AN OBJECT RECOGNIZING SYSTEM FOR INDUSTRIAL APPLICATIONS Marcelo Kleber Felisberto Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial – CPGEI Centro Federal de Educação Tecnológica do Paraná – CEFET-PR Av. Sete de Setembro n° 3165, Curitiba, PR, Brazil mkf@cpgei.cefetpr.br Tania Mezzadri Centeno Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial – CPGEI Centro Federal de Educação Tecnológica do Paraná – CEFET-PR Av. Sete de Setembro n° 3165, Curitiba, PR, Brazil mezzadri@cpgei.cefetpr.br Lúcia Valéria Ramos de Arruda Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial – CPGEI Centro Federal de Educação Tecnológica do Paraná – CEFET-PR Av. Sete de Setembro n° 3165, Curitiba, PR, Brazil arruda@cpgei.cefetpr.br Abstract. On several automatic systems for manufacturing and assembly, specially on Flexible Manufacturing Systems (FMS), identifying mechanical parts produced and detecting their position and orientation are important issues related to the necessity of handling these parts by industrial robots. Currently, the recognizing process is performed by human check, but it may lead to increasing errors and accident probabilities. So the implementation of an effective automatic system, in order to recognize the parts, would not only avoid these risks but it would also improve the process velocity as well as its reliability This work presents an automatic system for free-form objects recognizing which identifies mechanical parts produced by a FMS. This system can recognize objects in a monochromatic image, captured by a charge couple device (CCD) camera. In addition to this, the system can be easily enabled to verify the parts orientation. This work used the concept of behavior vector, from the image indexing techniques, as a solution for the objects representation. Then, during the recognizing process, at least one hypothesis are generated by a backpropagation neural network trained to recognize the pattern vectors (known objects). Finally, the hypotheses are evaluated through a final verification process. As a result, the system offers quick and correct answers and also flexibility to be applied in other applications. Keywords. object recognition, image indexing, image analysis, backpropagation neural network, FMS. 1. Introduction The goal of an object recognition vision system (ORVS) is to find objects on images taken from the real world, using object models which are known a priori (Jain et. all, 1995). Currently, this task is easily performed by humans, but it is surprisingly difficult to unable machines to recognize objects with the same efficiency (Pope, 1994). Artificial intelligence technologies have leaded to advances on computer vision researches, including new automatic object recognition approaches. However, the object recognition technology is not common on the industry nowadays, even on repetitive tasks (Orth, 1998). One reason is the limiting performance of some object recognizing systems when applied on real time applications. Another limiting factor is the complexity of the object shapes and the capacity of the system to learn or archive a vast number of patterns keeping the same efficiency and agility (Jain et. all, 1995). The development and implementation of effective automatic systems, in order to recognize the parts, are an important industrial necessity to diminish human errors and speed up the production (Centeno and Bagatelli, 2000). Besides, the automatic object recognition can support inspection tasks and allow robots and machines to manipulate parts and tools correctly (Rudek et. all, 2001). This paper makes a briefly review about solution on automatic object recognition approaches and present a new system for object recognition task in order to support industrial manufacturing processes. The system is based on image indexing technique and neural networks. 2. The object recognition problem A generic model of an automatic ORVS is shown on Fig. (1). The block diagram schemed represents the elementary components of an ORVS and how the information flows through the system during the recognition process. Basically, any automatic object recognition systems are functionally equivalent to this model (Jain et. all, 1995). Generally, the systems differ one from each others by the solution present on each component for the information processing as well as the kind of information that is processed and passed way. The components represented by each block are: ABCM Symposium Series in Mechatronics - Vol. 1 - pp.95-104 Copyright © 2004 by ABCM