Multi-Dimensional K-Means Image Segmentation for Off-Road Autonomous Navigation

E.M. Amendt and F.W. DePiero (USA)


Digital Image Processing, Image Segmentation, Image Sequence Processing, K-means, Autonomous Ground Vehicle


This paper examines a single vision camera system used for off-road environment autonomous navigation. It develops an adaptive road segmentation algorithm based on the multi-dimensional properties of the k-means algorithm and incorporates temporal data to increase algorithm stability. Algorithm stability is optimized by minimizing variations in segmentation from frame to frame. A metric to quantify these transitions is also defined.

Important Links:

Go Back