Research On Quality Detection Method And Equipment Of Toughened Glass Insulator

Abstract:As an important safety device in high-tension transmission line, toughen-ed glass insulator is widely used in power system. Its quality and performance are related to secure and steady operation of the power systems. Artificial detection method with eyes depends on the testing experience of quality inspector, with low accuracy and low speed. In order to solve these problems, some technologies are adopted in the project, such as binocular stereo vision, moving target detection and machine intelligence study technology. The purpose of the research is replacing artificial vision with computer vision and imitating artificial defect judgment by machine learning. In this paper, defect detection methods were studied. At the same time, a new efficient automatic quality detection device for toughened glass insulator was designed according to the national quality inspection standard.The project started from the design of detection device for toughened glass insulator and the research for quality detection method. Firstly, the detection device was designed, which could work in three detection modes. They were off-line detection, monocular movement defect detection and binocular stereo defect detection. The three modes were designed respectively according to three different detecting conditions. Secondly, the detection method was researched, including image segmentation, feature extraction, defect recognition etc.The precondition of off-line detection is not concerned with the detection speed and illumination conditions. Different image processing algorithms were contracted in order to obtain a suitable detection algorithm which was the most efficient.Monocular dynamic defect detection was studied for on-line detection. The toughened glass insulator detection device can implement integrated function such as defect recognition, quality testing and inferior insulator separation when it is built in the production line. Monocular dynamic defect detection is one of the main contents of this paper. This paper mainly studied the following:firstly, these methods were researched, such as three frames subtraction method for moving target detection and arc interference removal algorithm based on single circle fitting method. Secondly, the detection device was designed to match the production line using the software called SolidWorks. The device included rotation control test bench and dam type LED light source. Besides, actual environment and several hypotheses were simulated by the dynamic simulation software.As the CCD camera and the determinand stay in fixed positions, the dead angle of camera may cause error judgment. Binocular stereo defect detection was studied to solve this problem. Binocular stereo defect detection used binocular parallax principle to achieve defect three-dimensional positioning in real space. In this system, the second image of the suspected defect was filmed by the third CCD camera in order to improve the detection accuracy. For binocular stereo defect detection, the principle and detection methods were studied. The research content included binocular parallax principle, calibration method of binocular visual system, stereo matching,3D reconstruction and so on.Experiment has showed that both of the off-line detection and dynamic detection of defects can detect Bubble, crack and breakage efficiently. The detection device for toughened glass insulator can work independently or cooperate with production lines.