Plane detector based on P-N learning structure for video satellite

With the rapidly development of video satellites, which provide sequential remote sensing images, amounts of data with high quality information are hunted to disposition. Object detection is one of the useful applications for video satellite and we focus on plane detection in this paper. Previous object detection methods focus on spatial structure rather than temporal information because they handle single image instead of video. We introduce a P-N learning structure dedicated to sensing video, which is firstly adopted in remote sensing video multi-object detection. We adapt a temporal management as P-expert and adapt a unique cascade classifier as N-expert. Our method use both structure information and temporal information, which make it more effective for sensing video. We detail each module and present an experiment to show the validity.