Vision-based Navigation System for Autonomous Transportation Vehicle

Autonomous navigation using global positioning system (GPS) or geomagnetic direction sensor (GDS) to detect absolute position is widely used. An alternative system that detects boundary lines and intersections information can be achieved using machine vision based navigation systems. The objectives of this research are the development of the intersection detection method using machine vision and the approaching/turning at the intersection using dead reckoning method. Intersection guidance markers (IGM) were installed on farm road intersections. The method of intersection recognition involved the search to detect the IGM instead of the actual intersection, which would be difficult because the latter changes appearance through the season. Hough Transform and coordinate transformation was used for this purpose. A dead reckoning system for approaching and turning on the intersection was developed. Rotary encoders installed on both sides of the crawler sprocket are used as guidance sensors. RTK-DGPS and fiber optic gyroscope (FOG) were used for evaluating the track of autonomous traveling. The result of field tests (the speed was 0.8 m/s) showed the rms of intersection detection error and the dead reckoning error were 0.12 m at 8.64 Hz processing speed and 0.063 m respectively.