[4Yin2] Interactive session 2

[4Yin2-03] Mapping with Monocular Visual Odometry using Depth Maps

〇Kentaro Taguchi¹ (1.I-NET CORP.)

Keywords:simultaneous localization and mapping, monocular visual odometry, depth map

Visual SLAM, Lidar SLAM have been proposed for simultaneously localization and mapping. Among these, Visual SLAM using a monocular camera does not require special sensors, is lightweight, inexpensive, and suitable for constructing a measurement system, and can be used for mapping. The objective of this research is to make a map of the real-world scale by estimating the scale using depth maps. These depth maps are obtained by convolutional neural network from input images from a monocular camera. The scale and trajectory were estimated using a dataset of vehicles driving outdoors, and the issues of using depth maps for Visual Odometry were discussed.