Feature-based Real-time Object Identification in Industrial Processes

Successful segmentation is a primary requirement for many industrial applications of image processing. This paper presents a study on existing image segmentation and registration algorithms. Also, the paper proposes an improved image registration algorithm for extraction of desired features for high speed automatic sorting applications for a small group of objects. The sorting is based on object´s feature extraction and identification. This involves pixel intensity information in addition to physical features for example shape, color and texture (depending on application). The algorithm mainly involves four different parts firstly extraction of entry features by using pixel intensity or histogram methods, secondly matching and comparing those features with a knowledgebase and finally extracting desired features.