Image Processing, Analysis, and Machine Vision
Image Processing, Analysis, and Machine Vision
| Hi-index | 0.00 |
This pgper deals with methods of creating entities used for a geometric description of robot environment. These entities (primitives) are extracted from scene-depth measurements gathered by g mobile robot sensor system. The overall goal is to achieve efficient data fusion through extraction of specific geometric features - boundaries of obstacles - from sensor data. Although the geometric level of abstraction belongs to low level representations, it offers efficient reduction of data amount. This contribution overviews two different approaches. The first method has been designed for processing probabilistic sensor-based world models which integrate data from multiple sensors gad multiple- sensor positions. The other approgch handles cases of religble ranging when navigating g nearby obstacle. The main features of both methods are discussed with respect to on-line updating of the global world model and simple sensor-based planning and control of the robot. The approaches presented are optimized towards performance robustness and are accompanied by experimental results with the GLbot.