CN104899898B - Pavement detection method based on multidimensional information probabilistic model - Google Patents
Pavement detection method based on multidimensional information probabilistic model Download PDFInfo
- Publication number
- CN104899898B CN104899898B CN201510284119.0A CN201510284119A CN104899898B CN 104899898 B CN104899898 B CN 104899898B CN 201510284119 A CN201510284119 A CN 201510284119A CN 104899898 B CN104899898 B CN 104899898B
- Authority
- CN
- China
- Prior art keywords
- mrow
- road surface
- probabilistic model
- pavement
- mtd
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30256—Lane; Road marking
Abstract
The invention discloses a kind of pavement detection method based on multidimensional information probabilistic model, comprise the following steps:Step 1:Establish road surface Sample Storehouse;Step 2:Comprehensive road surface colourity and color character, establish gaussian probability model;Step 3:Pavement texture feature is extracted, establishes the optimal loose probabilistic model of pavement texture feature mode;Step 4:Lane detection;Step 5:With reference to the multidimensional information of colourity, color, texture, motion and lane line, pavement detection probabilistic model is established.The present invention has the pavement detection being applied under a variety of environment, and accuracy is higher, the advantages that robustness stronger and good application prospect.
Description
Technical field
The present invention relates to a kind of pavement detection technology, belong to Computer Image Processing field, it is more particularly to a kind of based on more
The pavement detection method of informational probability model is tieed up, the pavement detection method is a kind of based on road surface colourity, color, texture, motion
Multidimensional information synthesis lane detection establishes the Road Detection Algorithm of probabilistic model.
Background technology
With expanding economy, the progress of society, automobile turns into the main means of transport of people's trip, and intelligent transportation also obtains
Rapid development is arrived.Vehicle control system and traffic surveillance and control system are the chief components of intelligent transportation field.Vehicle control
System processed is autonomous driving vehicle, is had great significance for the burden for mitigating driver;Traffic surveillance and control system is for helping
Traffic dynamic monitors and recorded road vehicle and personal information, raising road safety managerial skills play important work
With.And pavement detection is vehicle control system and the intelligentized basis of traffic surveillance and control system, for searching the effective driver's compartment of vehicle
Domain, the scope for reducing other road traffic infomation detections, reduce the complexity of computing, realize that the real-time of detection is most important.
The complexity of environment, such as pavement markers such as road vehicles, sky and the interference of trees, light sudden change, lane lines
Influence, the defects of flase drop on pavement detection imperfect, non-road surface can be caused.The method on detection road surface is mainly based upon color at present
The methods of degree is with the template matches of saturation degree or color, but problems be present in these methods:Due to the stationarity of stencil-chosen
To illumination variation lack robustness, easily obscure sky etc. and object similar in the colourity of road surface, simultaneously for road surface marker not
It can detect, so as to cause to detect the fracture on road surface.These problems reduce the accuracy of pavement detection, are art technologies
Personnel's technical problem anxious to be resolved.
The content of the invention
The shortcomings that it is an object of the invention to overcome prior art and deficiency, there is provided one kind is based on multidimensional information probabilistic model
Pavement detection method, the pavement detection method solves the problems, such as existing algorithm to that can not detect complete road surface, and builds
The probabilistic model for meeting road surface colourity, color and texture feature is found.
The purpose of the present invention is achieved through the following technical solutions:A kind of pavement detection side based on multidimensional information probabilistic model
Method, comprise the following steps:
(1) road surface Sample Storehouse is established;
(2) comprehensive road surface colourity and color character, establish gaussian probability model;
(3) pavement texture feature is extracted, establishes the optimal loose probabilistic model of mode;
(4) lane detection;
(5) colourity, color, texture, motion, the multidimensional information of lane line are combined, establishes pavement detection probabilistic model.
In step 1, road surface characteristic of several sections under several light conditions is collected, to establish road surface sample
Storehouse.
In step 2, the method for establishing the gaussian probability model comprises the following steps:
Step 21, the R passages for obtaining image;
Step 22, image is transformed into YCrCb color spaces, extracts Cr chrominance channels and Cb chrominance channels;
Step 23, Kolmogorov-Smirnov inspections are carried out, verify Cr passages, Cb passages and the R passages of road surface sample
Meet Gaussian Profile;
Step 24, with reference to Cr, Cb, R passage, establish gaussian probability model.
In step 3, the method for building up for establishing the optimal loose probabilistic model of pavement texture feature mode includes following
Step:
Step 31, the gray level co-occurrence matrixes for establishing road surface sample image;
Step 32, the matrix of angle two (ASM) and contrast sub-matrix (IDM) for solving pavement texture;
Step 33, establish the optimal loose probabilistic model applicable elements of mode:The matrix of angle two and contrast of 60% sample above
Scoring matrix value is that mode is most to be worth, and characteristic value and the mode absolute distance of remaining sample are less than 5 times of standard deviations;Definition:Mode is most
Excellent loose probabilistic model meets the relaxation probability of sample to describe tested value x (probability can be less than zero):
Wherein, σ is the standard deviation of sample, and m is the mode (i.e. most be worth) in sample, and μ is the most value of non-mode in sample, k
For slackness (generally 3), the scope of the bigger expression detections of k is bigger, and precision is smaller.When μ is minimum value, (μ ± k σ), which takes, subtracts
Number, otherwise plus sige is taken, | | a-b | | represent a to b absolute distances;
Step 34, the probabilistic model using the optimal loose probabilistic model solution pavement texture feature of mode;Using mode most
The matrix of angle two and contrast scoring matrix value in excellent loose probabilistic model calculating region to be detected meet the probability that road surface is distributed and are:
Wherein, A represents region to be detected.
In step 4, the detection method of lane detection comprises the following steps:
Step 41, image is converted into gray level image;
Step 42, Canny operator extractions image border;
Step 43, scanning entire image simultaneously extract the line segment for meeting lane line feature;
Wherein, l is line segment length, and α is the angle of line segment and vertical direction, and v and θ are corresponding threshold value.
In steps of 5, the method for building up for establishing the multidimensional information probabilistic model comprises the following steps:
Step 51, the i-th two field picture in traffic video is set to fi(x, y), the first two field picture are f1(x, y), then meet road surface
The Probability p of feature is:
Wherein, P1(fi(x, y)) it is that present frame (x, y) pixel establishes Gauss model acquisition using colourity, color information
Probability, P2(fi(x, y)) it is that present frame (x, y) pixel establishes the optimal loose probabilistic model acquisition of mode using textural characteristics
Probability, P (fi-1(x, y)) be movable information using traffic video, i.e., previous frame be detected as road surface part have it is certain general
Rate is also road surface in present frame, particularly important when pavement detection is applied to video monitoring system;
Parameter alpha, β, γ are the weighted value (alpha+beta+γ=1) of corresponding probabilistic information, and 85%, 10% and is taken under general scene
5%;When pavement detection is applied to video monitoring system, γ proportion suitably increases, it can thus be concluded that pavement detection probabilistic image,
By selecting rational threshold value, you can judge road surface region;
Step 52, parameterTo be repaired with lane detection result to missing inspection region, method for repairing and mending is as follows:
Whether using lane detection result, search on the left of lane line has enough pixels in (right side) certain limit
Road surface is detected as, in this way, then fills lane line and the pixel between it;
When video image is the first frame, formula (5) deteriorates to:
Wherein, alpha+beta=1, by establishing multidimensional information probabilistic model, the accuracy and robustness of pavement detection are enhanced.
The operation principle of the present invention:The present invention is a kind of Road Detection Algorithm based on multidimensional information probabilistic model, the calculation
The multidimensional information such as method synthesis road surface colourity, color, texture, motion and lane line establish probabilistic model to improve pavement detection
Accuracy and robustness.In the algorithm, road surface Sample Storehouse is initially set up, is then based on YCrCb color spaces and RGB color
Cr, Cb, R Path Setup gaussian probability model in space, enhance the accuracy of pavement detection.Propose afterwards based on road surface line
The optimal loose probabilistic model of mode of feature is managed, the probability that road surface is detected as based on pavement texture feature is obtained, for not
There is stronger robustness with illumination condition.The information of road surface such as colourity, color, texture, motion are finally combined, establish multidimensional information
Probabilistic model, and missing inspection region is made up using the result of lane detection, so as to detect complete road surface.The present invention is applied to
Pavement detection under a variety of environment, accuracy is higher, and robustness is stronger, has a good application prospect.
The present invention is had the following advantages relative to prior art and effect:
1. proposing the optimal loose probabilistic model of mode, the detection model for meeting pavement texture feature is established, is enhanced
The robustness of pavement detection.
2. the CrCb chrominance channels on YCrCb color spaces road surface are demonstrated with meeting Gauss in RGB color R passages
Distribution, establishes the gaussian probability model based on CbCrR, enhances the accuracy of pavement detection.
3. combining the road surface characteristics such as colourity, color, texture, motion, lane line, multidimensional information probabilistic model checking is established
Road surface, improve the integrality and robustness of pavement detection.
Brief description of the drawings
Fig. 1 is the flow chart of the inventive method.
Fig. 2 is comprehensive road surface colourity and color character, establishes gaussian probability model and extraction pavement texture feature, foundation crowd
The flow chart of the optimal loose probabilistic model of number.
Embodiment
With reference to embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are unlimited
In this.
Embodiment
As shown in figure 1, a kind of utilize colourity, color, texture, motion and the car on road surface based on multidimensional information probabilistic model
The information such as diatom establish the algorithm that probabilistic model extracts complete road surface, as shown in Fig. 2 the algorithm initially sets up road surface sample
Storehouse, in YCrCb and Cr, Cb, R passage of RGB color separation road surface sample, verify that it meets Gaussian Profile and establishes Gauss
Probabilistic model;Afterwards using gray level co-occurrence matrixes extraction pavement texture feature, the optimal loose probabilistic model of mode is established;Then enter
Driveway line detects.Movable information is finally combined, by the gaussian probability model of Cr, Cb, R passage composition with utilizing pavement texture spy
The result for levying the optimal loose probabilistic model of mode established and previous frame image detection assigns different weights, sets up multidimensional letter
Probabilistic model is ceased, and using the result supplement road surface missing inspection region of lane detection, it is final to obtain accurately road surface region.The party
Method solves the problems, such as road surface region flase drop, missing inspection under complex environment.Road surface multidimensional information is utilized simultaneously, is extracted more accurate
Road surface region.
Comprise the following steps that:
1. the foundation of road surface Sample Storehouse;
Road surface picture or video of the different sections of highway under different illumination conditions are collected, road surface sample is intercepted (no with wicket
The road surface identification thing such as including lane line), establish road surface Sample Storehouse.
2. the foundation of road surface colourity, color character gaussian probability model;
In YCbCr color spaces, luminance signal is separate with carrier chrominance signal, has robust to the background under different illumination
Property, road surface sample carrier chrominance signal Cr, Cb is separated based on YCbCr color spaces, and obtained sample data is standardized, enter
Row Kolmogorov-Smirnov is examined, i.e., by the empirical distribution function of sample compared with gauss of distribution function, infers
Road surface chroma sample whether the totality from gauss of distribution function.Test result indicates that road surface Cr, Cb chromatic component obeys height
This distribution.But the Colour of the chaff interferences such as the chromatic component feature due to road surface and sky is close, easily causes flase drop.Traffic
Main chaff interference in video is trees, sky, and green (G), blue (B) color component are than great, and red (R) color component
Composition is smaller in chaff interference, while demonstrates the R component Gaussian distributed of road surface sample.Based on YCrCb color spaces and
RGB color establishes the gaussian probability model of Cr, Cb, R passage:
Wherein x is Cr, Cb, R passage vector of sample to be tested, and mean is the vectorial sample average, and Covariance is
The sample covariance matrix of Cr, Cb, R passage.I.e.:
X=[x (Cr) x (Cb) x (R)]T, (2)
Mean=[mean (Cr) mean (Cb) mean (R)]T, (3)
The elementary probability for being detected as road surface of sample to be tested can be obtained by formula (1), by introducing color space
R passages, greatly reduce flase drop region.
3. the foundation of the optimal loose probabilistic model of pavement texture feature mode;
In order to strengthen the robustness of pavement detection, learn the textural characteristics of road surface sample using gray level co-occurrence matrixes, and carry
Go out the probability distribution of the optimal loose probabilistic model analog texture feature of mode.
The gray level co-occurrence matrixes of road surface sample image are initially set up, it has reacted road surface Luminance Distribution feature, and related
The position distribution feature of luminance pixel, it is the basis for defining image texture characteristic.If f (x, y) is gray level image, size be M ×
N, grey level GN, then gray level co-occurrence matrixes M (i, j) be:
M (i, j)=Number of { f (x1,y1)=i, f (x2,y2)=j }, (5)
Condition that and if only if (6) is set up:
(x1,y1),(x2,y2) ∈ M × N, 0≤i, j≤GN, (6)
Wherein pixel (x1,y1) and (x2,y2) be distributed in diagonal.
The matrix of angle two (ASM) and contrast sub-matrix (IDM) of pavement texture, wherein angle two are calculated by gray level co-occurrence matrixes
Matrix is a measurement of intensity profile uniformity coefficient and texture thickness, when image texture is more careful, intensity profile is uniform, energy
Value is larger, conversely, smaller.Contrast sub-matrix reflects the readability of texture and regular degree, clean mark, it is regular compared with
By force, it is easy to description, value is larger;It is rambling, it is difficult to which that description, value is smaller.The two computational methods is as follows:
ASM=∑ M (i, j) × M (i, j), (7)
By the study of road pavement Sample Storehouse textural characteristics, it is found that the matrix of angle two of road surface sample and contrast scoring matrix value are equal
Meet following feature:The matrix of angle two and contrast scoring matrix value of 60% sample above are that mode is most to be worth, the feature of remaining sample
Value is less than 5 times of standard deviations with mode absolute distance.The optimal loose probabilistic model of mode is proposed accordingly to describe tested value (x)
Meet the relaxation probability of sample (probability can be less than zero):
Wherein σ is the standard deviation of sample, and m is the mode (being most worth) in sample, and μ is the most value of non-mode in sample, and k is
Slackness (generally 3), the scope of the bigger expression detections of k is bigger, and precision is smaller.When μ is minimum value, (μ ± k σ) takes minus sign,
Otherwise take plus sige.| | a-b | | represent a to b absolute distance.
The matrix of angle two of region (A) to be detected is calculated using the optimal loose probabilistic model of mode and contrast scoring matrix value meets
Road surface distribution probability be:
So as to obtain the probability that road surface is detected as based on pavement texture feature, its for different illumination conditions have compared with
Strong robustness.
4. lane detection;
Lane detection part obtains mainly in combination with the association attributes of Boundary extracting algorithm and lane line.
Because lane line can regard continuous closely-spaced bright (white or yellow) long straight line as roughly, while lane line exists
It is vertical line or camber line rather than horizontal line in image.Using this characteristic, set first and image is done into binary conversion treatment, and carry out figure
As edge extracting.Length (l) threshold value is set on this basis, chosen more than this threshold value and (α) angled with horizontal line
Long line segment carry out color mark.
On Edge extraction, HOUGH is compared, SOBEL, GABOR scheduling algorithms, the present invention have finally chosen Canny
Operator carries out computing, because the robustness and real-time of this operator are all well and good.Canny edge detection algorithms are in 1986
Itd is proposed by John Canny, which employs the method for first smooth rear derivative processing, specially following steps:
Filtering:Denoising is carried out to image using gaussian filtering, Gaussian kernel is as follows:
Practical engineering experience shows that the core that Gaussian function determines can be accurately positioned it in anti-noise jamming and rim detection
Between preferable half-way house is provided;
Enhancing:The basis at enhancing edge is to determine the changing value of image each point field intensity, and the point that will have significant change
Show especially out, the present invention carries out convolution algorithm with canny operators, and calculates gradient magnitude.Its operator is:
Detection is with being connected:According to gradient magnitude detected edge points, usual gradient magnitude is meant to be greatly the possibility at edge
It is bigger.But in order to avoid choosing the big gradient magnitude point to non-edge, take non-maxima suppression method.Maximum is chosen,
Abandon non-maximum.On this basis, canny sets dual-threshold voltage, and big threshold value is used for controlling the segmentation at edge, and small threshold
Value then controls edge connection.
By above step, the effect for extracting edge line well can be reached.In combination with lane line association attributes, compared with
Accurately detect track line position.
5. the foundation of multidimensional information probabilistic model;
In order to improve the accuracy of pavement detection and robustness, based on the study of road surface Sample Storehouse, with reference to colourity, color,
The road surface characteristics such as texture, motion, lane line, establish multidimensional information probabilistic model checking road surface.
If the i-th two field picture is f in traffic videoi(x, y), the first two field picture are f1(x, y), then meet the general of road surface characteristic
Rate p is:
Wherein P1(fi(x, y)) it is that present frame (x, y) pixel establishes Gauss model acquisition using colourity, color information
Probability, P2(fi(x, y)) it is that present frame (x, y) pixel establishes the optimal loose probabilistic model acquisition of mode using textural characteristics
Probability, P (fi-1(x, y)) be movable information using traffic video, i.e., previous frame be detected as road surface part have it is certain general
Rate is also road surface in present frame, particularly important when pavement detection is applied to video monitoring system.
Parameter alpha, β, γ are the weighted value (alpha+beta+γ=1) of corresponding probabilistic information, and 85%, 10% and is taken under general scene
5%.When pavement detection is applied to video monitoring system, γ proportion suitably increases.It can thus be concluded that pavement detection probabilistic image,
By selecting rational threshold value, you can judge road surface region.
ParameterTo be as follows with repairing of the lane detection result to missing inspection region, specific practice:Utilize lane detection
As a result, whether (right side) certain limit in have enough pixel be detected as road surface, in this way, then fill out if searching on the left of lane line
Fill lane line and the pixel between it.
When video image is the first frame, formula (15) deteriorates to:
Wherein, alpha+beta=1.
By establishing multidimensional information probabilistic model, strengthen the accuracy and robustness of pavement detection.
Above-described embodiment is the preferable embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any Spirit Essences without departing from the present invention with made under principle change, modification, replacement, combine, simplification,
Equivalent substitute mode is should be, is included within protection scope of the present invention.
Claims (5)
- A kind of 1. pavement detection method based on multidimensional information probabilistic model, it is characterised in that comprise the following steps:Step 1:Establish road surface Sample Storehouse;Step 2:Comprehensive road surface colourity and color character, establish gaussian probability model;Step 3:Pavement texture feature is extracted, establishes the optimal loose probabilistic model of pavement texture feature mode;Wherein described foundation The method for building up of the optimal loose probabilistic model of pavement texture feature mode comprises the following steps:Step 31, the gray level co-occurrence matrixes for establishing road surface sample image;Step 32, the matrix A SM of angle two and contrast sub-matrix IDM for solving pavement texture;Step 33, establish the optimal loose probabilistic model applicable elements of mode:The matrix of angle two and inverse differential square of 60% sample above Battle array value is that mode is most to be worth, and characteristic value and the mode absolute distance of remaining sample are less than 5 times of standard deviations;Definition:The optimal pine of mode Relaxation probabilistic model describes the relaxation probability that tested value x meets sample:<mrow> <mi>P</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mfrac> <mrow> <mo>|</mo> <mo>|</mo> <mi>x</mi> <mo>-</mo> <mrow> <mo>(</mo> <mi>&mu;</mi> <mo>&PlusMinus;</mo> <mi>k</mi> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <mo>|</mo> <mi>m</mi> <mo>-</mo> <mrow> <mo>(</mo> <mi>&mu;</mi> <mo>&PlusMinus;</mo> <mi>k</mi> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> </mrow> </mfrac> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mo>|</mo> <mo>|</mo> <mi>x</mi> <mo>-</mo> <mrow> <mo>(</mo> <mi>&mu;</mi> <mo>&PlusMinus;</mo> <mi>k</mi> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> <mo>&le;</mo> <mo>|</mo> <mo>|</mo> <mi>m</mi> <mo>-</mo> <mrow> <mo>(</mo> <mi>&mu;</mi> <mo>&PlusMinus;</mo> <mi>k</mi> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mo>|</mo> <mo>|</mo> <mi>x</mi> <mo>-</mo> <mrow> <mo>(</mo> <mi>&mu;</mi> <mo>&PlusMinus;</mo> <mi>k</mi> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> <mo>></mo> <mo>|</mo> <mo>|</mo> <mi>m</mi> <mo>-</mo> <mrow> <mo>(</mo> <mi>&mu;</mi> <mo>&PlusMinus;</mo> <mi>k</mi> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow>Wherein, σ is the standard deviation of sample, and m is the mode in sample, and μ is the most value of non-mode in sample, and k is slackness, works as μ For minimum value when (μ ± k σ) take minus sign, otherwise take plus sige, | | a-b | | represent a to b absolute distance;Step 34, the probabilistic model using the optimal loose probabilistic model solution pavement texture feature of mode;Using the optimal pine of mode The matrix of angle two and contrast scoring matrix value in relaxation probabilistic model calculating region to be detected meet the probability that road surface is distributed and are:<mrow> <msub> <mi>P</mi> <mn>2</mn> </msub> <mrow> <mo>(</mo> <mi>A</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mfrac> <mrow> <mi>P</mi> <mrow> <mo>(</mo> <mi>A</mi> <mi>S</mi> <mi>M</mi> <mo>(</mo> <mi>A</mi> <mo>)</mo> <mo>)</mo> </mrow> <mo>+</mo> <mi>P</mi> <mrow> <mo>(</mo> <mi>I</mi> <mi>D</mi> <mi>M</mi> <mo>(</mo> <mi>A</mi> <mo>)</mo> <mo>)</mo> </mrow> </mrow> <mn>2</mn> </mfrac> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <msub> <mi>P</mi> <mn>2</mn> </msub> <mrow> <mo>(</mo> <mi>A</mi> <mo>)</mo> </mrow> <mo>&GreaterEqual;</mo> <mn>0</mn> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <msub> <mi>P</mi> <mn>2</mn> </msub> <mrow> <mo>(</mo> <mi>A</mi> <mo>)</mo> </mrow> <mo><</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow>Wherein, A represents region to be detected;Step 4:Lane detection;Step 5:With reference to the multidimensional information of colourity, color, texture, motion and lane line, pavement detection probabilistic model is established.
- 2. the pavement detection method according to claim 1 based on multidimensional information probabilistic model, it is characterised in that in step In 1, road surface characteristic of several sections under several light conditions is collected, to establish road surface Sample Storehouse.
- 3. the pavement detection method according to claim 1 based on multidimensional information probabilistic model, it is characterised in that in step In 2, the method for establishing the gaussian probability model comprises the following steps:Step 21, the R passages for obtaining image;Step 22, image is transformed into YCrCb color spaces, extracts Cr chrominance channels and Cb chrominance channels;Step 23, Kolmogorov-Smirnov inspections are carried out, Cr passages, Cb passages and the R passages of checking road surface sample accord with Close Gaussian Profile;Step 24, with reference to Cr, Cb, R passage, establish gaussian probability model.
- 4. the pavement detection method according to claim 1 based on multidimensional information probabilistic model, it is characterised in that in step In 4, the detection method of lane detection comprises the following steps:Step 41, image is converted into gray level image;Step 42, Canny operator extractions image border;Step 43, scanning entire image simultaneously extract the line segment for meeting lane line feature;Wherein, l is line segment length, and α is the angle of line segment and vertical direction, and v and θ are corresponding threshold value.
- 5. the pavement detection method according to claim 1 based on multidimensional information probabilistic model, it is characterised in that in step In 5, the method for building up for establishing the multidimensional information probabilistic model comprises the following steps:Step 51, the i-th two field picture in traffic video is set to fi(x, y), the first two field picture are f1(x, y), then meet road surface characteristic Probability p be:Wherein, P1(fi(x, y)) it is that present frame (x, y) pixel establishes the general of Gauss model acquisition using colourity, color information Rate, P2(fi(x, y)) it is that present frame (x, y) pixel establishes the general of the optimal loose probabilistic model acquisition of mode using textural characteristics Rate, P (fi-1(x, y)) for using the movable information of traffic video, i.e., there is certain probability the part that previous frame is detected as road surface Also it is road surface in present frame;Parameter alpha, β, γ are the weighted value of corresponding probabilistic information, wherein, alpha+beta+γ=1;When pavement detection is applied to video monitoring During system, γ proportion suitably increases, it can thus be concluded that pavement detection probabilistic image, by selecting rational threshold value, you can judge Road surface region;Step 52, parameterTo be repaired with lane detection result to missing inspection region, method for repairing and mending is as follows:Whether using lane detection result, search on the left of lane line has enough pixels to be detected as road in certain limit Face, in this way, then fill lane line and the pixel between it;When video image is the first frame, formula (5) deteriorates to:Wherein, alpha+beta=1, by establishing multidimensional information probabilistic model, the accuracy and robustness of pavement detection are enhanced.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510284119.0A CN104899898B (en) | 2015-05-28 | 2015-05-28 | Pavement detection method based on multidimensional information probabilistic model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510284119.0A CN104899898B (en) | 2015-05-28 | 2015-05-28 | Pavement detection method based on multidimensional information probabilistic model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104899898A CN104899898A (en) | 2015-09-09 |
CN104899898B true CN104899898B (en) | 2018-01-05 |
Family
ID=54032543
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510284119.0A Active CN104899898B (en) | 2015-05-28 | 2015-05-28 | Pavement detection method based on multidimensional information probabilistic model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104899898B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107705312B (en) * | 2017-08-30 | 2021-02-26 | 合肥工业大学 | Method for extracting welded seam edge points based on line scanning data |
CN108960055B (en) * | 2018-05-30 | 2021-06-08 | 广西大学 | Lane line detection method based on local line segment mode characteristics |
CN110084115A (en) * | 2019-03-22 | 2019-08-02 | 江苏现代工程检测有限公司 | Pavement detection method based on multidimensional information probabilistic model |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CA2217379A1 (en) * | 1996-10-02 | 1998-04-02 | Philippe Delagnes | Process for detecting surface defects on a textured surface |
CN101377853A (en) * | 2008-09-24 | 2009-03-04 | 上海大学 | Method for extracting vehicle from colorful video image |
CN102393902A (en) * | 2011-12-01 | 2012-03-28 | 昆山市工业技术研究院有限责任公司 | Vehicle color detection method based on H_S two-dimensional histogram and regional color matching |
CN102663357A (en) * | 2012-03-28 | 2012-09-12 | 北京工业大学 | Color characteristic-based detection algorithm for stall at parking lot |
CN103310006A (en) * | 2013-06-28 | 2013-09-18 | 电子科技大学 | ROI extraction method in auxiliary vehicle driving system |
CN103455820A (en) * | 2013-07-09 | 2013-12-18 | 河海大学 | Method and system for detecting and tracking vehicle based on machine vision technology |
CN104036246A (en) * | 2014-06-10 | 2014-09-10 | 电子科技大学 | Lane line positioning method based on multi-feature fusion and polymorphism mean value |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101014125B1 (en) * | 2008-12-04 | 2011-02-14 | 재단법인대구경북과학기술원 | Method for detecting traffic sign board in intelligent vehicle and system for executing the method |
US8345100B2 (en) * | 2010-07-06 | 2013-01-01 | GM Global Technology Operations LLC | Shadow removal in an image captured by a vehicle-based camera using an optimized oriented linear axis |
-
2015
- 2015-05-28 CN CN201510284119.0A patent/CN104899898B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CA2217379A1 (en) * | 1996-10-02 | 1998-04-02 | Philippe Delagnes | Process for detecting surface defects on a textured surface |
CN101377853A (en) * | 2008-09-24 | 2009-03-04 | 上海大学 | Method for extracting vehicle from colorful video image |
CN102393902A (en) * | 2011-12-01 | 2012-03-28 | 昆山市工业技术研究院有限责任公司 | Vehicle color detection method based on H_S two-dimensional histogram and regional color matching |
CN102663357A (en) * | 2012-03-28 | 2012-09-12 | 北京工业大学 | Color characteristic-based detection algorithm for stall at parking lot |
CN103310006A (en) * | 2013-06-28 | 2013-09-18 | 电子科技大学 | ROI extraction method in auxiliary vehicle driving system |
CN103455820A (en) * | 2013-07-09 | 2013-12-18 | 河海大学 | Method and system for detecting and tracking vehicle based on machine vision technology |
CN104036246A (en) * | 2014-06-10 | 2014-09-10 | 电子科技大学 | Lane line positioning method based on multi-feature fusion and polymorphism mean value |
Non-Patent Citations (4)
Title |
---|
Color vision-based multi-level analysis and fusion for road area detection;Wu Xiaowen等;《IEEE Intelligent Vehicles Symposium》;20081231;第602-607页 * |
Detection and segmentation of concealed objects in terahertz images;Shen X等;《IEEE Transactions on Image Processing》;20081231;第17卷(第12期);第2465-2475页 * |
一种基于视频图像的道路检测方法;侯德鑫等;《仪器仪表学报》;20061231;第27卷(第6期);第338-339页 * |
一种对不确定区域再分类的路面检测算法;邓强等;《计算机工程与应用》;20101101;第46卷(第31期);第152-156页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104899898A (en) | 2015-09-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103971128B (en) | A kind of traffic sign recognition method towards automatic driving car | |
CN102375982B (en) | Multi-character characteristic fused license plate positioning method | |
CN103198315B (en) | Based on the Character Segmentation of License Plate of character outline and template matches | |
CN112819094B (en) | Target detection and identification method based on structural similarity measurement | |
CN103186904B (en) | Picture contour extraction method and device | |
CN106651872A (en) | Prewitt operator-based pavement crack recognition method and system | |
CN102509086B (en) | Pedestrian object detection method based on object posture projection and multi-features fusion | |
CN110322522B (en) | Vehicle color recognition method based on target recognition area interception | |
CN103824081B (en) | Method for detecting rapid robustness traffic signs on outdoor bad illumination condition | |
CN106650770A (en) | Mura defect detection method based on sample learning and human visual characteristics | |
CN110516550A (en) | A kind of lane line real-time detection method based on FPGA | |
CN104766071B (en) | A kind of traffic lights fast algorithm of detecting applied to pilotless automobile | |
CN102663357A (en) | Color characteristic-based detection algorithm for stall at parking lot | |
CN105894503A (en) | Method for restoring Kinect plant color and depth detection images | |
CN110852323B (en) | Angular point-based aerial target detection method | |
CN106919910B (en) | Traffic sign identification method based on HOG-CTH combined features | |
CN105005766A (en) | Vehicle body color identification method | |
CN106887004A (en) | A kind of method for detecting lane lines based on Block- matching | |
CN104766096A (en) | Image classification method based on multi-scale global features and local features | |
CN104899898B (en) | Pavement detection method based on multidimensional information probabilistic model | |
CN104143077B (en) | Pedestrian target search method and system based on image | |
CN106815583A (en) | A kind of vehicle at night license plate locating method being combined based on MSER and SWT | |
CN114219773B (en) | Pre-screening and calibrating method for bridge crack detection data set | |
CN108090459A (en) | A kind of road traffic sign detection recognition methods suitable for vehicle-mounted vision system | |
CN115527170A (en) | Method and system for identifying closing fault of door stopper handle of automatic freight car derailing brake device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |