Moving cluster classification technique with lidar traffic monitoring application
Ka C. Cheok, Shinichi Nishizawa, Walt Young · 1998
A number of methods exist for grouping static data points into clusters. Very few methods, however, consider clustering techniques for grouping dynamic or moving data points. For the purpose of classifying dynamic data into sets of moving clusters, it is necessary to consider the dynamic states, such as position, velocity, acceleration, rotation, etc., of the data. The clustering algorithm must correctly classify the clusters, even if the moving data clusters cross paths and intersect with each other. The paper presents a dynamic clustering method that has been successfully developed and applied to moving laser radar data for an on-board automobile traffic monitoring application. The DCM employs multiple Kalman filters to track the dynamic states of the data points, and a cluster classification and predictor algorithm to identify objects in the information.