This research focuses on developing a system that can retrieval objects from large image database by exploring the types of image features necessary for recognition of common objects in scene. Then we make global representation for these features that can be used in learning. After that, we figure out a novel method for generic object detecting in still images with automatically choosing feature. Our method is simple, computationally efficient and bases on features that is easily seen by naked-eye and very close with natural detecting by human. The main advantage of this method is that it can automatically choose features which are best for detecting one type of object. We present experimental results for detecting many visual categories including side view car, front view car, bike, motorbike, train, aero plane, horse, and sheep. Results clearly demonstrate that the proposed method is robust and produces good detection accuracy rate.
展开▼