Aircraft target detection using multimodal satellite-based data

In recent years, remote sensing image processing on orbit has become an essential task to conduct rapid satellite response, where the target detection using multimodal remote sensing images has attracted much research attention. More specifically, the aircraft is an important type of targets and aircraft target detection on-orbit becomes much essential due to the safety reasons. In this work, we main focus on the detection of aircraft targets using multimodal satellite-based data. Our method consists of three main procedures. First, we estimate the position of the imaging target in the combination of satellite attitude and orbit information, and then we can select several candidate image patch from the original large scale image data. Then, the initial investigation of candidate target areas is detected based on the texture characteristics, where Gabor filtering, image binarization and regional Unicom are utilized and the candidate aircraft target can be roughly detected. Since the characteristics of aircrafts remain relatively stable, the candidate areas can be further refined based on circular-matched filtering. To evaluate the performance of the proposed method on aircraft target detection, we have conducted experiments on six large scale remote sensing images. Experiments show that our proposed method is able to achieve an accuracy of 90% for target detection and the detection time is less than 0.5 s, which can be used for real time applications.