Shared Features for Scalable Appearance-Based Object Recognition
WACV-MOTION '05 Proceedings of the Seventh IEEE Workshops on Application of Computer Vision (WACV/MOTION'05) - Volume 1 - Volume 01
| Hi-index | 0.00 |
The present work describes a recognition system of objects through the computer using color, forms and texture descriptors. In the classification of the objects the classifier "K nearest neighbor" was used in a sequential way for characteristics of color, forms and texture. In the classification by color the format RGB was used; in the classification by form was used the idiosyncrasy and the factor of compactness; finally, in the classification by texture was only used the energy descriptor. Initially, the recognition system was planned to help blind people in the recognition of objects around their environment; but due to the existent problem in the design of general algorithms for the recognition, A proposal was given for its implementation within a classroom, being limited only to the recognition of 10 small objects. With the method proposed in this work the percentage of success in the recognition is improved more than 94.58%. Using the traditional method "K nearest neighbor" in the classification, the distances of the vector are calculated by characteristic up to the point of classification; using the proposed method the number of classes in each classification is reduced, in this way the classifier is able to classify a bigger number of objects in each classification in a sequential manner and with a bigger certainty when there is not confusion of classes.