Adaptive Reconstruction Using Multiple Views: Results and Applications

摘要

This report describes a method to use information from a sequence of images togive a better description of objects partly occluded by other objects (e.g. tree tops, camouflage net) and where different parts of the object of interest can be seen from different aspect angles. The project aims to merge different parts of the object of interest into a complete description, although the parts have been seen from different aspect angles. Examples of results from depth segmented images are given in the report. Examples of applications for the method are: reconnaissance from low altitude with a traditional aeroplane or an UAV (Unmanned Aerial Vehicle) as platform, preprocessing for target recognition, passive range measurement using egomotion, passive nearfield navigation for robots, tanks and other vehicles, increasing the detail resolution in the processed image compared to the images in the time sequence, noise reduction and increasing the effective fillfactor for starring infrared focal plane array cameras (staring IR-cameras).