CN105160669B - The detection of defects of insulator and localization method in a kind of unmanned plane inspection transmission line of electricity image - Google Patents
The detection of defects of insulator and localization method in a kind of unmanned plane inspection transmission line of electricity image Download PDFInfo
- Publication number
- CN105160669B CN105160669B CN201510531559.1A CN201510531559A CN105160669B CN 105160669 B CN105160669 B CN 105160669B CN 201510531559 A CN201510531559 A CN 201510531559A CN 105160669 B CN105160669 B CN 105160669B
- Authority
- CN
- China
- Prior art keywords
- insulator
- image
- point
- component image
- value
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Landscapes
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of detection of defects of insulator in unmanned plane inspection transmission line of electricity image and localization method, belong to image zooming-out and identification technology field.The detection of the defects of insulator of the present invention and localization method, its step are:Step 1: the image for taking photo by plane to obtain is transformed into HSI colourity saturation degree brightness spaces by rgb color space, phase is with obtaining the preliminary profile bianry image of insulator chain after the H component images of extraction and S component images are carried out into binary conversion treatment respectively;Step 2: the profile of insulator monomer is extracted using the ant group algorithm based on particle group optimizing parameter;Step 3: carrying out ellipse fitting to insulator monomer profiles using least square method, and the defects of insulator, is positioned by detecting the spacing between insulator monomer profiles.By using the technical scheme in the present invention, the accuracy of detection to defects of insulator can be significantly improved, reduces the interference of background, and the speed of service is very fast.
Description
Technical Field
The invention belongs to the technical field of image extraction and recognition, and particularly relates to a method for detecting and positioning insulator defects in an unmanned aerial vehicle inspection power transmission line image.
Background
With the continuous and rapid development of national economy and the increasing expansion of urban construction scale, the rapid and intensive development of industries such as high-tech industry, finance, medical treatment and health and the like, the demand on electric power energy is increased day by day, and the economic development not only enables the load of urban and rural power grids to be rapidly increased, but also puts higher requirements on the reliability and the quality of power supply. Therefore, power companies need to regularly inspect power line equipment, especially power lines and power towers, to ensure stable and safe operation of power transmission systems and normal progress of social production and life. Electric power line corridors in China often need to pass through various complex geographic environments and frequently pass through lakes, reservoirs, chongshan mountains and the like, so that the power transmission line has the characteristics of large coverage area, wide distribution area, long transmission distance, complex and variable geographic conditions, remarkable influence of environmental climate and the like, and great challenges are brought to daily operation, maintenance and overhaul of the line.
The patrol of the transmission line in China generally adopts a manual patrol mode, the method is simple, but has low efficiency and long period, a large amount of optical equipment and patrol personnel with high quality and rich experience need to be equipped, and the requirements on manpower and financial resources are high. And when the pole tower is higher and the surrounding geographic environment is more complex, the manual line patrol is more difficult, faults are easily omitted, and the line patrol is not thorough, so that the manual line patrol mode is difficult to meet the operation and maintenance requirements of the high-voltage power grid gradually.
From the nineties of the last century, some developed countries in europe and the united states have tried to apply unmanned aerial vehicles to transmission line repair and other works, and the technology is relatively mature up to now. The helicopter patrols and examines transmission line technique, has safe swift, receives region restriction little, can discover advantages such as trouble fast. In recent years, china also increases the research and development investment of unmanned aerial vehicle line patrol technology, and in 2012, shandong electric power is used for leading unmanned aerial vehicle line patrol to be brought into line patrol normalized application in China. In 2013, the application research of the unmanned inspection technology for the power transmission lines in the high altitude areas, born by the national power grid Qinghai province electric power company overhaul company, smoothly passes the acceptance of the national power grid company and passes the identification of the Qinghai province science and technology hall.
The insulator is one of isolating electric appliances and plays a role in supporting a wire and preventing current from flowing back to the ground in a power transmission line. Due to long-term exposure to the atmosphere, working in severe environments such as strong electric fields, wind, rain, snow, fog, chemical substance corrosion and the like, and due to the fact that materials, manufacturing process levels, artificial damage and the like are used, electrical faults occur to the insulator inevitably. The electrical faults of the insulator mainly include flashover and spontaneous explosion, wherein the flashover occurs on the surface of the insulator, burn traces can be seen, the insulation performance is usually not lost, the spontaneous explosion is mostly caused by the insulator manufacturing process and the change of the surrounding natural environment, the safety and effectiveness of the operation of a power transmission line can be seriously influenced by the loss of the spontaneous explosion of the insulator, and inestimable loss can be caused. Therefore, how to accurately detect the insulator in time from the aerial image with a complex background and identify the electrical fault of the insulator, particularly identifying the self-explosion fault of the insulator, is particularly important. The research on the automatic routing inspection of the power transmission line by using the unmanned aerial vehicle is more at home and abroad, but the research on accurately and quickly extracting the insulator from the image of the power transmission line and detecting and positioning the electrical fault of the insulator is less, the existing detection method is not strong in comprehensiveness, due to the complexity of the erection environment of the power transmission line, weather influence and the like, the acquired image is wide in range, complex in background and more in target objects, and contains interference information such as vegetation, earth, iron towers and the like, the target objects in partial images can be overlapped and staggered with the interference factors, so that the detection difficulty of the defects of the insulator is further increased, the detection precision of the defects is difficult to guarantee, and the defects of missing detection or inaccurate detection exist. For example, document "glass insulator defect diagnosis based on color image" proposes a glass insulator defect diagnosis and removal method based on histogram matching criterion, which can rapidly remove a large number of normal glass insulator pictures and improve the fault detection efficiency to a certain extent, but the document does not study the positioning of specific fault points of insulators, so that the faulty insulators cannot be replaced in time. Also, as the document "Segmentation of Insulator Images Based on HSI Color Space" proposes a method of image Segmentation in HSI Space by using the maximum inter-class variance method, but it only studies the extraction of blue insulators and cannot be used for the defect detection of insulators.
Disclosure of Invention
1. Technical problem to be solved by the invention
The invention aims to overcome the defects that the detection precision of the electric fault of the insulator in the electric tower is relatively low and the phenomenon of missing detection or wrong detection exists when the unmanned aerial vehicle is used for automatically inspecting the electric transmission line in the prior art, so that the operation safety of the electric transmission line is influenced, and provides a method for detecting and positioning the defect of the insulator in the image of the electric transmission line inspected by the unmanned aerial vehicle. By using the method provided by the invention, the insulator can be quickly and accurately extracted from the power transmission line image with a complex background, the detection precision of the electrical fault of the insulator is higher, and the safe and reliable operation of the power transmission line is ensured.
2. Technical scheme
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
the invention discloses a method for detecting and positioning insulator defects in an unmanned aerial vehicle routing inspection power transmission line image, which is characterized by comprising the following steps of: the method comprises the following steps:
converting an image obtained by aerial photography from an RGB color space to an HSI chroma saturation brightness space, extracting an H component image and an S component image of the HSI chroma saturation brightness space, respectively carrying out binarization processing on the extracted H component image and the extracted S component image to obtain respective corresponding binary images, and then carrying out median filtering on the two binary images and carrying out phase comparison on the two binary images to obtain a preliminary outline binary image of the insulator string;
secondly, extracting the outline of the insulator monomer in the preliminary outline binary image of the insulator string by adopting an ant colony algorithm based on particle swarm optimization parameters;
and thirdly, carrying out ellipse fitting on the outlines of the insulator monomers by adopting a least square method, and positioning the defects of the insulator by detecting the distance between the outlines of the insulator monomers.
Furthermore, when the aerial image is converted from the RGB color space to the HSI chroma saturation luminance space in the step one, the H component and the S component of any pixel point are calculated by the following equations (1) and (2):
where H and S represent the hue component and the saturation component of the HSI chroma saturation luminance space, respectively, and R, G, B represent the red, green and blue components of the RGB color space, respectively.
Furthermore, in the step one, a maximum inter-class variance method is adopted to perform binarization processing on the extracted H component image and the extracted S component image, and the specific steps are as follows: traversing H-component images and S-component imagesThe gray value of each pixel point is taken out, the gray range of the pixel points in the H component image and the S component image is assumed to be 0-m-1, wherein m-1 is the maximum gray value of the pixel points in the H component image and the S component image, and the probability of the occurrence of the pixel point with the gray value i is p i The gray mean values of the H component image and the S component image in the gray range of 0-m-1 are both mu, and the insulator sub-target and the background in the H component image and the S component image are separated into G on the assumption that a gray threshold T exists 0 = 0 to T-1 and G 1 (ii) = { T m-1} two gray scale intervals, and G 0 The probability of occurrence is w 0 ,G 1 The probability of occurrence is w 1 Then G is 0 And G 1 Mean gray level mu in interval 0 、μ 1 And the between-class variance δ of the two intervals 2 (T) are respectively:
in the formula (3)And w 0 +w 1 =1,w 0 μ 0 +w 1 μ 1 =μ;
Gradually increasing the gray threshold T within the range of 0-m-1 gray, taking the gray threshold T to all values within the range of 0-m-1, and calculating the inter-class variance delta obtained in each circulation 2 (T), obtaining the maximum between-class variance max delta after the circulation is finished 2 (T), the T value at the moment is the optimal gray segmentation threshold, the gray value of the pixel point with the gray value larger than the T value in the H component image and the S component image is set as 1, the gray value of the pixel point with the gray value smaller than the T value is set as 0, and therefore the binary images of the H component image and the S component image are obtained.
Furthermore, the concrete steps of extracting the contour of the insulator monomer in the preliminary contour binary image of the insulator string by adopting the ant colony algorithm based on the particle swarm optimization parameters in the second step are as follows:
step 1, assuming that the size of an original aerial image is M x N, initializing different pixel points of (M/2) x (N/2) ants randomly distributed in a preliminary contour binary image of an insulator string by an algorithm;
and 2, performing direction selection movement on all (M/2) × (N/2) ants in the preliminary outline binary image of the insulator string according to a transition probability formula in the formula (4), namely moving all ants in the direction of the maximum probability calculated in the formula (4):
in the formula (4), t is iteration times, (m, n) is a pixel point where the ant is currently located, (l, f) is any pixel point in 3 x 3 neighborhood of the point (m, n),for the probability of ant transitioning from pixel (m, n) to pixel (l, f) in the t-th iteration loop, Ω (m, n) is the set of all pixels in the 3 × 3 neighborhood of point (m, n), η l,f For the heuristic function at point (l, f), it is calculated by equation (5):
in the formula (5), c is an amplification constant, and the numerical value is 1;the gray scale gradient value at the ant position (l, f) is obtained, and the gray scale value is obtained by traversing each pixel point in the image;
in the formula (4), τ (m,n)(l,f) (t) is the intensity of pheromone on the path from point (m, n) to point (l, f) in the t-th iteration, and the initial value is 0.001, and each pheromone is used once in each iterationThe ants move once and generate pheromones at new positions, so that the pheromone intensities of all pixel points are updated, the pheromone intensities and the position updates of all the pixel points after the iteration of the ant colony algorithm is completed each time are stored in the M x N pheromone intensity matrix image, and the formula for the iterative update of the pheromone intensities is as follows:
in formulae (6) to (8), τ (m,n)(l,f) (t-1) is the magnitude of the intensity of the pheromone on the path from point (m, n) to point (l, f) in the t-1 st iteration cycle,for the pheromone quantity left on the paths from (m, n) to (l, f) by the kth ant in the t-1 iteration loop, the pheromone quantity generated by all ants in one iteration is a given fixed normal number, and delta 1 τ (m,n)(l,f) (t-1) and Δ 2 τ (m,n)(l,f) (t-1) the total amount of pheromones left on the local optimal path and the local worst path from the point (m, n) to the point (l, f) in the t-1 th iteration cycle respectively, wherein the local optimal path from the point (m, n) to the point (l, f) is the shortest path from the pixel point (m, n) to the point (l, f), and the local worst path from the point (m, n) to the point (l, f) is the longest path from the pixel point (m, n) to the point (l, f); l is 1 And L 2 The lengths of the above local optimum path and local worst path, phi (t-1) andrespectively at the t-1 th iterationWhen the circulation is replaced, the above local optimal path L is walked 1 And local worst path L 2 Adding the number of ants; xi (t) is the volatilization rate of the pheromone, the initial value of which is 0.5, and the attenuation equation of the volatilization rate xi (t) is as follows as the iterative cycle progresses:
in the formula (9), xi (t) and xi (t-1) are the volatilization rates of pheromones in the t-th iteration and the t-1-th iteration respectively, cn is the current iteration frequency of the algorithm, and tau max And τ min Respectively the maximum value and the minimum value of the total amount of pheromones on paths from the point (m, n) to the point (l, f) of all ants in the t-1 iteration, wherein J is a volatile rate correction value;
in the formula (4), α and β are weight factors of pheromone intensity and heuristic function, initial values of α and β are randomly given, and are positive numbers, and with the circulation of the ant colony algorithm, the particle swarm algorithm is used for parameter training optimization of α and β, and the specific optimization process is as follows:
step a, initializing (M/2) × (N/2) random solution vectors theta i =(α i ,β i ) As random particles, each θ i =(α i ,β i ) Is regarded as the position of a point in a two-dimensional space, wherein the i-th random particle is assigned a random velocity vector v i =(v αi ,v βi );
Step b, calling f for each particle max And training a secondary ant colony algorithm, wherein the ith random particle iteratively updates the spatial position and speed of the ith random particle according to a formula (10) when the ant colony algorithm is called:
wherein f is the number of times of calling the ant colony algorithm, f max The numerical value of (2) is 5; v. of i (f) And theta i (f) Respectively the speed and position of the ith particle at the end of the f-th ant colony algorithm, v i (f-1) and θ i (f-1) respectively representing the speed and the position of the ith particle at the end of the f-1-th ant colony algorithm, wherein i is more than or equal to 1 and less than or equal to (M/2) x (N/2); w is an inertia weight, and a random number larger than 1 is taken; p is best (f-1) finding the optimal particle position for the ith random particle in the calling of the f-1 ant colony algorithm, G best (i-1) the optimal particle position found for the entire population of particles at that time; constant c 1 、c 2 Respectively determine a particle selection P best And G best All of them are [0-2 ]]Random numbers independent of each other; constant s 1 、s 2 Is [0-1 ]]Random numbers independent of each other;
step c, when the ith particle calls the ant colony algorithm to detect the contour of the insulator and the movement update reaches the preset ant colony algorithm cycle number f max Or stopping circulation when the result of the re-circulation updating is consistent with the result of the last circulation, and updating the variable G best ;
D, replacing the next particle, repeating the steps b and c until all the particles are iterated, and finally obtaining G best The optimal particle swarm position is obtained, and the optimal particle swarm position G at the moment is obtained best Updating alpha and beta in the formula (4);
step 3, when the calling times of the ant colony algorithm reach the set maximum cycle times, the operation of the ant colony algorithm is completed, and the maximum inter-class variance method is used for determining the optimal pheromone intensity segmentation threshold tau aiming at the pheromone intensity matrix image found by the whole particle swarm at the moment 0 And obtaining the contour of the insulator monomer after threshold segmentation.
Furthermore, the specific steps of positioning the defects of the insulator by detecting the distance between the outlines of the insulator monomers in the third step are as follows: traversing a coordinate point of the center of the elliptic contour of the insulator, storing the center coordinates of the insulators on the same straight line into the same array A [ x ] [ y ], respectively calculating the average values of the widths and the heights of all the insulators on the straight line, and respectively adopting the width average value D and the height average value H as reference values of the widths and the heights of the single insulator contours; arranging the x coordinates of the center points of the outlines of the insulators on the same straight line from small to large in sequence according to a bubbling sequencing method, and calculating the distance R between the adjacent insulators according to a formula (11):
in the above formula, R 0 Is the average distance between the contour centers of the above-mentioned adjacent insulators on the same line, if lambda>, 1, the adjacent insulator Ax 1 is described][y1]And A [ x0 ]][y0]And lambda insulators are missing, the central positions of the lambda pseudo-insulators between the two insulators are obtained according to the coordinates of the adjacent insulators, and the outline of the defective insulator is drawn in an original drawing according to the width reference value D and the height reference value H of the pseudo-insulators.
3. Advantageous effects
Compared with the prior art, the technical scheme provided by the invention has the following remarkable effects:
(1) According to the method for detecting and positioning the insulator defects in the image of the power transmission line inspected by the unmanned aerial vehicle, the proper color space is selected by combining the imaging characteristics of the insulator in the image of the power transmission line, and the image is segmented by combining the H component and the S component, so that the color information of the image can be fully utilized, the influence of the illumination intensity on the segmentation effect can be reduced, and the segmentation effect of the insulator and the background in the image is greatly improved; the invention further ensures the segmentation effect of the insulator and the background by selecting a matched space segmentation strategy, and provides favorable conditions for the detection of subsequent insulator monomers and faults thereof. The inventor extracts the contour of the insulator monomer in the initial contour binary image of the insulator string by adopting the ant colony algorithm based on the particle swarm optimization parameters through long-term theoretical analysis and practice, optimizes the parameters in the ant colony algorithm through the particle swarm optimization, combines the parameters with the parameters and further improves the parameters, greatly improves the extraction precision of the contour of the insulator monomer, thereby ensuring the detection precision of the insulator fault, effectively preventing the phenomena of missing detection or error detection and ensuring the safe and reliable operation of the power transmission line.
Drawings
FIG. 1 is a flow chart of a method for detecting and positioning insulator defects in an image of an unmanned aerial vehicle inspection power transmission line according to the invention;
FIG. 2 (a) is an actual aerial image of three strings of side-by-side insulators obtained in an embodiment of the invention;
FIG. 2 (b) is an actual aerial image of a single string of defective insulators obtained in an embodiment of the invention;
FIG. 3 (a) is a preliminary contour binary image of the three strings of side-by-side insulators in FIG. 2 (a);
FIG. 3 (b) is a preliminary contour binary image of a single string of insulators with a defect in FIG. 2 (b);
fig. 4 (a) is a contour diagram of an insulator monomer extracted from fig. 3 (a) by using an ant colony algorithm based on particle swarm optimization parameters;
fig. 4 (b) is a contour diagram of the insulator monomer extracted from fig. 3 (b) by using an ant colony algorithm based on particle swarm optimization parameters;
fig. 5 (a) is a profile view of the insulator monomer in fig. 4 (a) after ellipse fitting;
fig. 5 (b) is a profile view of the insulator cell of fig. 4 (b) after ellipse fitting;
fig. 6 (a) and 6 (b) are comparative images of insulator defect marking before and after marking in the embodiment of the present invention.
Detailed Description
For a further understanding of the present invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings and examples.
Example 1
The insulator mainly shows the following characteristics in aerial images: (1) the shape of the insulator monomer is an elliptical shape with uniform length and width; (2) the power transmission lines are arranged in an equidistant string shape; (3) its chroma and saturation values are higher compared to the background; (4) In the fixed voltage class transmission line, the number of insulators in the insulator string is certain; (6) The insulator is generally light green and semitransparent, and the color of the insulator is similar to the background of earth surface vegetation, greenish lake water and the like in aerial photography images.
The invention discloses a method for detecting and positioning insulator defects in an unmanned aerial vehicle routing inspection power transmission line image, which is developed by combining the characteristics of an insulator in an aerial image, the flow of the method is shown as figure 1, and the method comprises the following specific steps:
step one, as shown in fig. 2 (a) and fig. 2 (b), the actual aerial photography image of this embodiment is obtained, and in combination with the characteristic that the insulator is generally light green and translucent, and in the aerial photography image, the color of the aerial photography image is similar to the background of surface vegetation, greenish lake water, and the like, the aerial photography image is converted from an RGB color space to an HSI chroma saturation luminance space, and an H component image and an S component image of the HSI chroma saturation luminance space are extracted, so that the color information of the image can be fully utilized, the influence of illumination intensity on the segmentation effect is reduced, the segmentation effect of the insulator and the background in the image is greatly improved, and the image of the image segmentation effect due to factors such as seasons, weather changes, and the like is avoided. For any pixel point, the H component and the S component are calculated by the following formulas (1) and (2) respectively:
in the above formula, H and S represent a hue component and a saturation component of the HSI chroma saturation luminance space, respectively, and R, G, and B represent a red component, a green component, and a blue component of the RGB color space, respectively.
And performing binarization processing on the extracted H component image and S component image respectively to obtain corresponding binary images, performing median filtering on the two binary images, and performing phase comparison on the two binary images to obtain a preliminary contour binary image of the insulator string, wherein the preliminary contour binary images of the insulator string corresponding to fig. 2 (a) and fig. 2 (b) are respectively shown in fig. 3 (a) and fig. 3 (b). In this embodiment, a maximum inter-class variance method is adopted to perform binarization processing on the extracted H component image and S component image, and the specific steps are as follows: traversing each pixel point in the H component image and the S component image, taking out the gray value of each pixel point, assuming that the gray ranges of the pixel points in the H component image and the S component image are both 0-m-1, wherein m-1 is the maximum gray value of the pixel points in the H component image and the S component image, and the probability of the occurrence of the pixel point with the gray value of i is p i The gray mean values of the H component image and the S component image in the gray range of 0-m-1 are both mu, and the insulator sub-target and the background in the H component image and the S component image are separated into G on the assumption that a gray threshold T exists 0 = 0 to T-1 and G 1 (ii) = { T m-1} two gray scale intervals, and G 0 The probability of occurrence is w 0 ,G 1 The probability of occurrence is w 1 Then G is 0 And G 1 Mean gray level mu in interval 0 、μ 1 And the between-class variance δ of these two intervals 2 (T) are respectively:
in formula (3)And w 0 +w 1 =1,w 0 μ 0 +w 1 μ 1 =μ;
Gradually increasing the gray threshold T within the range of 0-m-1 gray, taking the gray threshold T to all values within the range of 0-m-1, and calculating the inter-class variance delta obtained in each circulation 2 (T), obtaining the maximum inter-class square after the circulation is finishedDifference max δ 2 (T), the T value at the moment is the optimal gray segmentation threshold, the gray value of the pixel point with the gray value larger than the T value in the H component image and the S component image is set to be 1, the gray value of the pixel point with the gray value smaller than the T value is set to be 0, and therefore the binary images of the H component image and the S component image are obtained.
Secondly, extracting the outline of the insulator monomer in the preliminary outline binary image of the insulator string by adopting an ant colony algorithm based on particle swarm optimization parameters, wherein the method comprises the following specific steps of:
and step 1, assuming that the size of the original aerial image is M x N, initializing (M/2) x (N/2) different pixel points of the initial contour binary image of the insulator string by using an algorithm, wherein the ants are randomly distributed in the initial contour binary image.
And 2, performing direction selection movement on all (M/2) × (N/2) ants in the preliminary outline binary image of the insulator string according to a transition probability formula in the formula (4), namely moving all ants in the direction of the maximum probability calculated in the formula (4):
in the formula (4), t is iteration times, (m, n) is a pixel point where the ant is currently located, (l, f) is any pixel point in 3 x 3 neighborhood of the point (m, n),in order to determine the probability that ants are shifted from pixel (m, n) to pixel (l, f) in the t-th iteration loop, Ω (m, n) is the set of all pixels in the 3 × 3 neighborhood of point (m, n). Eta l,f For the heuristic function at point (l, f), it is calculated by equation (5):
in the formula (5), c is an amplification constant, and the numerical value is 1;the gray scale gradient value at the ant position (l, f) is used in this embodiment, and the gray scale gradient value at the ant position (l, f) is used in this embodimentAs an heuristic function, the gray scale change of the edge part of the insulator outline is severe, so that the pixel points of the insulator elevator outline can be better distinguished from other pixel points, the detection precision of the insulator single body outline is improved, and the normal starting of an algorithm is ensured.
In the formula (4), τ (m,n)(l,f) (t) is the magnitude of pheromone intensity on the path from the time point (M, N) to the point (l, f) in the t-th iteration, the initial value is 0.001, each ant moves once every iteration, and pheromone is generated at a new position, so that the pheromone intensity of all pixel points is updated, in this embodiment, the pheromone intensity and the position update of each pixel point after the iteration of the ant colony algorithm is completed each time are stored in an M × N pheromone intensity matrix image, and the formula for the iterative update of the pheromone intensity is as follows:
in formulae (6) to (8), τ (m,n)(l,f) (t-1) is the time point (m, n) of the t-1 th iteration cycle) The magnitude of the intensity of the pheromone on the path to point (l, f),for the pheromone quantity left on the paths from (m, n) to (l, f) by the kth ant in the t-1 iteration loop, the pheromone quantity generated by all ants in one iteration is a given fixed normal number, and delta 1 τ (m,n)(l,f) (t-1) and Δ 2 τ (m,n)(l,f) (t-1) the total amount of pheromones left on the local optimal path and the local worst path from the point (m, n) to the point (l, f) in the t-1 th iteration cycle respectively, wherein the local optimal path from the point (m, n) to the point (l, f) is the shortest path from the pixel point (m, n) to the point (l, f), and the local worst path from the point (m, n) to the point (l, f) is the longest path from the pixel point (m, n) to the point (l, f); l is 1 And L 2 The lengths of the above local optimum path and local worst path, phi (t-1) andrespectively walking through the local optimal path L in the t-1 th iteration cycle 1 And local worst path L 2 The number of ants on the surface; in the embodiment, the pheromone released by the ants on the worst path is processed by adopting the elite strategy, and the concentration of the pheromone released by the ants on the optimal path is amplified, so that the condition that the search is trapped in the local optimal pseudo solution can be prevented, the amount of the pheromone on the contour detected by mistake is reduced to the minimum, and the accuracy of contour detection is increased. Xi is the volatilization rate of the pheromone, the initial value of the volatilization rate is 0.5, and the attenuation equation of the volatilization rate xi is as follows along with the progress of the iterative loop:
in the formula (9), ξ (t) and ξ (t-1) are respectively the t-th order and the t-th orderthe volatilization rate of pheromone during t-1 iteration, cn is the current iteration number of the algorithm, tau max And τ min Respectively the maximum value and the minimum value of the total pheromone quantity on the paths from the point (m, n) to the point (l, f) of all ants in the t-1 iteration, wherein J is a volatility correction value;
in the formula (4), alpha and beta are weight factors of pheromone strength and an elicitation function respectively, initial values of alpha and beta are given randomly and are positive numbers respectively, the alpha and beta are trained and optimized by using a particle swarm optimization algorithm along with the circulation of the ant colony optimization algorithm, and the specific optimization process is as follows:
step a, initializing (M/2) × (N/2) random solution vectors theta i =(α i ,β i ) As random particles, each θ is i =(α i ,β i ) Is regarded as the position of a point in a two-dimensional space, wherein the i-th random particle is assigned a random velocity vector v i =(v αi ,v βi )。
Step b, calling f for each particle max Training a secondary ant colony algorithm, wherein the ith random particle iteratively updates the spatial position and the speed of the ith random particle according to a formula (10) when the ant colony algorithm is called:
wherein f is the number of times of calling the ant colony algorithm, and f is subjected to a large number of experimental researches max Is optimized, the optimized determined f max The numerical value of (2) is 4-8, and the numerical value of (5) in the embodiment, so that the generation of a local optimal pseudo solution can be prevented, and the detection precision of the single body contour of the insulator is improved. v. of i (f) And theta i (f) Respectively the speed and position of the ith particle at the end of the f-th ant colony algorithm, v i (f-1) and θ i (f-1) respectively representing the speed and position of the ith particle at the end of the f-1 th ant colony algorithm, wherein i is more than or equal to 1 and less than or equal to (M/2) and N/2; w is an inertia weight, and a random number larger than 1 is taken; p best (f-1) invoking the ith random particle f-1 timesOptimal particle position, G, found during ant colony algorithm best (i-1) finding the optimal particle position of the whole particle swarm when the i-1 th random particle finishes calling the ant colony algorithm; constant c 1 、c 2 Respectively determine a particle selection P best And G best All are [0-2 ]]Random numbers independent of each other; constant s 1 、s 2 Is [0-1 ]]Random numbers independent of each other.
In this embodiment, the position P of the optimal particle found by the ith random particle with each call of the ant colony algorithm best Updating, and setting the position of the optimal particle found by the ith random particle at the end of the f-1 th ant colony algorithm as P best (f-1), the position θ of the particle at the end of the f-th ant colony algorithm i (f)=(α i (f),β i (f) Length L of path taken by ant when the particle calls f-th ant colony algorithm) f And find P best (f-1) the length L of the path taken by the ant in the ant colony algorithm called best (f-1) comparison, if L f ≤L best (f-1), then in theta i (f)=(α i (f),β i (f) As the optimal particle position P that the particle found at the f-th ant colony algorithm best (f) (ii) a If L is f >L best (f-1), then P remains best (f-1) as the optimal particle position P found by the particle at the f-th ant colony algorithm best (f) In that respect In this embodiment, the initial spatial position θ of the ith random particle is used i =(α i ,β i ) For which the optimum particle position P is found best Of (4) is calculated.
C, when the ith particle calls the ant colony algorithm to detect the insulator contour and the motion update reaches the preset ant colony algorithm cycle number f max Or the result of the re-circulation updating is consistent with the result of the last circulation, namely the P found in the circulation best And P found in last cycle best When the same, stopping circulation and searching the ith particle for the optimal particle position P best The length of the path taken by the ant in the time-called ant colony algorithm and the search G best (i-1) paths taken by ants in the called ant colony algorithmLength comparison, and updating variable G with short path of ant as the measure best 。
D, replacing the next particle, repeating the steps b and c until all the particles are iterated, and finally obtaining G best The optimal particle swarm position is obtained, and the optimal particle swarm position G at the moment is obtained best Updating alpha and beta in the formula (4).
Step 3, when the calling times of the ant colony algorithm reach the set maximum cycle times, the operation of the ant colony algorithm is completed, and the maximum inter-class variance method is used for determining the optimal pheromone intensity segmentation threshold tau aiming at the pheromone intensity matrix image found by the whole particle swarm at the moment 0 By using τ 0 The intensity matrix image of the pheromone found by the whole particle swarm is divided, and the intensity of the pheromone is larger than tau 0 The pixel points are pixel points on the contour of the insulator monomer, so that the contour of the insulator monomer can be obtained after threshold segmentation, as shown in fig. 4 (a) and 4 (b).
It should be noted that, although there are related reports on researches on combining the particle swarm optimization algorithm and the ant colony algorithm in the prior art, no research is made on applying the particle swarm optimization algorithm to the detection of the insulator defects in the power transmission line images of the unmanned aerial vehicle, and because the complexity of the imaging characteristics of the insulator and the interference of the insulator by similar color backgrounds such as surface vegetation, lake water and the like are serious, the existing improved ant colony algorithm based on particle swarm optimization cannot be directly applied to the detection of the insulator defects, the detection precision is low, the phenomena of false detection and missed detection are easy to occur, and the running time is long, which also becomes a problem that puzzles the inventor for a long time. The inventor obtains the technical scheme of the invention only by analyzing the imaging characteristics of the insulator, combining the difference between background imaging and insulator imaging in the image and carrying out a large amount of long-term research and experimental simulation and further improving the existing improved ant colony algorithm based on particle swarm parameter optimization, so that the improved ant colony algorithm is suitable for extracting the insulator monomer contour in the initial contour binary image of the insulator string obtained after segmentation and can adapt to the imaging characteristics of the insulator, thereby greatly improving the detection precision of the insulator defect, reducing the interference of the background and improving the running speed.
And step three, traversing the coordinates of the central point and the length of the long and short axes of each insulator outline, performing ellipse fitting on the insulator single body outlines by adopting a least square method because the insulator single body outlines are in an elliptical shape with uniform length and width in the image and are arranged in a power transmission line in a string shape at equal intervals, and setting variables for calculating the number of connected domains in the image, wherein the fitted insulator single body outlines are as shown in the figure 5 (a) and the figure 5 (b). The method comprises the following steps of positioning the defects of the insulator by detecting the distance between the outlines of the insulator monomers, wherein the specific steps of positioning the defects of the insulator are as follows: traversing a coordinate point of the center of the elliptic contour of the insulator, storing the center coordinates of the insulator on the same straight line into the same array A [ x ] [ y ], respectively calculating the average values of the widths and the heights of all the insulator contours on the straight line, and respectively adopting the width average value D and the height average value H as reference values of the widths and the heights of the single insulator contours. Arranging the x coordinates of the contour center points of the insulators on the same straight line from small to large in sequence according to a bubbling sorting method, and calculating the distance R between the adjacent insulators according to a formula (11):
in the above formula, R 0 Is the average distance between the contour centers of the above-mentioned adjacent insulators on the same line, if lambda>, 1, the adjacent insulator Ax 1 is described][y1]And A [ x0 ]][y0]And lambda insulators are missing, the central positions of the lambda pseudo insulators between the two adjacent insulators are obtained according to the coordinates of the two adjacent insulators, and the outline of the defective insulator is drawn in an original drawing according to the width reference value D and the height reference value H of the lambda pseudo insulators.
As shown in fig. 6 (a) and fig. 6 (b), which are comparison diagrams before marking and after marking the insulator defect in the present embodiment, the number of insulators and the number of missing insulators in the current image are counted as 35 and 1, respectively. The platform of the embodiment is developed based on Qt software, the algorithm running time is about 20ms, the platform can be transplanted to an embedded platform for flexible use, and the platform can also be applied to a video real-time processing system after secondary development.
Claims (4)
1. A method for detecting and positioning insulator defects in power transmission line images by unmanned aerial vehicle inspection is characterized in that: the method comprises the following steps:
converting an aerial image from an RGB color space to an HSI chroma saturation brightness space, extracting an H component image and an S component image of the HSI chroma saturation brightness space, respectively carrying out binarization processing on the extracted H component image and the extracted S component image to obtain binary images corresponding to the H component image and the S component image, and carrying out median filtering on the two binary images and then carrying out phase comparison on the two binary images to obtain a preliminary outline binary image of the insulator string;
step two, extracting the outline of the insulator monomer in the preliminary outline binary image of the insulator string by adopting an ant colony algorithm based on particle swarm optimization parameters, wherein the steps are as follows:
step 1, assuming that the size of an original aerial image is M x N, initializing different pixel points of (M/2) x (N/2) ants randomly distributed in a preliminary contour binary image of an insulator string by an algorithm;
and 2, performing direction selection movement on all (M/2) × (N/2) ants in the preliminary outline binary image of the insulator string according to a transition probability formula in the formula (4), namely moving all ants in the direction of the maximum probability calculated in the formula (4):
in the formula (4), t is iteration times, (m, n) is a pixel point where the ant is currently located, (l, f) is any pixel point in 3 x 3 neighborhood of the point (m, n),in order to determine the probability of ant transfer from pixel (m, n) to pixel (l, f) in the t-th iteration cycle, Ω (m, n) is in the 3 × 3 neighborhood of point (m, n)Set of all pixels, η l,f For the heuristic function at point (l, f), it is calculated by equation (5):
in the formula (5), c is an amplification constant, and the numerical value is 1;the gray scale gradient value at the ant position (l, f) is obtained by traversing each pixel point in the image;
in the formula (4), τ (m,n)(l,f) (t) is the magnitude of pheromone intensity on the path from the point (M, N) to the point (l, f) in the t-th iteration, the initial value is 0.001, each ant moves once every iteration, pheromones are generated at new positions, the pheromone intensity of all pixel points is updated, the pheromone intensity and the position update of each pixel point after the iteration of the ant colony algorithm is completed every time are stored in an M x N pheromone intensity matrix image, and the formula for the iterative update of the pheromone intensity is as follows:
in formulae (6) to (8), τ (m,n)(l,f) (t-1) is the time point (m) in the t-1 st iteration cycleN) the magnitude of the intensity of the pheromone on the path to point (l, f),for the pheromone quantity left on the paths from (m, n) to (l, f) by the kth ant in the t-1 iteration loop, setting the pheromone quantity generated by all ants in one iteration to be a given fixed normal number, delta 1 τ (m,n)(l,f) (t-1) and Δ 2 τ (m,n)(l,f) (t-1) the total amount of pheromones left on the local optimal path and the local worst path from the point (m, n) to the point (l, f) in the t-1 th iteration cycle respectively, wherein the local optimal path from the point (m, n) to the point (l, f) is the shortest path from the pixel point (m, n) to the point (l, f), and the local worst path from the point (m, n) to the point (l, f) is the longest path from the pixel point (m, n) to the point (l, f); l is a radical of an alcohol 1 And L 2 The lengths of the above local optimum path and local worst path, phi (t-1) andrespectively walking through the local optimal path L in the t-1 th iteration cycle 1 And local worst path L 2 Adding the number of ants; xi is the volatilization rate of the pheromone, the initial value of the volatilization rate is 0.5, and the attenuation equation of the volatilization rate xi is as follows along with the progress of the iterative loop:
in the formula (9), xi (t) and xi (t-1) are the volatilization rates of pheromones in the t-th iteration and the t-1-th iteration respectively, cn is the current iteration frequency of the algorithm, and tau max And τ min Respectively the maximum value and the minimum value of the total pheromone quantity on the paths from the point (m, n) to the point (l, f) of all ants in the t-1 iteration, wherein J is a volatility correction value;
in the formula (4), α and β are weight factors of pheromone intensity and heuristic function, initial values of α and β are randomly given, and are positive numbers, and with the circulation of the ant colony algorithm, the particle swarm algorithm is used for parameter training optimization of α and β, and the specific optimization process is as follows:
step a, initializing (M/2) × (N/2) random solution vectors theta i =(α i ,β i ) As random particles, each θ is i =(α i ,β i ) Is regarded as the position of a point in a two-dimensional space, wherein the i-th random particle is assigned a random velocity vector
Step b, calling f for each particle max Training a secondary ant colony algorithm, wherein the ith random particle iteratively updates the spatial position and the speed of the ith random particle according to a formula (10) when the ant colony algorithm is called:
wherein f is the number of times of calling the ant colony algorithm, f max The numerical value of (2) is 5; v. of i (f) And theta i (f) Respectively the speed and position of the ith particle at the end of the f-th ant colony algorithm, v i (f-1) and θ i (f-1) respectively representing the speed and position of the ith particle at the end of the f-1 th ant colony algorithm, wherein i is more than or equal to 1 and less than or equal to (M/2) and N/2; w is an inertia weight, and a random number greater than 1 is taken; p best (f-1) finding the optimal particle position of the ith random particle in the f-1 th ant colony algorithm, G best (i-1) the optimal particle position found for the entire population of particles at that time; constant c 1 、c 2 Respectively determine a particle selection P best And G best All are [0-2 ]]Random numbers independent of each other; constant s 1 、s 2 Is [0-1 ]]Random numbers independent of each other;
step c, when the ith particle calls the ant colonyThe algorithm detects the contour of the insulator and the motion update reaches the preset ant colony algorithm cycle number f max Or stopping circulation when the result of the re-circulation updating is consistent with the result of the last circulation, and updating the variable G best ;
D, replacing the next particle, repeating the step b and the step c until all the particles finish iteration, and finally obtaining G best Namely the optimal particle swarm position, according to the optimal particle swarm position G at the moment best Updating alpha and beta in the formula (4);
step 3, when the calling times of the ant colony algorithm reach the set maximum cycle times, the operation of the ant colony algorithm is finished, and the optimal pheromone intensity segmentation threshold tau is determined by using a maximum between-class variance method aiming at the pheromone intensity matrix image found by the whole particle swarm at the moment 0 Obtaining the outline of the insulator monomer after threshold segmentation;
and thirdly, carrying out ellipse fitting on the outlines of the insulator monomers by adopting a least square method, and positioning the defects of the insulator by detecting the distance between the outlines of the insulator monomers.
2. The method for detecting and positioning the insulator defects in the unmanned aerial vehicle inspection power transmission line image according to claim 1, characterized in that: in the first step, when the aerial image is converted from an RGB color space to an HSI chroma saturation brightness space, H components and S components of any pixel point are calculated through formulas (1) and (2) respectively:
where H and S represent the hue component and the saturation component of the HSI chroma saturation luminance space, respectively, and R, G, B represent the red, green and blue components of the RGB color space, respectively.
3. The method for detecting and positioning the insulator defects in the unmanned aerial vehicle inspection power transmission line image according to claim 2, characterized in that: in the first step, a maximum inter-class variance method is adopted to carry out binarization processing on the extracted H component image and S component image respectively, and the specific steps are as follows: traversing each pixel point in the H component image and the S component image, taking out the gray value of each pixel point, assuming that the gray ranges of the pixel points in the H component image and the S component image are both 0-m-1, wherein m-1 is the maximum gray value of the pixel points in the H component image and the S component image, and the probability of the occurrence of the pixel point with the gray value of i is p i The gray level mean values of the H component image and the S component image in the gray level range of 0-m-1 are both mu, and the insulating sub-targets and the background in the two images are separated into G by assuming that a gray level threshold T exists 0 = 0 to T-1 and G 1 (ii) = { T m-1} two gray scale intervals, and G 0 The probability of occurrence is w 0 ,G 1 The probability of occurrence is w 1 Then G is 0 And G 1 Mean gray level mu in interval 0 、μ 1 And the between-class variance δ of the two intervals 2 (T) are respectively:
in the formula (3)And w 0 +w 1 =1,w 0 μ 0 +w 1 μ 1 =μ;
Gradually increasing the gray threshold T within the range of 0-m-1 gray level, taking the gray threshold T to all values within the range of 0-m-1, and calculating the inter-class variance delta obtained by each circulation 2 (T), obtaining the maximum between-class variance max delta after the circulation is finished 2 (T), the T value at the moment is the optimal gray segmentation threshold, the gray value of the pixel point with the gray value larger than the T value in the H component image and the S component image is set as 1, and the gray value smaller than the T value is set asAnd setting the gray value of the pixel point as 0, thereby obtaining respective binary images of the H component image and the S component image.
4. The method for detecting and positioning the insulator defects in the unmanned aerial vehicle inspection power transmission line image according to any one of claims 1-3, characterized in that: the specific steps of positioning the defects of the insulator by detecting the distance between the outlines of the insulator monomers in the third step are as follows: traversing coordinate points of the center of the elliptic contour of the insulator, storing the center coordinates of the insulator on the same straight line into the same array A [ x ] [ y ], respectively calculating the average values of the width and the height of all the insulator contours on the straight line, and respectively adopting the width average value D and the height average value H as reference values of the width and the height of a single insulator contour; arranging the x coordinates of the center points of the outlines of the insulators on the same straight line from small to large in sequence according to a bubbling sequencing method, and calculating the distance R between the adjacent insulators according to a formula (11):
in the above formula, R 0 Is the average distance between the contour centers of the adjacent insulators on the same straight line, if lambda>, 1, the adjacent insulator Ax 1 is described][y1]And A [ x0 ]][y0]And lambda insulators are missing, the central positions of the lambda pseudo-insulators between the two insulators are obtained according to the coordinates of the adjacent insulators, and the outline of the defective insulator is drawn in an original drawing according to the width reference value D and the height reference value H of the pseudo-insulators.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510531559.1A CN105160669B (en) | 2015-08-21 | 2015-08-21 | The detection of defects of insulator and localization method in a kind of unmanned plane inspection transmission line of electricity image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510531559.1A CN105160669B (en) | 2015-08-21 | 2015-08-21 | The detection of defects of insulator and localization method in a kind of unmanned plane inspection transmission line of electricity image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105160669A CN105160669A (en) | 2015-12-16 |
CN105160669B true CN105160669B (en) | 2018-01-09 |
Family
ID=54801511
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510531559.1A Active CN105160669B (en) | 2015-08-21 | 2015-08-21 | The detection of defects of insulator and localization method in a kind of unmanned plane inspection transmission line of electricity image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105160669B (en) |
Families Citing this family (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105957081B (en) * | 2016-04-28 | 2019-01-08 | 华北电力大学(保定) | A kind of glass insulator falls to go here and there fault detection method |
CN106161204B (en) * | 2016-06-08 | 2019-05-10 | 苏州大学 | Data transmission method in mobile social network based on group intelligence |
CN106408025B (en) * | 2016-09-20 | 2019-11-26 | 西安工程大学 | Aerial Images insulator classifying identification method based on image procossing |
CN107886096A (en) * | 2016-09-29 | 2018-04-06 | 成都思晗科技股份有限公司 | A kind of insulator of transmission line of electricity comes off defect inspection method |
CN106780438B (en) * | 2016-11-11 | 2020-09-25 | 广东电网有限责任公司清远供电局 | Insulator defect detection method and system based on image processing |
CN107194923B (en) * | 2017-05-22 | 2021-09-03 | 同济大学 | Ultraviolet image diagnosis method for defect inspection of contact network power equipment |
CN107742283B (en) * | 2017-09-16 | 2021-08-06 | 河北工业大学 | Method for detecting defect of uneven thickness of grid line on appearance of battery piece |
CN108108682B (en) * | 2017-12-14 | 2020-05-15 | 华北电力大学(保定) | Insulator flashover fault positioning method and system |
CN108470141B (en) * | 2018-01-27 | 2021-08-10 | 天津大学 | Statistical feature and machine learning-based insulator identification method in distribution line |
CN109785285B (en) * | 2018-12-11 | 2023-08-08 | 西安工程大学 | Insulator damage detection method based on ellipse characteristic fitting |
CN111220619B (en) * | 2019-12-05 | 2023-05-05 | 河海大学常州校区 | Insulator self-explosion detection method |
CN111415357B (en) * | 2020-03-19 | 2023-04-07 | 长光卫星技术股份有限公司 | Portable shadow extraction method based on color image |
CN111429419B (en) * | 2020-03-19 | 2023-04-07 | 国网陕西省电力公司电力科学研究院 | Insulator contour detection method based on hybrid ant colony algorithm |
CN113192027B (en) * | 2021-04-29 | 2023-03-24 | 华南理工大学 | Detection method and application of high-power LED module packaging defects |
CN113342046B (en) * | 2021-06-22 | 2022-06-03 | 国网湖北省电力有限公司宜昌供电公司 | Power transmission line unmanned aerial vehicle routing inspection path optimization method based on ant colony algorithm |
CN115318759B (en) * | 2022-07-29 | 2024-04-09 | 武汉理工大学 | Unmanned aerial vehicle-based insulator laser cleaning method and system |
CN116797604B (en) * | 2023-08-28 | 2023-12-26 | 中江立江电子有限公司 | Glass insulator defect identification method, device, equipment and medium |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7397970B2 (en) * | 2004-12-07 | 2008-07-08 | Lockheed Martin Corporation | Automatic scene correlation and identification |
CN104639914A (en) * | 2015-02-16 | 2015-05-20 | 国网安徽省电力公司铜陵供电公司 | Device and method for high-voltage power line insulator imaging and contaminant detection |
-
2015
- 2015-08-21 CN CN201510531559.1A patent/CN105160669B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN105160669A (en) | 2015-12-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105160669B (en) | The detection of defects of insulator and localization method in a kind of unmanned plane inspection transmission line of electricity image | |
CN113379703B (en) | Photovoltaic panel dark spot defect detection method based on Yolo-v4 network structure | |
CN113284124B (en) | Photovoltaic panel defect detection method based on unmanned aerial vehicle vision | |
CN108734689B (en) | Method for detecting scattered strands of conducting wires based on region growth | |
Wang et al. | Recognition and drop-off detection of insulator based on aerial image | |
CN109376606B (en) | Power inspection image tower foundation fault detection method | |
CN112435208B (en) | Insulator region detection method for significant edge guidance | |
CN107492094A (en) | A kind of unmanned plane visible detection method of high voltage line insulator | |
CN113592822B (en) | Insulator defect positioning method for electric power inspection image | |
CN107185854A (en) | The algorithm of photovoltaic cell acetes chinensis and color classification based on RGB channel | |
CN110797792B (en) | Power transmission tower bird nest positioning and fault early warning method | |
CN112070715A (en) | Transmission line small-size hardware defect detection method based on improved SSD model | |
CN106340006B (en) | Icing degree evaluation method based on insulator image umbrella extension | |
CN104166999B (en) | Cloud cluster extracting method based on strength layering of foundation cloud pictures | |
CN102831393A (en) | Rapid image recognizing method of power tower pole outline | |
CN114241364A (en) | Method for quickly calibrating foreign object target of overhead transmission line | |
CN111220619B (en) | Insulator self-explosion detection method | |
CN107179479A (en) | Transmission pressure broken lot defect inspection method based on visible images | |
Wang et al. | Insulator defect recognition based on faster R-CNN | |
CN112150412B (en) | Insulator self-explosion defect detection method based on projection curve analysis | |
CN109523543A (en) | A kind of wire strand breakage detection method based on Edge Distance | |
CN109949209A (en) | A kind of rope detection and minimizing technology based on deep learning | |
CN113902792A (en) | Building height detection method and system based on improved RetinaNet network and electronic equipment | |
Hao et al. | Detection of bird nests on power line patrol using single shot detector | |
CN117036825A (en) | Solar cell panel detection method, medium and system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CP02 | Change in the address of a patent holder |
Address after: 243000 R & D Building 1, north side of Beijing Avenue, Ma'anshan demonstration park, Anhui Province Patentee after: MAANSHAN AHUT INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTE Co.,Ltd. Address before: 243071 No. 578 Taibai Road, Ma'anshan economic and Technological Development Zone, Anhui Patentee before: MAANSHAN AHUT INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTE Co.,Ltd. |
|
CP02 | Change in the address of a patent holder |