Uav Remote Sensing Image Fast Seamless

Unmanned Aerial Vehicle(UAV) has been applied widely in battle field scouts, forest fire disaster monitoring, evaluations of natural disaster region, etc, due to four advantages of the low altitude remote sensing data from UAV: high resolution, good flexibility, high efficiency and low cost. However, UAV images possess features like small range, large amount, heavy shooting slope, irregularity overlap, which lead to bad image mosaic results based on the traditional image mosaic algorithm. So it becomes the key of successful UAV application to develop an effective image mosaic algorithm according to UAV images features.To aim at solving problems in image mosaic such as heavy work burden and low accuracy resulted from various UAV images features, the image mosaic algorithm based on features is improved, and the process is optimized, which enhance the speed and accuracy of the image mosaic algorithm.First, a phase correlation method used to calculate the overlap degree between reference images and registering images is realized. And a new method to carry out feature extraction and match only within the overlap region is proposed, which reduces the computational complexity dramatically, prevents non-overlap image region's data from interfering the algorithm and increases the accuracy of the mosaic algorithm.Second, feature points are extracted within the overlap regions between reference images and registering images by adopting the improved Harris angle point detection algorithm. And obvious unmatched points are deleted from the two feature points groups, according to the distance between reference images and registering images calculated from overlap degree. The matched feature points group can be certified finally through the similarity function. Comparing with the conventional method to match feature points before deleting, this method can reduce incorrect match points in a large amount while guarantee low computational complexity.Third, the images registration is implemented by using the matched feature points group to perform model transformation for matching images. Moreover, a gradated in-and-out amalgamation algorithm is adopted to fuse the registered images, which makes images linking together smoothly and seamlessly and improve the visual effect of the final matched image.Finally, based on a great amount of remote sensing data acquired from disaster region by UAV, the quality appraisements of the image registration, image fusion in the improved algorithm are conducted. Various kinds of experimental data demonstrate the improved image mosaic algorithm can obtain good mosaic result, possesses features such as high accuracy rate, good robustness, and has great application value.