Abstract
The aim of this work is to interpret the image for vision-based logic control of manufacturing plant. The images. which had been taken by an overhead fixed camera, were transformed with feature value by image preprocessing. A multi-layered neural network was used to recognize randomly selected images of metallic pegs and plastic rings. Images, which are not recognized by the neural network, will be rejected and consequently the actuator will allow the objects to pass through. On the other hand, if the images are recognized by the neural network then the logic controller generates different type of action and the actuator pushes the object down.