Automatic POS Supported Tie Point Transferring On Multi-angular Image

Nowadays, due to the frequent changes in surface morphology and various natural disasters, the update of geospatial data and emergency response are becoming particularly important. Responsive to the needs of development, three significant progresses were achieved in the photogrammetry, they are the rise of low-altitude UAV remote sensing platforms, oblique aerophotogrammetry and POS supported aerophotogrammetry. In this paper, the multi-angular image contains the UAV low-altitude image and the oblique image. Then we analyze the advantage of multi-angle images as well as the challenges brought to the aerial triangulation tie point transferring. On this basis, we attempt to achieve the rapid processing of tie point transferring with the assistant of the POS data. In order to achieve this target, the research works in this paper mainly include as follows:1. The calculation of the initial relationship between images is the key of fully automatic feature matching. So when facing the regular strips images, we recover the relationship of the strips. For oblique images, we use the POS data and the topographic information to get the relationship matrix of all images. After that, we can do a fully automatic feature extraction and matching procedure.2. According to the large rotation angle, large scale differences and large geometric distortion existing in the multi-angle images, we did a research on how to achieve the fully automatic process of tie point transferring of multi-angle images with the combination of the theory and algorithms in computer vision. Then a quantitative analysis of the existing four typical feature extraction operators including SIFT, SURF, MSER, ORB in multi-angle images automatic tie point transferring has been conducted, which will provide proof to the automatic extraction of the tie point. And in order to analyze the comparison results, three specific indicators have been proposed.3. In order to reduce the influence which was brought by the large rotation angle, large scale differences and large geometric distortion existing in the oblique images, we proposed a POS supported pre-corrected SIFT feature matching method to achieve a better result of tie point transferring.4. An automatic POS supported multi-angle images tie point transferring prototype system has been developed with Visual Studio2010. In order to prove the flexibility of the tie point transferring method in this paper, it has been used to do many tests on a lot of multi-angle image data.