真实环境中的视觉搜索是人和动物赖以生存的重要能力。现有理论侧重于探讨影响视觉搜索的自上而下的认知因素，而将自下而上的影响归结为搜索物的影像显著性。但在真实环境中，除了有物体表面投射的影像信息，观察者或观察对象的运动还会产生光流信息。光流信息与运动状态对应，标示了运动主体的物理属性，如大小、质量、弹性系数等。利用光流，观察者可区分、识别物体，进而完成搜索任务。本项目通过三个子课题探索静态影像和动态光流信息如何影响真实环境中的视觉搜索。子课题一研究静止或运动的观察者如何搜索三维物体。子课题二研究静止或运动的观察者如何搜索运动的人。子课题三探索提高搜索效率的训练方法。我们将结合心理物理法、贝叶斯模型、光流量化算法等方法，和虚拟现实、裸眼3D显示等技术，梳理视觉信息在搜索中的作用，提出并验证基于视觉信息和注意引导的理论模型，并应用到对人的视觉搜索干预和对智能搜索的设计中，提升真实环境中的搜索表现。

Visual search is a ubiquitous task and a critical skill for men and animals. Existing studies on visual search mainly focus on attentional guidance and the top-down cognitive influences on search effectiveness. The bottom-up influence on visual search is, rather crudely, simplified as objects’ image saliency. However, when searching in real world, where the observer and/or objects move, both static image information (the saliency of which has been considered in existing search models) and dynamic optic flow information are available. Optic flow is generated by the relative motions between an observer and world objects. So by detecting flow patterns, observers get to know the kinematic properties of events (which is defined as objects in motion) and hence perceive the physical properties of constituent objects, such as the mass, size and frictional coefficient etc.. These physical properties distinguish objects and allow the observer to search for a particular one. We integrate dynamical perceptual information (i.e. optic flow) into existing search models. Specifically, in two studies, we test how combined dynamical and static perceptional information affect visual search for three-dimensional objects and for moving people, when the observer is stationary or moving. Furthermore, we attempt to develop a training protocol that improves search effectiveness in real world. In this project, we will use research methods such as psychophysics, Bayesian estimation, and optic flow quantifying algorithms to formally analyze search behaviors, and apply technologies such as virtual reality and naked eye 3D display, to balance experimental control and ecological validity. Findings from this project will bring forth new theories for understanding visual search in real world, and have direct applications on personnel training and intelligent search designs. Ultimately, these are immensely valuable to science and the society.