CN102930535A - Power line contour extraction method for detecting online foreign object intrusion - Google Patents

Abstract

The invention discloses a power line contour extraction method for detecting online foreign object intrusion. According to the method, on the basis of a fact that the same memory space in a computer can be redistributed and reutilized, the defect that a great number of calculation and memory spaces are required in a standard Hough transform algorithm is overcome, linear parameters are extracted by means of two dot pairs on a line, and a great amount of unnecessary calculation is avoided, the memory spaces are saved, and the efficiency of extracting power line contours is greatly improved at the same time by technologies of dynamically allocating memory and selecting special feature points. The method is developed on the basis of a Qt+OpenCV platform, can be directly invoked by a power line foreign object intrusion detection module in an intelligent video monitoring system of a power system, and specific positions (polar coordinates or rectangular coordinates) of a power line in the image are returned to a main system. The method has the advantages that the algorithm complexity is low; the method is easy to implement; and the operating rate is high.

Description

A kind of power line profile extracting method that is applied to foreign body intrusion detection on the line

Technical field

The present invention relates to a kind of power line profile extracting method that foreign body intrusion detects on the line that is applied to, belong to the extraction identification that is applied to process to realize by image power line profile, providing basic support for follow-up realization line of electric force foreign body intrusion detects, is the key method of realizing foreign body intrusion measuring ability on the intelligent video monitoring line.

Background technology

The line of electric force foreign body intrusion very easily causes the alternate discharge short of line of electric force and line of electric force to shaft tower or to the lightning conducter discharge short, cause the tripping operation power outage, have a strong impact on safe operation and the power supply reliability of electrical network, effectively detect the line of electric force foreign body intrusion and have actively great meaning.

In the Power System Intelligent video monitoring system, the identification of the critical components such as the line of electric force of processing based on image, insulator is the precondition that realizes that foreign body intrusion detects.The Hough conversion provides in a kind of efficient detected image the method by straight line, circular arc, quafric curve even shape that arbitrary curve consists of, and the method is widely used in target detection identification.

The ultimate principle of Hough conversion is to utilize some the duality with line, the given curve negotiating curve representation form in original image space is become a point of parameter space, so just the test problems of given curve in the original image is converted into the spike problem of seeking in the parameter space, also namely be converted into the detection local characteristics detecting overall permanence, such as straight line, ellipse, circle, camber line etc.

The Hough conversion is in the situation that parameter space is no more than two dimension, and transform effect is desirable, and is with the obvious advantage.But the Hough conversion is mapped as corresponding curve in the parameter space with each marginal point in the image, in case parameter space increases, calculated amount just can sharply rise, and the parameter space that need to prestore, and expends huge storage space, also thereupon surge consuming time.The deficiency of conventional H ough conversion mainly contain following some:

1, parameter is by two parameters of straight line, and namely intercept and slope rise to three, i.e. central coordinate of circle and radius, and each point is mapped to a curved surface of parameter space, be one to many mappings, thereby calculated amount sharply increases;

2, need take a large amount of memory headrooms, consuming time for a long time, real-time is poor;

3, the image in the reality generally all is subject to the interference of outside noise, signal to noise ratio (S/N ratio) is lower, the performance of conventional H ough conversion will sharply descend this moment, be difficult to determine owing to suitable threshold value when carrying out the search of parameter space maximum value, the problem at " empty peak " and " undetected " often occur.

For the Hough change detection of straight line, because pseudo-peak value can appear in image, also brought difficulty to detection, limited its application.

Why standard Hough mapping algorithm (SHT algorithm) needs a large amount of storage spaces, mainly because numeral is to disperse in computing machine.When we realize the SHT algorithm in computing machine, have a fact usually to be left in the basket, that is exactly that storage space identical in the computing machine can be reallocated and utilize.

Summary of the invention

The present invention will solve the shortcoming of above-mentioned prior art, has overcome the shortcoming that the SHT algorithm needs a large amount of calculating storage spaces, and a kind of power line profile extracting method that foreign body intrusion detects on the line that is applied to is provided.

The present invention solves the technical scheme that its technical matters adopts: this power line profile extracting method that is applied to foreign body intrusion detection on the line may further comprise the steps:

(1) select Seed Points, for the binary edge map of a width of cloth N * N, each set D={ (x that forms from k marginal point in order _i, y _i) | i=1,2 ..., take out a unique point among the k} and calculate as Seed Points;

(2) selected characteristic point after determining Seed Points, is chosen from the remaining unique point of set D successively and is a bit participated in computing, and it is right that selecteed unique point and Seed Points will form a point, thus the straight line in unique definite image;

(3) calculating parameter detects and passes two unique points to { (x _i, y _i), (x _j, y _j) (the i index value that is Seed Points in set D wherein, j satisfy the straight line of i＜j≤k), can get its Hough transformation parameter by formula (1), (2), and the present invention has only used parameter θ in computation process, therefore parameter ρ can calculate, work as y _i-y _j=0 o'clock, θ _{I, j}=0;

${\mathrm{\θ}}_{i,j}={\tan }^{-1}\left(\frac{x_i-x_j}{y_i-y_j}\right)---\left(1\right)$

$\left|{\mathrm{\ρ}}_{i,j}\right|=\frac{|x_i{y}_i-x_j{y}_j|}{\sqrt{{(x_i-x_j)}^2+{(y_i-y_j)}^2}}---\left(2\right)$

(4) add up at parameter space, for Seed Points (x _i, y _i), calculate different θ by formula (1) _{I, j}, set an one dimension totalizer array H (θ), for it distributes specific totalizer cell unit H (θ _{I, j}), and it is initialized as 1, if as the θ that calculate _{I, j}(j=i+1, i+2 ..., k) with already present θ _{I, m}(m=i+1, i+2 ..., k) equate or the two difference less than the fault tolerance δ that the user sets, then make corresponding cell accumulator element H (θ _{I, m}) increasing by 1, step (2) ~ (4) have consisted of a little iteration cycle process;

(5) peak value in detected parameters space, program detects the number of unit among the totalizer array H (θ) through after the above-mentioned iteration once, and predetermined threshold value T is if H is (θ _{I, j}) T, then by point (x _i, y _i) and parameter θ _{I, j}Can from image, extract straight line;

(6) result of calculation is kept at other storage space, according to the result of step (5), can be with Seed Points (x _i, y _i) corresponding parameter (ρ _{I, j}, θ _{I, j}) be kept at an other array H ^*In (ρ, θ);

(7) remove the unique point that is positioned on the straight line that is detected, will be arranged in the unique point that step (6) obtains on the straight line and all be deleted from set D, avoid in continuous loop iteration producing unnecessary double counting;

(8) reinitialize parameter space, with Seed Points (x _i, y _i) release of corresponding totalizer array H (θ) space, by ensuing iteration is reused, step (1) ~ (8) have formed a loop iteration process, incite somebody to action again selected seed point after iteration finishes, and begin a new loop iteration process;

(9) EOP (end of program), the preservation parameter θ that finally obtains and the array H of ρ ^*(ρ, θ) is the straight line parameter array that detects.

As preferably, to determine after the Seed Points, choose in the unique point of being left from set D successively and a bit participate in computing, it is right that selecteed unique point and Seed Points will form a point, thus the straight line in unique definite image.

As preferably, two straight lines that unique point is right are passed in detection, only calculate its Hough transformation parameter θ, add up at parameter space, certain unit outnumbers predetermined threshold value T in totalizer array H (θ), then think to detect straight line, and with this to Parameter storage at other parameter array H ^*In (ρ, θ).

As preferably, will be arranged in the unique point that step (6) obtains on the straight line and all be deleted from set D, avoid in continuous loop iteration, producing unnecessary double counting.

As preferably, with Seed Points (x _i, y _i) release of corresponding totalizer array H (θ) space, by ensuing iterative computation is reused.

As preferably, after the EOP (end of program), the preservation parameter θ that finally obtains and the array H of ρ ^*(ρ, θ) is the straight line parameter array that detects.

Inventing useful effect is: this invention is based on the characteristic that reuses of storage space in the computing machine, utilize two points on the straight line to extracting the straight line parameter, by dynamic assigning memory and the technology of choosing special unique point, thereby avoided a large amount of unnecessary calculating, save memory headroom, greatly improved the efficient that straight line extracts simultaneously.This invention will lay the foundation for the quick identification of the critical components such as the ground wire processed based on image, line of electric force, has extraordinary application prospect.

Description of drawings

Fig. 1 is line of electric force image-recognizing method software flow;

Embodiment

The invention will be further described below in conjunction with accompanying drawing:

With reference to accompanying drawing: the inventive method is based on developing under the Qt+OpenCV platform, can directly offer the line detection algorithm that electric system patrols and examines in the detection system calls, the particular location (polar coordinates or rectangular coordinate) of line of electric force in image returned to main system, as shown in Figure 1.

This power line profile extracting method that is applied to foreign body intrusion detection on the line may further comprise the steps:

(1) selects Seed Points.For the binary edge map of a width of cloth N * N, k marginal point (the marginal point pixel value is 1, and background is 0) arranged.K the synthetic set of group of edge points D={ (x _i, y _i) | i=1,2 ..., k} in this set formation, takes out an Edge Feature Points as Seed Points at every turn in order, after processed the finishing of this point, will be deleted from set, chooses next point and again participates in calculating as Seed Points.

(2) remaining unique point is chosen successively.After determining Seed Points, from remaining unique point, choose successively and a bit participate in computing.For example, Seed Points needs to choose successively m+1 so for m unique point in the set, m+2 ..., until k unique point.It is right that selecteed unique point and Seed Points will form a point, by changing unique point, and the straight line in can unique definite image.

(3) calculating parameter.Detection is passed two unique points to { (x _i, y _i), (x _j, y _j) (the i index value that is Seed Points in set D wherein, j satisfy i＜j≤k), calculate the Hough transformation parameter of respective straight.Suppose that the Seed Points coordinate is (x _i, y _i), the coordinate of another unique point is (x _j, y _j), got by standard Hough conversion:

$\left\{\begin{array}{c}{\mathrm{\ρ}}_{i,j}=x_i\cos {\mathrm{\θ}}_{i,j}+y_i\sin {\mathrm{\θ}}_{i,j}\\ {\mathrm{\ρ}}_{i,j}=x_j\cos {\mathrm{\θ}}_{i,j}+y_j\sin {\mathrm{\θ}}_{i,j}\end{array}\right.$

Wherein parameter θ is identical with the definition in the SHT algorithm with the ρ definition.

Can further be drawn by above formula:

${\mathrm{\θ}}_{i,j}={\tan }^{-1}\left(\frac{x_i-x_j}{y_i-y_j}\right)---\left(1\right)$

$\left|{\mathrm{\ρ}}_{i,j}\right|=\frac{|x_i{y}_i-x_j{y}_j|}{\sqrt{{(x_i-x_j)}^2+{(y_i-y_j)}^2}}---\left(2\right)$

In fact, the present invention has only used parameter θ in computation process, therefore parameter ρ can calculate.Work as y _i-y _j=0 o'clock, θ _{I, j}=0.

(4) add up at parameter space.For Seed Points (x _i, y _i), set an one dimension totalizer array H (θ), for the different θ that calculated by formula (1) _{I, j}, for it distributes specific totalizer cell unit H (θ _{I, j}), and it is initialized as 1.If the θ that calculates _{I, j}(j=i+1, i+2 ..., k) equal the θ that existed _{I, m}(m=i+1, i+2 ..., k), a perhaps fault tolerance δ setting less than the user of the difference of the two, so corresponding cell accumulator element H (θ _{I, m}) increase by 1.For example, there is H (θ _{I, m}), if | θ _{I, j}-θ _{I, m}|≤δ then has H (θ _{I, m})+1=H (θ _{I, m}).

(5) peak value in detected parameters space.Step 2 ~ 4 have consisted of a little iteration cycle process.Above program through after iteration, detect the number of unit among the totalizer array H (θ).Preset threshold value T, if H is (θ _{I, j}) T, so by point (x _i, y _i) and parameter θ _{I, j}, just can from image, extract straight line.

(6) result of calculation is kept at other storage space.By the result of step (5), straight line available parameter θ and ρ represent, for example ρ _{I, j}=x _iCos θ _{I, j}+ y _iSin θ _{I, j}, (x wherein _i, y _i) be exactly the Seed Points that in this iteration, sets.Corresponding (the ρ that calculates _{I, j}, θ _{I, j}), it is kept at an other array H ^*In (ρ, θ).Array H ^*(ρ, θ) is a new storage space of opening up in internal memory.Along with the growth of extracting the straight line number, array H ^*(ρ, θ) be dynamic growth thereupon also.

(7) remove the unique point that is positioned on the straight line that is detected.Detect the straight line that satisfies condition by step (6), the unique point that will be arranged on this straight line is all deleted from set D, can effectively avoid producing in continuous loop iteration unnecessary double counting.

(8) reinitialize parameter space.After a loop iteration is finished, Seed Points (x _i, y _i) corresponding totalizer array H (θ) will not re-use, and most important parameter (ρ in the array _{I, j}, θ _{I, j}) it has been stored in new array H ^*In (ρ, θ), thus H (θ) space is discharged, and reused by ensuing iteration.

(9) EOP (end of program).Step 1 ~ 8 have formed a large loop iteration process.Above a few step program will restart selected seed point after finishing, and begin a new loop iteration process.The final utilization is kept at array H ^*Parameter ρ in (ρ, θ) and θ value can be extracted existing straight line in the original image.

In addition to the implementation, the present invention can also have other embodiments.All employings are equal to the technical scheme of replacement or equivalent transformation formation, all drop on the protection domain of requirement of the present invention.

Claims (6)

1. one kind is applied to the power line profile extracting method that foreign body intrusion detects on the line, may further comprise the steps:

(1) select Seed Points, for the binary edge map of a width of cloth N * N, each set D={ (x that forms from k marginal point in order _i, y _i) | i=1,2 ..., take out a unique point among the k} and calculate as Seed Points;

(2) selected characteristic point after determining Seed Points, is chosen from the remaining unique point of set D successively and is a bit participated in computing, and it is right that selecteed unique point and Seed Points will form a point, thus the straight line in unique definite image;

(3) calculating parameter detects and passes two unique points to { (x _i, y _i), (x _j, y _j) (the i index value that is Seed Points in set D wherein, j satisfy the straight line of i＜j≤k), can get its Hough transformation parameter by formula (1), (2), and the present invention has only used parameter θ in computation process, therefore parameter ρ can calculate, work as y _i-y _j=0 o'clock, θ _{I, j}=0;

${\mathrm{\θ}}_{i,j}={\tan }^{-1}\left(\frac{x_i-x_j}{y_i-y_j}\right)---\left(1\right)$

$\left|{\mathrm{\ρ}}_{i,j}\right|=\frac{|x_i{y}_i-x_j{y}_j|}{\sqrt{{(x_i-x_j)}^2+{(y_i-y_j)}^2}}---\left(2\right)$

(4) add up at parameter space, for Seed Points (x _i, y _i), calculate different θ by formula (1) _{I, j}, set an one dimension totalizer array H (θ), for it distributes specific totalizer cell unit H (θ _{I, j}), and it is initialized as 1, if as the θ that calculate _{I, j}(j=i+1, i+2 ..., k) with already present θ _{I, m}(m=i+1, i+2 ..., k) equate or the two difference less than the fault tolerance δ that the user sets, then make corresponding cell accumulator element H (θ _{I, m}) increasing by 1, step (2) ~ (4) have consisted of a little iteration cycle process;

(5) peak value in detected parameters space, program detects the number of unit among the totalizer array H (θ) through after the above-mentioned iteration once, and predetermined threshold value T is if H is (θ _{I, j}) T, then by point (x _i, y _i) and parameter θ _{I, j}Can from image, extract straight line;

(6) result of calculation is kept at other storage space, according to the result of step (5), can be with Seed Points (x _i, y _i) corresponding parameter (ρ _{I, j}, θ _{I, j}) be kept at an other array H ^*In (ρ, θ);

(7) remove the unique point that is positioned on the straight line that is detected, will be arranged in the unique point that step (6) obtains on the straight line and all be deleted from set D, avoid in continuous loop iteration producing unnecessary double counting;

(8) reinitialize parameter space, with Seed Points (x _i, y _i) release of corresponding totalizer array H (θ) space, by ensuing iteration is reused, step (1) ~ (8) have formed a loop iteration process, incite somebody to action again selected seed point after iteration finishes, and begin a new loop iteration process;

(9) EOP (end of program), the preservation parameter θ that finally obtains and the array H of ρ ^*(ρ, θ) is the straight line parameter array that detects.

2. a kind of power line profile extracting method that foreign body intrusion on the line detects that is applied to according to claim 1, it is characterized in that: determine after the Seed Points, choose in the unique point of from set D, being left successively and a bit participate in computing, it is right that selecteed unique point and Seed Points will form a point, thus the straight line in unique definite image.

3. a kind of power line profile extracting method that foreign body intrusion on the line detects that is applied to according to claim 1, it is characterized in that: detect and pass two straight lines that unique point is right, only calculate its Hough transformation parameter θ, add up at parameter space, certain unit outnumbers predetermined threshold value T in totalizer array H (θ), then think to detect straight line, and with this to Parameter storage at other parameter array H ^*In (ρ, θ).

4. a kind of power line profile extracting method that foreign body intrusion on the line detects that is applied to according to claim 1, it is characterized in that: will be arranged in the unique point that step (6) obtains on the straight line and all be deleted from set D, and avoid in continuous loop iteration producing unnecessary double counting.

5. a kind of power line profile extracting method that foreign body intrusion detects on the line that is applied to according to claim 1 is characterized in that: with Seed Points (x _i, y _i) release of corresponding totalizer array H (θ) space, by ensuing iterative computation is reused.

6. a kind of power line profile extracting method that foreign body intrusion on the line detects that is applied to according to claim 1 is characterized in that: after the EOP (end of program), and the preservation parameter θ that finally obtains and the array H of ρ ^*(ρ, θ) is the straight line parameter array that detects.

Images