Research Of IC Chip Pin Inspect Methods Based On Machine Vision

During the chip packaging process, flaw inspection of components is the necessary foundation of correct packaging. It’s difficult to achieve high speed and precision inspection of IC by use traditional inspection methods. Computer image inspect technology is a measuring technique which has advantages of non-contact, real time, accuracy and high reliability. Traditional test technology is lack of these advantages. This paper analyzes the current situation and development of machine vision inspection technology, introduced the working principle of the machine vision inspection system, and has done some researches focus on several key technology in vision inspection of IC pins. Achieved inspect of IC pin’s number, interval and size by use matlab.This paper conducts an investigation in the image preprocessing algorithm. Through experimental analysis of some classical filtering algorithm such as median filtering, the results prove that all these algorithms will smooth on every pixel, which will result in image blurring and then affect the precise of the inspection. So this paper proposes an improved switch median filter based direction information. Experiment result shows that this filter can reserve image edge information well. In order to move away some impurity that appeared after the algorithms processing of this paper, the image morphological operation could be a available method in post-processing of the experiment result.Through analyzing the principle of some edge detection algorithm and comparing the result of their processing on the experiment image, it proves that Canny edge detection algorithm Show better performance in the object edge detection. However Gaussian filter algorithm that the Canny operator uses may cause image over smooth and the edge of the fuzzy, and it also be sensitive to pulse noise. In order to make the edge more completely, to remove noise effectively and to improve the accuracy rating of result, so this paper combines the Canny edge detection algorithm with the improved filter algorithm and it get a well performance.Paper discusses a method of search the eight neighborhood pixel of image edge points to get the outlines of the inspect objects, the experiment result shows that it can segment de object contour completely. The two template matching methods discuss in this paper, one is based on gray correlation and the other is based on the image geometric moment. The advantage of the first template is the large computation and time high complexity, and the advantage of the second template is based on the objects in the region. So this paper put out a method of recognize the pin contour by calculating contour moments. The experiment result shows that this method is fast and can reach the accuracy requirements of visual inspection. This paper introduced the implementation of the IC pins vision inspection methods in detail, and display the experiment data and analysis.