of unwanted straight line like this, only leave roadside straight line.In a kind of implementation method, as shown in Figure 12 A, utilize this relation, it can get rid of straight line 1,6,7, only leaves roadside straight line 2,3,4,8.
By the principle of stereoscopic vision shown in Figure 14 A, for the some P in real world, we can obtain following relational expression:
Here, b is parallax range, and Z is the degree of depth, and f is the focal length of camera, and d is parallax value.If have two some P and Q in real world, they have the same degree of depth, and in anaglyph, pixel distance in the horizontal direction (pixel wide) Dwidth between corresponding pixel points has relation as shown in Figure 14B actual distance Rwidth(actual width between them so):
Downward projection is carried out in the region, road surface of anaglyph, obtain U disparity map.In U disparity map, parallax is d place, between pixel wide Dwidth and the angle of inclination in roadside between two roadside straight lines, just like the relation shown in figure below 15A:
Here, θ 1 is roadside straight line l1 and the angle of horizontal line on U disparity map; θ 2 is roadside straight line l2 and the angle of horizontal line on U disparity map.
When anaglyph is carried out to downward projection, the inclination angle beta of kerb line in former anaglyph, and between the tilt angle theta of kerb line projection line on U disparity map, just like Figure 15 b) corresponding relation shown in:
tg(θ)=cos(α)·tg(β)
Here, α is the dihedral angle between road surface and U disparity map.Merge above-mentioned several formula, we obtain road actual width and the corresponding relation of roadside straight line in anaglyph:
Here, β 1 is roadside straight line l1 and the angle of horizontal line on former disparity map; β 2 is roadside straight line l2 and the angle of horizontal line on former disparity map.Because road surface equation is obtaining before, so α is a value of having determined.According to the relevant regulations of a national communication aspect, the width between lane line, fixes.In China, the chances are 2 meters wide.Therefore, we can utilize road actual width and the roadside straight line above-mentioned corresponding relation in anaglyph, obtain region, roadside.Meeting the line of above-mentioned corresponding relation, is kerb line; Not meeting the line of above-mentioned relation, is not kerb line, should give removal.Had relational expression (1), for every pair of straight line in disparity map, we calculate the actual distance between them.So for every straight line, the actual distance between it and other straight lines, has just formed a vector.Each vector is removed the component that is wherein not equal to lane line width.The corresponding straight line of vector that residual components number is maximum so, just elects benchmark kerb line as.Other and benchmark kerb line distance are the straight line of lane line width, also elect kerb line as.Figure 12 B has shown by the estimated region, roadside out of above-mentioned angle restriction.Actual width between straight line 4 and 5 is not lane line width, and in like manner, straight line 5 and 6 neither.By calculating, we determine that straight line 4 is for benchmark kerb line because it is the straight line that has maximum number lane line width, with the actual width of straight line 3 and straight line 8 be all lane line width.Then determining straight line 4 and straight line 8 is kerb line.Finally, the roadside based on determining, we obtain region, roadside, with purple rectangle, have represented out.Roadside region growing module 2214 utilizes the extensibility that roadside makes progress to come region, growth source roadside, in Figure 12 C, above shown in row's purple little rectangle.
Shown in Fig. 9 is according to the process flow diagram of the road boundary detection process of road boundary detection system of the present invention.Pixel generation module 231 to be selected scans region, roadside one time, utilizes the difference of parallax value between pixel, only leaves the pixel on useful roadside.The edge of white line, curb stone or irrigation canals and ditches is all to occur in pairs substantially.Width between every edge is also fixed substantially.The difference of the parallax value on one-tenth edge between two pixels at same y coordinate place, for dissimilar roadside, has following relation:
Dis (irrigation canals and ditches) <dis (white line) <dis (curb stone) <dis (other)
Here, dis represents the difference of parallax value.By scanning region, roadside once, we can utilize above-mentioned relation to remove pixel useless, only retain the pixel on the roadside that meets this relation.Shown in Figure 10, it is the process flow diagram of pixel generation module 231 to be selected.In conjunction with following schematic diagram, detailed process is as follows:
1. from the middle of region, roadside, scan to the left and right respectively.
2. when pixel p runs into the pixel q that meets above-mentioned parallax value difference, leave pixel p, calculate
Dis (p, q), and by the value of ds (p, q), judged the roadside type of pixel p.
3. remove other pixels on sweep trace.
Shown in Figure 13 A, it is the result after pixel generation module 231 to be selected executes.We find, the number of pixel has reduced, and this is just next module, and straight line generation module 2212 has been saved the time.Each pixel staying has the roadside type under it.Straight line generation module 2212, by certain generation method of straight line, carries out matching to the pixel staying, and obtains the position of straight line.Here adoptable approximating method comprises, the method for Hough transformation, RANSAC algorithm, least-square fitting approach etc.Roadside type judging module 232 is by the position in the roadside of determining, and the roadside type of each pixel on roadside, selects the maximum roadside type of pixel number on roadside, is the type in whole piece roadside, obtains final road boundary detection result.Shown in Figure 13 B, be final road boundary detection result, here, white line goes out with No. 4 line drawings, and curb stone goes out with 3 and No. 8 line drawings.
The sequence of operations illustrating in instructions can be carried out by the combination of hardware, software or hardware and software.When carrying out this sequence of operations by software, computer program wherein can be installed in the storer in the computing machine that is built in specialized hardware, make computing machine carry out this computer program.Or, computer program can be installed in the multi-purpose computer that can carry out various types of processing, make computing machine carry out this computer program.
For example, can computer program is pre-stored to hard disk or ROM(ROM (read-only memory) as recording medium) in.Or, can be temporarily or for good and all storage (record) computer program in removable recording medium, such as floppy disk, CD-ROM(compact disc read-only memory), MO(magneto-optic) coil, DVD(digital versatile disc), disk or semiconductor memory.So removable recording medium can be provided as canned software.
The present invention has been described in detail with reference to specific embodiment.Yet clearly, in the situation that not deviating from spirit of the present invention, those skilled in the art can carry out change and replace embodiment.In other words, the present invention is open by the form of explanation, rather than is limited to explain.Judge main idea of the present invention, should consider appended claim.
Claims (10)
1. adopt anaglyph to detect the method in roadside, comprising:
In anaglyph, estimate region, road surface denoising;
The corresponding relation in anaglyph by road actual width and angle of inclination, roadside, determines region, roadside; And
By the parallax difference between related pixel point in judgement anaglyph, detect type and the position in roadside.
2. employing anaglyph as claimed in claim 1 detects the method in roadside, describedly in anaglyph, estimates that region, road surface denoising comprise:
Utilize boundary operator, on the basis of former disparity map, generate edge image;
Road surface equation by estimating on V disparity map is partitioned into region, road surface in edge disparity map; And
Denoising module is utilized the monotonicity that on roadside, pixel is worth on edge disparity map, removes noise.
3. employing anaglyph as claimed in claim 2 detects the method in roadside, wherein said by road actual width and angle of inclination, roadside the corresponding relation in anaglyph determine that region, roadside comprises:
By in past n frame, road actual width and angle of inclination, the roadside corresponding relation in anaglyph, set up region, roadside estimation model;
Region, roadside estimation model and Common Prediction Method based on setting up, dope the road boundary detection region of present frame; And
Update module, by the road boundary detection result of present frame and the trusted area accumulated result of past m frame, is upgraded region, roadside estimation model.
4. employing anaglyph as claimed in claim 3 detects the method in roadside, and wherein said region, the roadside estimation model of setting up comprises:
Select the area-of-interest on image to reduce computing time;
Adopt certain computing method, by the location positioning of above-mentioned zone cathetus out;
The corresponding relation in anaglyph by road actual width and roadside straight line, determines the region, roadside in area-of-interest; And
Utilize roadside extensibility to come region, growth source roadside.
5. employing anaglyph as claimed in claim 1 detects the method in roadside, and wherein said type and the position that detects roadside comprises:
Scan region, roadside one time, utilize the difference of parallax value between pixel, only leave the pixel on useful roadside;
Adopt certain computing method, by the location positioning of above-mentioned zone cathetus out; And
According to the roadside type of pixel on straight line, thus the roadside type of definite whole piece straight line
6. adopt anaglyph to detect the system in roadside, comprising:
Anaglyph pretreatment module is estimated region, road surface denoising in anaglyph;
Region, roadside estimation module, the corresponding relation in anaglyph by road actual width and angle of inclination, roadside, determines region, roadside; And
Road boundary detection module, by judging the parallax difference between related pixel point in anaglyph, thereby detects type and the position in roadside.
7. employing anaglyph as claimed in claim 6 detects the system in roadside, and described anaglyph pretreatment module comprises:
Edge disparity map generation module, utilizes boundary operator, on the basis of former disparity map, generates edge image;
Road surface Region Segmentation module, the road surface equation by estimating on V disparity map is partitioned into region, road surface in edge disparity map; And
Denoising module, utilizes the monotonicity that on roadside, pixel is worth on edge disparity map, removes noise.
8. employing anaglyph as claimed in claim 6 detects the system in roadside, and region, described roadside estimation module comprises:
Region, roadside MBM, by past n frame, road actual width and angle of inclination, the roadside corresponding relation in anaglyph, set up region, roadside estimation model;
Roadside regional prediction module, region, roadside estimation model and Common Prediction Method based on setting up, dope the road boundary detection region of present frame; And
Region, roadside estimation model update module, by the road boundary detection result of present frame and the trusted area accumulated result of past m frame, upgrades region, roadside estimation model.
9. employing anaglyph as claimed in claim 8 detects the system in roadside, and region, described roadside MBM comprises:
Area-of-interest is selected module, selects the area-of-interest on image to reduce computing time;
Straight line generation module, adopts certain computing method, by the location positioning of above-mentioned zone cathetus out;
Angle restriction area judging module, the corresponding relation in anaglyph by road actual width and roadside straight line, determines the region, roadside in area-of-interest; And
Roadside region growing module, utilizes roadside extensibility to come region, growth source roadside.
10. employing anaglyph as claimed in claim 6 detects the system in roadside, and described road boundary detection module comprises:
Pixel generation module to be selected, scans region, roadside one time, utilizes the difference of parallax value between pixel, only leaves the pixel on useful roadside;
Straight line generation module, adopts certain computing method, by the location positioning of above-mentioned zone cathetus out; And
Roadside type judging module, according to the roadside type of pixel on straight line, thus the roadside type of definite whole piece straight line.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210342374.2A CN103679121B (en) | 2012-09-14 | 2012-09-14 | Method and system for detecting roadside using visual difference image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210342374.2A CN103679121B (en) | 2012-09-14 | 2012-09-14 | Method and system for detecting roadside using visual difference image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103679121A true CN103679121A (en) | 2014-03-26 |
| CN103679121B CN103679121B (en) | 2017-04-12 |
Family
ID=50316614
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210342374.2A Expired - Fee Related CN103679121B (en) | 2012-09-14 | 2012-09-14 | Method and system for detecting roadside using visual difference image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103679121B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105551016A (en) * | 2015-12-02 | 2016-05-04 | 百度在线网络技术（北京）有限公司 | Method and device of road edge identification on the basis of laser-point cloud |
| CN105868687A (en) * | 2015-02-09 | 2016-08-17 | 丰田自动车株式会社 | Traveling road surface detection apparatus and traveling road surface detection method |
| CN106096512A (en) * | 2016-05-31 | 2016-11-09 | 上海美迪索科电子科技有限公司 | Utilize the detection device and method that vehicles or pedestrians are identified by depth camera |
| CN108229406A (en) * | 2018-01-11 | 2018-06-29 | 海信集团有限公司 | A kind of method for detecting lane lines, device and terminal |
| CN108399360A (en) * | 2018-01-22 | 2018-08-14 | 海信集团有限公司 | A kind of continuous type obstacle detection method, device and terminal |
| CN113959398A (en) * | 2021-10-09 | 2022-01-21 | 广东汇天航空航天科技有限公司 | Vision-based distance measurement method and device, travelable equipment and storage medium |
| CN115143888A (en) * | 2022-09-06 | 2022-10-04 | 无锡沃尔德轴承有限公司 | Measuring system for position change of bearing base |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1548685A2 (en) * | 2003-12-26 | 2005-06-29 | Fuji Jukogyo Kabushiki Kaisha | Vehicle driving assist system |
| CN101999972A (en) * | 2010-11-24 | 2011-04-06 | 上海理工大学 | Stereoscopic vision based auxiliary walking device for blindmen and auxiliary method thereof |
| CN102668540A (en) * | 2009-12-25 | 2012-09-12 | 株式会社理光 | Imaging device, on-vehicle imaging system, road surface appearance detection method, and object detection device |
-
2012
- 2012-09-14 CN CN201210342374.2A patent/CN103679121B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1548685A2 (en) * | 2003-12-26 | 2005-06-29 | Fuji Jukogyo Kabushiki Kaisha | Vehicle driving assist system |
| CN102668540A (en) * | 2009-12-25 | 2012-09-12 | 株式会社理光 | Imaging device, on-vehicle imaging system, road surface appearance detection method, and object detection device |
| CN101999972A (en) * | 2010-11-24 | 2011-04-06 | 上海理工大学 | Stereoscopic vision based auxiliary walking device for blindmen and auxiliary method thereof |
Non-Patent Citations (2)
| Title |
|---|
| JUN ZHAO ET AL.: "Detection of Non-flat Ground Surfaces Using V-Disparity Images", 《THE 2009 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS》 * |
| ZHENCHENG HU ET AL.: "U-V-Disparity：An efficient algorithm for Stereovision Based Scene Analysis", 《IEEE PROCEEDINGS.INTELLIGENT VEHICHLES SYMPOSIUM》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105868687A (en) * | 2015-02-09 | 2016-08-17 | 丰田自动车株式会社 | Traveling road surface detection apparatus and traveling road surface detection method |
| CN105868687B (en) * | 2015-02-09 | 2019-05-31 | 丰田自动车株式会社 | Track detection device and track detection method |
| CN105551016A (en) * | 2015-12-02 | 2016-05-04 | 百度在线网络技术（北京）有限公司 | Method and device of road edge identification on the basis of laser-point cloud |
| CN105551016B (en) * | 2015-12-02 | 2019-01-22 | 百度在线网络技术（北京）有限公司 | A kind of curb recognition methods and device based on laser point cloud |
| CN106096512A (en) * | 2016-05-31 | 2016-11-09 | 上海美迪索科电子科技有限公司 | Utilize the detection device and method that vehicles or pedestrians are identified by depth camera |
| CN106096512B (en) * | 2016-05-31 | 2020-08-25 | 上海美迪索科电子科技有限公司 | Detection device and method for recognizing vehicle or pedestrian by using depth camera |
| CN108229406A (en) * | 2018-01-11 | 2018-06-29 | 海信集团有限公司 | A kind of method for detecting lane lines, device and terminal |
| CN108399360A (en) * | 2018-01-22 | 2018-08-14 | 海信集团有限公司 | A kind of continuous type obstacle detection method, device and terminal |
| CN108399360B (en) * | 2018-01-22 | 2021-12-24 | 海信集团有限公司 | Continuous obstacle detection method, device and terminal |
| CN113959398A (en) * | 2021-10-09 | 2022-01-21 | 广东汇天航空航天科技有限公司 | Vision-based distance measurement method and device, travelable equipment and storage medium |
| CN113959398B (en) * | 2021-10-09 | 2024-04-12 | 广东汇天航空航天科技有限公司 | Distance measurement method and device based on vision, drivable equipment and storage medium |
| CN115143888A (en) * | 2022-09-06 | 2022-10-04 | 无锡沃尔德轴承有限公司 | Measuring system for position change of bearing base |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103679121B (en) | 2017-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105206109B (en) | A kind of vehicle greasy weather identification early warning system and method based on infrared CCD | |
| CN103617412B (en) | Real-time lane line detection method | |
| CN103679121A (en) | Method and system for detecting roadside using visual difference image | |
| JP6442834B2 (en) | Road surface height shape estimation method and system | |
| EP2605185B1 (en) | Detection of obstacles at night by analysis of shadows | |
| CN103714538B (en) | road edge detection method, device and vehicle | |
| JP5787024B2 (en) | Three-dimensional object detection device | |
| CN108629292B (en) | Curved lane line detection method and device and terminal | |
| US20050015201A1 (en) | Method and apparatus for detecting obstacles | |
| CN105550665A (en) | Method for detecting pilotless automobile through area based on binocular vision | |
| CN105667518A (en) | Lane detection method and device | |
| CN108280450A (en) | A kind of express highway pavement detection method based on lane line | |
| CN104902261B (en) | Apparatus and method for the road surface identification in low definition video flowing | |
| CN101408978B (en) | Method and apparatus for detecting barrier based on monocular vision | |
| Jiang et al. | Lane and obstacle detection based on fast inverse perspective mapping algorithm | |
| Liu et al. | Development of a vision-based driver assistance system with lane departure warning and forward collision warning functions | |
| JPH07129898A (en) | Obstacle detecting device | |
| CN104751151A (en) | Method for identifying and tracing multiple lanes in real time | |
| CN108171695A (en) | A kind of express highway pavement detection method based on image procossing | |
| CN110197173B (en) | Road edge detection method based on binocular vision | |
| Petrovai et al. | A stereovision based approach for detecting and tracking lane and forward obstacles on mobile devices | |
| CN112115889B (en) | Intelligent vehicle moving target detection method based on vision | |
| CN109241855B (en) | Intelligent vehicle travelable area detection method based on stereoscopic vision | |
| CN109791607A (en) | It is detected from a series of images of video camera by homography matrix and identifying object | |
| Li et al. | Lane marking quality assessment for autonomous driving |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170412 |