Multi locus estimation from local search space using distributed and cooperative processing

在本项目中，从以下三个方面对基于分布式协作处理的多目标轨迹推定问题进行了研究：1）开发基于通过动态分配的观测推定缩小规则的摄像机资源技术：将多目标轨迹问题与多摄像机资源分配问题结合起来，并在不完全约束充分性问题的框架下研究了求解该问题的技术。因此，利用多个摄像机捕获（目标轨迹推定）移动对象和分配（资源分配）摄像机成为可能。2）多目标轨迹推定问题应用的开发：人体跟踪是将多目标轨迹推定问题应用于实际世界的应用。然后，在这个项目中，我们研究了以下两点。2.1）这是从多个摄像机的人脸区域检测系统的开发：其中基于投票的人脸检测算法是使用多个摄像机稳定的。2.2）根据多个摄像机的合作处理的人检测系统的开发：通过多个摄像机之间的协作，统一了由时滞量获得的运动区域信息，并引入了时序约束，提出了一种稳定、高效的运动人体检测算法，并进行了静态系统的实现。

In this project, examination was performed from the following three sides about the multi-object locus presumption problem by distributed cooperation processing1)Development of the technique of being the camera resources based on the rule narrowed down with presumption for observation assigned dynamically : The multi-object locus problem was caught with the resources assignment problem of multi cameras, And the technique of solving this problem by the framework of an incomplete restrictions sufficiency problem was examined. Consequently, capture (objective locus presumption) with multi cameras of a move object and assignment (resources assignment) of a camera became possible.2) Development of multi-object locus presumption problem application : Human tracking is an application which applies a multi-object locus presumption problem to the actual world. Then, in this project we examined the following two points.2.1) It is development of the face region detection system from multi cameras : The voting based face detection algorithm of which is stabilized using multi cameras.2.2) Development of the person detection system according to cooperation processing of multi cameras : By cooperating, and unifying the moving region information acquored by the amount of time lag among multi cameras, and also introducing the time sequence restriction, the algorithm which detect a move person stably and efficiently was proposed, and the still actual system was also performed.