基于改进型混合高斯模型的嵌入式实时运动检测系统

Embedded Real-Time Detection System of Moving Object Based on Improved Gaussian Mixed Model

混合高斯模型由于其计算量大,算法结构复杂,难以在嵌入式系统中实现运动物体的实时检测,为解决此问题,文中提出了一种基于改进型混合高斯模型的实时运动检测方案,对混合高斯模型进行简化和结构调整,同时进行了C语言层面和CPU层级的优化,使其更合适于嵌入式平台,并详细分析了DM6446平台的软硬件设计,介绍了该算法在DM6446平台上的实现过程;实验结果表明:该系统能够有效克服外界环境变化带来的干扰,能够实时检测,可以实现多目标跟踪。

It is difficult to achieve real-time detection of moving object on embedded CPU,because Mixed Gauss model is large amount of computation and its structure is complex,so it proposes a real-time embedded detection scheme based on Mixed Gauss Model,which simplifies the model.The optimized algorithm on C code level and CPU level has made the model more suitable for embedded CPU,also introduces the realization of the algorithm on the DM6446 platform and the system design of software and hardware.The experimental results show that the system can effectively overcome the interference of external environment change and realize real-time detection and multiple target tracking.