CN116188838A - Artificial intelligence-based external damage hidden danger point interference judging method - Google Patents

Abstract

In order to solve the problems of inaccurate identification of external broken hidden danger points and easy generation of false leakage detection in the prior art, the invention provides an artificial intelligence-based external broken hidden danger point interference judging method, and the technical scheme for solving the technical problems comprises the following steps of; 1. inputting an image, namely acquiring an unmanned aerial vehicle and a monitoring photo; 2. feature extraction, namely calculating the acquired picture by using a pre-trained feature matrix to obtain a feature result; 3. feature coding, namely coding a feature result in the second step; 4. after the result classification is finished, comparing the result classification with samples in a knowledge base to finish classification; 5. and outputting a result, and decoding and outputting corresponding early warning information according to the classification type. According to the invention, the images or the monitoring images shot in the unmanned aerial vehicle inspection process are analyzed and processed by an artificial intelligence method, so that the accurate judgment of hidden danger points and the grading judgment are realized, and a more reliable basis is provided for the subsequent hidden danger point inspection and repair.

Description

Artificial intelligence-based interference judgment method for hidden danger points

technical field

The invention belongs to the technical field of identifying hidden danger points of external breakage of power transmission lines, in particular to an artificial intelligence-based interference judgment method for hidden danger points of external breakage.

Background technique

In the context of the current era, the safety of transmission lines is a necessary condition for ensuring industrial and domestic electricity consumption. During the actual operation of transmission lines, they are easily damaged by external forces and accidents occur. Among them, the hidden dangers of external damage are mainly reflected in the following aspects Two aspects: 1. Theft, 2. Prevention problems, 3. Caused by illegal buildings.

With the development of society, drones are often used for automatic inspection of power transmission lines. This method of operation greatly improves the speed of inspection and improves the efficiency of inspection, but there are hidden dangers to external damage in inspections of drones Point identification is not accurate, and it is easy to cause missed inspections. Therefore, how to improve the accuracy of identifying hidden danger points during inspections has always been a problem that needs attention.

Contents of the invention

In order to solve the problem of inaccurate identification of hidden danger points in the prior art and easy to cause false detection and missed detection, the present invention provides a method for judging the interference of hidden danger points based on artificial intelligence. The images captured in the camera or monitoring images are analyzed and processed to achieve accurate judgment of hidden danger points and grading judgments, providing a more reliable basis for subsequent inspection and maintenance of hidden danger points.

The invention provides a method for judging interference of hidden danger points based on artificial intelligence, and its technical solution for solving technical problems includes the following steps;

Step 1. Image input, collecting drones and surveillance photos;

Step 2, feature extraction, use the pre-trained feature matrix to calculate the collected pictures, and obtain the feature result;

Step 3, feature encoding, encoding the feature results in step 2;

Step 4. Classify the results. After the encoding is completed, compare with the samples in the knowledge base to complete the classification;

Step 5, output the result, decode and output the corresponding early warning information according to the classification category.

Preferably, the features in the second step include linear features and nonlinear features.

Preferably, the linear feature is the result of a linear transformation between the image bit and its two-dimensional Euclidean space neighborhood.

Preferably, a linear filter is used to perform feature extraction on the linear feature, and the specific algorithm is:

Among them, dst(x, y) is the target image, which is the result of linear transformation;

Kernel is the filter kernel involved in the calculation,

Cols indicates the row length (number of columns) of the filter kernel;

rows indicates the column length (number of rows) of the filter kernel;

src is the original image;

x, y are bits;

x', y' are the bits currently calculated;

The anchor is the anchor point involved in the calculation, and the anchor determines the stepping distance of the kernel.

Preferably, according to different filter kernels, the linear features can be extracted into edge features and shape features.

Preferably, the nonlinear features are extracted by GLCM, and the nonlinear features include contrast, correlation, entropy, and homogeneity.

Preferably, the feature encoding is performed based on the GBDT classification algorithm in the step 3: for the input feature matrix and prior results, a classifier is set to satisfy the following relationship:

Among them: N is the number of nodes of the classifier, i is the state of the classifier;

Let the total number of iterations be M, and the current iteration cycle be m, then:

求解：/>

得到结果，并更新：

Solving: />

Get the result, and update:

得到最终解，

get the final solution,

Finally, f _i (x) is the feature encoder, and its output is the feature value.

Preferably, the comparison in step 4 is by comparing the feature value with the target value, wherein in the comparison process, different information is compared using different vector values, and the value with the smallest variance can be selected.

In summary, using the technical solution of the present invention has at least the following beneficial effects:

1. It can use the feature matrix to extract the features of the pictures captured during the UAV inspection process, and can convert the captured pictures into feature information, which is convenient for subsequent analysis and processing;

2. By extracting the linear features and nonlinear features of the image separately, the subsequent comparison is more accurate;

3. By assigning different alarm information to the target value, different levels of settings can be realized according to different characteristic values.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can also be obtained based on these drawings without creative effort.

Fig. 1 is the flow chart of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

The present invention provides an artificial intelligence-based method for judging the interference of hidden danger points, see accompanying drawing 1, comprising the following steps;

Step 1. Image input, collecting drones and surveillance photos; during the inspection process, take pictures by manipulating the drone or the automatic cruise of the drone, or use surveillance photos in some areas to collect the captured pictures /image for input;

Step 2, feature extraction, use the pre-trained feature matrix to calculate the collected pictures, and obtain the feature result;

Step 3, feature encoding, encoding the feature results in step 2;

Step 4. Classify the results. After the encoding is completed, compare with the samples in the knowledge base to complete the classification;

Step 5, output the result, decode and output the corresponding early warning information according to the classification category.

It should be noted that the features in step 2 of the present invention include linear features and nonlinear features, and the linear features and nonlinear features are extracted respectively by corresponding algorithms. The linear feature is the result of a linear transformation between the image bit and its two-dimensional Euclidean space neighborhood. In the specific calculation, the linear filter is used to extract the linear features, and the specific algorithm is:

Among them, dst(x, y) is the target image, which is the result of linear transformation;

Kernel is the filter kernel involved in the calculation, that is, a two-dimensional vector, and different filter kernels extract different features

information;

Cols indicates the row length (number of columns) of the filter kernel;

rows indicates the column length (number of rows) of the filter kernel;

src is the original image;

x, y are bits;

x', y' are the bits currently calculated;

The anchor is the anchor point involved in the calculation. The anchor determines the stepping distance of the kernel and determines the density of feature extraction.

It should be noted that in the process of linear feature extraction in the present invention, edge features and shape features can be extracted from linear features according to different filter kernels.

The nonlinear features in the present invention are extracted by GLCM, wherein the nonlinear features mentioned in the present invention mainly include contrast, correlation, entropy, and homogeneity. When performing specific calculation and extraction, proceed as follows:

For a grayscale image of a color image, its grayscale co-occurrence matrix (GLCM, Gray LevelCo-occurrence Matrix) is expressed as follows:

Where (Δx, Δy) is a set of offsets, and the offset is a vector with length and direction;

C is a grayscale co-occurrence matrix, I is an input grayscale image, and x, y are bits;

After the GLCM is extracted, the GLCM is calculated, and the following commonly used eigenvalues can be obtained (p(i, j) is

bits in GLCM):

Contrast:

Correlation (μ is correlation coefficient):

entropy:

homogeneity:

The above four eigenvalues are the nonlinear features to be extracted in the present invention and their calculation precautions.

It should be noted that the feature encoding in Step 3 of the present invention is based on the GBDT classification algorithm, as follows: For the input feature matrix and prior results, a classifier is set to satisfy the following relationship:

Among them: N is the number of nodes of the classifier, i is the state of the classifier;

Let the total number of iterations be M, and the current iteration cycle be m, then:

求解：/>

Solving: />

Get the result, and update:

得到最终解，

get the final solution,

Finally, f _i (x) is the feature encoder, and its output is the feature value.

The comparison in step 4 of the present invention is to compare the obtained eigenvalue with the target value, because different early warning information is represented by different eigenvalues, the output result is compared with the prior eigenvalue, and the selection variance is the smallest Different information is compared using different vector values, so that the identification of the specific area (prior interest area) in the infrared image can be realized, and its category can be identified.

In the present invention, the orthographic image is first determined to be the only data source for identifying external damage, and on the basis of intelligently identifying the hidden dangers of external damage construction, the image is processed by obtaining the altitude information of the picture and the data of the pole and tower account, combined with the focal length of the UAV pan/tilt dividing the area, assigning pixel values to the target position according to the difference between the similarity matrix of the target position and the similarity matrix of the neighborhood position, and obtaining a grayscale image; performing region segmentation on the grayscale image, and obtaining a region segmentation image , which can be divided into the hidden defect area within a certain range of the line channel, the safety prevention area a little further away, and the concern area a little further away, and then the risk points are graded and judged.

The above description is a preferred embodiment of the present invention, and it should be pointed out that for those skilled in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, and these improvements and modifications are also considered Be the protection scope of the present invention.

Claims (8)

1. The method for judging the interference of the hidden danger points based on the artificial intelligence is characterized by comprising the following steps of;

step one, inputting an image, and collecting an unmanned aerial vehicle and a monitoring photo;

step two, extracting features, namely calculating the acquired pictures by using a feature matrix trained in advance to obtain feature results;

step three, feature coding, namely coding the feature result in the step two;

step four, classifying results, namely comparing the results with samples in a knowledge base after encoding is completed, and completing classification;

and fifthly, outputting a result, and decoding and outputting corresponding early warning information according to the classification type.

2. The method for determining the disturbance of the hidden danger point based on artificial intelligence according to claim 1, wherein the features in the second step include linear features and nonlinear features.

3. The artificial intelligence based method of claim 2, wherein the linear features are the result of a linear transformation of the image bits and the two-dimensional euclidean space neighborhood in which they reside.

4. The method for judging the external damage hidden trouble point interference based on artificial intelligence according to claim 3, wherein the linear characteristics are extracted by using a linear filter, and the specific algorithm is as follows:

wherein dst (x, y) is the target image, i.e. the linear transformation result;

kernel is a filter kernel that participates in the computation,

cols represents the line length (column number) of the filter kernel;

rows represents the column length (number of rows) of the filter kernel;

src is the original image;

x, y are bits;

x ', y' are currently calculated bits;

anchor is the anchor point of participating in calculation, and anchor has decided the step distance of kernel.

5. The method for judging the disturbance of the hidden danger point of external damage based on artificial intelligence according to claim 3, wherein the linear features can be extracted into edge features and shape features according to different filter kernels.

6. The method for judging the external damage hidden trouble point interference based on artificial intelligence according to claim 1, wherein the nonlinear features are extracted by GLCM, and the nonlinear features comprise contrast, correlation, entropy and homogeneity.

7. The method for determining the disturbance of the hidden danger point of external damage based on artificial intelligence according to claim 1, wherein in the third step, feature coding is performed based on a GBDT classification algorithm: for the input feature matrix and the prior result, a classifier is set to satisfy the following relation:

wherein: n is the number of nodes of the classifier, i is the state of the classifier;

let the total iteration number be M, the current iteration period be M, then there are:

solving: />

Obtaining a result and updating:

a final solution is obtained and the solution is obtained,

finally f _i (x) The output of the feature encoder is a feature value.

8. The method for determining the interference of the hidden danger point based on artificial intelligence according to claim 1, wherein the comparison in the fourth step is to compare the characteristic value with the target value, wherein in the comparison process, different information is respectively compared by using different vector values, and the value with the smallest variance is selected.

Images