语义与统计相结合的智能行为分析关键技术研究  

Key Technology Study on Intelligent Action Analysis of Integrated Semantics and Statistics

作　　者：吴月娥[1] 边后琴[1] 

机构地区：[1]上海电力学院,上海200090

出　　处：《现代电子技术》2009年第22期123-125,129,共4页Modern Electronics Technique

基　　金：上海市重点学科建设资助项目(P1303);上海高校选拔优青教师科研专项基金资助项目(Z-2009-03)

摘　　要：智能行为分析是计算机视觉领域的核心问题,行为分析的最大难点在于表示的基本元素跨度极大,而且物体的图像又是千变万化,因此在智能行为分析系统中必须结合语义与统计信息。针对建立智能行为分析系统的需求,首先建立既能表达用户语义,又能表达千变万化的场景和物体统计信息的"图像语法";然后通过将学习过程理解为迭代投影过程,建立一个统一的机器学习方法,实现较复杂的场景和行为的分析和识别。Intelligent action analysis is the key problem in the area of computer vision. The biggest problem of action analysis is the unlimited span of the basic elements represented and the ever - changing of the images of objects. So in the system of intelligent action analysis, semantic and statistic information should be integrated. Constructing the image grammar for requirement of the intelligent action analysis system, which can represent not only the user semantics, but also the everchanging scene and statistic information of object. Then constructing uniform machine learning method by learning process to be interpreted as iterative projection process. Finally complex scenes and actions are analyzed and recognized.

关 键 词：智能视频监控 语义信息 统计学习 图像语法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]