Based on the analysis of partial microscopic moving characteristics of the internal smoke, a kind of smoke moving direction extracting algorithm is designed in this paper. The algorithm adopts Meanshift kernel function histogram to set up the model to describe the smoke moving characteristic and uses the Bhattachyarya coefficient to estimate the smoke block direction of the motion vector. In order to reduce the noise and the interference of rigid body, the algorithm counts moving data of each data block by the appropriate time window. This algorithm estimates the each smoke suspected area proportion in main movement and finally discriminates the smoke motion direction. The experimental results show that this algorithm performances well at extracting the smoke movement direction and reduces average searching point and improves searching precision, the features under a highly interference condition caused by thick black smoke and body moving, at the same time, it has a better ability to eliminate interference and provides a criterion for further smoke discrimination algorithm research.%通过烟雾运动分析发现在烟雾内部会呈现出局部小运动特征，基于此提出了一种新颖的烟雾运动方向快速提取算法。算法通过Meanshift核函数直方图构建烟雾运动特征模型，并利用Bhattachyarya系数值估计烟雾块的运动矢量方向。同时，为了抑制噪声和刚性物体的干扰，设计了时间窗对每个数据块的运动进行统计，通过估算出每个烟雾疑似区域主运动所占的比率实现对烟雾内部运动方向的识别。实验结果表明，该算法很好地提取了烟雾的运动方向特征，降低了运动估计的运算复杂度和提高搜索精度，并对干扰有很强的排除能力，为进一步的烟雾的识别算法研究提供了判据。
展开▼
