作者: Shyam Adhikari , Changju Yang , Krzysztof Slot , Hyongsuk Kim
DOI: 10.3390/S18010178
关键词:
摘要: 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 mountains, using combination deep neural network dynamic programming. The (DNN) concept has recently emerged as very effective tool for processing sensor signals. A patch-based DNN is trained with supervised data classify fixed-size image patches into “trail” “non-trail” categories, reshaped fully convolutional architecture produce trail segmentation map arbitrary-sized input images. As non-trail do not exhibit clearly defined shapes or forms, classifier prone misclassification, produces sub-optimal maps. Dynamic programming introduced find an optimal on output map. Experimental results showing accurate detection real-world datasets captured head mounted system are presented.