Mosaicing of aerial images using feature point extraction for use in variation detection

Abstract

Our aim in this project is specifically to design a system that will automatically mosaic aerial images. These images may be from a satellite or any other source such as an Unmanned Aerial Vehicle (UAV). Any such vehicle when sent into air to photograph a certain area takes several pictures as it flies above. For proper analysis these images need to be mosaiced in order to create a single bigger picture of the whole area. The images have areas which overlap or are present in more than one image. Before all the images are combined these overlaps need to be automatically detected the images be joined in such a way that the image contains no doubling. The number of images may be quite large and the matching process may take quite some time. Efficient algorithms need to be designed and implemented in order to piece together all the pictures correctly in a short time without much error. In order to resolve the above difficulties, the techniques of Feature Point Extraction are used. This relies on accurate detection of image features such as detecting corners. After the feature points have been extracted we can use the analyzed data to help us mosaic the images. The resulting mosaic can be used in a number of ways. Suppose that the same area is photographed once again. The new pictures can again be analyzed in the same way to form another mosaic. The two mosaics can then be compared using mathematical techniques to detect any changes in the area and report to the user. Some examples of these changes can be construction of a building or the digging of a bunker etc