Region of interest detection based fast and robust geo-spatial object recognition

In this paper a novel computationally efficient algorithm to detect objects automatically from high definition satellite imagery with high performance is presented. The proposed algorithm has three main steps supporting each other: Filtering, shape based and appearance based object detection. A region of interest indicating the possible regions that may have the objects to be detected is determined in a very short time via filtering step. In the remaining steps, the objects are extracted from that region and then target ones are detected. While shape based detection provides high precision; appearance based detection improves the recall by using the detected objects by shape based detector. The proposed method is superior with its computational efficiency and high performance. Computational efficiency mainly comes from the filter which is the main contribution of this study to determine region of interest and high detection performance is by means of using shape and appearance based approaches to support each other, which is another contribution as well. The method can detect objects of different scales and rotations with high detection and certain localization performance in high resolution satellite images.