With the advent of a ordable drone systems, imagery from airborne sensors has become available for addressingmany di erent tasks in various fields of application. For some of these tasks the imagery has to come with ageoreference satisfying certain accuracy requirements. If we want to perform such a task and the accuracy ofGPS and INS sensors onboard the sensor platform cannot match accuracy requirements or location informationis faulty or unavailable, we need to establish a georeference or improve the inaccurate existing one.We do this with our image registration workow. It matches the contours of objects present both in theimagery and a reference image which comes with a georeference satisfying the accuracy requirement of the taskto be performed. This approach has proven to be both feasible and robust to appearance unsimilarity betweenthe image and the reference image, enabling to use a reference that is rather unsimilar in appearance to theimage.The workow comprises four steps, namely extracting the objects, extracting their contours, reducing theamount of contour points and finally matching them. To improve the performance of our workow, we aspire toimprove the performance of each of the four steps individually.While previous work has focussed on the netuning of the three latter steps keeping the object extractingmethod and thus step one xed for the time being, the scope of this work is the implementation of a novelobject extraction method and its evaluation in the context of the workow. Long line shaped objects such asroad networks are likely to be present both in the image and the reference despite their possible unsimilarity inappearance. The method extracts such objects after growing them by merging smaller individual line-shapedobjects if certain merge criteria is met.
展开▼
