CAMSHIFT IMPROVEMENT WITH MEAN-SHIFT SEGMENTATION, REGION GROWING, AND SURF METHOD

Abstract

CAMSHIFT algorithm has been widely used in object tracking. CAMSHIFT utilizes
color features as the model object. Thus, original CAMSHIFT may fail when the object color is
similar with the background color. In this study, we propose CAMSHIFT tracker combined with
mean-shift segmentation, region growing, and SURF in order to improve the tracking accuracy.
The mean-shift segmentation and region growing are applied in object localization phase to extract
the important parts of the object. Hue-distance, saturation, and value are used to calculate the
Bhattacharyya distance to judge whether the tracked object is lost. Once the object is judged lost,
SURF is used to find the lost object, and CAMSHIFT can retrack the object. The Object tracking
system is built with OpenCV. Some measurements of accuracy have done using frame-based
metrics. We use datasets BoBoT (Bonn Benchmark on Tracking) to measure accuracy of the
system. The results demonstrate that CAMSHIFT combined with mean-shift segmentation, region
growing, and SURF method has higher accuracy than the previous methods.