Landmark Extraction using Corner Detection and k-Means Clustering for Autonomous Leader-Follower Caravan

A.B. Nevin and D.M. Bevly (USA)


computer vision, landmark extraction, kmeans clustering, corner detection


Extracting a landmark in a digital video sequence is vi tal for visual navigation. For the application of a leader follower caravan, a database of man-made or natural land marks is created. This research focuses on the challenge of the follower-vehicle losing line-of-sight with the leader ve hicle. An experiment is conducted using data in an outdoor environment to test an algorithm based on corner detection and classifying them into subsets using k-means clustering techniques. Assuming all corners from a landmark will be grouped into the same cluster, the centroid of each cluster is established as a region of interest and is stored in an out put file. From the results, it is concluded that the optimal parameters for the algorithm are to extract the ten strongest features and cluster them into three subsets. These param eters yield a success rate of 98%. All regions of interest are used to build a database of visual landmarks or objects for the follower vehicle to use for visual navigation.

Important Links:

Go Back