Research And Implementation Of On-line Quality Detection System For Glass Insulator

Insulator is one of the key components of high voltage transmission line, whose performance may directly influence the operation security of the whole transmission line. Because of zero self-failure and esay to maitaian, insulator has been widely used in many fields. At present, artificial sampling statistics is considered as the commonly adopted method of detection and counting of glass insulator. Thanks to the error of manual observation and timeliness of statistics, manual test result can hardly meet the needs of the modern enterprise production quality management. It is necessary to improve the automatic production technology for high quality of the insulator and an on-line quality detection method for glass insulator based on machine vision will be strongly required.This paper design and realize the on-line quality detection system for glass insulator. Meanwhile, some methods based on machine vision are studied which include the method of taking insulator from the mesh belt background, the complete and broken insulator classification algorithm and multiple target tracking and counting algorthm. The research works in this paper can be summarized as follows:(1) Design of on-line quality detection system for glass insulator. According to the production process of glass insulator and the work environment, the structure of the detection system for glass insulator is proposed. The system device is designed with a perfect integration of lighting system, image acquisition system and computer detection system. The system software is designed to meet the enterprise needs. Furthermore, the technical index of the final system is determined.(2)The glass insulator detecting method based on the machine vision is mainly studied. This paper analyzed the the image features of insulator against mesh belt background, and moving insulators detection and classification algorithm. Because of the complex types of the broken insulator fragment distribution and the irregular shapes, this research mainly focuses on the overall extraction of broken insulator debris and broken feature recognition. First of all, the fast bilateral filtering is used for detection image preprocessing. Furthermore, the feature extraction method with the window location area of moving target detection algorithm is proposed. In addition to the basic feature such as color features and morphological characteristics, fragmentation characteristics and boundary diffusion features which based on the physical mode of broken insulator, are used for classification.(3) A tracking multi-object counting method of nearest window's characteristics combined with Kalman is proposed to reduce the targets missing that caused by the light and shelter disturbution during the moving and detecting procedure. The sequences of image that collected from the production site are tested by building the prototype equipment. Through simulation testing, the results show that the system can detect and identify for insulator on line, with a good satisfaction of automatic counting demands in the factory assembly line.