Computer vision based characterization of production phases for pastry goods

Abstract

Advances in computing technology to improve the automation of industrial production processes has grown the expectations for food products of high quality and safety standards. Computer vision and machine learning techniques provide an alternative for developing automated and cost-effective methods to accomplish the requirements in food industry. In this paper, we propose several computer vision algorithms to characterize the process of transformation of bakery goods during production. Features are extracted from images acquired from a camera mounted in the furnace of a manufacturing company that produces bakery foods. Radial Basis Function Neural Network (RBFNN) is used to model the transformation of appearance of a particular product of interest. The RBF Neural Network offers several advantages compared to other neural network architectures especially when a characterization of a transitional process is needed. RBFNN possesses the property of best approximation and the output of the network can be optimized by setting suitable values of the center and the spread of RBF. Experimental results indicate that gradient based features have higher recognition rate due to their ability to capture the texture.