Pattern recognition combining de-noising and linear discriminant analysis within a real world application

Computer aided systems based on image analysis have become popular in zoological systematics in the recent years. For insects in particular, the difficult taxonomy and the lack of experts greatly hampers studies on conservation and ecology. This problem was emphasized at the UN Conference of Environment, Rio 1992, leading to a directive to intensify efforts to develop automated identification systems for pollinating insects. We have developed a system for the automated identification of bee species which employs image analysis to classify bee forewings. Using the knowledge of a zoological expert to create learning sets of images together with labels indicating the species membership, we have formulated this problem in the framework of supervised learning. While the image analysis process is documented in [5], we describe in this paper a new model for classification that consists of a combination of Linear Discriminant Analysis with a de-noising technique based on a nonlinear generalization of principal component analysis, called Kernel PCA. This model combines the property of visualization provided by Linear Discriminant Analysis with powerful feature extraction and leads to significantly improved classification performance.