Published Paper Details:

Object detection is one of the essential applications and crucial tasks in remote sensing images analysis. With the enlargement of remote sensing knowledge, high-resolution (HR) optical remote sensing images are presented and the geospatial detail information of substances is more and more abundant. In this paper, rotation invariant parts based model to sense objects with difficult shape in high resolution remote sensing images. Explicitly, the geospatial matters with difficult shape are initially alienated into several main parts, and the assembly information among parts is labeled and controlled in polar coordinates to achieve the rotation invariance on configuration. In encoding the features of the rotated parts and objects, a new rotation invariant feature is projected by spreading histogram oriented gradients. Throughout the final recognition step, a clustering technique is introduced to discover the parts in objects, and that process can also be used to fuse the detection consequences