CN114241364A - A rapid calibration method for foreign object targets in overhead transmission lines - Google Patents

Abstract

The invention discloses a method for quickly calibrating foreign objects in overhead transmission lines. Aiming at the difficulty of traditional tools and methods to efficiently and safely remove foreign objects in overhead transmission lines, a rapid calibration method for high-altitude objects is proposed. First, acquire the foreign object video image in real time, process and identify the video image and select the target area, and use the image feature point-ORB feature point matching fusion algorithm to realize the rapid identification and calibration of the foreign object target, which greatly increases the accuracy of the foreign object target location. , reduce the operator's difficulty and human error, and can achieve close to 100% foreign object recognition and tracking effect.

Description

A rapid calibration method for foreign object targets in overhead transmission lines

technical field

The invention belongs to the field of power transmission safety, in particular to a method for quickly calibrating a foreign object target in an overhead power transmission line.

Background technique

Electricity is the foundation of the national economy and social development. As the energy pillar supporting the development of the national economy, the power system plays an irreplaceable role in today's social development and is related to the stable development of the entire society. The safe and stable operation of the power grid is not only related to the Economic development is closely related, and it is also the foundation for people to live and work in peace and contentment and social stability. With the rapid development of my country's power grid, the scale of transmission lines is getting larger and larger, and the number of transmission line equipment is increasing. Line inspection and maintenance are difficult, and the technical content is high.

In such a large-scale power transmission grid, power system failures are inevitably caused by various reasons. In various accidents caused by the power grid, non-insulated objects such as kites, long cloth strips, and plastic bags are suspended between the transmission lines. Short circuit between lines, tripping the switch of the substation. Live removal of foreign objects is the item with the highest safety risk in live work. There are two traditional main methods for removing foreign objects hooked on high-voltage wires: one is to remove them after a power outage; the other is to remove them during equipotential live work. Both of these two methods require a lot of human and material resources, and the operation procedures are complex, time-consuming, labor-intensive, and safety and reliability are low. At the same time, due to the variety of types and winding methods of foreign objects, the workers basically rely on the “follow-up” to deal with foreign objects during live work; some foreign objects that are tightly entangled with the transmission line components are highly dangerous and can only be treated by power outages, but the power outage treatment directly It reduces the reliability of power supply, causes social and economic losses, and affects people's lives.

Therefore, it is necessary to innovate and integrate new technologies to improve the efficiency and quality of transmission line inspections, adopt efficient methods that keep pace with the times and remove foreign objects on overhead transmission lines in a timely manner, and solve hidden safety hazards of transmission lines in advance.

As a new type of foreign matter removal method, laser has been widely used in all walks of life. Different from ordinary physical shearing, hooking and other methods, the laser removes the foreign object by increasing the temperature of the surface of the foreign object for burning. Most of the floating foreign objects such as plastic films and kites are wrapped around the transmission line. These materials are usually plastic, nylon, polyester and other flammable substances with low ignition points, and their ignition points are generally below 300 °C. The method of heating and fusing foreign objects in the transmission line will not cause substantial damage to the transmission line.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for quickly calibrating foreign objects in overhead transmission lines, which solves the problem of target identification of entangled foreign objects on overhead transmission lines, and increases the flexibility, mobility and efficiency in the process of removing foreign objects in overhead transmission lines, so as to achieve better Good control needs.

The technical solution for realizing the present invention is: a method for quickly calibrating foreign objects in overhead transmission lines, comprising the following steps:

Step 1. Obtain the video image of the target area in real time, process the image, identify the current video image information frame by frame, when the target object is identified, set the frame picture as the initial frame, and enlarge the video image of the target area;

Step 2, further accurately identify the enlarged video image, solve the image information, obtain the coordinate information of the foreign object target, and the coordinate information of the foreign object target at the winding position of the overhead transmission line;

Step 3: Detecting the picture image information of the initial frame, performing automatic tracking processing of the foreign object target, and identifying the position state information of the foreign object target in the video image, including image feature points, centroid position and movement trajectory;

Step 4. Track the foreign body and cut it by laser;

Step 5: Determine whether there is still a foreign object target, if so, return to Step 1, otherwise end.

A rapid calibration system for foreign object targets in overhead transmission lines, comprising the following modules:

Image processing module: used to obtain the video image of the target area in real time, and process the image. When the target object is recognized, the frame will be set as the initial frame, and the video image of the target area will be enlarged;

Foreign object coordinate acquisition module: used to further accurately identify the enlarged video image, solve the image information, obtain the coordinate information of the foreign object target, and the coordinate information of the foreign object target in the winding position of the overhead transmission line;

Foreign body tracking module: used to track foreign body targets, identify the position status information of foreign body targets in the video image, including image feature points, centroid position and movement trajectory;

Foreign body processing module: Use laser to cut foreign body and judge whether there is still foreign body target.

Compared with the prior art, the significant advantages of the present invention are:

(1) The technical scheme of the present invention adopts an improved image recognition and tracking algorithm to improve the accuracy and precision of foreign object target recognition;

(2) The technical solution of the present invention uses the trained foreign object target data training set, combined with ORB feature points and Mean-Shift operator prediction, can quickly detect and identify the foreign object target to be detected.

The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

Description of drawings

FIG. 1 is a flow chart showing the realization of the method for identifying foreign objects in overhead transmission lines of the present invention.

FIG. 2 is a mapping diagram of the grayscale histogram equalization algorithm in the present invention.

FIG. 3 is a schematic diagram for comparison and comparison of ORB feature point identification in an embodiment of the present invention.

Detailed ways

A method for quickly calibrating a foreign object target in an overhead transmission line, comprising the following steps:

Step 1. Obtain the video image of the target area in real time, process the image, and identify the current video image information frame by frame. When the target object is identified, the frame image is set as the initial frame, and the video image of the target area is enlarged, specifically: :

Step 1-1. Obtain a video image, perform grayscale conversion on each frame of the obtained image, establish a random mathematical model of a digital grayscale image, map each pixel on the original image, and obtain a grayscale-enhanced image, specifically: :

Step 1-1-1, construct a random mathematical model of digital grayscale image for the acquired image, count the grayscale values of all pixels, and obtain a grayscale distribution histogram, that is, obtain the grayscale level of the original image;

Step 1-1-2. In the original grayscale image X, according to the segmentation threshold _rc , the image is divided into two sub-images without intersection, and then the histogram equalization algorithm is applied to the two sub-images respectively, and the gray distribution is carried out. Normalized probability calculation:

Among them, p _Ld (d _j ) and P _Ud (d _j ) represent the gray-scale normalization probability of the low-gray sub-image and the high-gray sub-image, respectively; where d _k =0,1,2,..., L-1, L is the maximum value of the gray level, P _d (d _j ) is the gray distribution probability of the gray level d _j in the original image, and P _d (d _k ) is the gray level of the gray level d _k in the original image. degree distribution probability;

Step 1-1-3, construct the mapping relationship between the gray level of the original image and the gray level of the target image, construct the gray distribution probability density function of the continuous variable, and perform the low gray area and high gray area of the image according to the gray distribution probability density function. By dividing the gray area, the low gray area and the high gray area of the image are obtained respectively, and the gray image of the high gray area is obtained, that is, the image after gray enhancement.

Step 1-2, use the Gaussian blur method to denoise the grayscale-enhanced image through filtering to obtain a denoised image;

Step 1-3: Perform Gaussian filtering on the denoised image to obtain a smooth image, specifically:

Using the Gaussian function value of the discrete point as the weight, the window template is collected in the denoised image, and the weighted average of each pixel value in the grayscale matrix in the collected window template is performed to eliminate the Gaussian noise:

Set the size of the window template to (2k+1)×(2k+1), then the weight coefficient of each element in the template is:

where (i,j)(i,j=1,2,...,2k+1) represents the relative coordinates of each pixel in the window template.

Steps 1-4, use the Canny edge detection algorithm to detect the foreign object on the obtained smooth image, when it is detected that the foreign object is in the center of the image, update the frame image to the initial frame, and zoom in on the video image of the area where the object is located.

Step 2, further accurately identify the enlarged video image, solve the image information, obtain the coordinate information of the foreign object target, and the coordinate information of the foreign object target winding position on the overhead transmission line, specifically:

Step 2-1. Use the Canny edge detection algorithm to detect the target edge of the enlarged target area, detect the horizontal, vertical and two diagonal edges of the target area in four directions, calculate its gradient strength and direction, and Noise removal is performed on it, and the Gaussian blur method is used to convert the image information into Gaussian cells to eliminate the influence of noise on recognition, specifically:

Step 2-1-1. Divide the image gradient direction into four directions according to the angle, namely 0°, 45°, 90°, and 135° edges, and compare the center value in its corresponding gradient direction in each 3×3 neighborhood. The size of the other two values, if the central value is the maximum value, it is retained, otherwise the central value is suppressed to remove false edge points;

Step 2-1-2. Set two gradient size thresholds: the high threshold is m _thh and the low threshold m _thl , and the image is divided into the following two edge maps:

E _H ={e(x,y)|if M(x,y) _{≥m thh} , then e(x,y)=1; else e(x,y)=0}

E _L ={e(x,y)|if M(x,y) _{≥m thl} , then e(x,y)=1; otherwise e(x,y)=0}

Among them, M(x, y) represents the edge gradient value at the (x, y) coordinate in the image before segmentation, and e(x, y) represents the edge gradient value at the current (x, y) coordinate after segmentation;

Step 2-1-3, mark the corresponding positions of all pixels in the high threshold map E _H as edge points, then traverse all the points in E _H , and use the 8-connection relationship to determine the final edge point in the low threshold map E _L , which enhances the edge continuity and completes the target edge detection.

Step 2-2: Convert the image information of the enlarged target area into an image recognition matrix, process the image recognition matrix by Hough transform detection, and obtain the position coordinates of the foreign object target and the winding point of the overhead transmission line in the image.

Step 3: Detect the picture image information of the initial frame, perform automatic tracking processing of the foreign object target, and identify the position state information of the foreign object target in the video image, including the image feature point, the position of the centroid and the movement trajectory, specifically:

Step 3-1. Use the foreign object area in the updated initial frame of the video image as the reference frame, use the Mean-Shift operator to predict the foreign object target area in the current frame, and use the Mean-Shift algorithm to solve the foreign object target feature points in the current frame. Offset mean of :

为当前基准点下的特征点偏移均值，C_k,d表示归一化常量，k(x)为轮廓函数，h为核函数带宽参数，d为矩阵维度，H是一个d*d维的带宽矩阵，x为基准点，x_i为d维欧式空间R^d的n个样本点，n为d维空间中的样本点数量；in,

is the mean value of feature point offset under the current reference point, C _k,d represents the normalization constant, k(x) is the contour function, h is the kernel function bandwidth parameter, d is the matrix dimension, and H is a d*d dimension Bandwidth matrix, x is the reference point, x _i is the n sample points in the d-dimensional Euclidean space R ^d , and n is the number of sample points in the d-dimensional space;

Step 3-2, update the current image, take the mean value of the foreign object feature point offset as the new reference point, and repeat the cycle update until the feature point offset distance threshold is met;

Step 3-3. Use the feature point offset distance threshold as the matching basis, use the Mean-Shift operator to predict the foreign body area of the current frame, extract the ORB feature points of the foreign body area in the current frame and the reference frame, and construct the foreign body area through feature point matching. feature set;

Step 3-4, extract and screen new feature points in the feature set of the foreign body area in the current frame, update the feature set of the foreign body area, remove the feature points that do not match, and improve the accuracy of foreign body detection by continuously updating the feature points in the feature set of the foreign body area Rate;

Steps 3-5, convert the space coordinates of the foreign object target in the image into angular coordinates in real time, and track the foreign object target.

Step 4. Track the foreign body and cut it by laser;

Step 5: Determine whether there is still a foreign object target, if so, return to Step 1, otherwise end.

A rapid calibration system for foreign object targets in overhead transmission lines, comprising the following modules:

Image processing module: used to obtain the video image of the target area in real time, and process the image. When the target object is recognized, the frame will be set as the initial frame, and the video image of the target area will be enlarged;

Foreign object coordinate acquisition module: used to further accurately identify the enlarged video image, solve the image information, obtain the coordinate information of the foreign object target, and the coordinate information of the foreign object target in the winding position of the overhead transmission line;

Foreign body tracking module: used to track foreign body targets, identify the position status information of foreign body targets in the video image, including image feature points, centroid position and movement trajectory;

Foreign body processing module: Use laser to cut foreign body and judge whether there is still foreign body target.

A computer device, comprising a memory, a processor and a computer program stored in the memory and running on the processor, the processor implements the following steps when executing the computer program:

Step 1. Obtain the video image of the target area in real time, process the image, identify the current video image information frame by frame, when the target object is identified, set the frame picture as the initial frame, and enlarge the video image of the target area;

Step 2, further accurately identify the enlarged video image, solve the image information, obtain the coordinate information of the foreign object target, and the coordinate information of the foreign object target at the winding position of the overhead transmission line;

Step 3: Detecting the picture image information of the initial frame, performing automatic tracking processing of the foreign object target, and identifying the position state information of the foreign object target in the video image, including image feature points, centroid position and movement trajectory;

Step 4. Track the foreign body and cut it by laser;

Step 5: Determine whether there is still a foreign object target, if so, return to Step 1, otherwise end.

A computer-storable medium having a computer program stored thereon, the computer program implementing the following steps when executed by a processor:

Step 1. Obtain the video image of the target area in real time, process the image, identify the current video image information frame by frame, when the target object is identified, set the frame picture as the initial frame, and enlarge the video image of the target area;

Step 2, further accurately identify the enlarged video image, solve the image information, obtain the coordinate information of the foreign object target, and the coordinate information of the foreign object target at the winding position of the overhead transmission line;

Step 3: Detecting the picture image information of the initial frame, performing automatic tracking processing of the foreign object target, and identifying the position state information of the foreign object target in the video image, including image feature points, centroid position and movement trajectory;

Step 4. Track the foreign body and cut it by laser;

Step 5: Determine whether there is still a foreign object target, if so, return to Step 1, otherwise end.

Example

With reference to Figure 1, a method for quickly calibrating foreign object targets in overhead transmission lines includes the following steps:

Step 1. Obtain the video image of the target area in real time, process the image, and identify the current video image information frame by frame. When the target object is identified, the frame image is set as the initial frame, and the video image of the target area is enlarged, specifically: :

Step 1-1. Obtain a video image, perform grayscale conversion on each frame of the obtained image, establish a random mathematical model of a digital grayscale image, map each pixel on the original image, and obtain a grayscale-enhanced image, specifically: :

Step 1-1-1, construct a random mathematical model of digital grayscale image for the acquired image, count the grayscale values of all pixels, and obtain a grayscale distribution histogram, that is, obtain the grayscale level of the original image;

Step 1-1-2. In the original grayscale image X, according to the segmentation threshold _rc , the image is divided into two sub-images without intersection, and then the histogram equalization algorithm is applied to the two sub-images respectively, and the gray distribution is carried out. Normalized probability calculation:

Among them, p _Ld (d _j ) and P _Ud (d _j ) represent the gray-scale normalization probability of the low-gray sub-image and the high-gray sub-image, respectively; where d _k =0,1,2,..., L-1, L is the maximum value of the gray level, P _d (d _j ) is the gray distribution probability of the gray level d _j in the original image, and P _d (d _k ) is the gray level of the gray level d _k in the original image. degree distribution probability;

Step 1-1-3, construct the mapping relationship between the gray level of the original image and the gray level of the target image, construct the gray distribution probability density function of the continuous variable, and perform the low gray area and high gray area of the image according to the gray distribution probability density function. By dividing the gray area, the low gray area and the high gray area of the image are obtained respectively, and the gray image of the high gray area is obtained, that is, the image after gray enhancement.

Step 1-2, use the Gaussian blur method to denoise the grayscale-enhanced image through filtering, and obtain a denoised image;

Step 1-3: Perform Gaussian filtering on the denoised image to obtain a smooth image, specifically:

Using the Gaussian function value of the discrete point as the weight, the window template is collected in the denoised image, and the weighted average of each pixel value in the grayscale matrix in the collected window template is performed to eliminate the Gaussian noise:

Set the size of the window template to (2k+1)×(2k+1), then the weight coefficient of each element in the template is:

where (i,j)(i,j=1,2,...,2k+1) represents the relative coordinates of each pixel in the window template.

Assuming that the window template is 3×3, and the value of σ is 1.5, Gaussian filtering is performed on the image of the embodiment, and the calculated weight matrix is shown in Figure 3. The result obtained by multiplying each point by the corresponding weight value is The Gaussian filter output value of the center point, repeat the above process for all pixels in the image, and get the Gaussian filtered image

Steps 1-4, use the Canny edge detection algorithm to detect the foreign object on the obtained smooth image, when it is detected that the foreign object is in the center of the image, update the frame image to the initial frame, and zoom in on the video image of the area where the object is located.

Step 2, further accurately identify the enlarged video image, solve the image information, obtain the coordinate information of the foreign object target, and the coordinate information of the foreign object target winding position on the overhead transmission line, specifically:

Step 2-1. Use the Canny edge detection algorithm to detect the target edge of the enlarged target area, detect the horizontal, vertical and two diagonal edges of the target area in four directions, calculate its gradient strength and direction, and Noise removal is performed on it, and the Gaussian blur method is used to convert the image information into Gaussian cells to eliminate the influence of noise on recognition, specifically:

Step 2-1-1. Divide the image gradient direction into four directions according to the angle, namely 0°, 45°, 90°, and 135° edges, and compare the center value in its corresponding gradient direction in each 3×3 neighborhood. The size of the other two values, if the center value is the maximum value, it is retained, otherwise the center value is suppressed to remove false edge points;

Step 2-1-2. Set two gradient size thresholds: the high threshold is m _thh and the low threshold m _thl , and the image is divided into the following two edge maps:

E _H ={e(x,y)|if M(x,y) _{≥m thh} , then e(x,y)=1; else e(x,y)=0}

E _L ={e(x,y)|if M(x,y) _{≥m thl} , then e(x,y)=1; otherwise e(x,y)=0}

Among them, M(x, y) represents the edge gradient value at the (x, y) coordinate in the image before segmentation, and e(x, y) represents the edge gradient value at the current (x, y) coordinate after segmentation;

Step 2-1-3, mark the corresponding positions of all pixels in the high threshold map E _H as edge points, then traverse all the points in E _H , and use the 8-connection relationship to determine the final edge point in the low threshold map E _L , which enhances the edge continuity and completes the target edge detection.

Step 2-2: Convert the image information of the enlarged target area into an image recognition matrix, process the image recognition matrix by Hough transform detection, and obtain the position coordinates of the foreign object target and the winding point of the overhead transmission line in the image.

Step 3: Detect the picture image information of the initial frame, perform automatic tracking processing of the foreign object target, and identify the position state information of the foreign object target in the video image, including the image feature point, the position of the centroid and the movement trajectory, specifically:

Step 3-1. Use the foreign object area in the updated initial frame of the video image as the reference frame, use the Mean-Shift operator to predict the foreign object target area in the current frame, and use the Mean-Shift algorithm to solve the foreign object target feature points in the current frame. Offset mean of :

为当前基准点下的特征点偏移均值，C_k,d表示归一化常量，k(x)为轮廓函数，h为核函数带宽参数，d为矩阵维度，H是一个d*d维的带宽矩阵，x为基准点，x_i为d维欧式空间R^d的n个样本点，n为d维空间中的样本点数量；in,

is the mean value of feature point offset under the current reference point, C _k,d represents the normalization constant, k(x) is the contour function, h is the kernel function bandwidth parameter, d is the matrix dimension, and H is a d*d dimension Bandwidth matrix, x is the reference point, x _i is the n sample points in the d-dimensional Euclidean space R ^d , and n is the number of sample points in the d-dimensional space;

Step 3-2, update the current image, take the mean value of the foreign object feature point offset as the new reference point, and repeat the cycle update until the feature point offset distance threshold is met;

Step 3-3. Use the feature point offset distance threshold as the matching basis, use the Mean-Shift operator to predict the foreign body area of the current frame, extract the ORB feature points of the foreign body area in the current frame and the reference frame, and construct the foreign body area through feature point matching. feature set;

Step 3-4, extract and screen new feature points in the feature set of the foreign body area in the current frame, update the feature set of the foreign body area, remove the feature points that do not match, and improve the accuracy of foreign body detection by continuously updating the feature points in the feature set of the foreign body area Rate;

Step 3-5: Convert the space coordinates of the foreign object target in the image into angular coordinates in real time, and track the foreign object target.

Step 4. Track the foreign body and cut it by laser;

Step 5: Determine whether there is still a foreign object target, if so, return to Step 1, otherwise end.

Table 1 Comparison of the matching number of ORB feature points

The correct rate of matching is calculated by the test data in Table 1, which shows that the correct rate of image recognition processing for the acquired video sequence, the use of Mean-Shift operator and ORB feature point matching principle to achieve foreign object tracking, the matching rate exceeds the tracking matching of the prior art Rate.

Figure 4 shows a schematic diagram of the comparison between ORB feature point recognition and ordinary recognition. Compared with ordinary recognition, ORB feature point recognition greatly eliminates the recognition error caused by background noise. Ordinary recognition recognizes the target by detecting the position and edge of the target, which is easily affected by different backgrounds, which is easy to cause misrecognition. ORB feature point detection, first eliminate the influence of background noise, improve the gray value of the recognition target, then frame the edge of the object through edge detection, perform ORB feature point matching within the target range of high gray value, and continuously extract and filter the feature set of the target area of foreign objects The new feature points in the foreign body area are updated, and the foreign body detection accuracy is improved.

The position of the foreign object target point is determined in real time through ORB feature point matching. The error of a few matching points does not affect the correct rate of the total number of matching points. The foreign object recognition algorithm is optimized to achieve a target recognition effect close to 100%.

Claims (10)

1. A method for quickly calibrating a foreign object target of an overhead transmission line is characterized by comprising the following steps:

step 1, acquiring a video image of a target area in real time, processing the image, identifying current video image information frame by frame, setting a frame picture as an initial frame when a target object is identified, and amplifying the video image of the area where the target is located;

step 2, further accurately identifying the amplified video image, resolving image information, and acquiring coordinate information of the foreign object target and coordinate information of the foreign object target at the winding position of the overhead transmission line;

step 3, detecting picture image information of an initial frame, carrying out automatic tracking processing on the foreign object target, and identifying position state information of the foreign object target in the video image, wherein the position state information comprises image characteristic points, a mass center position and a moving track;

step 4, tracking the foreign bodies, and cutting by laser;

and 5, judging whether the foreign object still exists, if so, returning to the step 1, and if not, ending.

2. The method for rapidly calibrating the foreign object target of the overhead transmission line according to claim 1, wherein the step 1 of acquiring the video image and processing the image specifically comprises the following steps:

step 1-1, acquiring a video image, performing gray level conversion on each acquired frame of image, establishing a digital gray level image random mathematical model, and mapping each pixel on an original image to obtain an image with enhanced gray level;

step 1-2, carrying out noise reduction on the image with enhanced gray scale by filtering by using a Gaussian blur method to obtain a noise-reduced image;

step 1-3, performing Gaussian filtering on the noise-reduced image to obtain a smooth image;

and 1-4, detecting a foreign object target on the obtained smooth image by using a Canny edge detection algorithm, updating the frame image into an initial frame when the foreign object target is detected to be positioned at the center of an image picture, and amplifying a video image of a region where the target is positioned.

3. The method for rapidly calibrating the foreign object target of the overhead transmission line according to claim 2, wherein the step 1-1 is to perform gray level conversion on each frame of the acquired image to acquire an image with enhanced gray level, and specifically comprises the following steps:

step 1-1-1, constructing a digital gray image random mathematical model for the acquired image, and counting gray values of all pixels to obtain a gray distribution histogram, namely the gray level of the original image;

step 1-1-2, according to a segmentation threshold r in an original gray level image X_cThe image is divided into two non-intersection subgraphs, and then a histogram equalization algorithm is applied to the two subgraphs respectively to carry out gray distribution normalization probability calculation:

wherein p is_Ld(d_j) And P_Ud(d_j) Respectively representing the gray normalization probabilities of the low-gray subgraph and the high-gray subgraph; wherein d is_kL-1, L being the maximum value of the grey level, P_d(d_j) For grey level d in the original image_jProbability of gray distribution of (P)_d(d_k) For grey level d in the original image_kThe gray level distribution probability of (1);

1-1-3, constructing a mapping relation from an original image gray level to a target image gray level, constructing a gray distribution probability density function of continuous variables, dividing a low gray area and a high gray area of an image according to the gray distribution probability density function, respectively obtaining the low gray area and the high gray area of the image, and obtaining a gray image of the high gray area, namely the image with enhanced gray.

4. The method for rapidly calibrating the foreign object target of the overhead transmission line according to claim 2, wherein the obtaining of the smooth image in the step 1-3 specifically comprises:

taking a Gaussian function value of a discrete point as a weight, carrying out window template collection on the noise-reduced picture, and carrying out weighted average on each pixel value in a gray matrix in the collected window template so as to eliminate Gaussian noise:

setting the size of the window template to be (2k +1) × (2k +1), then the weight coefficient of each element in the template is:

where (i, j) (i, j ═ 1, 2.., 2k +1) represents the relative coordinates of each pixel point within the window template.

5. The method for rapidly calibrating the foreign object target of the overhead transmission line according to claim 1, wherein the image information calculated in the step 2 specifically comprises:

step 2-1, performing target edge detection on the amplified target area by using a Canny edge detection algorithm, detecting edges in four directions including the horizontal direction, the vertical direction and two diagonal lines of the target area, calculating the gradient strength and the direction of the edges, performing noise elimination on the edges, converting image information into Gaussian cells by using a Gaussian blur method, and eliminating the influence of noise on identification;

and 2-2, converting the image information of the amplified target area into an image identification matrix, and processing the image identification matrix by using Hough transformation detection to obtain the position coordinates of the foreign object target and the winding point of the overhead transmission line in the image.

6. The method for rapidly calibrating the foreign object target of the overhead transmission line according to claim 5, wherein the target edge detection in the step 2-1 specifically comprises:

step 2-1-1, dividing the gradient direction of the image into four directions according to angles, wherein the four directions are respectively edges of 0 degrees, 45 degrees, 90 degrees and 135 degrees, comparing the sizes of other two values of the central value in the corresponding gradient direction in each 3 x 3 neighborhood, if the central value is the maximum value, retaining the central value, otherwise, inhibiting the central value and removing false edge points;

step 2-1-2, setting two gradient size thresholds: high threshold is m_thhLow threshold value m_thlThe image is divided into two edge maps as follows:

E_Hif M (x, y) ≧ M is ≧ e (x, y) |_thhIf e (x, y) is 1; otherwise e (x, y) ═ 0}

E_LIf M (x, y) ≧ M is ≧ e (x, y) |_thlIf e (x, y) is 1; otherwise e (x, y) ═ 0}

Wherein M (x, y) represents the edge gradient value at the (x, y) coordinate in the image before segmentation, and e (x, y) represents the edge gradient value at the current (x, y) coordinate after segmentation;

step 2-1-3, map E of high threshold_HMarking all the corresponding positions of the pixels as edge points, and traversing E_HAt all points in (1), using 8-way relation at the low threshold map E_LAnd determining the final edge point to enhance the edge continuity and finish the target edge detection.

7. The method for rapidly calibrating the foreign object target of the overhead transmission line according to claim 1, wherein the identifying the position state information of the foreign object target in the video image in the step 3 specifically comprises:

step 3-1, taking a foreign matter region in an initial frame after the video image is updated as a reference frame, predicting a foreign matter target region in a current frame by using a Mean-Shift operator, and solving the Shift average value of foreign matter target feature points in the current frame by using a Mean-Shift algorithm:

wherein,

shifting the mean value for the feature points under the current reference point, C_k,dRepresenting a normalized constant, k (x) being a profile function, H being a kernel bandwidth parameter, d being a matrix dimension, H being a bandwidth matrix in d x d dimensions, x being a reference point,x_iis d dimension European space R^dN is the number of sample points in the d-dimensional space;

3-2, updating the current image, taking the foreign matter characteristic point offset mean value as a new reference point, and repeatedly and circularly updating until the characteristic point offset distance threshold is met;

3-3, predicting a foreign matter region of the current frame by using a Mean-Shift operator by using a feature point offset distance threshold as a matching basis, extracting ORB feature points of the foreign matter region in the current frame and a reference frame, and constructing a foreign matter region feature set through feature point matching;

3-4, extracting and screening new feature points from the feature set of the foreign body region of the current frame, updating the feature set of the foreign body region, eliminating unmatched feature points, and improving the accuracy of foreign body detection by continuously updating the feature points in the feature set of the foreign body region;

and 3-5, converting the spatial coordinates of the foreign object target in the image into angle coordinates in real time, and tracking the foreign object target.

8. The utility model provides a quick calibration system of overhead transmission line foreign matter target which characterized in that includes following module:

an image processing module: the system comprises a frame, a target area video image acquisition unit, a target object acquisition unit, a display unit and a display unit, wherein the frame is used for acquiring a target area video image in real time, processing the image, setting a frame picture as an initial frame when the target object is identified, and amplifying the area video image of the target;

foreign object coordinate acquisition module: the system is used for further accurately identifying the amplified video image, resolving image information, and acquiring coordinate information of the foreign object target and coordinate information of the foreign object target at the winding position of the overhead transmission line;

a foreign matter tracking module: the system is used for tracking the foreign object target and identifying position state information of the foreign object target in a video image, wherein the position state information comprises image feature points, a mass center position and a moving track;

a foreign matter treatment module: and cutting the foreign matter by using laser, and judging whether the foreign matter target still exists.

9. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1-7 are implemented by the processor when executing the computer program.

10. A computer-storable medium having a computer program stored thereon, wherein the computer program is adapted to carry out the steps of the method according to any one of claims 1-7 when executed by a processor.

Images