CN106845558B - The method and apparatus of pavement detection - Google Patents
The method and apparatus of pavement detection Download PDFInfo
- Publication number
- CN106845558B CN106845558B CN201710107745.1A CN201710107745A CN106845558B CN 106845558 B CN106845558 B CN 106845558B CN 201710107745 A CN201710107745 A CN 201710107745A CN 106845558 B CN106845558 B CN 106845558B
- Authority
- CN
- China
- Prior art keywords
- image
- pavement
- gray scale
- gray
- dirt
- 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
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000001514 detection method Methods 0.000 title claims abstract description 54
- 230000011218 segmentation Effects 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims description 33
- 238000013528 artificial neural network Methods 0.000 claims description 26
- 230000006399 behavior Effects 0.000 claims description 13
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 230000008439 repair process Effects 0.000 claims description 6
- 238000000926 separation method Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 2
- 239000007787 solid Substances 0.000 claims 1
- 230000008569 process Effects 0.000 description 9
- 201000010099 disease Diseases 0.000 description 6
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 239000004568 cement Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005345 coagulation Methods 0.000 description 1
- 230000015271 coagulation Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000002945 steepest descent method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/758—Involving statistics of pixels or of feature values, e.g. histogram matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G06T5/77—
-
- G06T5/92—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
Abstract
The present invention relates to the method and apparatus of pavement detection.The described method includes: extracting the location information of road sign and/or dirt in pavement image；Road sign and/or dirt, the gray scale pavement image after being repaired are removed from the pavement image according to the positional information；The gray scale pavement image is split according to a variety of scales, calculates the gray scale pavement image corresponding texture invariant feature value under various multi-scale segmentations；The feature vector of the gray scale pavement image is obtained according to the texture invariant feature value under a variety of scales, pavement behavior is detected by described eigenvector.The present invention can be improved the accuracy of pavement detection.
Description
Technical field
The present invention relates to field of intelligent transportation technology, the device of method and pavement detection more particularly to pavement detection.
Background technique
Traditional pavement detection is that expert is needed to carry out manual identified, and detection efficiency is low, and takes time and effort.In recent years, lead to
It crosses computer progress Intelligent road to detect to obtain relatively broad popularization, after acquisition pavement image data, is based on road surface
Image data carries out the analysis of surface conditions, to carry out pavement disease classification.However, since the pavement image that actual acquisition arrives is logical
The influence that often will receive road sign or dirt, causes computer to be difficult to accurately identify the real conditions on road surface, is unfavorable for subsequent road surface
Disease classification.
Summary of the invention
Based on this, the embodiment of the invention provides the method and apparatus of pavement detection, can be improved the accurate of pavement detection
Property.
The method of one aspect of the present invention offer pavement detection, comprising:
Extract the location information of road sign and/or dirt in pavement image；
Road sign and/or dirt, the gray scale road after being repaired are removed from the pavement image according to the positional information
Face image；
The gray scale pavement image is split according to a variety of scales, calculates the gray scale pavement image in various scales
Corresponding texture invariant feature value under segmentation；
The feature vector of the gray scale pavement image is obtained according to the texture invariant feature value under a variety of scales, by described
Feature vector detects pavement behavior.
Another aspect of the present invention provides a kind of device of pavement detection, comprising:
Position detecting module, for extracting the location information of road sign and/or dirt in pavement image；
Image repair module is obtained for removing road sign and/or dirt from the pavement image according to the positional information
Gray scale pavement image after to reparation；
Separation calculation module calculates the gray scale for being split according to a variety of scales to the gray scale pavement image
Pavement image corresponding texture invariant feature value under various multi-scale segmentations；
Detection module, for obtaining the feature of the gray scale pavement image according to the texture invariant feature value under a variety of scales
Vector detects pavement behavior by described eigenvector.
The method and apparatus of the pavement detection provided based on the above embodiment, by extract pavement image in road sign and/or
The location information of dirt；Road sign and/or dirt are removed from the pavement image according to the positional information, after being repaired
Gray scale pavement image；The gray scale pavement image is split according to a variety of scales, calculates the gray scale pavement image each
Corresponding texture invariant feature value under kind multi-scale segmentation；The gray scale road is obtained according to the texture invariant feature value under a variety of scales
The feature vector of face image detects pavement behavior by described eigenvector.Thus road sign and/or the inspection of dirt road pavement can be eliminated
The influence of result is surveyed, the accuracy of pavement behavior detection is improved.
Detailed description of the invention
Fig. 1 is the schematic flow chart of the method for the pavement detection of an embodiment；
Fig. 2 is the schematic flow chart of the method for the pavement detection of another embodiment；
Fig. 3 is the structural schematic diagram of neural network employed in an embodiment；
Fig. 4 is the schematic diagram of the device of the pavement detection of an embodiment.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Fig. 1 is the schematic flow chart of the method for the pavement detection of an embodiment；As shown in Figure 1, the road in the present embodiment
Face detection method comprising steps of
S11 extracts the location information of road sign and/or dirt in pavement image.
There are road sign and dirt in pavement image as shown in Figure 2, road surface, and tradition can not by the method on COMPUTER DETECTION road surface
Road sign, dirt and road surface are distinguished, therefore cause pavement detection result inaccurate.Road sign, dirt and road under normal conditions
The inherent color in face is different, therefore corresponding pixel is also different, can pass through gray processing, binaryzation etc. line of reasoning face image recognition
Road sign and/or dirt out, and then extract road sign and/or dirt location information.
In one embodiment, gray processing processing can be carried out to original pavement image in advance, then in conjunction with adaptive threshold
Believe the position that binaryzation method and fixed threshold binaryzation method extract road sign and/or dirt from gray processing treated pavement image
Breath.
Wherein, the gray processing that the process that color image transforms into gray level image is image is handled.In color image
The color of each pixel there are tri- components of R, G, B to determine, and each component has 255 intermediate values desirable, and such a pixel can
To there is the variation range of the color of more than 1,600 ten thousand (255*255*255).And gray level image is the identical one kind of tri- components of R, G, B
Special color image, the variation range of one pixel is 255 kinds, so being transformed into image in Digital Image Processing
Gray level image can be such that subsequent image calculation amount becomes less.The description of gray level image still reflects as color image
The distribution and feature of the entirety of entire image and the coloration of part and brightness degree.Making gray processing processing to original pavement image can
It is realized using any feasible method, such as finds out the average value of tri- components of R, G, B of each pixel, then this is averaged
Value is given to three components of this pixel, obtains gray processing treated pavement image.Or the color space according to YUV
In, the physical significance of the component of Y is the brightness of point, reflects brightness degree by the value, according to the variation of RGB and YUV color space
It is corresponding with tri- color components of R, G, B that relationship can establish brightness Y: Y=0.3R+0.59G+0.11B, with the expression of this brightness value
The gray value of image obtains gray processing treated pavement image.
Wherein, image binaryzation is that the gray value of each pixel in gray processing treated image is further arranged
It is 0 or 255, that is, whole image is showed into black and white effect.The binary image obtained by choosing threshold value appropriate, still
It so can reflect pavement image entirety and local feature.Fixed threshold binaryzation method is referred to through same threshold value to grayscale image
As carrying out binaryzation, the pixel that all gray scales are greater than or equal to given threshold is judged as belonging to certain objects, by its gray scale
Value is set as 255, and otherwise these pixels are excluded other than object area, sets 0 for its gray value, indicate background or
The object area of exception.Adaptive threshold binaryzation method is that a kind of grey level histogram according to picture obtains one and is suitble to current figure
The binarization threshold of picture；Such as maximum variance between clusters (Da-Jin algorithm, OTSU), it is the gamma characteristic by image, by image point
At background and target two parts；Inter-class variance between background and target is bigger, illustrates that the two-part difference for constituting image is got over
Greatly；When partial target mistake is divided into background or part background mistake is divided into target and all two parts difference can be caused to become smaller, therefore make between class
The maximum segmentation of variance means misclassification probability minimum.
S12 removes road sign and/or dirt, the ash after being repaired from the pavement image according to the positional information
Spend pavement image.
In one embodiment, road sign and/or dirt can be filled up according to road sign and/or the peripheral image vegetarian refreshments of dirt position
Dirt, removes until by all road signs and/or dirt, and the gray scale pavement image after being repaired is equivalently employed without road sign and/or dirt
The pavement image of dirt.
S13 is split the gray scale pavement image according to a variety of scales, calculates the gray scale pavement image various
Corresponding texture invariant feature value under multi-scale segmentation.
S14 obtains the feature vector of the gray scale pavement image according to the texture invariant feature value under a variety of scales, passes through
Described eigenvector detects pavement behavior.
Under normal circumstances, no matter the pavement image on intact road surface is according to great multi-scale segmentation, the subgraph divided
Structural similarity variance under different scale can be more stable.And for the pavement image for having disease, in difference
The structural similarity variance of the subgraph obtained under multi-scale segmentation, which can exist, floats.Therefore using intact pavement image multiple dimensioned
Divide the stable feature of lower texture, by calculating gray scale pavement image corresponding texture invariant feature under various multi-scale segmentations
Value, can detect road surface whether there is disease.The program can be suitably used for the road surface (pitch, cement or concrete) of unlike material, Shandong
Stick is good, and accuracy rate is high, practical.
As shown in Fig. 2, being further detailed below by method of another embodiment to pavement detection of the invention.
S201, obtains pavement image, and road pavement image carries out gray processing processing；
S202, to gray processing, treated that image carries out OTSU binary conversion treatment and histogram equalization processing respectively.
S203, whether white spot area and the ratio of total image area are greater than the in the image after detecting OTSU binary conversion treatment
One setting ratio, if so, performing the next step suddenly, if it is not, according to image zooming-out road sign and/or dirt after OTSU binary conversion treatment
Location information, execute step S205.
In one embodiment, first setting ratio is set as 1/3, i.e., in the image after detection OTSU binary conversion treatment
Whether white spot area and the ratio of total image area are greater than 1/3, if so, performing the next step suddenly, if it is not, at according to OTSU binaryzation
The location information of image zooming-out road sign and/or dirt after reason executes step S205.
Threshold binarization treatment is fixed to the image after histogram equalization processing using preset threshold in S204, and
According to the location information of image zooming-out road sign and/or dirt after fixed threshold binary conversion treatment.
In one embodiment, the fixed threshold is set as 230, with fixed threshold 230 to histogram equalization processing after
Image threshold binarization treatment is fixed.Certainly, according to the actual situation, the fixed threshold may also be set to other values.
S205, the pixel on periphery fills up road sign and/or dirt according to the positional information, until by all road signs and/
Or dirt removal, the gray scale pavement image after being repaired.
S206 obtains preset initial segmentation scale n0, as current segmentation scale n=n0。
The gray scale pavement image is divided into the subgraph of several n × n sizes by S207.
S208, calculates separately the gray level co-occurrence matrixes of each subgraph, and calculate the gray level co-occurrence matrixes of each subgraph with
The structural similarity of the gray level co-occurrence matrixes of its four neighborhoods subgraph
Since texture is that occurred on spatial position by intensity profile repeatedly and formed, thus be separated by image space
There can be certain gray-scale relation between two pixels of certain distance, i.e., the spatial correlation characteristic of gray scale in image.Co-occurrence matrix is used
The joint probability density of the pixel of two positions defines, it not only reflects the distribution character of brightness, and also reflection has same bright
Degree or close to the position distribution characteristic between the pixel of brightness, is the second-order statistics feature of related image brightness variation；It is definition
The basis of one group of textural characteristics.Gray level co-occurrence matrixes are exactly that a kind of spatial correlation characteristic by studying gray scale describes texture
Common method.Gray level co-occurrence matrixes are counted to the situation for keeping two pixels of certain distance to be respectively provided with certain gray scale on image
It obtains.It takes any point (x, y) in image (N × N) and deviates its another point (x+a, y+b), if the gray value of the point pair
For (g1, g2).It enables point (x, y) move on entire picture, then can obtain various (g1, g2) values, if the series of gray value is r,
Then the combination of (g1, g2) shares r × r kind.For entire picture, the number of each (g1, g2) value appearance is counted, is then arranged
A square matrix is arranged into, then they are normalized to the probability P (g1, g2) occurred, such side by the total degree occurred with (g1, g2)
Battle array is known as gray level co-occurrence matrixes.
In one embodiment, these subgraphs are calculated with SSIM (Structural Similarity, structural similarity) algorithm
The structural similarity of the gray level co-occurrence matrixes of the gray level co-occurrence matrixes of picture and its four neighborhoods subgraph.Structural similarity is to weigh
The index for measuring two number bit image similarity degrees, when two images wherein one be undistorted image, another for distortion after
The structural similarity of image, the two can regard the image quality measurement index of distorted image as.Compared to used in tradition
Image quality measurement index, structural similarity can more meet judgement of the human eye to image quality in the measurement of image quality.
S209 calculates the variance of the structural similarity of all subgraphs, obtains the gray scale pavement image current
Texture invariant feature value under scale n segmentation
S210, judges whether the ratio of the picture altitude of current scale n and the gray scale pavement image is more than or equal to second
Setting ratio；If it is not, being updated according to setting step-length d to current scale n, so that n=n+d, return step S206, calculate institute
State gray scale pavement image corresponding texture invariant feature value under new multi-scale segmentation.If so, performing the next step rapid.
S211, acquisition divide obtained texture invariant feature value every time, obtain the feature vector of the gray scale pavement image
It is k=n/d that segmentation times, which can be obtained, according to current scale n, therefore obtains k texture invariant feature value, with the k
A texture invariant feature value obtains the feature vector of the gray scale pavement image
In one embodiment, if the height of the gray scale pavement image is 100, initial segmentation scale n0=10, setting step
Long d=10 (assuming that the unit of three is identical), second setting ratio are 1/2.If then first gray scale pavement image is divided into
The subgraph of dry 10 × 10 sizes, calculates the gray level co-occurrence matrixes of each subgraph, it is a little then to calculate this with SSIM algorithm
The gray level co-occurrence matrixes of image, and the structural similarity with the gray level co-occurrence matrixes of its four neighborhoods subgraph, finally calculate institute
There is the variance of the structural similarity of subgraph, the degree of stability of the gray scale pavement image texture is judged with the variance yields, referred to as
Texture invariant feature value.Ratio due to currently dividing the picture altitude of scale n and the gray scale pavement image is less than 1/2,
Therefore n=n+d=10+10=20 is updated.The gray scale pavement image is then divided into the subgraph of 20 × 20 sizes, is calculated
Texture invariant feature value of the gray scale pavement image under the multi-scale segmentation.And so on, divide the image into n × n size
Subgraph (until n >=50), calculate image texture invariant feature value at this time.Then this k=50/10=5 texture is steady
Determine texture invariant feature vector of the characteristic value as the gray scale pavement image.
Described eigenvector input is preselected trained neural network, according to the output knot of the neural network by S212
Fruit obtains pavement behavior testing result.
In one embodiment, as shown in figure 3, the neural network includes input layer, hidden layer and output layer.The mind of input layer
It is equal to the quantity of element in the texture invariant feature vector through the quantity of member, hidden layer there are 3 neurons, and output layer has 1 mind
Through member, whether the testing result exported by the neural network can have the degree value of disease or disease for road surface.It can manage
Solution, according to actually detected target, other structures form is also may be selected in the hidden layer and output layer of the neural network.
In one embodiment, in order to guarantee the accuracy detected, further include the steps that in advance being trained neural network.
Such as: the known pavement behavior of the texture invariant feature vector and its sample image of using multiple sample images is as training sample pair
Neural network is trained, until training result meets preset condition, the i.e. result and its input sample of neural network output
Error between corresponding known results is within the set range.By the corresponding feature vector input training of pavement image to be detected
Good neural network, can learn the pavement behavior of current road image out by the neural network, and the accuracy of detection is high.
In one embodiment, the neural network can be BP (BackPropagation) neural network, and RBF also can be used
(Radical Basis Function) neural network or other neural networks.BP neural network can learn and store a large amount of
Input-output mode map relationship, without disclosing the math equation for describing this mapping relations in advance.Its learning rules
It is constantly to adjust the weight and threshold value of network by backpropagation using steepest descent method, make the error sum of squares of network most
It is small.BP neural network is made of the forward-propagating of information and two processes of backpropagation of error.Each neuron of input layer is responsible for
It receives from extraneous input information, and passes to each neuron of middle layer；Middle layer is internal information process layer, is responsible for information
Transformation, according to the demand of information change ability, middle layer can be designed as single hidden layer or more hidden layer configurations；The last one hidden layer
Be transmitted to the information of each neuron of output layer, after further treatment after, the forward-propagating treatment process that once learns is completed, by defeated
Layer outwardly output information processing result out.When reality output and desired output are not inconsistent, into the back-propagation phase of error.
Error corrects each layer weight in the way of error gradient decline, to hidden layer, the layer-by-layer anti-pass of input layer by output layer.Zhou Erfu
The information forward-propagating and error back propagation process of beginning is process and neural network learning that each layer weight constantly adjusts
Trained process, the error that this process is performed until network output are reduced to acceptable degree or preset
Until learning number.RBF neural, that is, radial basis function neural network is a kind of efficient feed forward type neural network, RBF mind
It is three etale topology structures of single hidden layer through network, it has best approximation capability and global optimum's characteristic, and structure is simple, instruction
It is fast to practice speed.
The method of the pavement detection of above-described embodiment, can be accurate in conjunction with the binaryzation method and fixed threshold binaryzation method of OTSU
Road sign and/or dirt position are extracted, it is further to remove road sign and/or dirt, effective detection basis is provided for pavement detection；
And using intact pavement image, texture remains unchanged stable feature under multi-scale division, and it is good to carry out road surface by neural network
Bad judgement meets the mechanism that human eye judges pavement image quality, road surface (pitch, cement or the coagulation suitable for unlike material
Soil), robustness is good, and accuracy rate is high, practical.
It should be noted that for the various method embodiments described above, describing for simplicity, it is all expressed as a series of
Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the sequence of acts described, because according to
According to the present invention, certain steps can use other sequences or carry out simultaneously.In addition, also any group can be carried out to above-described embodiment
It closes, obtains other embodiments.
Based on thought identical with the method for the pavement detection in above-described embodiment, the present invention also provides the dresses of pavement detection
It sets, which can be used for executing the method for above-mentioned pavement detection.For ease of description, the structure of the Installation practice of pavement detection
In schematic diagram, part related to the embodiment of the present invention illustrate only, it will be understood by those skilled in the art that schematic structure is simultaneously
The not restriction of structure twin installation may include perhaps combining certain components or different than illustrating more or fewer components
Component layout.
Fig. 4 is the schematic diagram of the device of the pavement detection of one embodiment of the invention；As shown in figure 4, the present embodiment
Pavement detection device include: position detecting module 310, image repair module 320, separation calculation module 330 and detection
Module 340, details are as follows for each module:
The position detecting module 310, for extracting the location information of road sign and/or dirt in pavement image.
The position detecting module 310 in one embodiment carries out gray processing processing for road pavement image, in conjunction with adaptive
Threshold binarization method and fixed threshold binaryzation method is answered to extract road sign and/or dirt from gray processing treated pavement image
Location information
Described image repair module 320, for according to the positional information from the pavement image remove road sign and/or
Dirt, the gray scale pavement image after being repaired.
Described image repair module 320 in one embodiment, the pixel for periphery according to the positional information are filled up
Road sign and/or dirt, until the gray scale pavement image by all road signs and/or dirt removal, after being repaired.
The separation calculation module 330 calculates institute for being split according to a variety of scales to the gray scale pavement image
State gray scale pavement image corresponding texture invariant feature value under various multi-scale segmentations.
The separation calculation module 330 in one embodiment, if being n for currently dividing scale, by the gray scale road surface
Image segmentation is the subgraph of several n × n sizes；The gray level co-occurrence matrixes of each subgraph are calculated separately, and calculate each son
The structural similarity of the gray level co-occurrence matrixes of the gray level co-occurrence matrixes of image and its four neighborhoods subgraph；Calculate all subgraphs
The variance of the structural similarity obtains texture invariant feature value of the gray scale pavement image in the case where dividing current scale n.
The detection module 340, for obtaining gray scale road surface figure according to the texture invariant feature value under a variety of scales
The feature vector of picture detects pavement behavior by described eigenvector.
The detection module 340 in one embodiment, for described eigenvector input to be preselected trained nerve net
Network obtains pavement behavior testing result according to the output result of the neural network.
It should be noted that the information between each module is handed in the embodiment of the device of the pavement detection of above-mentioned example
Mutually, the contents such as implementation procedure, due to being based on same design, bring technical effect and sheet with preceding method embodiment of the present invention
Invention preceding method embodiment is identical, and for details, please refer to the description in the embodiment of the method for the present invention, and details are not described herein again.
In addition, the logical partitioning of each functional module is only to lift in the embodiment of the device of the pavement detection of above-mentioned example
Example explanation, can according to need in practical application, for example, for corresponding hardware configuration requirement or software realization convenience
Consider, above-mentioned function distribution is completed by different functional modules, i.e., divides the internal structure of the device of the pavement detection
At different functional modules, to complete all or part of the functions described above.Wherein each function mould can both use hardware
Form realize, can also be realized in the form of software function module.
It will appreciated by the skilled person that realizing all or part of the process in above-described embodiment method, being can
It is completed with instructing relevant hardware by computer program, the program can be stored in a computer-readable storage and be situated between
In matter, sells or use as independent product.When being executed, the complete of the embodiment such as above-mentioned each method can be performed in described program
Portion or part steps.Wherein, the storage medium can be magnetic disk, CD, read-only memory (Read-Only
Memory, ROM) or random access memory (Random Access Memory, RAM) etc..
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiments.
The embodiments described above only express several embodiments of the present invention, should not be understood as to the invention patent range
Limitation.It should be pointed out that for those of ordinary skill in the art, without departing from the inventive concept of the premise,
Various modifications and improvements can be made, and these are all within the scope of protection of the present invention.Therefore, the scope of protection of the patent of the present invention
It should be determined by the appended claims.
Claims (12)
1. a kind of method of pavement detection characterized by comprising
Extract the location information of road sign and/or dirt in pavement image；
Road sign and/or dirt, the gray scale road surface figure after being repaired are removed from the pavement image according to the positional information
Picture；
The gray scale pavement image is split according to a variety of scales, calculates the gray scale pavement image in various multi-scale segmentations
Under corresponding texture invariant feature value；The texture invariant feature value is for characterizing the gray scale pavement image according to various scales
Segmentation, the floating feature of the structural similarity variance for the subgraph divided；
The feature vector of the gray scale pavement image is obtained according to the texture invariant feature value under a variety of scales, passes through the feature
Vector detection pavement behavior.
2. the method for pavement detection according to claim 1, which is characterized in that detect road surface shape by described eigenvector
Condition, comprising:
Described eigenvector input is preselected into trained neural network, road surface is obtained according to the output result of the neural network
Situation detection result.
3. the method for pavement detection according to claim 1, which is characterized in that extract road sign and/or dirt in pavement image
Dirt location information include:
Road pavement image carries out gray processing processing, combining adaptive threshold binarization method and fixed threshold binaryzation method from gray processing
The location information of road sign and/or dirt is extracted in treated pavement image.
4. the method for pavement detection according to claim 3, which is characterized in that combining adaptive threshold binarization method and
Fixed threshold binaryzation method extracts the location information of road sign and/or dirt from gray processing treated pavement image, comprising:
OTSU binary conversion treatment is carried out to gray processing treated image, white point face in the image after detecting OTSU binary conversion treatment
Whether the long-pending ratio with total image area is greater than the first setting ratio；
If it is not, according to the location information of image zooming-out road sign and/or dirt after OTSU binary conversion treatment；
If so, using preset threshold, to gray processing, treated that threshold binarization treatment is fixed in image, according to fixed threshold
The location information of image zooming-out road sign and/or dirt after binary conversion treatment.
5. the method for pavement detection according to claim 4, which is characterized in that after being handled using preset threshold gray processing
Image threshold binarization treatment is fixed, comprising:
Histogram equalization processing is carried out to gray processing treated image, using preset threshold to histogram equalization processing after
Image threshold binarization treatment is fixed.
6. the method for pavement detection according to claim 1, which is characterized in that according to the positional information from the road surface
Road sign and/or dirt, the gray scale pavement image after being repaired are removed in image, comprising:
The pixel on periphery fills up road sign and/or dirt according to the positional information, goes until by all road signs and/or dirt
It removes, the gray scale pavement image after being repaired.
7. the method for pavement detection according to claim 1, which is characterized in that according to a variety of scales to the gray scale road surface
Image is split, and calculates the gray scale pavement image corresponding texture invariant feature value under various multi-scale segmentations, comprising:
If currently segmentation scale is n, the gray scale pavement image is divided into the subgraph of several n × n sizes；
The gray level co-occurrence matrixes of each subgraph are calculated separately, and calculate the gray level co-occurrence matrixes and its four neighborhood of each subgraph
The structural similarity of the gray level co-occurrence matrixes of image；
The variance for calculating the structural similarity of all subgraphs obtains the gray scale pavement image in segmentation current scale n
Under texture invariant feature value.
8. the method for pavement detection according to claim 7, which is characterized in that dividing obtaining the gray scale pavement image
After cutting the texture invariant feature value under current scale n, further includes:
Judge whether the ratio of the picture altitude of current scale n and the gray scale pavement image is more than or equal to the second setting ratio；
If it is not, being updated according to setting step-length d to current scale n, so that n=n+d, calculates the gray scale pavement image in new ruler
Corresponding texture invariant feature value under degree；
Divide obtained texture invariant feature value every time if so, obtaining, obtains the feature vector of the gray scale pavement image.
9. a kind of device of pavement detection characterized by comprising
Position detecting module, for extracting the location information of road sign and/or dirt in pavement image；
Image repair module is repaired for removing road sign and/or dirt from the pavement image according to the positional information
Gray scale pavement image after multiple；
Separation calculation module calculates the gray scale road surface for being split according to a variety of scales to the gray scale pavement image
Image corresponding texture invariant feature value under various multi-scale segmentations；The texture invariant feature value is for characterizing the gray scale road
Face image is according to various multi-scale segmentations, the floating feature of the structural similarity variance for the subgraph divided；
Detection module, for obtained according to the texture invariant feature value under a variety of scales the feature of the gray scale pavement image to
Amount, detects pavement behavior by described eigenvector.
10. the device of pavement detection according to claim 9, which is characterized in that
The position detecting module carries out gray processing processing for road pavement image, combining adaptive threshold binarization method and solid
Determine the location information that threshold binarization method extracts road sign and/or dirt from gray processing treated pavement image；
And/or
Described image repair module, the pixel for periphery according to the positional information fill up road sign and/or dirt, until inciting somebody to action
All road signs and/or dirt removal, the gray scale pavement image after being repaired；
And/or
The gray scale pavement image is divided into several n × n if being n for currently dividing scale by the separation calculation module
The subgraph of size；Calculate separately the gray level co-occurrence matrixes of each subgraph, and calculate the gray level co-occurrence matrixes of each subgraph with
The structural similarity of the gray level co-occurrence matrixes of its four neighborhoods subgraph；Calculate the side of the structural similarity of all subgraphs
Difference obtains texture invariant feature value of the gray scale pavement image in the case where dividing current scale n；
And/or
The detection module, for described eigenvector input to be preselected trained neural network, according to the neural network
Output result obtain pavement behavior testing result.
11. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In when the processor executes described program the step of any one of realization claim 1 to 8 the method.
12. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The step of claim 1 to 8 any the method is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710107745.1A CN106845558B (en) | 2017-02-27 | 2017-02-27 | The method and apparatus of pavement detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710107745.1A CN106845558B (en) | 2017-02-27 | 2017-02-27 | The method and apparatus of pavement detection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106845558A CN106845558A (en) | 2017-06-13 |
| CN106845558B true CN106845558B (en) | 2019-11-05 |
Family
ID=59133498
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710107745.1A Active CN106845558B (en) | 2017-02-27 | 2017-02-27 | The method and apparatus of pavement detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106845558B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109801282A (en) * | 2019-01-24 | 2019-05-24 | 湖北大学 | Pavement behavior detection method, processing method, apparatus and system |
| 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 |
| CN111783686A (en) * | 2020-07-03 | 2020-10-16 | 中国交通通信信息中心 | Asphalt pavement health state monitoring system and method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6958821B1 (en) * | 2000-11-21 | 2005-10-25 | Eastman Kodak Company | Analyzing images to determine third party product materials corresponding to the analyzed images |
| CN104851098A (en) * | 2015-05-22 | 2015-08-19 | 天津大学 | Objective evaluation method for quality of three-dimensional image based on improved structural similarity |
| CN105335749A (en) * | 2015-08-28 | 2016-02-17 | 浙江理工大学 | Gray-level co-occurrence matrix based method for extracting boundary line of lawn non-cutting region |
| CN105989571A (en) * | 2015-03-19 | 2016-10-05 | 英特尔公司 | Control of computer vision pre-processing based on image matching using structural similarity |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014070273A1 (en) * | 2012-11-02 | 2014-05-08 | Board Of Regents, The University Of Texas System | Recursive conditional means image denoising |
-
2017
- 2017-02-27 CN CN201710107745.1A patent/CN106845558B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6958821B1 (en) * | 2000-11-21 | 2005-10-25 | Eastman Kodak Company | Analyzing images to determine third party product materials corresponding to the analyzed images |
| CN105989571A (en) * | 2015-03-19 | 2016-10-05 | 英特尔公司 | Control of computer vision pre-processing based on image matching using structural similarity |
| CN104851098A (en) * | 2015-05-22 | 2015-08-19 | 天津大学 | Objective evaluation method for quality of three-dimensional image based on improved structural similarity |
| CN105335749A (en) * | 2015-08-28 | 2016-02-17 | 浙江理工大学 | Gray-level co-occurrence matrix based method for extracting boundary line of lawn non-cutting region |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106845558A (en) | 2017-06-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109584248B (en) | Infrared target instance segmentation method based on feature fusion and dense connection network | |
| CN109583425B (en) | Remote sensing image ship integrated recognition method based on deep learning | |
| Shen et al. | Detection of stored-grain insects using deep learning | |
| CN106920243B (en) | Improved ceramic material part sequence image segmentation method of full convolution neural network | |
| Woźniak et al. | Adaptive neuro-heuristic hybrid model for fruit peel defects detection | |
| Topouzelis et al. | Detection and discrimination between oil spills and look-alike phenomena through neural networks | |
| Sowmya et al. | Colour image segmentation using fuzzy clustering techniques and competitive neural network | |
| CN109740639B (en) | Wind cloud satellite remote sensing image cloud detection method and system and electronic equipment | |
| CN111695466B (en) | Semi-supervised polarization SAR terrain classification method based on feature mixup | |
| CN104392228A (en) | Unmanned aerial vehicle image target class detection method based on conditional random field model | |
| CN106845558B (en) | The method and apparatus of pavement detection | |
| CN108399420A (en) | A kind of visible light naval vessel false-alarm elimination method based on depth convolutional network | |
| CN112926652B (en) | Fish fine granularity image recognition method based on deep learning | |
| CN109543585A (en) | Underwater optics object detection and recognition method based on convolutional neural networks | |
| CN109685806A (en) | Image significance detection method and device | |
| CN112580662A (en) | Method and system for recognizing fish body direction based on image features | |
| CN116645592B (en) | Crack detection method based on image processing and storage medium | |
| CN108280469A (en) | A kind of supermarket's commodity image recognition methods based on rarefaction representation | |
| CN114863263B (en) | Snakehead fish detection method for blocking in class based on cross-scale hierarchical feature fusion | |
| Pramunendar et al. | New Workflow for Marine Fish Classification Based on Combination Features and CLAHE Enhancement Technique. | |
| Lei et al. | Sea-land segmentation for infrared remote sensing images based on superpixels and multi-scale features | |
| CN112598031A (en) | Vegetable disease detection method and system | |
| CN110363218A (en) | A kind of embryo's noninvasively estimating method and device | |
| CN113449811A (en) | Low-illumination target detection method based on MS-WSDA | |
| CN114863198A (en) | Crayfish quality grading method based on neural network |
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 |