High frame-rate tracking of multiple color-patterned objects

摘要

In this study, we develop a high frame-rate vision system that can execute color histogram-based tracking of multiple color-patterned objects in a 512 x 512 image at 2,000 fps by implementing an expanded cell-based labeling algorithm as the hardware logic. In the hardware implementation of the expanded cell-based labeling algorithm, the 16-bin hue-based color histograms of 1,024 color-patterned objects in an image can be extracted simultaneously by dividing the image into 8 9 8 cells concurrently, after calculating the 0th, 1st, and 2nd moment features to obtain the positions, areas, and orientation angles of multiple objects. We verified the effectiveness of our developed tracking system by performing several experiments using multiple color-patterned objects, which were always tracked even when they moved rapidly with occlusions in the camera views.