针对车辆跟踪过程中跟踪目标丢失或者失败的情况,提出一种改进型Camshift(Continuously Adaptive Mean Shift)算法和卡尔曼滤波相结合的跟踪方法。首先,利用卡尔曼滤波器实现跟踪目标的位置估计,以克服目标被遮挡造成的跟踪失败的问题,...针对车辆跟踪过程中跟踪目标丢失或者失败的情况,提出一种改进型Camshift(Continuously Adaptive Mean Shift)算法和卡尔曼滤波相结合的跟踪方法。首先,利用卡尔曼滤波器实现跟踪目标的位置估计,以克服目标被遮挡造成的跟踪失败的问题,然后再利用改进型Camshift算法依据目标距离搜索中心的位置,对H分量创建的颜色直方图中的每个像素位进行高斯模型核函数的加权处理,并自适应计算得到最优的搜索窗口,从而改善了传统Camshift不能直接抵制噪声干扰的缺点,解决了因跟踪目标在同色背景噪声干扰下出现的丢失问题。最后通过仿真实验表明:改进型Camshift算法和卡尔曼滤波的结合有效地提高了车辆跟踪的准确性和连续性。展开更多
This paper put forward the super-resolution image algorithm based on Gauss process regression sparse solution. We establish local Gauss process regression model, to solve the feasibility problem of regression super-re...This paper put forward the super-resolution image algorithm based on Gauss process regression sparse solution. We establish local Gauss process regression model, to solve the feasibility problem of regression super-resolution problem in solving Gauss process; further use sparse algorithm, not only it can optimize the super parameter of Gauss kernel function, but also to optimize the initial entry training, so as to obtain more accurate regression Gauss process. Experimental results show that： the paper proposed algorithm can does not reduce the image reconstruction results, and it can reduce the computational complexity.展开更多
文摘针对车辆跟踪过程中跟踪目标丢失或者失败的情况,提出一种改进型Camshift(Continuously Adaptive Mean Shift)算法和卡尔曼滤波相结合的跟踪方法。首先,利用卡尔曼滤波器实现跟踪目标的位置估计,以克服目标被遮挡造成的跟踪失败的问题,然后再利用改进型Camshift算法依据目标距离搜索中心的位置,对H分量创建的颜色直方图中的每个像素位进行高斯模型核函数的加权处理,并自适应计算得到最优的搜索窗口,从而改善了传统Camshift不能直接抵制噪声干扰的缺点,解决了因跟踪目标在同色背景噪声干扰下出现的丢失问题。最后通过仿真实验表明:改进型Camshift算法和卡尔曼滤波的结合有效地提高了车辆跟踪的准确性和连续性。
文摘This paper put forward the super-resolution image algorithm based on Gauss process regression sparse solution. We establish local Gauss process regression model, to solve the feasibility problem of regression super-resolution problem in solving Gauss process; further use sparse algorithm, not only it can optimize the super parameter of Gauss kernel function, but also to optimize the initial entry training, so as to obtain more accurate regression Gauss process. Experimental results show that： the paper proposed algorithm can does not reduce the image reconstruction results, and it can reduce the computational complexity.
