A stereo visual odometry system for material handling scenarios.

The author presents a study based on Visual Odometry (VO) for navigation ofautonomous vehicles. This type of technique has the capability to estimate theego-motion of an agent (e.g., vehicle, human and robot) based on its surrounding environment by acquiring only a stream of images captured by a single or multiple cameras connected to it. The use of this method is related to the fact that images can contain a lot of information (textures, contours, lines etc.) that can be very useful for estimating the motion of the agent on which the camera is fixed. In detail, through the acquired image stream, VO can estimate the motion (translation and rotation) by incrementally examining the changes that motion induces on the images. In this thesis, the author proposes the development of an algorithm, based on the acquisition from a stereo camera (Stereolabs ZED 2i), with the aim of verifying and validating the correct functioning of this technique in a material handling environment. The work was carried out in the R&D department of Toyota Material Handling company based in Bologna