Detection Of Core Rod Defects Based On Visual Saliency

Defects of composite insulator core rod will bring serious problems for high voltage lines, and nondestructive examination of defects is essential. Traditional detection methods are mainly based on human eye detection which has several problems such as poor reliability, low detection efficiency and complex operation. With the continuous development of modern computer vision technology, industrial defect detection technology based on machine vision has been widely concerned in various fields of modern industry in recent years, and become an important part of the industrial quality inspection solutions. Therefore,research on detection of core rod defects based on related technology of machine vision has important significance and application value.The thesis focuses on the research of core rod defects detection based on visual saliency. This thesis analyzes main saliency algorithms and makes algorithm improvement for the demand of defects detection. Meanwhile,a software with automatic online core rod defects detection function. The main work of this thesis is as follows:1) Make processing and analysis of the core rod defects images using typical visual saliency algorithm. On algorithm performance of visual saliency, The thesis processes defects images with 4 different classical saliency algorithms respectively, analyzes respective saliency maps and evaluates detection accuracy rate.2) According to the demand of core rod defects detection, design and implement a saliency detection algorithm based on texture and gray scale features. In order to solve the shortcomings of LC algorithm significantly on detecting the defects of local texture map for inspected images,respectively using Gabor filter to process and LC algorithm to calculate to fuse the images with Haar wavelet transform, and finally optimizing the binary defects maps with morphology. Through the contrast test, the algorithm further improves the accuracy of core rod defect detection.3) Achieve the real-time defects detection software system of multi camera, which has timing and off-line detection, multi camera images capturing, real-time display, automatic detection and preservation, defect monitoring and automatic alarm functions, to meet the needs of industrial applications of core rod defects detection.