Wafer Chip Count Based On Curve Fitting

With the even increasing development of Semiconductor Industry, Industrial upgrading has caused growing concern from the major domestic and international manufacturers.At present, When producing chips, the mainstream integrated circuit manufacturers often have to count the chip manually.It is not only inefficient but also costly. Therefore, it is not suitable for the integrated circuit industry to expand the needs of production.On the other hand, with the wide usage of computer communication network based on image and video, and the rapid development of multimedia technology, machine vision technology has always been a hot research topic in the semiconductor industry production. Machine vision has become a major subject in the field of computer science and engineering. Therefore, using the machine vision technology, we can realize the automatic counting of wafer chip which can effectively improve the degree of automation of wafer industrial production.Applying machine vision to the production of semiconductor,we must overcome the interference caused by many factors such as uneven illumination, image quality, low-contrast image and the change of viewing angle. To overcome these problems, in this paper, we propose a counting method based on curve fitting. First, we make some preprocessing to the original image and locate the position of the wafer in the same time, then we detect the straight lines and select them by using the random sample consensus algorithm,and in this way we filter out the reliable lines to estimate the cutting lines of the wafer as well as the wafer size and other information. By making use of the detected straight lines and the length and width of the chip and so on, we reconstruct the geometry model of the wafer through the curve fitting method, thus we simulate the internal structure of the wafer and eventually realize the automatic counting of wafers.In this paper, we study the image processing, by improving the algorithm and the technical parameters, we have completed the precision sub-pixel detection location of the image,further improving the accuracy in locating the wafer. In the meantime,detecting and screening the wafer cut lines, can acheieve the precise location of these lines,and automatically and effectively calculate the the number of the wafer chip finally. The experimental results show that the proposed algorithm has strong anti-noise ability,which can effectively deal with uneven illumination, image quality, low contrastive image and the change of viewing angle and many other issues. The technology can also avoid high cost arising from manual operation. In the meantime, the algorithm can effectively apply to the intelligent detection system of the wafer production and has a good developing prospect.