The interpretation of aerial images is difficult, especially for low-resolution images. Although solutions have been worked on for many years, performance of these systems is still not sufficient to be useful in practical applications. One potential solution is to create systems composed of many specialized modules. We introduce one such module, a Bayesian tracker for linear structures, such as pipelines and access roads. We show that the tracker can be used successfully to detect these structures in low-resolution images.
Rui Gao, Walter F. Bischof