Automatic detection method of dimension error for aluminum alloy parts
                                                 FENG Miao, ZHOU Chuan de, MENG Ming hui
  （College of Mechanical and Power Engineering, Chongqing University of Science and Technology, Chongqing 401331, China）

Abstract: Aiming at the problems of the reflection light interference and the difficulty of detecting large size aluminum alloy parts in the dimension errors detection process, the machine vision system, image synthesis, and image processing were studied.Then,an automatic detection method based on machine vision system was proposed. The high precision images for large size parts were obtained by using a coaxial parallel light system and a two axis motion platform. In order to obtain a higher detection accuracy, the local high precision images were captured and then spliced together to rebuild the whole image of the aluminum alloy part, which allows this method to detect large size parts without size limitation. The preprocess and feature extraction of the obtained images were performed by visual inspection technology. The part edges were extracted by using Canny operator combined with bilinear interpolation method, which can obtain sub pixel precision for the part edges. The dimension errors of aluminum alloy parts were determined by matching these processed images with the CAD drawings. The results indicate that the detection accuracy is higher than 0.02mm, which can meet the requirement of the automatic detection for the aluminum alloy part manufacture.

Key words: machine vision system; aluminum alloy parts; image synthesis; template matching