A real-time stitching method of aerial images based on tile map

Abstract

Aiming at the problems of small speed and large memory resource consumption for aerial images got by Unmanned Aerial Vehicle (UAV), which caused by the high pixel, high precision and large size, a real-time image stitching method integrating feature information transfer and image slicing is proposed. It improves the speed of image stitching and reduces the memory consumption in the stitching process. The position data is used to calibrate images firstly. Then image feature extraction is done by the Speeded Up Robust Features (SURF) algorithm. The gradual in and out method is used for image fusion. After each stitching is completed, the feature points of the stitched image are no longer re-extracted, on the contrary, the feature information of the previous image is transferred to realize continuous frame stitching. The stitched image using N images is sliced and stored, and the system memory is freed. The experiments are carried on real aerial images of a certain place. The results have shown that the stitching time of each ten images is 7.935 seconds, and the memory consumption is below 700MB stably. The results verified that the proposed method can get good performance in speed and memory.