Inspection system for detecting defects in a transistor using Artificial neural network (ANN)

A machine vision system based on ANN for identification of defects occurred in transistor fabrication is presented in this paper. The developed intelligent system can identify commonly occurring errors in the transistor fabrication. The developed machine vision and ANN module is compared with the commercial MATLAB^® software and found results were satisfactory. This work is broadly divided into four stages, namely intelligent inspection system, machine vision module, ANN module and Inspection expert system. In the first a system with a camera is developed to capture the various segments of the transistor. The second stage is the image processing stage, in this the captured bitmap format image of the transistor is filtered and its size is altered to an acceptable size for the developed ANN using Set Partitioning Hierarchical Tree (SPIHT). These modified data are given as input to the ANN in the third stage. Generalized ANN with Back propagation algorithm is used to inspect the transistor. The ANN is trained and the weight values are updated in such a way that the error in identification is the least possible. The output of ANN is the inspected report. The developed system is explained with a real time industrial application. Thus, the developed algorithms will solve most of the problems in identifying defects in a transistor.