Airplane detection in remote sensing images based on Object Proposal
Qinhan Luo, Zhenwei Shi · 2016
Automatic detection of airplanes in remote sensing images (RSIs) remains a challenge. Its primary problem is how to locate the airplanes from the huge searching space of the image in an efficient way. In this paper, we utilize a simple but effective technology, Object Proposal, for airplane locating. The main objective of the technology is to generate a relatively small set of bounding boxes that most likely contain objects of interest. In our approach, a small set of bounding boxes that most likely contain the airplanes are first generated by the Object Proposal algorithm. Afterwards, a SVM classifier is trained on the HOG features to detect the airplanes. Finally, the trained object detector is applied to those bounding boxes instead of exhaustive search to complete the detection task. Experiments show that our Object Proposal method is effective in its ability of producing good quality proposals. It can be further utilized for the detection task to reduce the computation cost.