摘要
为了提高传统Mean Shift算法在目标快速运动和被大面积遮挡两种情况下跟踪的效果,对Mean Shift跟踪算法进行了3点改进:采用Kalman滤波器预测运动目标轨迹,以提高算法对快速运动目标的鲁棒性;提出了一种融合Kalman滤波器残差和Bhattacharyya系数的遮挡处理机制,以提高目标被大面积遮挡时的跟踪效果;提出了一种基于自适应更新因子的目标模型更新机制,以提高动态适应能力。对比实验结果表明,改进算法能有效提高在上述两种情况下的跟踪效果,并且在遮挡情况下具有较好的鲁棒性。
To improve the tracking effect of traditional Mean Shift algorithm in fast motion and being blocked in large areas of moving targets,the Mean Shift algorithm was improved in three aspects in this paper. Firstly,a Kalman filter was used to predict the target tracks to improve the robustness to fast targets.Secondly,an occlusion-handling mechanism combining Kalman filter residuals and Bhattacharyya coefficients was put forward to improve the tracking effect of largely-occluded targets. Thirdly,an update mechanism based on an adaptive update factor was proposed to improve the dynamic adaptability of target models. At last,the comparison experiments results showed that the developed algorithm could improve the tracking effect in both cases and had good robustness to occlusion.
出处
《兵器装备工程学报》
CAS
2016年第2期127-130,共4页
Journal of Ordnance Equipment Engineering
