Vision Recognition System For Opening Of Train Tank Based On Smart Camera

As a machine vision solution with non-contact, high reliability and easy implementation, smart camera is widely applied in fields such as working condition surveillance, product testing and quality control etc. In our country, this field is just at starting stage, and research of this thesis has applicable meaning for application and promotion of machine vision in the industry field in China.This thesis mainly studies development methods and performances of VC4018 smart camera and circle detection problem under uneven light condition etc. This thesis taking VC4018 as development platform studies three kinds of circle detection methods including Hough transform circle detection, least-square circle fitting and edge matching. As edge image is not sensitive to light change, usual edge detection technologies are discussed, and circle detections are all based on edge. Hough transform circle detection is a classical method, with good robust and high precision, but with defect of large computational complexity, and it is suitable for scenes with not high requirement for speed; least-square circle fitting has fast speed and high precision, adaptive to edge loss caused by light, and can finish detection fast and accurately under low light; edge matching is adaptive to false edge, not suitable for serious edge loss, and needs proper light supplement under low light; since matching method had some problems in real-time operation of VC4018, this thesis studies program optimization.