Study Of Automatic Visual Inspection Of The Three-dimensional Geometry

The automatic visual inspection of the industrial parts is one of the main problems for computer vision. Under corporately motivated by industry and academia, AVI has been introduced into the various branches of society such as food, industry, agriculture, aviation and so on, and greatly developed in recent years. Visual inspection usually involves measurement of specific part features such as geometric dimension, part identification, surface finish and seam. In this paper, we specially analyze the theory and method involved in three-dimensional visual inspection, and implement a profile inspection of freight according to real environment which is a useful attempt for AVI's application in our country.Inspection system, characterized by real time and accuracy, requires to flexibly adopt the novel technique in computer vision in order to design the perfect visual system for user. This paper discusses the theory of a vision based inspection system and two key techniques involved, i.e. the acquisition of range data and modeling of sensed data. Two efficient methods, a motion vision of multiple baseline and range imaging of structure light, are presented. At the same time, a rapid octree construction from perspective range image is developed.In workpiece site, industrial parts usually are carried by conveyor from one working procedure to another in pipeline manner. It is desirous how to perform an online inspection during transport other than interrupt the production. This paper implements an inspection system of freight's profile with low cost and high throughput for Hong Kong Air Cargo Terminals Ltd. The profile data of object are firstly obtained by analyzing all surfaces that are scanned and sampled by means of optoelectronic devices. Then, the sensor data is compared with standard contour to determine whether or not the packed fright measures up. We study and present the algorithms about automatic inspection, which analytical and experimental results demonstrate their effectiveness and feasibility.A special emphasis is put on the following in this dissertation:1. A multiple sensors based model combining the stereo or structure light with infrared device is presented in order to solve the profile inspection...