Research On Appearance Detection Method Of Starter Based On Machine Vision

The powerless electronic starter is a device for controlling the start circuit of the compressor and the secondary coil to be disconnected in the refrigerator,to avoid unnecessary starting of the compressor,and the appearance defect is serious,which may affect the compressor to be disconnected from the coil in time.At present,the appearance quality inspection of the starter usually adopts the manual detection method,the production efficiency is seriously reduced,and it is easy to be misdetected and missed.The appearance defect characteristics of the starter were studied.Combining machine vision technology and automatic control technology,a set of starter appearance defect detection system was researched and designed,which is efficient,stable and reliable.The specific research contents are as follows:(1)Overall design of the inspection system.The characteristics of the starter defect were analyzed,and the first set of machine vision inspection system for the appearance defect detection of the starter was designed.The mechanical and control platform and the Matlab-based starter defect detection software design were completed.(2)Research on key technologies based on shape feature detection.Six kinds of image edge detection operators such as Canny operator and LOG operator are studied.The advantages and disadvantages of various operators are analyzed and compared through experiments.At the same time,the deficiencies of the traditional Canny operator are analyzed,and the improved Canny operator is proposed.The detection effect is better and the noise suppression ability is stronger.(3)Research on key technologies based on spatial relationship feature detection method.The flow of positioning of the top cover character area of the launcher is designed.The tilt angle of the top cover image and the image tilt correction based on Hough transform and Radon transform are studied.At the same time,three kinds of character region localization methods such as texture feature character localization method,projection analysis character localization method and morphological character localization method are studied,and the top cover character area of the starter is located.Four kinds of character segmentation methods such as edge segmentation method,connected domain method,vertical projection method and template matching method are studied,and the top cover character of the initiator is segmented.(4)Research and experiment on the detection method of appearance defects.The defect characteristics of the starter are analyzed.The defects include shape defects,gray scale defects,spatial relationship defects and color defects.The detection methods based on defect features are studied separately.In order to verify the effectiveness of the detection system related algorithms,and to check whether the system identification rate,defect classification rate and detection efficiency meet the requirements of design technical indicators,a single verification experiment and system performance testing experiments are designed.