Inertial-Kinect Fusion for Outdoor 3D Navigation

The lightweight and low-cost 3D sensing device, such as Microsoft Kinect, has gained much attention in computer vision and robotics community. Although quite promising and successful for indoor applications, its outdoor usage has been significantly hampered by its short detection range (around 4 meters) coupled with ambient infrared interference. This paper addresses the theoretical and practical development of an Inertial-Kinect fused SLAM framework that can handle the 3D to 2D degeneration in Kinect sensing, called a depth dropout problem. The vision node is designed to provide either full 6DOF or partial 5DOF vehicle pose measurements depending on the depth availability, whilst the low-cost inertial system designed in house (less than $40AUD) enables continuous metric mapping and navigation. Indoor and outdoor experiment results are provided, demonstrating the robustness of the proposed approach in a challenging environmental conditions.