CN-121540717-B - Surface oxidation detection method and system for overhead stranded wire

CN121540717BCN 121540717 BCN121540717 BCN 121540717BCN-121540717-B

Abstract

The invention belongs to the technical field of image processing, and particularly relates to a surface oxidation detection method and system for overhead stranded wires, wherein the method comprises the steps of acquiring gradient field data of an overhead stranded wire image; the method comprises the steps of determining local main direction entropy of pixel points based on the difference of included angles between gradient directions and local main directions of all neighborhood pixel points of the pixel points, determining entropy thresholds through clustering to screen anchor points, constructing gradient section lines along the local main directions of the anchor points, determining local rust width of the anchor points according to gradient amplitude changes of the pixel points on the gradient section lines, acquiring self-adaptive structural element parameters based on the local rust width and the local main directions of the anchor points, executing morphological top hat transformation to calculate response values of the anchor points, and marking the anchor points with response values larger than the response thresholds as oxidation defects. The method and the device remarkably improve the accuracy of identifying the oxidized rust on the surface of the overhead stranded wire by self-adaptively matching the local geometric form of the rust.

Inventors

GUO XIONGTAO
YAO WANG
WANG ZHANSHENG
LIU BINGJI

Assignees

淳化昆仑优佳电缆有限公司

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (8)

1. A surface oxidation detection method for an overhead stranded wire, comprising: Acquiring gradient field data of an overhead stranded wire image, wherein the gradient field data comprises gradient amplitude values and gradient directions of each pixel point; Acquiring a local main direction of any pixel point, wherein the local main direction is equal to a gradient direction corresponding to a peak value of gradient directions of all neighborhood pixel points of the pixel point; determining the local main direction entropy of the pixel according to the difference between the gradient directions of all the neighborhood pixel points of the pixel and the local main direction, including: Calculating the difference between the gradient direction of each neighborhood pixel point of any pixel point and the local main direction of the pixel point, and counting the probability of each type of difference , In the formula (I), in the formula (II), Is the first Local main directional entropy of each pixel point; is the first The first pixel point in all the neighborhood pixels of each pixel point Probability of class angle difference; is the first Index values and total numbers of categories with different included angles in all neighborhood pixel points of the pixel points; As a logarithmic function; counting the distribution condition of local main direction entropy of all pixel points, and determining an entropy threshold value, wherein the distribution condition comprises the following steps: The method comprises the steps of clustering local main direction entropy of all pixel points by using a DBSCAN clustering algorithm to obtain a clustering result, calculating the average value of the local main direction entropy of the pixel points in each clustering cluster, and taking the minimum local main direction entropy in the clustering cluster corresponding to the maximum value of the average value as an entropy threshold value; extracting pixel points with local main direction entropy smaller than an entropy threshold value as anchor points; Constructing a gradient section line along the local main direction of each anchor point, and determining the local rust width of each anchor point according to the gradient amplitude change of the pixel points on the gradient section line, wherein the local rust width is the distance between the left edge point and the right edge point of the anchor point, and the left edge point and the right edge point are the pixel points corresponding to the initial local maximum value of the gradient amplitude when searching along the anchor point to the two sides of the gradient section line; And obtaining parameters of the self-adaptive structural elements for each anchor point, wherein the parameters comprise the vertical direction of the local main direction, the local rust width and the experience length, executing morphological top hat transformation by using the parameters, calculating the response value of the anchor point, and marking the anchor point with the response value larger than the response threshold value as an oxidation defect.
2. The surface oxidation detection method for overhead strands according to claim 1, wherein the method for obtaining the neighborhood pixel points comprises the following steps: Defining a center of each pixel A neighborhood window, wherein all pixel points in the neighborhood window are used as neighborhood pixel points of the pixel points, and the neighborhood window comprises a plurality of pixel points, wherein the pixel points are the neighborhood pixel points of the pixel points Is a preset positive odd number.
3. A surface oxidation detection method for an overhead strand according to claim 1, wherein said constructing gradient section lines along the local main direction of each anchor point comprises: A line segment is defined along its local main direction centered at each anchor point as the gradient cross-section line of that anchor point.
4. The surface oxidation detection method for an overhead stranded wire according to claim 1, wherein the method for acquiring the pixel point corresponding to the primary local maximum value of the gradient amplitude comprises: And when searching along the anchor points to the two sides of the gradient section line respectively, extracting the pixel points with the gradient amplitude value larger than that of the adjacent two pixel points for the first time.
5. A surface oxidation detection method for an overhead strand according to claim 1, wherein the empirical length is equal to the local rust width of the anchor point Multiple times, and Taking 3.
6. The method for detecting surface oxidation of an overhead stranded wire according to claim 1, wherein the calculating the response value of the anchor point comprises: And performing open operation on the overhead stranded wire image by using the self-adaptive structural element parameters of each anchor point to obtain an operation result image of the anchor point, and extracting the gray value of the anchor point in the operation result image to serve as the operation gray value of the anchor point, wherein the response value of the anchor point is equal to the normalization result of the absolute difference value of the gray value and the operation gray value of the anchor point.
7. The surface oxidation detection method for overhead strands according to claim 1, wherein the gradient field data acquisition method comprises the step of calculating gradient amplitude and gradient direction of each pixel point in an overhead strand image by using a Sobel operator.
8. A surface oxidation detection system for an overhead strand, comprising a processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a surface oxidation detection method for an overhead strand according to any one of claims 1-7.

Description

Surface oxidation detection method and system for overhead stranded wire Technical Field The invention relates to the technical field of image processing. More particularly, the invention relates to a surface oxidation detection method and system for overhead strands. Background The overhead stranded wire is used as a key bearing and conductive part of a power transmission system, the surface quality state before leaving the factory is directly related to the long-term stability and safety of the power grid operation, in the links of storage and quality inspection, the surface of the stranded wire is easy to generate oxidized rust with different forms due to environmental factors, if the oxidized rust cannot be found and processed in time, the product quality is seriously affected, and even potential safety hazards are buried, so that the development of an automatic image detection method for replacing low-efficiency manual visual inspection is urgent for the industry. In the field of automatic detection, aiming at tiny and linear oxidized rust on the surface of an overhead stranded wire, a core technical problem is complex variability of rust forms, and under different storage conditions and oxidation degrees, the width, length, curvature and local trend of the rust show obvious non-uniformity, so that the adaptability and accuracy of a detection algorithm form a serious challenge. In the prior art, a morphological top hat transformation algorithm is often adopted, a twisted line image is subjected to morphological operation through preset fixed structural elements (Structuring Element, SE), bright features with sizes smaller than those of the structural elements in the image are extracted to serve as defect areas, due to the variability of the actual rust form, if the actual width or local trend of the rust deviates greatly from the preset structural element parameters, matching failure is caused, defect detection is missed, otherwise, when the texture of a twisted line which happens to exist in the background is similar to the fixed structural elements in morphology, the false detection is easily recognized as a target, and the reliability and accuracy of a detection result are seriously affected. Disclosure of Invention In order to solve the technical problem that the structural elements cannot be adaptively matched with rust with changeable forms, so that the detection accuracy is low, the invention provides the following aspects. In a first aspect, the invention provides a surface oxidation detection method for overhead strands, which comprises the steps of obtaining gradient field data of an overhead strand image, wherein the gradient field data comprise gradient amplitude values and gradient directions of each pixel point, obtaining local main directions of any pixel point, wherein the local main directions are equal to gradient directions corresponding to peak values of gradient directions of all neighborhood pixel points of the pixel point, determining local main directions entropy of the pixel point according to included angle differences between the gradient directions of all neighborhood pixel points of the pixel point and the local main directions, counting distribution conditions of the local main directions entropy of all pixel points, determining an entropy threshold value, extracting pixel points with the local main directions smaller than the entropy threshold value as anchor points, constructing a gradient section line along the local main directions of each anchor point, determining local rust width of each anchor point according to gradient amplitude change of the pixel point on the gradient section line, wherein the local rust width is the distance between anchor points at left edge point and right edge point of the anchor point, the left edge point and the right edge point of the anchor point is the anchor point, determining the local main direction entropy of the pixel point along the gradient section line, calculating the local main direction entropy corresponding to the maximum value of the gradient, and performing self-adaption response to the local main direction parameter value of the anchor point, and the self-adaption value is obtained by using the gradient main direction as a response parameter, and the response value of the local main direction to the anchor point is high-adaption value. According to the method, gradient field data of an overhead stranded wire image are acquired, the order of gradient directions is measured by utilizing local main direction entropy, anchors with linear characteristics are effectively screened out, interference of a disordered texture background is primarily eliminated, global fixed structural elements used in traditional morphological detection are abandoned, gradient section lines are constructed along the local main directions of the anchors, the local rust width of each anchor is accurately determined by analyzing gradient amplitude chan