Mapping Of Forestry Mobile Robots Based On Binocular Stereo Vision

Autonomous mobility is the primary issue to be concerned for forestry robots that are to be applied in autonomous operations in forests and is an inevitable trend of development on them as well. Nevertheless, there are two fundamental problems to implement autonomous mobility of forestry robots: self-localization and concurrent mapping. This dissertation focused on research on mapping of forestry mobile robots based on binocular stereo vision, and it mainly included establishment of the Voyager-Ⅱ tracked autonomous mobile robot system equipped with the Bumblebee(?)2 binocular stereo vision, realization of self-localization for the Voyager-Ⅱ Robot based on binocular stereo vision, and implementation of both indoor and outdoor environment map building.The main research contents and related results are as follows:1. Using the Visual Studio 2010 Integrated Development Environment, the Motion Estimation Software System was designed and also developed through the C++ Programming Language to establish both communication and control between the Voyager-Ⅱ Robot and the airborne notebook computer, connect the Bumblebee(?)2 binocular stereo cameras, grab and show stereo images in real time, execute the stereo vision processing, record journal logs, and implement the concrete self-localization and concurrent mapping algorithms.2. For the self-localization of the Voyager-Ⅱ robot, a new approach of based on binocular stereo vision was proposed. Utilizing the Bumblebee(?)2 binocular stereo cameras, it is possible to exploit the nonholonomic constraints of tracked vehicles in order to promote a restrictive motion model which allows us to parameterize the motion with only 1 feature correspondence. Then, combine RANdom SAmple Consensus (RANSAC) framework plus Kalman Filter (KF) through using the available prior probabilistic information from the KF in the RANSAC model hypothesis stage. Thus, recover parameters of the above motion model with RANSAC integrated into KF to localize the Voyager-Ⅱ Robot. Experiments were performed in both indoor and outdoor environments and related results indicated that the proposed approach was respectively effective, and lived up to the real-time processing speed of 30 frames per second.3. The task of building environment map was accomplished based on the Bumblebee(?)2 binocular stereo cameras, consisting of three-dimensional measurement and mapping. The three-dimensional measurement part was responsible for pre-processing of stereo images, obtaining the depth image from pre-processed stereo images and corresponded error analysis for binocular stereo vision measurement. The mapping part was mainly about the concrete environment map building approach including two-dimensional grid map and three-dimensional environment map. Experiments on the Voyager-Ⅱ Robot were performed in both indoor and outdoor environments and related results indicated that the proposed approach was respectively effective and robust.