Embedded Vision for Self-Driving on Forest Roads

TL;DR

This paper presents RovisLab AMTU, an embedded vision robotic system for autonomous off-road navigation and inspection of forest roads, combining deep learning and traditional computer vision techniques for real-time environment perception.

Contribution

Introduction of a novel autonomous mobile inspection unit with a multi-task DNN and handcrafted vision methods optimized for real-time off-road terrain analysis.

Findings

Successful real-time scene and object segmentation

Effective localization and mapping with combined DNN and handcrafted features

Demonstrated system performance on a test track

Abstract

Forest roads in Romania are unique natural wildlife sites used for recreation by countless tourists. In order to protect and maintain these roads, we propose RovisLab AMTU (Autonomous Mobile Test Unit), which is a robotic system designed to autonomously navigate off-road terrain and inspect if any deforestation or damage occurred along tracked route. AMTU's core component is its embedded vision module, optimized for real-time environment perception. For achieving a high computation speed, we use a learning system to train a multi-task Deep Neural Network (DNN) for scene and instance segmentation of objects, while the keypoints required for simultaneous localization and mapping are calculated using a handcrafted FAST feature detector and the Lucas-Kanade tracking algorithm. Both the DNN and the handcrafted backbone are run in parallel on the GPU of an NVIDIA AGX Xavier board. We show experimental results on the test track of our research facility.