Abstract
This paper presents a vision sensor-based solution to the challenging problem of detecting and following trails in highly unstructured natural environments like forests, rural areas and mountains, using a combination of a deep neural network and dynamic programming. The deep neural network (DNN) concept has recently emerged as a very effective tool for processing vision sensor signals. A patch-based DNN is trained with supervised data to classify fixed-size image patches into “trail” and “non-trail” categories, and reshaped to a fully convolutional architecture to produce trail segmentation map for arbitrary-sized input images. As trail and non-trail patches do not exhibit clearly defined shapes or forms, the patch-based classifier is prone to misclassification, and produces sub-optimal trail segmentation maps. Dynamic programming is introduced to find an optimal trail on the sub-optimal DNN output map. Experimental results showing accurate trail detection for real-world trail datasets captured with a head mounted vision system are presented.
Highlights
Autonomous navigation in highly unstructured environments like man-made trails in forests or mountains is an extremely challenging problem for robots
In this paper we propose a two-stage pipeline using a combination of deep neural network (DNN) and dynamic programming to detect and follow trails in natural environments
A deep neural network is composed of a series of non-linear processing layers stacked on top of
Summary
Autonomous navigation in highly unstructured environments like man-made trails in forests or mountains is an extremely challenging problem for robots. [17] used a variant of DNN to map an input image to several driving indicators like distance to lane markings, angle of vehicle with respect to the lane etc. The task of trail detection is related to the task of road (lane) detection, majority of methods developed for the latter case rely heavily on road image models that involve several prior knowledge clues, such as presence of expected road markings, road/lane geometry constraints or temporal consistency [5,19] These priors are utilized to cope with occlusions, shadows, under- and over-exposure or glare, i.e., factors that are common in traffic situations, yet are not necessarily relevant for trail detection.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
