一种基于视频运动分析的智能码流方法

采用迭代重加权最小二乘（IRLS）方法计算连续两帧图像的光流，根据光流计算的结果估计视频监控场景的运动状态，将场景分为静止、一般运动、剧烈运动，并分别制定不同的码流存储策略，实验表明，该方法能对监控场景的常见运动行为识别并分类，具有一定的鲁棒性，应用该方法可以减少冗余视频帧对存储资源的占用，扩充敏感监控数据的存储能力。

An intelligent stream method based on vedio motion analysis

Iteratively reweighted least squares (IRLS) method is used to calculate two consecutive image optical flow, According to the results, estimated the motion state of the video surveillance scenes, and divide them into static, general sports and strenuous exercise. Then different bit stream storage strategy is used to save the different scenes. Experiments show that the method can recognize and classify the common motor behavior of the monitoring scene with robustness. This method can reduce the redundant video frame storage resource occupation, and expand the storage capacity of sensitive video data .