CN108109156B - SAR image Approach for road detection based on ratio feature - Google Patents
SAR image Approach for road detection based on ratio feature Download PDFInfo
- Publication number
- CN108109156B CN108109156B CN201711415934.1A CN201711415934A CN108109156B CN 108109156 B CN108109156 B CN 108109156B CN 201711415934 A CN201711415934 A CN 201711415934A CN 108109156 B CN108109156 B CN 108109156B
- Authority
- CN
- China
- Prior art keywords
- pixel
- road
- image
- value
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20076—Probabilistic image processing
-
- 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
- G06T2207/30181—Earth observation
- G06T2207/30184—Infrastructure
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of SAR image Approach for road detection based on ratio feature, mainly solves the problems, such as that the prior art is high to road edge position inaccurate, false alarm rate.Its realization includes: 1) to carry out dropping spot to SAR image and extract 9 kinds of textural characteristics, and 2) it is screened from 9 kinds of textural characteristics according to Pasteur's distance and contributes maximum 3 kinds of textural characteristics to classifying；3) ratio feature R1 and similar ratio feature R2 is compared in image after node-by-node algorithm drop spot；4) with the sample architecture road dictionary D 2) with 3) result1With background dictionary D2；5) each pixel is solved respectively and road dictionary D1Mean difference E1With with background dictionary D2Mean difference E2, classified by difference to pixel, obtain Preliminary detection result；6) Preliminary detection result is optimized, obtains final having arrived testing result.The present invention can than more completely, clearly detect road in SAR image, the road suitable for detecting different directions, different in width SAR image.
Description
Technical field
The invention belongs to digital image processing techniques fields, the in particular to Approach for road detection of SAR image, can be used for ground
Figure update, transportation logistics and urban planning.
Background technique
In recent years, with the continuous development of synthetic aperture radar key technology, the continuous improvement of SAR imaging resolution, signal
Processing capacity constantly enhances, message transmission rate is continuously increased, equipment volume constantly reduces, quality constantly reduces, and SAR image can
It is widely used in information gathering, battlefield surveillance, attack guidance, Hitting Effect Evaluation etc..SAR image is born till now more than 50 from it
The time in year technically has been achieved for rapid progress and development.In SAR image analysis, line feature has very important
Meaning, because certain objects itself have linear structure, such as road, bridge, river, coastline in image.Utilize computer
The Linear feature information that road etc is automatically extracted from SAR image is the hope of people's many years.Line feature for passing more
Sensor image registration, cartography application and image segmentation and target identification etc..For there is the object of one fixed width range, accurately
Contour line facilitate the segmentation of different objects and the identification of target.
Road is the major part for constituting modern traffic system, has important geography, political affairs as important artificial atural object
It controls, various meanings such as economy.Since synthetic aperture radar SAR system has many advantages, such as round-the-clock, round-the-clock, scheme from SAR
Road is extracted as in be paid more and more attention and apply.In a large amount of High Resolution SAR Images of acquisition, the extraction of road was usually both
It is the pilot process of SAR interpretation, can also be used as a kind of interpretation result.The extraction of SAR image road is that academic circles at present is universal
The hot spot of concern.Meanwhile SAR image road extraction has important answer in the fields such as military information interpretation and civilian urban planning
With value.
General SAR image Road Detection algorithm is first to determine road seeds according to unit detection, is then examined using line
The method that road seeds are connected into the propositions such as road line segment, such as Tupin, Kartartzis etc. and Jeon by method of determining and calculating.Often
Rule these unit detection methods exist to road edge position inaccurate, false alarm rate is high the problems such as.
Summary of the invention
It is an object of the invention to be directed to the deficiency of above-mentioned prior art, a kind of SAR image based on ratio feature is proposed
Approach for road detection improves the accuracy positioned to edge to reduce false alarm rate.
To achieve the above object, implementation of the invention includes the following:
(1) SAR image to be detected is read in, and drop spot pretreatment is carried out to SAR image；
(2) gray level co-occurrence matrixes are constructed point by point to the image after drop spot, and with gray level co-occurrence matrixes calculate energy, entropy,
Contrast, mean value, variance, correlation, non-similarity, unfavourable balance away from 9 kinds of textural characteristics of uniformity；
(3) the 9 kinds of textural characteristics extracted above are optimized and is screened according to Bhattacharyya range index, only
Choosing can be respectively effectively to Road Detection in SAR image and maximum 3 kinds of classification contribution of feature: mean value, variance and right
Degree of ratio；
(4) ratio feature is extracted:
(4a) extracts two kinds of ratio features to the image after drop spot point by point, i.e., the comparison ratio between road and two side areas
Similar ratio feature R2 between feature R1 and both sides of the road region；
The 3 kinds of textural characteristics extracted in step (3) and step (4a) are extracted two kinds of ratio features and are normalized by (4b),
Obtain 5 dimensional feature vectors of each pixel；
(5) pixel of part tape label is randomly choosed as sample, including road waypoint and non-rice habitats point, with the sample structure
Produce road dictionary D1With background dictionary D2；
(6) Preliminary detection:
(6a) solves each pixel and road dictionary D respectively1Difference E1With each pixel and background dictionary D2Difference
Value E2；
(6b) is according to the two differences E1And E2Classify to pixel: if E1-E2< 0 indicates the pixel of test sample
The Atomic Correlations of point and road area are stronger, then are tentatively judged to pixel and belong to road area；If E1-E2>=0, indicate test
The pixel of sample and the Atomic Correlations of background area are stronger, then are tentatively judged to pixel and belong to background area；
(7) Preliminary detection that obtains for step (6) is as a result, leptosomatic feature further according to road, with area perimeter
The optimization algorithm of ratio excludes false-alarm region, obtains the final result of Road Detection.
Compared with the prior art, the invention has the following advantages:
1. a kind of calculation method of the present invention due to defining new ratio feature, the algorithm have fully considered road and two
Difference is big between side region, and the two sides similitude of road is strong, overcomes lacking for conventional unit detection operator edge position inaccurate
Two kinds of ratio features of point, extraction accurately can carry out edge positioning to road；
2. the present invention is for, there are redundancy, proposition screens feature with Bhattacharyya distance between textural characteristics
Rule, since Bhattacharyya distance has fully considered the correlation between the mean value, variance and standard drawing of feature, not only
The complexity for reducing feature also improves the operational efficiency of algorithm；
3. the present invention feature high to requirement of real-time for Road Detection, proposes and utilizes difference point by construction dictionary
The design of class, this has been fully considered, and similitude between similar is strong, the big feature of difference between inhomogeneity, relative to some complexities
Classifier high, practicability is low, difference classifier of the invention simplify calculating, can satisfy the requirement of real-time；
4. the optimization algorithm proposed by the present invention with area perimeter ratio, can effectively exclude false-alarm region.Due to preliminary
There are false-alarms for testing result, it is known that road all has the characteristics that leptosomatic, the i.e. constant closed curve of a length, according to integral
The characteristics of principle, shape more level off to slender type, and area is smaller, this more meets road.
Detailed description of the invention
Fig. 1 is Road Detection general flow chart of the invention；
Fig. 2 is the sub-process figure of the optimization algorithm in the present invention；
Fig. 3 is the road area schematic diagram of three kinds of different in width in the present invention；
Fig. 4 is the test image that the present invention uses；
Fig. 5 is image after the denoising in the present invention；
Fig. 6 is the Road Detection PRELIMINARY RESULTS figure in the present invention；
Fig. 7 is that the present invention uses the result figure after optimization algorithm removal false-alarm；
Fig. 8 is the standard drawing of test image.
Specific embodiment
With reference to the accompanying drawing, the present invention is described in detail:
Referring to Fig.1, steps are as follows for realization of the invention:
Step 1: input data simultaneously carries out drop spot.
The inhibition of SAR coherent spot is the key that the premise of the work such as SAR image processing and subsequent image identification.
The SAR image of input is first carried out drop spot using PPB algorithm by the present invention, obtains the image that road is more clear, with
Weaken the influence of noise, increases the contrast between road and background, beaten in order to accurately carry out the feature of Road Detection
Basis is descended.
Step 2: texture feature extraction.
SAR image includes textural characteristics abundant, if these texture informations are with reasonable, it would be beneficial in Road Detection.
Gray level co-occurrence matrixes are constructed point by point according to the image after denoising, then extract 9 kinds of textural characteristics with gray level co-occurrence matrixes, respectively
Be: energy, entropy, contrast, non-similarity, unfavourable balance away from, uniformity, mean value, variance and correlation.Textural characteristics have embodied road
The detailed information on road has apparent difference with background, can be distinguished road and background by textural characteristics.
(2a) generates the image elder generation normalized after drop spot by the probability density function of gray level co-occurrence matrixes GLCM
Gray level co-occurrence matrixes GLCM, the SAR image after normalization is described with the gray level co-occurrence matrixes GLCM of generation,
The probability density function of gray level co-occurrence matrixes GLCM is as follows:
G (i, j)={ (x1,y1),(x2,y2)∈M×Nf(x1,y1)=i, f (x2,y2)=j },
Wherein, g (i, j) indicates the frequency of element in SAR image, and also referred to as GLCM probability density function, M are SAR image f
The transverse dimensions of (x, y), N are longitudinal dimension of SAR image f (x, y), (x1,y1) and (x2,y2) it is image SAR image f (x, y)
Two pixels, f (x1,y1) and f (x2,y2) it is SAR image (x1,y1) and (x2,y2) gray value, i and j are to indicate pixel f
(x1,y1) and f (x2,y2) gray value size numerical value；
(2b) according to gray level co-occurrence matrixes probability density function g (i, j) and normalization after SAR image at a distance of (Δ x,
The number that two gray-scale pixels when Δ y) occur simultaneously, generates gray level co-occurrence matrixes p (i, j), if grey level is divided into w
Grade, can obtain the gray level co-occurrence matrixes of a w × w；
(2c) calculates separately 9 kinds of common textural characteristics according to the gray level co-occurrence matrixes p (i, j) of generation:
(2c1) calculates energy feature Ene:
Energy is the parameter for measuring gray-value variation stable state in image, it can describe the journey of image texture precision
Degree and pixel gray value uniform state, calculation formula are as follows:
Wherein p (i, j) gray level co-occurrence matrixes, since gray level co-occurrence matrixes value majority is 1, Shao Shuowei in homogeneous region
0, thus energy value is big, shows that current texture is a kind of texture that rule variation is relatively stable；Due to ash in non-homogeneous region
Spending co-occurrence matrix value majority is 0, and minority is 1, thus energy value is smaller, shows that current texture is a kind of to change unstable line
Reason.
(2c2) calculates entropy feature Ent:
Entropy is to describe the measurement that random information is how many in image, and calculation formula is as follows:
When all element value is distributed with the matrix element value in gray level co-occurrence matrixes p (i, j) relatively or in matrix
It is maximum uncertain when, entropy is bigger, shows that image is disorderly and unsystematic, texture is more many and diverse；
(2c3) calculates contrast metric Con:
Contrast is to measure in image the how many parameter of varied number locally occur, and calculation formula is as follows:
The place for occurring localized variation in image is more, and contrast value is bigger, and the construction area in SAR image is due to mirror surface
It is mingled with blackspot between the speck that reflection generates, so its contrast value is larger；
(2c4) calculates non-similarity feature Dis:
Similitude is to measure in image the how many parameter of varied number locally occur, and calculation formula is as follows:
If local contrast is higher, non-similarity is also higher；
(2c5) calculates unfavourable balance moment characteristics Idm:
Inverse difference moment: being to investigate matrix element value in gray level co-occurrence matrixes to be gathered in the how many ginseng of leading diagonal bigger numerical
Number, calculation formula are as follows:
If inverse difference moment value is bigger, illustrate that image is uniform；
(2c6) calculates homogeneity Hom:
Uniformity is to investigate matrix element value in gray level co-occurrence matrixes to be gathered in the how many ginseng of leading diagonal bigger numerical
Number, calculation formula are as follows:
If uniformity value is bigger, illustrate that image is uniform；
(2c7) calculates characteristics of mean Mea:
Mean value: being the parameter for describing matrix element value in gray level co-occurrence matrixes, and calculation formula is as follows:
Mean value reflects the central tendency of element value in gray level co-occurrence matrixes, usually can be used as square in gray level co-occurrence matrixes
Array element one estimated value of element.
(2c8) calculates Variance feature Var:
Variance: being the parameter for describing matrix element value in gray level co-occurrence matrixes and deviateing equal extent value, and calculation formula is as follows:
Variance and mean value reflect the uniformity coefficient of image jointly；
(2c9) calculates correlative character Cor:
Correlation: being the parameter for describing correlation between each pixel gray value in image, from line direction and column direction
The similarity degree between gray level co-occurrence matrixes element is investigated respectively, can distinguish that the grain direction in image, calculation formula are as follows:
Step 3: Feature Selection
According to the imaging characteristics of road area in SAR image, only select to be suitble to the feature of this kind of SAR image as with generation
On the one hand the feature of table can reduce interference of the redundancy garbage to useful feature information, inspection on the other hand can be improved
Survey efficiency.In the present invention, the optimization and screening of feature are carried out according to Bhattacharyya distance, this selection feature
Method be it is simple and quick and effective, define Bhattacharyya distance it is as follows:
Wherein, μ1、σ1Respectively indicate the mean value and variance of first kind atural object pixel value on same texture template image；μ2、σ2
The mean value and variance of the second class atural object pixel value on same texture template image are respectively indicated, BD value is bigger, it was demonstrated that this feature area
Divide the ability of these two types of atural objects stronger.
Since there are redundancies between textural characteristics, calculated to simplify, this example is according to Bhattacharyya range index
The optimization and screening of feature are carried out, in concrete operations, road and background can be distinguished by selecting from 9 class textural characteristics
3 kinds of best features are mean value u, variance var and contrast cor respectively.
Step 4: extracting ratio feature
Textural characteristics cannot completely express the characteristics of road area in SAR image, fixed to road edge in order to extract
The accurate feature in position, the present invention extracts two kinds of ratio features to the image after drop spot point by point, i.e., between road and two side areas
Compare the similar ratio feature R2 between ratio feature R1 and both sides of the road region.Extraction step is as follows:
4a) extract 15 × 15 image block Q point by point to the image after drop spot；
The comparison ratio feature R1 of central pixel point 4b) is calculated in the center image block P of image block Q center extraction 7 × 7:
R1=min (max (u3/u1,u1/u3),max(u2/u1,u1/u2))
Wherein, u1Indicate the mean value of pixel in region 1, u2Indicate the mean value of pixel in region 2, u3It indicates in region 3
The mean value of pixel.
Referring to Fig. 3, the case where calculating 3 kinds of different roads width, simultaneously retains maximum value；According to the resolution ratio and reality of SAR image
The width of border road, road are usually exactly the width of 1-3 pixel in SAR image, and wherein Fig. 3 (a) indicates road width
The case where for 1 pixel, Fig. 3 (b) indicate the case where road width is 2 pixels, and Fig. 3 (c) indicates that road width is 3
The case where pixel, region 1 are the road assumed, and region 3 is the background area on the left of road, and region 2 is the back on the right side of road
Scene area.The average ratio value between region 1 and region 3, region 1 and region 2 is calculated separately in the present invention, i.e., maximum is first taken to take again
Minimum, this example are that the mutual ratios of 3 mean value of 1 mean value of region and region take maximum, 2 mean value of 1 mean value of region and region it is mutual
Ratio takes maximum；Then it is taken from above-mentioned maximum the two minimum to get to the comparison ratio feature of road area and background area；
It calculates separately the value of the comparison ratio feature R1 in the case of 3 kinds of roads width and is maximized, temporarily as in center image block P
The comparison ratio feature of imago vegetarian refreshments；
4c) 22.5 degree are rotated for image block Q counterclockwise centered on central pixel point, execute step 4b), obtain image
Block central pixel point corresponding comparison ratio feature R1 in the angle, in order to have found that closest to the direction of road, one
Corotating 7 times, 8 direction comparison ratio features have always been obtained；
Contrast 4d) according to road and two side areas is larger, and comparison ratio feature R1 value is bigger, and pixel is road
The bigger feature of a possibility that putting, this example consider 8 directions altogether, compare 8 comparison ratio feature R1 and retain most
Big value, as the comparison ratio feature R1 of image block Q central pixel point, since center image block P is extracted from image block Q
Out, thus R1 be also center image block P central pixel point comparison ratio feature.It is special to retain comparison ratio in calculating process
Image block Q when R1 maximum is levied, using the direction of image block Q as closest to the direction of road.
7 × 7 center image block P 4e) is extracted in the image block Q retained from step 4d)；
4f) calculate the similar ratio feature R2 of the central pixel point of image block P:
R2=max (u3/u2,u2/u3)
Wherein, u2Indicate the mean value of pixel in region 2, u3Indicate the mean value of pixel in region 3.
Similar ratio feature R2 represents the size of both sides of the road region similitude, and referring to Fig. 3, Fig. 3 indicates that road width is
The case where 1-3 pixel, calculates separately the average ratio value in region 2 and region 3 in the present invention and takes big；Calculate separately 3 kinds of roads
Similar ratio feature R2 in the case of degree of having a lot of social connections retains minimum value, as the similar ratio feature of image block central point, the value of R2
A possibility that central pixel point of closer 1, image block P are road waypoint is bigger.
Step 5: construction dictionary.
To the image zooming-out mean value of input, variance, contrast, comparison ratio feature R1 and totally 5 kinds of similar ratio feature R2
Feature is normalized, and obtains a series of samples with 5 dimensional feature vectors, these samples are with markd, wherein one
Belong to the road area in image, another part belongs to the background area in image.Go out road word with these sample architectures
Allusion quotation D1With background dictionary D2；
Step 6: Preliminary detection.
Each pixel and road dictionary D are solved respectively1Difference E1With each pixel and background dictionary D2Difference
E2, according to the two differences E1And E2Classify to pixel, tentatively judges that pixel belongs to road area or background area
Domain.
It is solved respectively and road dictionary D by sample of the following formula to input1Difference E1With with background dictionary D2Difference
Value E2:
Wherein, x is the test sample of input,For road dictionary D1K-th of atom, p indicate dictionary D1In atom
Number；For background dictionary D2First of atom, q indicate dictionary D2In atom number；
Classify by comparing the size of two differences to the pixel of test sample, obtains the preliminary knot of Road Detection
Fruit:
If E1-E2The Atomic Correlations of < 0, the pixel and road area that indicate test sample are stronger, then are tentatively judged to picture
Vegetarian refreshments belongs to road area；
If E1-E2>=0, the Atomic Correlations of the pixel and background area that indicate test sample are stronger, then are tentatively judged to picture
Vegetarian refreshments belongs to background area；
The Preliminary detection that this step obtains is as shown in Figure 6.
Step 7: result optimizing.
By the general outline of the available road of Preliminary detection, how true road is therefrom determined, and exclude void
Alert interference, is a very crucial step.The characteristics of present invention is according to road itself, i.e., road is leptosomatic, and profile is one
The characteristics of closed curve, for the shape of this curve closer to slender type, area is smaller, more meets road, it is former according to integral
Reason proposes the optimization algorithm of following area perimeter ratio, effectively to exclude false-alarm region.
Referring to Fig. 2, this step is accomplished by
A Preliminary detection result figure I to be optimized 7a) is inputted, referring to Fig. 6, size is m × n, with Preliminary detection knot
The center that fruit schemes I is fixed point, and it is (m+2) × (n+2) that the size of image, which is expanded, will expand the boundary come out and is both configured to 0, i.e.,
For background, the image after expansion isIt initializes in connected domain as the number S of the pixel of road*=0 and connected domain perimeter
L*=0；
7b) to the image after expansionStart to scan from pixel (2,2), skip all boundary points, until encountering
The pixel (x, y) that one pixel value is 1, enables S*=1, the pixel value of pixel (x, y) is set as 3；
It 7c) takes its 4 neighborhoods to pixel (x, y), i.e., in the window of the 3*3 put centered on (x, y), searches up and down
The number m for the point that pixel value is 0 in 4 neighborhoods, uses L*The value of+m updates L*；
Its 8 neighborhoods 7d) are taken to pixel (x, y), i.e., search 8 neighborhood of surrounding in the window of the 3*3 put centered on (x, y)
The number n for the point that interior pixel value is 1, uses S*The value of+n updates S*, this n pixel value is changed to 3 from 1 one by one, and record these pictures
Position where vegetarian refreshments；
7e) to the n pixel recorded in step 7d), step 7c is repeated one by one) and step 7d), until pixel
It can not find the point that pixel point value is 1 in (x, y) and its 8 neighborhoods, i.e., a connected domain traversed one time, recorded in this connected domain
For the number S of the pixel of road*And the perimeter L of connected domain*If the side length of pixel is set as 1, pixel number
S*It is equivalent to the area of road in connected domain；
Threshold T 7f) is set, with updated S*With updated L*It calculates road and approaches rateAgain by road
It approaches rate to be compared with threshold value: if road P*> T is then to be changed to inner the recorded pixel value of step 7d) those of 3 o'clock
0, it is otherwise constant, reinitialize S*=0, L*=0；
The setting of threshold value is the value of a floating, according to the requirement flexible choice of detection, if the value setting of T is smaller,
The effect for removing false-alarm is also better, but omission factor can be higher simultaneously, if the value setting of T is larger, the road of detection is relatively more
Add whole, but the removal effect of false-alarm is relatively poor, this example sets threshold T as 0.35；
7g) to the image after expansionStart to scan from pixel (x+1, y+1), repeat step 7b) arrive step
7f), until the last one point pixel (m, n) of Preliminary detection result images；The image after all expansionsMiddle pixel value is
3 point value is changed to 1 again, and the boundary after expansion is removed, after just obtaining the optimization of image to be detected Road Detection as a result, as schemed
7。
Effect of the invention can be further illustrated by following simulation result.
1. simulated conditions
The present invention is under the programmed environment of Visual Studio 2010 and MFC technology is combined to carry out emulation experiment.
It is 4 meters that the size for being detected SAR image, which is 329 × 329, resolution ratio, as shown in Figure 4.
2. emulation content
Emulation 1 carries out PPB drop spot processing with present invention SAR image shown in Fig. 4, as a result such as Fig. 5.It can be bright from Fig. 5
Aobvious to find out, the image after spot drops in the present invention is more clear, and roadway characteristic is more obvious, this extracts next step can be accurate
The feature for representing road is laid a good foundation；
Emulation 2 is classified with each pixel of the present invention to image such as Fig. 5 after drop spot, obtains Preliminary detection knot
Fruit figure, such as Fig. 6.In terms of Fig. 6 testing result, tentatively the general outline of road and trend detected, but still deposit
By non-rice habitats point error detection at road waypoint the problem of, so taking post-processing to Preliminary detection result for optimization algorithm
It is necessary；
The characteristics of emulation 3, present invention combination road itself, i.e., road is all leptosomatic linear target, with area perimeter
The optimization algorithm of ratio optimizes Preliminary detection result figure as shown in FIG. 6, be effectively removed isolated noise spot and
Block-like false-alarm region, as a result such as Fig. 7.Testing result from the result of Fig. 7 as can be seen that after optimization algorithm eliminates
Some noise spots of Preliminary detection result figure and block-like false-alarm region；
Fig. 7 is compared with the standard drawing of corresponding test image shown in Fig. 8, it can be seen that road inspection of the invention
It surveys result and the degree that is consistent of standard drawing is very high, show that the present invention has good Road Detection effect.
Claims (6)
1. a kind of SAR image Approach for road detection based on ratio feature, comprising:
(1) SAR image to be detected is read in, and drop spot pretreatment is carried out to SAR image；
(2) gray level co-occurrence matrixes are constructed to the image after drop spot point by point, and calculate energy, entropy, comparison with gray level co-occurrence matrixes
Degree, mean value, variance, correlation, non-similarity, unfavourable balance away from 9 kinds of textural characteristics of uniformity；
(3) the 9 kinds of textural characteristics extracted above are optimized and is screened according to Bhattacharyya range index, only chosen
Can be respectively: mean value, variance and comparison effectively to Road Detection in SAR image and maximum 3 kinds of the feature of classification contribution
Degree；
(4) ratio feature is extracted:
(4a) extracts two kinds of ratio features to the image after drop spot point by point, i.e., the comparison ratio feature between road and two side areas
Similar ratio feature R2 between R1 and both sides of the road region；
The 3 kinds of textural characteristics extracted in step (3) and step (4a) are extracted two kinds of ratio features and are normalized by (4b), are obtained
5 dimensional feature vectors of each pixel；
(5) pixel of part tape label is randomly choosed as sample, including road waypoint and non-rice habitats point, is gone out with the sample architecture
Road dictionary D1With background dictionary D2；
(6) Preliminary detection:
(6a) solves each pixel and road dictionary D respectively1Difference E1With each pixel and background dictionary D2Difference E2；
(6b) is according to the two differences E1And E2Classify to pixel: if E1-E2< 0, indicate test sample pixel and
The Atomic Correlations of road area are stronger, then are tentatively judged to pixel and belong to road area；If E1-E2>=0, indicate test sample
Pixel and background area Atomic Correlations it is stronger, then be tentatively judged to pixel and belong to background area；
(7) Preliminary detection that obtains for step (6) is as a result, leptosomatic feature further according to road, with area perimeter ratio
Optimization algorithm excludes false-alarm region, obtains the final result of Road Detection.
2. according to the method described in claim 1, wherein in step (3) according to Bhattacharyya range index to step (2)
The 9 kinds of textural characteristics extracted optimize screening, are to optimize screening to every kind by following range formula:
Wherein, BD indicates Bhattacharyya range index, μ1、σ1Respectively indicate first kind atural object on same texture template image
The mean value and variance of pixel value；μ2、σ2Respectively indicate the mean value of the second class atural object pixel value and side on same texture template image
Difference.
3. being carried out as follows according to the method described in claim 1, wherein calculating comparison ratio feature R1 in step (4):
4a) extract 15 × 15 image block Q point by point to the image after drop spot；
4b) in the center image block P of image block Q center extraction 7 × 7, the feelings that road width is 1,2,3 pixel are calculated separately
Condition obtains 3 comparison ratio features, and takes maximum value therein, temporarily special as the comparison ratio of image block P central pixel point
Sign；
22.5 degree 4c) are rotated counterclockwise centered on central pixel point to image block Q, return step 4b), a corotating 7 times obtains
To 8 comparison ratio features；
Maximum value 4d) is found out from 8 comparison ratio features, the comparison ratio feature R1 as image block P central pixel point:
R1=min (max (u3/u1,u1/u3),max(u2/u1,u1/u2)),
Wherein, u1Indicate the mean value of pixel in region 1, u2Indicate the mean value of pixel in region 2, u3Indicate pixel in region 3
The mean value of point；
The image block Q in corresponding direction when retaining comparison ratio feature R1 maximum simultaneously, as closest to the direction of road.
4. being carried out as follows according to the method described in claim 1, wherein calculating likelihood ratio value tag R2 in step (4):
7 × 7 center image block P 4e) is extracted in the image block Q retained from step (4d)；
4f) calculate separately road width be 1,2,3 pixel the case where, obtain 3 similar ratio features, and take it is therein most
Small value, the similar ratio feature R2 as image block P central pixel point:
R2=max (u3/u2,u2/u3),
Wherein, u2Indicate the mean value of pixel in region 2, u3Indicate the mean value of pixel in region 3.
5. according to the method described in claim 1, wherein solving each pixel and road dictionary D1Difference E1With each pixel
Point and background dictionary D2Difference E2, it is carried out respectively with following formula:
Wherein, x is the test sample of input,For road dictionary D1K-th of atom, p indicate dictionary D1In atom number；For background dictionary D2First of atom, q indicate dictionary D2In atom number.
6. according to the method described in claim 1, wherein excluding false-alarm area with the optimization algorithm of area perimeter ratio in step (7)
Domain carries out as follows:
Preliminary detection result images I 7a) is inputted, it is fixed point with the center of Preliminary detection result images I that size, which is m × n,
It is (m+2) × (n+2) that the size of image, which is expanded, will expand the boundary come out and is both configured to 0, as background, the image after expansion
ForIt initializes in connected domain as the number S of the pixel of road*=0 and connected domain perimeter L*=0；
7b) to image after expansionStart to scan from pixel (2,2), skip all boundary points, until encountering first pixel
The pixel (x, y) that value is 1, enables S*=1, the pixel value of pixel (x, y) is set as 3；
Its 4 neighborhoods 7c) are taken to pixel (x, y), are searched the number m for the point that pixel value is 0 in 4 neighborhoods, are used L*The value of+m updates
L*；
Its 8 neighborhoods 7d) are taken to pixel (x, y), are searched the number n for the point that pixel value is 1 in 8 neighborhoods, are used S*The value of+n updates
S*, the pixel value of this n point is changed to 3 one by one, and record the position where these pixel values；
Step 7c 7e) is repeated to above-mentioned n point) and step 7d), it can not find until in pixel (x, y) and its 8 neighborhoods
The point that pixel value is 1 obtains in this connected domain as the number S of the pixel of road*And the perimeter L of connected domain*；
Threshold T 7f) is set, with updated S*With updated L*It calculates road and approaches rateRoad is approached again
Rate is compared with threshold value: if road P*> T is then to be changed to 0 inner the recorded pixel value of step 7d) those of 3 o'clock, no
It is then constant, reinitialize S*=0, L*=0；
7g) to image after expansionStart to scan from pixel (x+1, y+1), repeat step 7b) arrive step 7f), Zhi Daochu
The last one pixel (m, n) for walking detection result image I, image after all expansionsThe value weight for the point that middle pixel value is 3
It newly is changed to 1, removes image after expansionBoundary, obtain the final optimization pass result of Road Detection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711415934.1A CN108109156B (en) | 2017-12-25 | 2017-12-25 | SAR image Approach for road detection based on ratio feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711415934.1A CN108109156B (en) | 2017-12-25 | 2017-12-25 | SAR image Approach for road detection based on ratio feature |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108109156A CN108109156A (en) | 2018-06-01 |
| CN108109156B true CN108109156B (en) | 2019-10-11 |
Family
ID=62212580
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711415934.1A Active CN108109156B (en) | 2017-12-25 | 2017-12-25 | SAR image Approach for road detection based on ratio feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108109156B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110188889A (en) * | 2019-04-28 | 2019-08-30 | 温州市兴海市政建设有限公司 | A kind of urban road surfaces maintenance management system based on preventative maintenance |
| CN117809052A (en) * | 2024-03-01 | 2024-04-02 | 海豚乐智科技(成都)有限责任公司 | Block target detection and feature extraction method, device and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102592114A (en) * | 2011-12-26 | 2012-07-18 | 河南工业大学 | Method for extracting and recognizing lane line features of complex road conditions |
| CN102915441A (en) * | 2012-08-30 | 2013-02-06 | 杭州电子科技大学 | Method for detecting unstructured road on basis of Haar texture |
| CN104392212A (en) * | 2014-11-14 | 2015-03-04 | 北京工业大学 | Method for detecting road information and identifying forward vehicles based on vision |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7869667B1 (en) * | 2007-01-19 | 2011-01-11 | Google Inc. | Method for automatic alignment of raster data with vector data in a geographic information system |
| CN101739667B (en) * | 2009-12-04 | 2012-06-20 | 西安电子科技大学 | Non-downsampling contourlet transformation-based method for enhancing remote sensing image road |
| CN101982835B (en) * | 2010-11-12 | 2012-02-08 | 西安电子科技大学 | Level set method for edge detection of SAR images of airport roads |
| CN104331873A (en) * | 2013-07-22 | 2015-02-04 | 浙江大学 | Method for detecting road from single image |
-
2017
- 2017-12-25 CN CN201711415934.1A patent/CN108109156B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102592114A (en) * | 2011-12-26 | 2012-07-18 | 河南工业大学 | Method for extracting and recognizing lane line features of complex road conditions |
| CN102915441A (en) * | 2012-08-30 | 2013-02-06 | 杭州电子科技大学 | Method for detecting unstructured road on basis of Haar texture |
| CN104392212A (en) * | 2014-11-14 | 2015-03-04 | 北京工业大学 | Method for detecting road information and identifying forward vehicles based on vision |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108109156A (en) | 2018-06-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110210463B (en) | Precise ROI-fast R-CNN-based radar target image detection method | |
| Lv et al. | Land cover change detection with heterogeneous remote sensing images: Review, progress, and perspective | |
| Henry et al. | Road segmentation in SAR satellite images with deep fully convolutional neural networks | |
| Lechner et al. | Remote sensing of small and linear features: Quantifying the effects of patch size and length, grid position and detectability on land cover mapping | |
| CN111476159B (en) | Method and device for training and detecting detection model based on double-angle regression | |
| CN108596055B (en) | Airport target detection method of high-resolution remote sensing image under complex background | |
| CN110378308B (en) | Improved port SAR image near-shore ship detection method based on fast R-CNN | |
| CN110009010B (en) | Wide-width optical remote sensing target detection method based on interest area redetection | |
| CN112183432B (en) | Building area extraction method and system based on medium-resolution SAR image | |
| Lei et al. | Multiscale superpixel segmentation with deep features for change detection | |
| CN109784392B (en) | Hyperspectral image semi-supervised classification method based on comprehensive confidence | |
| CN111080629A (en) | Method for detecting image splicing tampering | |
| CN109409263A (en) | A kind of remote sensing image city feature variation detection method based on Siamese convolutional network | |
| CN108898065B (en) | Deep network ship target detection method with candidate area rapid screening and scale self-adaption | |
| CN109871875B (en) | Building change detection method based on deep learning | |
| CN111639587B (en) | Hyperspectral image classification method based on multi-scale spectrum space convolution neural network | |
| CN107622239B (en) | Detection method for remote sensing image specified building area constrained by hierarchical local structure | |
| Peng et al. | Object-based change detection from satellite imagery by segmentation optimization and multi-features fusion | |
| Lei et al. | End-to-end change detection using a symmetric fully convolutional network for landslide mapping | |
| CN111898627B (en) | SVM cloud microparticle optimization classification recognition method based on PCA | |
| CN108492298A (en) | Based on the multispectral image change detecting method for generating confrontation network | |
| CN108109156B (en) | SAR image Approach for road detection based on ratio feature | |
| CN112348750B (en) | SAR image change detection method based on threshold fusion and neighborhood voting | |
| Huang et al. | Superpixel-based change detection in high resolution sar images using region covariance features | |
| Xiong et al. | Pixel-Level patch detection from full-scale asphalt pavement images based on deep learning |
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 |