中文摘要：

提出一种视频中群体状态演进的预报方法。该方法在连续介质流体动力学模型格子Boltzmann模型的基础上增加表现群体运动目的驱使项,使该模型更能描述高密度人群粒子向目标位置聚集的特点。模型还能预报高密度人群场景的风险最高位置。由于该模型只需输入视频初始时某帧的速度场,就可演进出场景未来的状态,所以,该方法是对高密度人群视频未来状态的预报方法。实验证明,该模型预报的速度、密度场准确,定位的场景最高风险位置也与场景本身的最高风险位置接近。

英文摘要：

An addition which represent the people＇s purpose drive is added to the lattice Boltzmann model,which is also known as a continuous fluid dynamic model. With the addition,particles will move to its destination more precisely. Besides that,the model can find the most risky position of a crowd video scene. As the model can estimate the future state of a crowd with only the first frame as input,the model is considered to be a forecasting method for crowd state. Experiments show that the future velocity and density fields can be estimated with great accuracy and the risky position found by the proposed model,which is close to the real risky position of crowd scene.