Data Collection Based On AUTOLINE

As the increasing requirement of chip production in China, the chip productionmachine develops quickly. During the postgraduate time, the author joined the secondcampus project organized by Motorola(China) Electronics Ltd. In Tianjin assembleand test factory, the author participated in the research of chip image recognition anddefection detection system.This article firstly gives the brief introduction of today's chips production inChina. Then demonstrates the necessity of this research and the technical guideline, aswell as the requirement of this experiment.After that, the article discusses the chips position detection system in details inthe following aspects: the hardware of the system, the image pre-process before imagedetection, fast position detect of chips and chip's defection detection.In order to restrain the affection of noise, this thesis discusses some methods ofreducing noise, analyses their virtues and defects and chooses the optional algorithmicof reducing noise based on action.This paper refer to a one dimension projection template matching method fornon-accurate locating object based on analysing two dimension template matchingmethod, so accelerating the recognition speed and reducing computation. Thepositioning and defection of chips are controlled in 100ms.The chips defection detection part discusses the base image procession, regiongrowth method and edge detection. The edge detection discusses in first orderderivation and second order derivation and evaluates conclusive their performance.Using Roberts's arithmetic operator can not only protect the edge data but also meetneed of timely performance.