基于Kalman滤波和边缘直方图的实时目标跟踪

Edge-color-histogram and Kalman filter-based real-time object tracking

为提高光照变化和背景混淆下的实时跟踪效果,提出了基于Kalman滤波和边缘直方图特征(edge orientationhistogram,EOH)的实时目标跟踪方法。边缘方向直方图对光照变化不敏感,且对颜色变化依赖程度低,适合背景混淆和光照变化场合下跟踪。Kalman滤波用于预测目标位置和角度,通过预测调整边缘方向直方图的计算得以快速找到跟踪目标,提高边缘方向直方图的匹配能力。经过实际复杂背景和光照场景下与颜色直方图的多组对比测试,颜色直方图很快会失去跟踪目标,EOH能快速正确跟踪,说明EOH和Kalman滤波能在这种场景下快速正确跟踪目标。

An objeet tracking algorithm based on Kalman filter and edge orientation histogram （EOH） was used to improve the tracking performance with scene illumination variations or complicated backgrounds. Since the EOH is not sensitive to light or color variation, this method works well when tracking object in such cases. The Kalman filter is used to estimate the new position and angle and then a new estimate of the EOH is calculated to quickly and efficiently match the target. Tests show that the color histogram method easily losts objects while the EOH follows them very well. The tests show that the EOH and the Kalman filter can track objects in real-time when the illumination varies or background clutter is present.