| Scene matching area selection is a technique that selects some reference areas in the flight path according to certain requirements or criteria. These areas which contain obvious feature and rich information are often easy to match and used to guide the aircraft to scheduled target position. How to build a criteria that holds the capability of flexibility and jamming resistance is the key point for the selection of suitable matching area.Nowadays, many aircrafts navigate automatically based on optical sensors which are easily influenced by bad weather and lighting condition. Therefore, the guidance system works with a low efficiency. To overcome this shortcoming, Synthetic Aperture Radar (SAR) is used more and more often. As an active imaging sensor,SAR does not only provide images which resolution are as high as optical ones', but also work effectively without being influenced by bad weather condition. Because of these advantages, SAR is a very important imaging sensor for unmanned aircraft navigation and guidance.The core problem of SAR subimage selection for mission planning of unmanned aircraft navigated by scene matching was studied detailedly in this dissertation. And a new method based on machine learning is presented. The main contents of this method includ (1) principle of scene matching system; (2) SAR image feature analysis; (3) theory of Supporting Vector Machine; (4) selecting subimage for target recognition based on SVM.Besides, this paper studied some mathching methods suitable for SAR image, and adopted the multi-subimage matching algorithm based on spatial relation constrain to improve the performance of correlation matching. What was studied in this paper does a great help to improve the matching probability of SAR image matching and lower the error of navigation. So the work is very meaningful. At the last chapter, a summary of what has been done in the this paper is made and what should be studied further in the future is pointed out. |
PDF Full Text Request