Peter W. Pachowicz, Sung W. Baik
This paper presents and validates a method for adaptive object recognition in image sequences under dynamic perceptual conditions, and consequently, under changing object characteristics. The approach builds a close-loop interaction between object recognition and model modification systems. Object recognition applies a modified RBF classifier in order to recognize objects on a current image of a sequence. The feedback reinforcement generation mechanism evaluates the classification results when compared to the previous images and activates classifier modification, if needed. Classifier modification selects a strategy and employs four behaviors in adapting the classifier's structure and parameters. These behaviors include accommodation, translation, generation, and extinction applied to selected classifier components. Accommodation modifies the component's boundary/spread. Translation shifts a given component over the feature space. Generation creates a new component of the RBF classifier. Extinction eliminates components that are no longer in use. The evolved RBF model is verified in order to confirm applied model modifications. Experimental results are presented for indoor and outdoor image sequences. The approach is validated and compared with traditional non-adaptive methods for object recognition. This validation tracks an error rate, classifier complexity, and the utilization of model modification behaviors over the image sequences.
Keywords: Adaptive systems, Vision, Machine learning, Adaptive object recognition
Citation: Peter W. Pachowicz, Sung W. Baik: On-Line Model Modification for Adaptive Object Recognition. In W.Horn (ed.): ECAI2000, Proceedings of the 14th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2000, pp.668-672.