CN104008555B - A kind of curve detection method cumulative based on backtracking - Google Patents
A kind of curve detection method cumulative based on backtracking Download PDFInfo
- Publication number
- CN104008555B CN104008555B CN201410271291.8A CN201410271291A CN104008555B CN 104008555 B CN104008555 B CN 104008555B CN 201410271291 A CN201410271291 A CN 201410271291A CN 104008555 B CN104008555 B CN 104008555B
- Authority
- CN
- China
- Prior art keywords
- curve
- point
- backtracking
- psd
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 14
- 230000001186 cumulative effect Effects 0.000 title claims abstract 7
- 238000000034 method Methods 0.000 claims description 16
- 238000009825 accumulation Methods 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 11
- 239000000284 extract Substances 0.000 abstract description 3
- 238000010276 construction Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 8
- 238000005381 potential energy Methods 0.000 description 5
- 238000004088 simulation Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of curve detection method cumulative based on backtracking, by analyzing the characteristic measure of the gray feature curve construction of curve;Structure energy function makes the energy minimum of curve point;Any chosen in curve scans for obtaining characteristic image as starting point, use shortest path first;Each cumulative obtaining of backtracking that click on searched is recalled cumulative figure;Selected threshold in the cumulative figure of backtracking, extracts and obtains final curve.The present invention can the curve extracted in image rapidly and efficiently in the case of having only to use a starting point.
Description
Technical Field
The invention relates to the technical field of information science, in particular to a curve detection method based on backtracking accumulation.
Background
At present, the shortest path technology has a good application in detecting simple open curves in images. The shortest path technique can be used to search a path and minimize the total energy of the search, and can be solved by the following mathematical problem:
wherein,is shown as starting point psMinimum energy to an arbitrary point p, A (p)sP) denotes the starting point psSet of paths to an arbitrary point p in the image, E (l) energy of curve l, l(s) ∈ RnIs a parameterized form at arc length s, P (l(s)) Representing potential energy, and omega is a correction term.
Short path techniques were first used for active edge models, extracting curves by minimizing internal and external energies, due to the use of energy minimization criteria, but active edge models are sensitive to the location of the initial active edge. Later, other scholars obtained curves by solving Eikonal's equation to minimize the total energy, setting the start and end points of the curve, but were more susceptible in the continuous region due to the correlation of this method with the active edge. Recently, some scholars have proposed methods using less a priori information, such as a shortest path method called keypoint detection, which only needs to provide a starting point to search forward the whole curve, but still needs to know the length of the curve to terminate the whole search process.
It is easy to find that these methods require a lot of a priori knowledge, such as the start and end points of a given curve, and the length of the curve as a priori condition if only one start point is given. When the curve has many branches, one end point is required for each branch to obtain a good detection result.
Therefore, the main research in the aspect of curve detection at present is to obtain a good curve extraction effect on the premise of only needing little prior knowledge.
Disclosure of Invention
The invention aims to solve the technical problem of overcoming the dependence of the existing curve detection method on prior knowledge, and provides a curve detection method based on backtracking accumulation, which can efficiently extract a curve in an image on the premise of only acquiring one starting point. The backtracking accumulation means that: after the shortest path search, after each point is traced back to the starting point or a certain number of points in the reverse direction, the curve point is visited with high frequency, but the non-curve point is visited less, so that the curve point is successfully separated from the non-curve point, and finally the curve contour is drawn.
In order to achieve the purpose, the invention adopts the technical scheme that:
the invention relates to a curve detection method based on backtracking accumulation, which comprises the following steps:
step 1: analyzing the gray scale feature of any point p in the image I to be processed, constructing a feature metric value M (p) of the curve feature of the description point p, and constructing a curve point pcCharacteristic metric value of (M) (p)c) Greater than the non-curve point pnon-cCharacteristic metric value of (M) (p)non-c) I.e. M (p)c)>M(pnon-c);
Step 2: constructing an energy function P (p) by means of the constructed characteristic metric values M (p), the constructed energy function corresponding to the curve point pcEnergy P (P) ofc) Less than the non-curve point pnon-cEnergy P (P) ofnon-c) I.e. P (P)c)<P(pnon-c);
And step 3: selecting any point p on the curvesAs starting point, ps∈pcSearching the image I to be processed by using the shortest path algorithm, and searching the passing point ppsdCharacteristic metric value of (M) (p)psd) Form a characteristic image If;
And 4, step 4: in the obtained characteristic image IfRecord the point p passed by the searchpsdFor each point p through which the search passespsdBacktracking and accumulating to obtain a backtracking and accumulating graph
And 5: for backtracking accumulation chartSelecting a threshold lambdafSaid threshold value λfIn the range of [10-3,103]So thatIs greater than threshold lambdafIs a curve point pcAll curve points pcForm the finally obtained graph Ic。
Further, the constructed energy function p (p) is the inverse of the characteristic metric m (p).
Further, the shortest path algorithm searches the image to be processed I without specifying a search end point.
Further, for the backtracking accumulation graphAt any point p inbkFrom the characteristic image IfPoint p searched inpsdBacktracking to the starting point by pbk,pbkAccumulating graphs in backtrackingValue of (A)For all backtracking passes pbkP of (a)psdCharacteristic metric value of (M) (p)psd) And (4) summing.
Compared with the prior art, the invention has the beneficial effects that:
the method comprises the steps of firstly constructing curve feature expression according to the gray level distribution features of the curve, further representing potential energy of each point in an image, then searching under the condition of giving a starting point by using a shortest path algorithm, backtracking each point according to a feature graph obtained by searching the shortest path, accumulating the feature value of each point backtracked to the starting point, and finally giving a proper threshold value to filter out the points with smaller accumulated values, thereby obtaining the final curve. The invention can quickly and effectively detect and extract the curve in the image on the premise of only needing one starting point and not needing any other prior information.
Drawings
FIG. 1 is a diagram of a simulation of an image to be processed according to the present invention;
FIG. 2 is a feature image of a simulated graph obtained using a shortest path algorithm in accordance with the present invention;
FIG. 3 is a graph of curve I extracted from a simulation graph based on a backtracking accumulation method according to the present inventionc;
FIG. 4 is a diagram of an actual crack of an image to be processed according to the present invention;
FIG. 5 is a feature image of an actual crack map obtained using the shortest path algorithm of the present invention;
FIG. 6 is a graph of an actual fracture map extracted by the backtracking accumulation-based method of the present invention.
Detailed Description
The present invention is further illustrated by the following examples, which are intended to be purely exemplary and are not intended to limit the scope of the invention, as various equivalent modifications of the invention will occur to those skilled in the art upon reading the present disclosure and fall within the scope of the appended claims.
As shown in fig. 1, a graph of a simulation of an image to be processed is used to illustrate a specific trace-back accumulation-based curve detection method, which includes the following steps:
step 1: analyzing the difference of the gray scale characteristics of the curve and the background in the analog image to be processed in fig. 1, and defining a curve characteristic metric m (p) for describing the curve characteristics of any point p in fig. 1; the feature metric M (p) is constructed using the difference of the curve gray feature and the background gray feature such that the curve point pcCurve characteristic metric value of M (p)c) Greater than the non-curve point pnon-cThe curve characteristic measure of, i.e. M (p)c)>M(pnon-c). The gray level of the curve point is lower than the average gray level of the background;the curve is a thin line, the proportion of curve points in a certain window is small, and the background area does not have the characteristic, so that the curve points and the background area can be separated.
Further, the window width is set to w for a known arbitrary point pi,jWith pi,jCentered, within a window of w × w, p is calculatedi,jInverse ratio (p) of the sum of the gray scale differences with other points in the windowi,j) I.e. byFor curve point pcWith ratio (p)c)>ratio(pnon-c) (ii) a Wherein m, n are two variables, I (p)i,j) Represents pi,jOf the pixel, correspondingly, I (p)i+m,j+n) Representing other points p within the windowi+m,j+nThe gray value of (a); the curve feature metric value is then defined as m (p) i (p) ratio (p).
Step 2: constructing an energy function P (p) by the curve characteristic metric value M (p) to make a curve point pcEnergy of (4) is minimal;
from M (p)c)>M(pnon-c) In order to make the curve point pcThe potential energy P (p) of any point p in the constructed image is the reciprocal of the characteristic measure M (p), namely the potential energy P (p) is easier to search
And step 3: selecting any point p in the curve which is knownsAs starting point, ps∈pcUsing the shortest pathThe image I to be processed is searched by the path algorithm, and the passing point p is searchedpsdCurve characteristic metric value of M (p)psd) Form a characteristic image If;
Specifically, as shown in FIG. 1, a, b, and c in FIG. 1 are all points on the curve, i.e., a ∈ pc,b∈pc,c∈pcTaking a point a as a starting point, performing shortest path search by using Dijkstra algorithm in the shortest path algorithm, performing the search in a neighborhood taking the point a as a center, recording the accumulated cost from the point to the starting point when a certain point is searched, and selecting a point with the minimum accumulated cost for continuous calculation each time; searching through points ppsdCurve characteristic metric value of M (p)psd) The feature image is formed, as shown in fig. 2, the point with the gray value greater than 0 in fig. 2 is the searched point ppsdE.g. points b, d, e; and the point with the gray value of 0 is the point p which is not searchednon-psdE.g., point f in the figure.
And 4, step 4: in the obtained characteristic image IfFrom each search through a point ppsdStarting, backtracking and accumulating to the starting point to obtain a backtracking and accumulating graph
Specifically, for the characteristic image I obtained in step 3f(FIG. 2), since we describe the curve characteristics, the curve points b and c in FIG. 1 have larger characteristic metric M (p)c) Value and smaller potential energy P (P)c) (ii) a For p accessed from step 3psdPoints, such as points b, d, e in fig. 2, are traced back with their characteristic metric value M (p)psd) Is accumulated to the point p passed in the backtracking processbkGet the backtracking accumulation chartNamely, it is
Wherein, the backtracking passes through the point pbkBelonging to an arbitrary point p searched forpsd. E.g., point d, backward search back to the starting point a1、d2、d3When the point b is reached, the point a is returned, and similarly, the point e passes through the point e when backtracking to the starting point a1、e2、e3、e4When the point b is reached, the point a is returned, so that the curve points a and b are more frequently visited than the non-curve points d and e during the backtracking, and therefore
And 5: accumulating graphs in backtrackingMiddle selection threshold lambdafSaid threshold value λfIn the range of [10-3,103],Is greater than threshold lambdafIs a curve point pcAll curve points pcForm the finally obtained graph IcAs shown in fig. 3.
In particular, becauseI.e. the curve point pcOfThe value will be larger, setting the threshold lambdafSo thatThus at a given threshold λfObtaining a final curve point set graph IcAs indicated by the black curve in FIG. 3, wherein
Evaluation of effects
Firstly, a simulation graph with a curve is used, as shown in fig. 1, to extract the curve, wherein the average gray value of the curve in the simulation graph is lower than the background value, and the background is generated randomly, and the curve graph I is obtained by the method of the inventioncAs shown in fig. 3, the detected curve is marked with a black line in fig. 3. In order to verify the application effect of the invention in practical situations, a real crack diagram is used, for example, an experiment is carried out in fig. 4, and besides the properties of the simulated diagram, the crack diagram has the influence of multiple textures, shadows, illumination and the like, so that the crack diagram is more challenging. With a starting point p as shown in fig. 4sUsing the constructed energy function P (p) as a starting point, and searching by using a shortest path algorithm to obtain a characteristic image IfAs shown in fig. 5; next, for the characteristic image I shown in FIG. 5fEach point in the graph is backtracked and accumulated to obtain a backtracking accumulation graphSetting a certain threshold lambdafSo thatTo obtain the final graph IcAs shown in fig. 6, fig. 6 indicates the detected curve using a black line. Experimental results show that the method disclosed by the invention has the extraction efficiency and quality in both the synthetic diagram and the crack diagram.
Visual evaluation: by observing the extraction results of fig. 3 and 6, we can see that the curve in the graph can be extracted quickly and effectively by using the method of the present invention under the condition of only providing one starting point.
Claims (4)
1. A curve detection method based on backtracking accumulation is characterized in that the backtracking accumulation refers to the following steps: after the shortest path is searched, after each point is reversely traced to a starting point or a certain point number, the curve point can be accessed at high frequency, and the non-curve point is accessed less, so that the curve point and the non-curve point are successfully separated, and the curve outline is finally drawn; the method comprises the following steps:
step 1: analyzing the gray scale feature of any point p in the image I to be processed, constructing a curve feature metric value M (p) for describing the curve feature of the point p,and construct a curve point pcCurve characteristic metric value of M (p)c) Greater than the non-curve point pnon-cCurve characteristic metric value of M (p)non-c) I.e. M (p)c)>M(pnon-c);
Step 2: constructing an energy function P (p) by means of the constructed curve characteristic metric values M (p), the constructed energy function corresponding to the curve points pcEnergy P (P) ofc) Less than the non-curve point pnon-cEnergy P (P) ofnon-c) I.e. P (P)c)<P(pnon-c);
And step 3: selecting any point p on the curvesAs starting point, ps∈pcSearching the image I to be processed by using the shortest path algorithm, and searching the passing point ppsdCurve characteristic metric value of M (p)psd) Form a characteristic image If;
And 4, step 4: in the obtained characteristic image IfRecord the point p passed by the searchpsdFor each point p through which the search passespsdBacktracking and accumulating to obtain a backtracking and accumulating graph
And 5: for backtracking accumulation chartSelecting a threshold lambdafSaid threshold value λfIn the range of [10-3,103]So thatIs greater than threshold lambdafIs a curve point pcAll curve points pcForm the finally obtained graph Ic。
2. The method of claim 1, wherein the constructed energy function p (p) is the inverse of the curve feature metric value m (p).
3. The curve detection method according to claim 1, wherein the shortest path algorithm searches the image to be processed I without specifying a search end point.
4. The curve detection method according to claim 1, wherein the backtracking cumulative graph is obtained by performing a back-tracking cumulative graphAt any point p inbkFrom the characteristic image IfPoint p searched inpsdBacktracking to the starting point by pbk,pbkAccumulating graphs in backtrackingValue of (A)For all backtracking passes pbkP of (a)psdCurve characteristic metric value of M (p)psd) And (4) summing.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410271291.8A CN104008555B (en) | 2014-06-17 | 2014-06-17 | A kind of curve detection method cumulative based on backtracking |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410271291.8A CN104008555B (en) | 2014-06-17 | 2014-06-17 | A kind of curve detection method cumulative based on backtracking |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104008555A CN104008555A (en) | 2014-08-27 |
CN104008555B true CN104008555B (en) | 2016-09-21 |
Family
ID=51369195
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410271291.8A Active CN104008555B (en) | 2014-06-17 | 2014-06-17 | A kind of curve detection method cumulative based on backtracking |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104008555B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112950665B (en) * | 2021-02-01 | 2022-09-02 | 武汉大学 | Semi-automatic curve-shaped target extraction method and system |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103268611A (en) * | 2013-05-27 | 2013-08-28 | 广西工学院 | Method for detecting accurate real-time curve in complex screen |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9030559B2 (en) * | 2012-08-28 | 2015-05-12 | Palo Alto Research Center Incorporated | Constrained parametric curve detection using clustering on Hough curves over a sequence of images |
-
2014
- 2014-06-17 CN CN201410271291.8A patent/CN104008555B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103268611A (en) * | 2013-05-27 | 2013-08-28 | 广西工学院 | Method for detecting accurate real-time curve in complex screen |
Non-Patent Citations (1)
Title |
---|
《CENTERLINE CONSTRAINED MINIMAL PATH PROPAGATION FOR VESSEL》;Yang chen et al;《2014 IEEE 11th International Symposium on Biomedical Imaging(ISBI)》;20140429;794-797 * |
Also Published As
Publication number | Publication date |
---|---|
CN104008555A (en) | 2014-08-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Rudolph et al. | Same same but differnet: Semi-supervised defect detection with normalizing flows | |
CN108734151B (en) | Robust long-range target tracking method based on correlation filtering and depth twin network | |
Fernandes et al. | Pavement pathologies classification using graph-based features | |
CN103729842A (en) | Fabric defect detection method based on local statistical characteristics and overall significance analysis | |
CN103745453B (en) | Urban residential areas method based on Google Earth remote sensing image | |
CN110532914A (en) | Building analyte detection method based on fine-feature study | |
CN109584203A (en) | Reorientation image quality evaluating method based on deep learning and semantic information | |
CN105550703A (en) | Image similarity calculating method suitable for human body re-recognition | |
CN107504908A (en) | White light reflection dynamic measurement film thickness method based on wavelet modulus maxima | |
Lv et al. | Novel automatic approach for land cover change detection by using VHR remote sensing images | |
CN115546113A (en) | Method and system for predicting parameters of tunnel face crack image and front three-dimensional structure | |
CN104008555B (en) | A kind of curve detection method cumulative based on backtracking | |
CN113743455A (en) | Target retrieval method, device, electronic equipment and storage medium | |
CN107024532B (en) | Method for extracting defect positions of magnetic leakage pipeline based on window features | |
CN109583626B (en) | Road network topology reconstruction method, medium and system | |
CN113345588A (en) | Rapid attribute reduction method for incomplete data set | |
Wijaya et al. | Building crack due to lombok earthquake classification based on glcm features and svm classifier | |
CN115184998A (en) | Rayleigh wave frequency dispersion curve automatic extraction method based on improved U-net neural network | |
CN104573633B (en) | A kind of matching and recognition methods referring to the bionical texture of vein and linear texture | |
CN110222632A (en) | A kind of waterborne target detection method of gray prediction auxiliary area suggestion | |
KR101415309B1 (en) | Determination method and apparatus of the optimum distance of upward continuation for stripping | |
CN115861324B (en) | Pavement damage maintenance cost prediction method and system | |
CN105139863B (en) | Audio frequency domain continuity graph calculation method | |
Yao et al. | Enhanced Deeplab Network for Infrared Small Target Detection | |
Yang et al. | Deep feature inpainting for unsupervised visual anomaly detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220125 Address after: 510507 1801, No. 623, yueken Road, Tianhe District, Guangzhou, Guangdong Patentee after: Guangzhou yilianzhong ruitu Information Technology Co.,Ltd. Address before: 210096 No. four archway, 2, Jiangsu, Nanjing Patentee before: SOUTHEAST University |
|
TR01 | Transfer of patent right |