Abstract
In path planning of autonomous agricultural vehicles, obstacles detection and identification, and taking appropriate collision avoidance measures are critical for safe operation. As such, the objectives of this research are to detect the hay bales in an image, and to estimate the position of a hay bale with an omnidirectional camera. Approximately 90% of bales within the 40-m maximum detection distance could be identified using developed image processing technique. Then, the estimation of hay bale's position was further improved based on local discriminate analysis. The estimation of the hay bales distance from the camera generated a root mean squares (RMS) error of 1.1m; it was concluded that this accuracy was more than sufficient to enable the autonomous navigation of a robot tractor in the meadow.
