The novel method proposed in this article is a content classification method. Using techniques like Content Based Image Retrieval and Bag Of Words, we classified several scenes captured by a Kinect camera. The method is simple, robust and provides high accuracy according to the tests. We tested the method using various classifiers such as SVMs, decision trees, KNN. The results demonstrate that the method is invariant to rotation, scale or illumination. Even if the objects change position in the scene, the classification is little influenced by that.
展开▼
