| Περίληψη: | This work presents a number of state-of-the-art methods for road recognition applied to adverse off-road scenarios. In general, reliable deployment of unmanned ground vehicles in these kinds of environments depends on the combined use of expensive sophisticated sensors, in order to guarantee the system’s robustness. Of the available sensors, cameras present the most inexpensive and approachable option. Thus, this work recreates and evaluates computer vision methods, developed for road following off-road or not, when applied on more challenging scenarios than those originally tested on, dictated by road texture and colour, obstructive foliage and terrain and inconsistent lighting conditions. The conclusions do not intent to identify a method better than the others, but rather evaluate the strengths and weaknesses of each method, while keeping an open mind on possible combinations and additions that would prove the methods more effective.
|
|---|
