Research On Chip Positioning System Of Automatic Gold Ball Welder

The automatic gold wire bonder is one of the modern high technology, The image recognition system is one of the main key technologies to realizing the automatic welding chip. To break the technology blockade of developed countries,developing our own automatic gold wire bonder technology is imperative. At present, there already have been a lot of ways to locating the chip in image, However for accurately positioning the image and real-time to recognition the chip of the automatic gold wire bonder of the chip still need further study in its accuracy and real-time performance.This paper mainly introduce the preprocessing method of enhace the image which includes two aspects; space domain and frequency domain, the advantages and disadvantages of these two methods have been compared. Firstly, for image enhacement, two filtering methods (homomorphism filter and gray histogram modification)are presented. According to the experiment result, we choose the method based on histogram matching to enhace the contrast of image. Second, when extracting the edge of the image, two methods which are based on searching and zero crossing are introduced. These methods include Roberts Cross, Prewitt, Sobel, Kirsch, Marr-Hildreth, Canny and Laplacian. A improved Canny edge detector method has been presented in this paper. Finally, the algorithms of extracting the feature have been presented. They include Moravec, Harris and Sift. For the matching algorithm, this paper mainly introduces the Sift algorithm, its high accuracy of translation, rotation, scaling has also been proved. It has important meaning for realizing the automatic gold wire bonder based on image processmg.