CN105261011A - Extraction method for extracting insulator out of complex background figure during routing inspection and aerial photograph process of unmanned plane - Google Patents

Abstract

The invention discloses an extraction method for extracting an insulator out of a complex background figure during the routing inspection and aerial photograph process of an unmanned plane. According to the method, firstly, an original figure is converted into an HSI figure and the soil is extracted by utilizing the variable H. Meanwhile, the vegetation is extracted out of the original figure based on a vegetation extraction formula specified in Ecognition and a power tower is extracted through spectrum value segmentation process based on the significance of the power tower. After that, through the two-dimensional OSTU threshold segmentation process, the original figure is segmented and extracted, so that the figure of an insulator can be obtained. At the same time, the figure of the insulator figure, the figure of the vegetation, the soil binary figure and the power tower are subjected to the arithmetic of substraction, and then a crude insulator binary figure can be obtained. The main direction of each connected domain in the binary figure is found out, and the traversal over the normal directions of the main directions is conducted to obtain a distance distribution histogram. If the distance distribution histogram has its regularity, the crude insulator binary figure is judged to contain an insulator. Finally, the extracted insulator figure is ensured to be good in effect.

Description

A kind of unmanned plane patrols and examines the extracting method of insulator in complex background image of taking photo by plane

Technical field

The invention belongs to remote sensing image technical field, especially relate to the extracting method that a kind of unmanned plane patrols and examines insulator in complex background image of taking photo by plane.

Background technology

Transmission line of electricity is an important ingredient in national grid, mainly bear the conveying of electric energy, wherein ultra high-tension transmission line is as the basic routing line of electrical network, plays vital effect, directly affect the stable development of national economy to the safety of electrical network, reliability service.Most power transmission line is all exposed in outdoor environment, experiencing for a long time expose to the weather, the thunderbolt of sleet, lighting, dirt move and the infringement of the external environment such as depression, also subjects the internal pressure of electric load and mechanical load simultaneously.Above-mentioned factor all can cause power line components in various degree aging even damaged, if can not Timeliness coverage and the hidden danger eliminated in these power transmission lines, bring larger harm just may to link of transmitting electricity, form serious potential threat to the safe operation of national grid.Therefore, making regular check on power transmission line is an important process effectively ensureing power transmission line safety, normal conveying, is national grid to the reliable supply of electric energy and the basis of safe transmission and prerequisite.

The maintenance model of Traditional Man prophylactic repair and correction maintenance power transmission line, is difficult to meet the daily maintenance requirement of high voltage transmission line, causes huge hidden danger to the safe operation of national grid, be difficult to the demand for development meeting the current power transmission network of China.The technology of helicopter line walking is developed so far existing more than 20 years in China, and power department has carried out a large amount of experimental studies to helicopter line walking detection technique, and some power departments have also carried out the application of helicopter line walking.In the data acquisition of helicopter line walking, adopt the equipment such as comfort, Digital Video, high-resolution telescope, visible ray video recorder as observation instrument more.Utilize helicopter routing inspection electrical network, not only can eliminate the danger that artificial work high above the ground exists, and the omnidirectional shooting Detection results of aircraft is more more comprehensive than manual detection, can Timeliness coverage problem it is solved.

Insulator is the important component part of overhead transmission line, and its effect is support wire and prevents electric current from returning ground.Because insulator will stand wind, Exposure to Sunlight for a long time, drench with rain, add oneself body mechanical fatigue, therefore insulator occurs broken, and crack equivalent damage, makes insulator not play a role normally.Utilize helicopter line walking mode per hourly can collect the image/video data of hundreds of million, in identification, judge in power transmission line view data process, if staff adopts naked eyes mode to judge whether there is abnormal occurrence in image, a large amount of manpower, material resources and financial resources will be needed, and the artificial interpretation time one is long, easily cause visual fatigue, easily erroneous judgement or situation of failing to judge occur, be difficult to the potential safety hazard accurately finding that transmission line of electricity exists.Therefore utilize the method for image procossing and the method for pattern-recognition, from image, identify the inexorable trend that insulator is development fast.Extract at thresholding method in the process of insulator at present, only utilize a certain variable in HSI color space to be used as the basis split, and be not suitable for most insulator.Adopt simple spectral signature to carry out classification of remote-sensing images and easily produce the different spectrum phenomenon of jljl, the color similarity of such as insulator and power tower and background is complicated, in partial picture, so utilize spectral signature not ideal to the effect being partitioned into insulator.Just under this background, correlative study has been carried out for the identification of insulator in unmanned plane filmed image.

Summary of the invention

The problem of carrying out under being aimed at simple background for existing many insulator extraction algorithms, the invention provides a kind of unmanned plane and patrols and examines the insulator extracting method of taking photo by plane in complex background image more.

The technical solution adopted in the present invention is: a kind of unmanned plane patrols and examines the extracting method of insulator in complex background image of taking photo by plane, and it is characterized in that, comprises the following steps:

Step 1: carry out color space conversion to original remote sensing image, becomes HSI image by RGB video conversion, to the HSI Extraction of Image S variable after conversion, utilizes two-dimentional OSTU thresholding method to process S variable two-value image, obtains soil binary map;

Step 2: process original remote sensing image, obtains vegetation binary map;

Step 3: utilize two-dimentional OSTU thresholding method to process original remote sensing image, obtains the preliminary insulator binary map after splitting;

Step 4: have spectrum saliency according to power tower in raw video, extracts power tower binary map by spectral value segmentation;

Step 5: according to soil binary map, vegetation binary map and power tower binary map, removes the soil in preliminary insulator binary map, vegetation and power tower, obtains thick insulator binary map;

Step 6: connected domain principal direction mark is carried out to thick insulator binary map; First calculate the pixel shape index of each connected domain pixel, after statistics, obtain this connected domain principal direction; Then utilize the shape index of principal direction and pixel to calculate the main shaft pixel of this connected domain, and its pixel coordinate is preserved; Each main shaft pixel of traversal pixel coordinate, main shaft pixel is done a normal perpendicular to principal direction, is preserved by maximum distance crossing with connected domain for normal;

Step 7: according to the size property of middle insulator of taking photo by plane, if the maximum distance that obtains in step 6 exist be greater than 60 or be less than 10 element, just connected domain is removed;

Step 8: further process is done to the maximum distance obtained in step 7; According to the style characteristic of middle insulator of taking photo by plane, ask for the mean value of maximum distance, more each numerical value of maximum distance and mean value are subtracted each other, this process makes the gear feature of insulator more remarkable;

Step 9: the maximum distance in step 8 after processing further is added up, calculate adjacent current maximum distance and be more than or equal to 0, and the distance length that next distance is less than 0, if identical distance length accumulation number of times occurs being greater than 3, then can judge that this connected domain is as insulator, finally obtains accurate insulator binary map.

As preferably, be utilize RGB vegetation in Ecognition software to extract formula in step 2, original remote sensing image is processed.

Relative to prior art, the invention has the beneficial effects as follows: complicated in background, insulator is close with power tower color and shooting angle is changeable, from aerial images, extract insulator.

Accompanying drawing explanation

Fig. 1: be the method flow diagram of the embodiment of the present invention.

Embodiment

Understand for the ease of those of ordinary skill in the art and implement the present invention, below in conjunction with drawings and Examples, the present invention is described in further detail, should be appreciated that exemplifying embodiment described herein is only for instruction and explanation of the present invention, is not intended to limit the present invention.

Ask for an interview Fig. 1, a kind of unmanned plane provided by the invention patrols and examines the extracting method of insulator in complex background image of taking photo by plane, and comprises the following steps:

Step 1: carry out color space conversion to original remote sensing image, becomes HSI image by RGB video conversion, to the HSI Extraction of Image S variable after conversion, utilizes two-dimentional OSTU thresholding method to process S variable two-value image, obtains soil binary map;

The spectral value of soil near power tower pedestal in original remote sensing image and insulator comparatively close, when carrying out two-dimentional OSTU thresholding method, soil is always extracted by as insulator, and, due to the remote sensing image of unmanned plane shooting, together with insulator is always obscured with soil, the shape facility of insulator is weakened by soil, and is unfavorable for the accurate extraction of insulator below.Therefore utilize the value of soil S variable in HSI image to be always no more than the feature of certain threshold value, from raw video, extract soil.

$\left\{\begin{array}{c}H\left\{\begin{array}{cc}\mathrm{\θ}& B\≤G;\\ 360-\mathrm{\θ}& B>G;\end{array}\right.\\ S=1-\frac{3*\min (R,G,B)}{R+G+B}\\ I=\frac{R+G+B}{3}\end{array}\right.$

Wherein be expressed as according to the feature of soil in variable S, be handled as follows and obtain soil binary map.

$F=\left\{\begin{array}{cc}255& S<0.392;\\ 0& S\&GreaterEqual;0.392;\end{array}\right.$

Step 2: utilize RGB vegetation in Ecognition software to extract formula, original remote sensing image is processed, obtains vegetation binary map;

In remote sensing image, vegetation large area exists, and only relying on threshold segmentation method to carry out process to raw video thoroughly cannot split vegetation.The formula that wherein in Ecognition software, vegetation is extracted is as follows:

VI′＝(2*G′-R′-B′)-(1.4*R′-G′)

Wherein $G^{\&prime;}=\frac{G}{R+G+B},R^{\&prime;}=\frac{R}{R+G+B},B^{\&prime;}=\frac{B}{R+G+B};$

Step 3: utilize two-dimentional OSTU thresholding method to process original remote sensing image, obtains the preliminary insulator binary map after splitting;

Two dimension OSTU thresholding method is the expansion of varimax between class, is asked and is extended for from two one dimension distribution maximum between-cluster variances the maximal value solving inter _ class relationship matrix trace.The method comparatively differs from method to the segmentation effect of raw video between maximum kind well a lot.Therefore Selection utilization two dimension OSTU thresholding method carries out Threshold segmentation.The key step of two dimension OSTU thresholding method:

(1) (total L gray level, each probability of occurrence is p) to set up image grey level histogram

$N=\underset{i=0}{\overset{L-1}{\&Sigma;}}{n}_i$

p _i＝n _i/N

(2) calculate the probability of occurrence of background and target, computing method are as follows:

$p_A=\underset{i=1}{\overset{t}{\&Sigma;}}{p}_i,p_B=1-p_A$

(3) inter-class variance calculating A and B two regions is as follows:

${\mathrm{\&omega;}}_A={\mathrm{\&Sigma;}}_{i=0}^{t}i*p_i/p_A,{\mathrm{\&omega;}}_B={\mathrm{\&Sigma;}}_{i=t+1}^{L-1}i*p_i/p_B;$

ω ₀＝p _A*ω _A+p _B*ω _B；

σ ²＝p _A*(ω _A-ω ₀) ²+p _B*(ω _B-ω ₀) ²；

First expression formula calculates the average gray value in A and B region respectively;

Second expression formula calculates the average gray value of the gray level image overall situation;

3rd expression formula calculates the inter-class variance in A, B two regions.

(4) several step more than has calculated the inter-class variance on single gray-scale value, therefore optimal segmentation threshold value should be can make A and B in image class between the gray-scale value of gray variance.

Step 4: have spectrum saliency according to power tower in raw video, extracts power tower binary map by spectral value segmentation;

Step 5: according to soil binary map, vegetation binary map and power tower binary map, removes the soil in preliminary insulator binary map, vegetation and power tower, obtains thick insulator binary map;

In raw video, power tower is always obvious compared with other atural objects.According to the significant characteristics of power tower, obtained maximal value and the mean value of three wave bands by traversal raw video, then use formulas Extraction power tower binary map below.

$T=\left\{\begin{array}{cc}255& \left\{\begin{array}{c}{R}_{(i,j)}>R_{\max }-\frac{R_{\max }-R_{aver}}{2};\\ G_{(i,j)}>G_{\max }-\frac{G_{\max }-G_{aver}}{2};\\ B_{(i,j)}>B_{\max }-\frac{B_{\max }-B_{aver}}{2};\end{array}\right.\\ 0& \left\{\begin{array}{c}{R}_{(i,j)}\&le;R_{\max }-\frac{B_{\max }-B_{aver}}{2};\\ G_{(i,j)}\&le;G_{\max }-\frac{G_{\max }-G_{aver}}{2};\\ B_{(i,j)}\&le;B_{\max }-\frac{B_{\max }-B_{aver}}{2};\end{array}\right.\end{array}\right.$

Wherein R _max, G _max, B _maxrepresent the maximal value of RGB tri-wave bands, R _aver, G _aver, B _averbe expressed as the mean value of RGB tri-wave bands.

Step 6: connected domain principal direction mark is carried out to thick insulator binary map; First calculate the pixel shape index of each connected domain pixel, after statistics, obtain this connected domain principal direction; Then utilize the shape index of principal direction and pixel to calculate the main shaft pixel of this connected domain, and its pixel coordinate is preserved; Each main shaft pixel of traversal pixel coordinate, main shaft pixel is done a normal perpendicular to principal direction, is preserved by maximum distance crossing with connected domain for normal;

The principal direction of connected domain is mainly determined by the connected domain that length on some directions is the longest.According to pixel shape index, calculate the length of each direction line of each pixel of each connected domain, the record extreme length at pixel center and the direction of this length, the direction i.e. principal direction of this connected domain that the extreme length of this connected domain is corresponding.The pixel of main shaft element to be the direction of pixel center extreme length in connected domain be principal direction.

Step 7: according to the size property of middle insulator of taking photo by plane, if the maximum distance that obtains in step 6 exist be greater than 60 or be less than 10 element, just connected domain is removed;

Step 8: further process is done to the maximum distance obtained in step 7; According to the style characteristic of middle insulator of taking photo by plane, ask for the mean value of maximum distance, more each numerical value of maximum distance and mean value are subtracted each other, this process makes the gear feature of insulator more remarkable;

Step 9: the maximum distance in step 8 after processing further is added up, calculate adjacent current maximum distance and be more than or equal to 0, and the distance length that next distance is less than 0, if identical distance length accumulation number of times occurs being greater than 3, then can judge that this connected domain is as insulator, finally obtains accurate insulator binary map.

Should be understood that, the part that this instructions does not elaborate all belongs to prior art.

Should be understood that; the above-mentioned description for preferred embodiment is comparatively detailed; therefore the restriction to scope of patent protection of the present invention can not be thought; those of ordinary skill in the art is under enlightenment of the present invention; do not departing under the ambit that the claims in the present invention protect; can also make and replacing or distortion, all fall within protection scope of the present invention, request protection domain of the present invention should be as the criterion with claims.

Claims (2)

1. unmanned plane patrols and examines an extracting method for insulator in complex background image of taking photo by plane, it is characterized in that, comprises the following steps:

Step 1: carry out color space conversion to original remote sensing image, becomes HSI image by RGB video conversion, to the HSI Extraction of Image S variable after conversion, utilizes two-dimentional OSTU thresholding method to process S variable two-value image, obtains soil binary map;

Step 2: process original remote sensing image, obtains vegetation binary map;

Step 3: utilize two-dimentional OSTU thresholding method to process original remote sensing image, obtains the preliminary insulator binary map after splitting;

Step 4: have spectrum saliency according to power tower in raw video, extracts power tower binary map by spectral value segmentation;

Step 5: according to soil binary map, vegetation binary map and power tower binary map, removes the soil in preliminary insulator binary map, vegetation and power tower, obtains thick insulator binary map;

Step 6: connected domain principal direction mark is carried out to thick insulator binary map; First calculate the pixel shape index of each connected domain pixel, after statistics, obtain this connected domain principal direction; Then utilize the shape index of principal direction and pixel to calculate the main shaft pixel of this connected domain, and its pixel coordinate is preserved; Each main shaft pixel of traversal pixel coordinate, main shaft pixel is done a normal perpendicular to principal direction, is preserved by maximum distance crossing with connected domain for normal;

Step 7: according to the size property of middle insulator of taking photo by plane, if the maximum distance that obtains in step 6 exist be greater than 60 or be less than 10 element, just connected domain is removed;

Step 8: further process is done to the maximum distance obtained in step 7; According to the style characteristic of middle insulator of taking photo by plane, ask for the mean value of maximum distance, more each numerical value of maximum distance and mean value are subtracted each other, this process makes the gear feature of insulator more remarkable;

Step 9: the maximum distance in step 8 after processing further is added up, calculate adjacent current maximum distance and be more than or equal to 0, and the distance length that next distance is less than 0, if identical distance length accumulation number of times occurs being greater than 3, then can judge that this connected domain is as insulator, finally obtains accurate insulator binary map.

2. unmanned plane according to claim 1 patrols and examines the extracting method of insulator in complex background image of taking photo by plane, and it is characterized in that: be utilize RGB vegetation in Ecognition software to extract formula in step 2, process original remote sensing image.