Neural networks and fourier descriptors for part positioning using bar code features in material handling systems

Bar codes have been widely used in many industrial products for automatic identification in data collection and inventory control purposes. This paper presents an effective method to utilize the specific graphic feature of bar codes for positioning parts on plain conveyor belts without work carriers. First, a Fourier descriptor based method is used to obtain the rotation and field-depth information of a part and to detect the four corners of a bar code on the part. Then, by feeding the detected corners to a feedforward neural network, the horizontal and vertical translation of the part with respect to a calibrated location can be obtained. The proposed part positioning system has been successfully implemented in a laboratory setting. It shows that this economical device is capable of guiding an automated bar code scanner for bar code registering and robot arms or other automated material handling devices for part transferring without using any part presenting device.