利用自适应组合模型实现车辆跟踪

针对视频序列图像目标车辆跟踪中经常因场景光照变化、目标旋转、遮挡等因素导致丢失问题,提出了基于颜色自适应的改进CamShift算法;通过建立凸函数组合模型,利用多目标规划最优求解算法获取自适应颜色识别最佳组合,提高了算法抗干扰能力;利用目标倾角预测识别目标发生形变和旋转,构造多变量状态信息预测目标发生遮挡和瞬间消失,并通过IIR滤波器快速预测目标在下一时刻的运动方式。实验表明,本算法跟踪精度高,鲁棒性强。

Vehicle Tracking Implemented by Auto-adaptive Combination Model

Aiming at problems caused by illumination changing of the scene, goal revoking and masking during the video sequence image target tracking, the paper proposes the improved color auto-adaptive CamShift algorithm. Through establishing the convex function combination model, the multi-objective programming optimal algorithm is used to gain the auto-adaptive color and to recognize the best combination. The anti-jamming ability of the algorithm is improved. The target dip is used to forecast the target＇ s deformation and revolving. The muhivariable status messages are constructed to forecast the masking and instantaneous vanishing of the goal. IIR filter is employed to forecast quickly mode of motion of goal in next moment. The massive experiments indicate that this algorithm has a high tracking accuracy and strong robust.