CN-121981894-A - Coal mine robot inspection image acquisition and analysis system based on machine vision

CN121981894ACN 121981894 ACN121981894 ACN 121981894ACN-121981894-A

Abstract

The invention relates to the technical field of image processing, in particular to a coal mine robot inspection image acquisition and analysis system based on machine vision, which comprises the steps of obtaining different image blocks of an inspection image according to brightness difference characteristics of pixel points in the inspection image, obtaining weights of neighborhood pixel points according to position distribution characteristics of target pixel points and neighborhood pixel points in the image blocks, obtaining comprehensive feature descriptors of the target pixel points according to gradient difference characteristics, weights and discrete characteristics of the neighborhood pixel points, judging an image enhancement method according to the comprehensive feature descriptors and comprehensive feature descriptors of preset sample images, and obtaining gradient feature evaluation values according to gradient characteristics and the comprehensive feature descriptors of the suspected edge pixel points. According to the invention, image enhancement and iteration are carried out according to the brightness characteristic and gradient characteristic evaluation value of the image block, so that an enhanced inspection image is obtained, and the image enhancement effect and the image analysis accuracy are improved.

Inventors

DONG KANGJUN
Yuan Jiacong
FENG YANG
Yang Heirui

Assignees

渭南师范学院

Dates

Publication Date: 20260505
Application Date: 20260401

Claims (10)

1. The system for collecting and analyzing the inspection images of the coal mine robot based on the machine vision is characterized by comprising the following modules: the image acquisition module is used for acquiring a patrol image of the coal mine robot; The image analysis module is used for partitioning the inspection image according to the brightness difference characteristics of the pixel points in the inspection image to obtain different image blocks, obtaining the weight of the neighborhood pixel points according to the position distribution characteristics of the target pixel points and the neighborhood pixel points in the image blocks, obtaining a first characteristic descriptor according to the gradient difference characteristics of the target pixel points and the neighborhood pixel points and the weight, obtaining a second characteristic descriptor according to the discrete characteristics of the neighborhood pixel points, and obtaining a comprehensive characteristic descriptor of the target pixel points according to the first characteristic descriptor and the second characteristic descriptor; The image enhancement module is used for preliminarily judging an image enhancement method according to a comprehensive feature descriptor of a target pixel point in an image block and a comprehensive feature descriptor of a preset sample image, acquiring a suspected edge pixel point, acquiring a gradient feature evaluation value according to a gradient feature of the suspected edge pixel point and the comprehensive feature descriptor; The enhancement iteration module is used for judging the enhancement effect according to the gradient characteristics and the comprehensive characteristic descriptors after the image blocks are enhanced, carrying out image enhancement iteration to obtain final enhanced image blocks, and obtaining enhanced inspection images according to all the final enhanced image blocks.
2. The machine vision-based coal mine robot inspection image acquisition and analysis system according to claim 1, wherein the step of partitioning the inspection image according to the brightness difference characteristics of the pixels in the inspection image to obtain different image blocks comprises: Wherein K represents an adjustment coefficient, An exponential function based on a natural constant is represented, Y represents a variation coefficient of brightness values of all pixels of the inspection image in the Lab color model, Represents the average value of the brightness of all the pixel points, Represents the maximum value of the brightness of the pixel point, Representing the brightness minimum value of the pixel point; Wherein C represents an adaptive blocking parameter, The representation is rounded down and up, Representing a preset minimum partitioning parameter, Represents a preset maximum block parameter, K represents an adjustment coefficient, Representing an adjustment reference, dividing the inspection image into an average of Image blocks.
3. The machine vision-based coal mine robot inspection image acquisition and analysis system as claimed in claim 1, wherein the step of obtaining the weight of the neighborhood pixel according to the position distribution characteristics of the target pixel and the neighborhood pixel in the image block comprises: Taking a pixel point with gradient amplitude not being constant 0 in the image block as a target pixel point, wherein Representing the weight of the nth neighborhood pixel of the target pixel, A linear normalization is represented and, Representing the Euclidean distance between the target pixel point and the nth neighborhood pixel point, connecting the nth neighborhood pixel point with the target pixel point, And the sine value of the included angle between the connecting line corresponding to the nth neighborhood pixel point and the gradient direction of the target pixel point is represented, and N represents the number of the neighborhood pixel points.
4. The machine vision-based coal mine robot inspection image acquisition and analysis system as set forth in claim 1, wherein the step of obtaining the first feature descriptors according to the gradient difference features of the target pixel and the neighborhood pixel and the weights includes: In the following A first feature descriptor representing a target pixel, N representing the number of neighboring pixels, Represents an exponential function with a natural constant as a base, F represents the gradient magnitude of the target pixel point, Representing the gradient magnitude of the n-th neighborhood pixel, And the weight of the nth neighborhood pixel point is represented.
5. The machine vision-based coal mine robot inspection image acquisition and analysis system according to claim 1, wherein the step of obtaining the second feature descriptors from the discrete features of the neighborhood pixel points comprises: In the following A second feature descriptor representing the target pixel point, selecting a preset number of neighborhood pixel points as neighborhood key points according to the sequence from big to small of the first feature descriptors of the neighborhood pixel points, calculating Euclidean distances between the neighborhood key points and other nearest neighborhood key points to obtain nearest neighbor distances, Representing the average of the nearest neighbor distances of all neighboring keypoints of the target pixel, Representing the maximum value of the nearest neighbor distance, Representing the minimum of the nearest neighbor distance.
6. The machine vision-based coal mine robot inspection image acquisition and analysis system according to claim 1, wherein the step of obtaining the comprehensive feature descriptors of the target pixel points according to the first feature descriptors and the second feature descriptors comprises: And calculating the product of the first feature descriptor and the second feature descriptor to obtain the comprehensive feature descriptor of the target pixel point.
7. The machine vision-based coal mine robot inspection image acquisition and analysis system as set forth in claim 1, wherein the steps of primarily judging the image enhancement method and acquiring the suspected edge pixel points according to the comprehensive feature descriptors of the target pixel points in the image block and the comprehensive feature descriptors of the preset sample image include: when the gradient magnitudes of the pixel points in the image block are all constant 0, the image block does not carry out image enhancement; When gradient amplitude non-uniformity of pixel points in the image block is constant 0, calculating an average value of comprehensive feature descriptors of edge pixel points in a preset sample image to obtain a judging threshold value, if the comprehensive feature descriptors of target pixel points in the image block are smaller than the judging threshold value, carrying out Gaussian filtering processing on the image block, otherwise, carrying out self-adaptive image enhancement on the image block, and taking the target pixel points with the comprehensive feature descriptors not smaller than the judging threshold value as the suspected edge pixel points.
8. The machine vision-based coal mine robot inspection image acquisition and analysis system according to claim 1, wherein the step of obtaining the gradient feature evaluation value according to the gradient feature and the comprehensive feature descriptor of the suspected edge pixel point comprises the following steps: Wherein R represents the gradient characteristic evaluation value of the image block, M represents the number of suspected edge pixels, A comprehensive feature descriptor representing the mth suspected edge pixel point, Representing the gradient magnitude of the mth suspected edge pixel.
9. The machine vision-based coal mine robot inspection image acquisition and analysis system according to claim 1, wherein the step of performing image enhancement according to the brightness characteristics and the gradient characteristic evaluation values of the image block, the gradient characteristics and the brightness characteristics of the preset sample image comprises: The method comprises the steps of calculating an average value of gradient amplitude values of edge pixel points of a preset sample image to obtain a gradient threshold value, when the gradient characteristic evaluation value is not lower than the gradient threshold value, not carrying out image enhancement on an image block, calculating an average value of brightness values of all pixel points of the preset sample image to obtain a brightness threshold value, when the gradient characteristic value is lower than the gradient threshold value, if the average value of the brightness values of the pixel points of the image block is lower than or equal to the brightness threshold value, increasing the brightness of the image block, otherwise, decreasing the brightness of the image block, increasing the contrast parameter by default, selecting brightness parameters and contrast parameters in a linear brightness contrast enhancement model according to a brightness adjustment direction through a particle swarm optimization algorithm, and carrying out image enhancement on the image block according to brightness parameter and contrast parameter results.
10. The machine vision-based coal mine robot inspection image acquisition and analysis system as claimed in claim 9, wherein the step of judging the enhancement effect according to the gradient characteristics and the comprehensive characteristic descriptors after the enhancement of the image block and performing image enhancement iteration to obtain the final enhanced image block comprises the following steps: where W represents an enhanced quality assessment value, Representing the gradient characteristic evaluation value after image block enhancement, G representing the number of pixels whose gradient amplitude is not constant zero and does not belong to the pixel points of the suspected edge after image block enhancement, Representing the gradient magnitude of the g-th pixel, A comprehensive feature descriptor representing a g-th pixel point; And stopping image enhancement and obtaining a final enhanced image block by the image block if the enhanced gradient characteristic evaluation value is not lower than the gradient threshold value after the image block completes one-time image enhancement, and stopping enhancement by the image block and taking the enhanced result of the previous round as the final enhanced image block if the enhanced quality evaluation value is smaller than the enhanced quality evaluation value after the previous round of enhancement.

Description

Coal mine robot inspection image acquisition and analysis system based on machine vision Technical Field The invention relates to the technical field of image processing, in particular to a coal mine robot inspection image acquisition and analysis system based on machine vision. Background The coal mine production environment has the characteristics of complex space structure, high safety risk and the like, is used for reducing the manual inspection intensity and improving the inspection efficiency and safety, is gradually applied to underground operation, and automatically inspects and acquires images on roadway environment, equipment running state and potential safety hazards by carrying visual sensors. The existing inspection system mainly adopts a visible light camera to acquire images and combines a deep learning algorithm to realize target identification and state analysis, but the underground light source distribution is uneven, the illumination condition is limited and other factors influence, the acquired images are easy to have the problems of insufficient brightness, partial overexposure or obvious shadow and the like, so that the condition of missing inspection or false inspection is easy to generate in the process of carrying out abnormal identification and detection on the acquired images, and therefore, the images acquired in an inspection scene need to be processed, and the image quality is improved. In the process of processing the inspection image, the prior method mostly adopts a fixed image processing flow to uniformly execute operations such as enhancement, denoising and the like on the acquired image, but in an actual underground inspection scene, the image quality difference of different time and different positions is possibly larger, not all inspection images need to be enhanced, and the characteristics such as brightness, definition distribution and the like of the same image in different areas are also uneven. Meanwhile, the inspection robot is affected by illumination change, attitude change, environmental interference and the like in the moving process, image characteristics show dynamic change, if a uniform and fixed processing flow and processing intensity are adopted, the problem that extra noise or overexposure is introduced due to excessive enhancement of images with good quality is easy to occur, and the problem that images with poor quality are possibly insufficient in enhancement, so that characteristic identification is inaccurate is solved. Therefore, the fixing processing mode is difficult to improve the enhancement effect of different inspection images, and further the accuracy of image analysis and identification is affected. Disclosure of Invention In order to solve the technical problems, the invention aims to provide a coal mine robot inspection image acquisition and analysis system based on machine vision, which adopts the following technical scheme: the image acquisition module is used for acquiring a patrol image of the coal mine robot; The image analysis module is used for partitioning the inspection image according to the brightness difference characteristics of the pixel points in the inspection image to obtain different image blocks, obtaining the weight of the neighborhood pixel points according to the position distribution characteristics of the target pixel points and the neighborhood pixel points in the image blocks, obtaining a first characteristic descriptor according to the gradient difference characteristics of the target pixel points and the neighborhood pixel points and the weight, obtaining a second characteristic descriptor according to the discrete characteristics of the neighborhood pixel points, and obtaining a comprehensive characteristic descriptor of the target pixel points according to the first characteristic descriptor and the second characteristic descriptor; The image enhancement module is used for preliminarily judging an image enhancement method according to a comprehensive feature descriptor of a target pixel point in an image block and a comprehensive feature descriptor of a preset sample image, acquiring a suspected edge pixel point, acquiring a gradient feature evaluation value according to a gradient feature of the suspected edge pixel point and the comprehensive feature descriptor; The enhancement iteration module is used for judging the enhancement effect according to the gradient characteristics and the comprehensive characteristic descriptors after the image blocks are enhanced, carrying out image enhancement iteration to obtain final enhanced image blocks, and obtaining enhanced inspection images according to all the final enhanced image blocks. Further, the step of blocking the inspection image according to the brightness difference characteristics of the pixel points in the inspection image to obtain different image blocks includes: Wherein K represents an adjustment coefficient, An exponential function based on a natur