为满足嵌入式系统对图像处理的实时性、可靠性,满足工程对图像质量的需求,提出了一种基于GMM的动态物体提取方法.首先详细介绍了混合高斯模型、去阴影模型、平滑模型的算法原理和实现方法,然后针对DM6437评估板硬件的特点,阐述了DSP算法设计与实现的优化技术,编程实现算法功能,按照要求整合算法,在不影响视频画面质量的前提下,给出了通过优化程序结构和简化模块运算的方法,并对其进行浮点转定点等相关平台移植操作,最终在DM6437上实现了混合高斯模型实时动态物体提取算法.实验结果表明,该算法运算速度快,处理实时性强,图像处理效果明显,具有广泛的实用价值.%In order to satisfy the embedded system to the imagery processing timeliness, the reliability, satisfies the project to the picture quality demand, a dynamic object extraction method based on the GMM is proposed. First introduced in detail mixes the Gauss model, the shadow remove model, the smooth model algorithm principle and realization method, then in view of the DM6437 appraisal board hardware's characteristic, elaborated the DSP algorithm design with the optimization techniques which realizes, the programming realizes the algorithm function, according to request conformity algorithm, in keeping the video frequency picture quality under the premise, has given through the optimization program structure and the simplified module operation method, and carries on the floating point to transfer the fixed point and so on to be related the platform transplant operation, realized finally on DM6437 mixed the Gauss model real-time dynamic object extraction algorithm. The experimental result indicated that this algorithm operating speed is quick, processing timely, the imagery processing effect is obvious, has the widespread use value.
展开▼
