CN-121982090-A - Machine vision-based extraction type switch cabinet copper bar machining size detection method

CN121982090ACN 121982090 ACN121982090 ACN 121982090ACN-121982090-A

Abstract

The invention relates to the technical field of image processing, in particular to a machine vision-based extraction type switch cabinet copper bar processing size detection method which comprises the steps of collecting copper bar images containing processing areas and dividing local windows, constructing an optical ladder inversion index of the copper bar images, wherein the optical ladder inversion index is used for representing the deviation degree of the product of the normalized brightness and the normalized gradient of the local windows in the copper bar images, and constructing a compensation space unbalance index of the copper bar images, wherein the compensation space unbalance index is used for quantifying the structural risk of edge penetration in illumination estimation. In order to reduce the problems of serious uneven illumination, excessive smoothness of hole edges and positioning deviation caused by specular reflection of a metal surface after the processing of a copper bar of a switch cabinet, the invention constructs an optical ladder inversion index, a compensation space unbalance index and an iterative compensation intensity index, and accurately quantifies structural risks caused by space heterogeneous illumination and Gaussian penetration.

Inventors

WANG XINHUA
JIA MIAOMIAO
LI JUNLIANG
LI CHENGLIANG
Lei Ruojing
FU ZHIQIANG
WANG DALIN

Assignees

陕西中昊电气集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260327

Claims (10)

1. The method for detecting the processing size of the copper bar of the extraction type switch cabinet based on machine vision is characterized by comprising the steps of collecting copper bar images containing processing areas and dividing local windows; Constructing an optical ladder inversion index of the copper bar image, wherein the optical ladder inversion index is used for representing the deviation degree of the product of the normalized brightness and the normalized gradient of a local window in the copper bar image; constructing an iterative compensation intensity index of the copper bar image, wherein the iterative compensation intensity index is positively correlated with a compensation space unbalance index of the copper bar image and is positively correlated with an average space aggregation degree of a highlight region; and performing iterative enhancement processing on the copper bar image, and in each iteration, performing Retinex illumination compensation on the image by adopting an adaptive compensation weight based on the iterative compensation intensity index, and performing variation regularization by adopting a fractional order determined based on the optical ladder inversion index to obtain an enhanced copper bar image.
2. The machine vision-based extraction type switch cabinet copper bar processing size detection method according to claim 1, wherein the construction of the optical ladder inversion index specifically comprises the following steps: And comparing the product of the difference value of the gray average value of each local window and the minimum value and the corresponding gradient average value with the product of the maximum value and the global arithmetic average value to obtain the optical ladder inverse index.
3. The machine vision-based extraction type switch cabinet copper bar processing size detection method according to claim 1, wherein the calculation method of the compensating space unbalance index is as follows: multiplying the steppe inversion index by a spatial adjacency ratio equal to the number of highlight low gradient windows adjacent to the high gradient window divided by the total number of highlight low gradient windows to obtain a compensated spatial imbalance index.
4. The machine vision-based extraction type switch cabinet copper bar processing size detection method according to claim 1, wherein the average space aggregation degree of the highlight region is obtained by the following steps: for each window in the set of highlighting low gradient windows, statistics are made of it And (3) connecting the number of windows belonging to the high-brightness low-gradient window set in the adjacent region, and calculating the average value of the number of windows belonging to the high-brightness low-gradient window set in all windows in the high-brightness low-gradient window set to be used as the average space aggregation degree of the high-brightness region.
5. The machine vision-based extraction type switch cabinet copper bar processing size detection method according to claim 3 or 4, wherein the method for determining the highlight low gradient window set and the highlight high gradient window set comprises the following steps: the local window with the gray average value larger than the global gray average value and the gradient average value not larger than the global gradient average value is defined as a high-brightness low-gradient window, and the local window with the gradient average value larger than the global gradient average value is defined as a high-gradient window.
6. The machine vision-based extraction type switch cabinet copper bar processing size detection method according to claim 1, wherein the fractional order determined based on the optical ladder inversion index comprises: And linearly mapping the optical ladder inversion index to obtain the fractional order.
7. The machine vision-based extraction type switch cabinet copper bar processing size detection method as claimed in claim 6, wherein the calculation formula for performing linear mapping on the optical ladder inversion index is as follows: In the formula (I), in the formula (II), For the fractional order corresponding to the current image, Is the optical ladder inversion index of the current image.
8. The machine vision-based extraction type switch cabinet copper bar processing size detection method according to claim 1, wherein the calculation formula of the self-adaptive compensation weight is as follows: ; In the formula, Is the first The compensation weights of the round of iteration, For the total number of iterations, Is the first The iteration of the image compensates for the intensity index.
9. The method for detecting the machining dimension of the copper bar of the extraction type switch cabinet based on machine vision according to claim 1, further comprising the steps of carrying out edge detection and Hough transformation on the enhanced copper bar image to extract geometric features of copper bar contours and punching holes, and converting the geometric features from pixel dimensions to physical dimensions by means of camera parameters calibrated in advance.
10. The machine vision-based extraction type switch cabinet copper bar processing size detection method according to claim 1, further comprising, after acquiring the copper bar image: and preprocessing the copper bar image, wherein the preprocessing comprises distortion correction, filtering and extraction of a region of interest on the copper bar image.

Description

Machine vision-based extraction type switch cabinet copper bar machining size detection method Technical Field The invention relates to the technical field of image processing. More particularly, the invention relates to a machine vision-based method for detecting the machining size of a copper bar of a draw-out type switch cabinet. Background The extraction type switch cabinet is an important power distribution device commonly applied in a power system, and the precision of the machining size of the copper bar serving as a core conductive member directly relates to the assembly matching quality and the energizing operation safety of the whole switch cabinet. In recent years, with the continuous improvement of industrial automation level, a non-contact detection technology based on machine vision is increasingly applied to the field of automatic measurement of switch cabinet parts due to the advantages of high efficiency, no damage, objectivity and the like, and becomes an important means for judging the processing quality of copper bars. In an actual production and manufacturing scene, after the copper bar is subjected to mechanical processing (such as punching, cutting and other working procedures), the geometric dimensions of the copper bar, such as the outline, the aperture size, the hole spacing and the like, need to be strictly checked by a quality inspection link. The conventional machine vision detection system generally erects an industrial camera and a light source on a positioning platform, performs image acquisition on a copper bar to be detected on a production line or a detection platform, then extracts the outer contour of the copper bar and the edge of each punched hole by using an image processing algorithm, and finally calculates specific physical dimensions and compares the physical dimensions with drawing tolerances so as to automatically judge whether a product is qualified. The fractional order variational image enhancement model is an important method which is raised in the field of image processing in recent years, and is characterized in that the image is enhanced and denoised by minimizing an energy functional comprising a fractional order differential regularization term, a fractional order differential operator gives consideration to low-frequency smoothing and high-frequency edge preservation, compared with an integer order variational method, the fractional order variational image enhancement model is more accurate in texture detail and edge expression, an iterative optimization framework has good convergence and interpretation, and regularization parameters and fractional orders provide flexible regulation and control space. However, the surface of the copper bar has obvious metal specular reflection characteristics after mechanical processing, under the irradiation of structured light or linear array light sources, the local illumination intensities of different areas in an image are highly unevenly distributed, the high-light areas are locally overexposed due to pixel saturation, and the shadow areas have serious insufficient contrast. In the original fractional order variational model, the fractional orderAnd regularization parametersThe global unified value is taken on the whole image, the spatial heterogeneity illumination distribution caused by the specular reflection on the copper bar surface cannot be adaptively responded, the edge of the overexposed area is excessively smoothed, the edge of the shadow area is enhanced insufficiently, the edge positioning deviation of key dimensions such as the aperture, the pitch and the like of the copper bar is finally caused, and systematic measurement errors are introduced. Disclosure of Invention The invention provides a machine vision-based extraction type switch cabinet copper bar processing size detection method, which aims to solve the problem that in an original fractional order variation model in the related art, the fractional order isAnd regularization parametersThe method is characterized in that global unified value is obtained on the whole image, spatial heterogeneity illumination distribution caused by specular reflection on the surface of the copper bar cannot be adaptively responded, so that the edge of an overexposed area is excessively smoothed, the edge of a shadow area is enhanced insufficiently, and finally, the problem of systematic measurement errors is introduced due to edge positioning deviation of key dimensions such as the aperture, the pitch and the like of the copper bar. The invention provides a machine vision-based extraction type switch cabinet copper bar machining size detection method which comprises the steps of collecting copper bar images containing machining areas and carrying out local window division, constructing an optical ladder inversion index of the copper bar images, wherein the optical ladder inversion index is used for representing the deviation degree of the product of normalized bri