Adaptive image processing system for object classification

摘要

The system consists of an unit (1) for the automatic image recording, an unit (2) for extracting features from the image for producing image features, and a feature database (3) for the storage of the extracted image features for all patterns recorded in the learning phase. An unit (4) is provided for normalizing the image features, and another unit (5) for the automatic reduction of the amount of patterns stored in the feature database. A further unit (6) is provided for the automatic reduction of the image features to those features relevant for the classification. An unit (7) for the automatic classification and/or assignment of an unknown pattern to a class is provided, whereby a classificator is automatically modeled using patterns which are typical for respective decision classes. The outputs of the feature reduction and the classification are led back by a learning unit (9) to the feature database and/or to the feature extraction in loops (10,11), to optimize the amount of the patterns and the features.