帧间逼近法用于大噪声红外图像的运动检测

提出一种基于像素邻域的切比雪夫逼近的方法实现运动检测。文中假定红外视频序列背景在相邻帧之间在区域上存在线性关系，由于噪声存在，这种线性方程不相容。通过在Banach空间上建立切比雪夫逼近，求出两帧之间的逼近系数，对逼近系数设定阈值，实现复杂背景的运动物体探测。通过对雨雪天气的两组数据仿真实验，并与简单帧间差模型在同等阈值下比较噪信比，发现该法鲁棒性显著提高。实验表明，该法在提取出主体运动同时，能够有效抑雨、雪等运动背景构成的噪声，具有一定应用价值。

Motion Detection by Temporal Approach Method in Large Noise Infrared Image

To detect salient motion in a snow or rain weather in outer-door condition,and avoid the large noise,this issue present a new method as temporal approach in infra video images.Assume there exist a linear relationship between difference temporal frames in video queue,and the equations are inconsistent because of the attending of noise.Through Chebyshev approximation in Banach space,coefficients be get to the equations.By setting adequate threshold on the solution coefficients,we can not only detect the salient motion but eliminate the noise motion.Exert the algorithm to real situation,the experiment result shows that the method is more precise and better robustness than simple temporal subtraction.