Defect Inspection of LED Chips Using Generalized Regression Neural Network

Abstract

The inspection of the defects in LED chip has become a critical task for manufacturers in order to enhance product quality. In this paper, a new approach for the defect inspection of LED chip is presented, which uses both the features of defects and the generalized regression neural networks. The approach consists of following three steps. First of all, preprocess of LED chip image is performed by using the image operations such as image enhancement. Secondly, the chip image is divided into a lot of sub-regions, the features of each sub-region are extracted, the database of features is built. Thirdly, an initial structure of generalized regression neural network is constructed, then the neural network is trained by using the features in database. The generalized regression neural network has the ability to converge to the underlying function of the data with only few training samples available, and the additional knowledge needed to input by the user is relatively small. The experimental results show that the defect inspection approach in this paper can effectively identify the LED chips with defects.