Development Of Position Feedback Sensor Based On Vision Using Neural Network

Loading...
Thumbnail Image
Date
2009-12
Authors
Mohd Yunos, Yushazaziah
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Measurement of linear position is fundamental in many industrial processes especially as positioning feedback. Optical encoder is one of the frequently used position sensor. Unfortunately, there are some disadvantages in the usage of this type of encoder. This proposed work presents an approach to build a position sensor using image classification. The image was classified by using supervised backpropagation neural network. Input image that was fed to the classifier is a grayscale image of the surrounding environment (up view) which was captured to represent the position. The features in the image are not clear and thus feature extraction is difficult to be performed to extract statistical data. The image was rescaled and fed into the network as one vector. Series of images were captured at various positions with each series having different distance to each other. As the interval between images becomes closer, there is more overlapping between images.
Description
Keywords
Position Feedback Sensor , Vision Using Neural Network
Citation