High-robustness vehicle collision detection method

The invention discloses a high-robustness vehicle collision detection method. The method comprises the following steps: optimizing kernel function shape parameters a and b; the method comprises the following steps: inputting a video frame, and carrying out noise reduction preprocessing by using a 1Shot-MaxPol algorithm; a Farneback algorithm is adopted to obtain an optical flow field {vxi, vyj}; calculating an interaction field F (x, y) by adopting a formula; and accumulating the area difference St of the positive and negative regions of the F (x, y), obtaining an abnormal value through quadratic regression, and judging that collision occurs. 一种高鲁棒性的车辆碰撞检测方法，包括以下步骤：优化核函数形状参数a和b；输入视频帧，并以1Shot-MaxPol算法进行降噪预处理；采用Farneback算法得到光流场{νxi,νyj}；采用公式计算相互作用场F(x,y)；累加F(x,y)的正负区域的面积差St，通过二次回归得到异常值当判定为碰撞发生。