CAJ | 학술논문

利用混合高斯模型描述序列图像背景是地面复杂背景下红外搜索跟踪系统中的关键技术.针对背景更新算法的收敛性和估计准确性的问题,提出了一种自适应混合高斯模型的背景更新框架.该背景更新框架通过建立混合模型迭代过程中学习因子与时间、模型状态的关系并且自适应选择混合模型中高斯分布的个数,有效地提高了背景估计的精确程度和算法的收敛速度.通过真实序列图像详细分析本文背景更新框架性能,同时研究设计了基于 TMS320C6678DSP 芯片的背景更新图像处理系统,实现了视频的传输、处理等系列功能,并与其他算法作比较,从数据分析和实验结果等方面表明本文算法的优越性.
이용혼합고사모형묘술서렬도상배경시지면복잡배경하홍외수색근종계통중적관건기술.침대배경경신산법적수렴성화고계준학성적문제,제출료일충자괄응혼합고사모형적배경경신광가.해배경경신광가통과건립혼합모형질대과정중학습인자여시간、모형상태적관계병차자괄응선택혼합모형중고사분포적개수,유효지제고료배경고계적정학정도화산법적수렴속도.통과진실서렬도상상세분석본문배경경신광가성능,동시연구설계료기우 TMS320C6678DSP 심편적배경경신도상처리계통,실현료시빈적전수、처리등계렬공능,병여기타산법작비교,종수거분석화실험결과등방면표명본문산법적우월성.
It is a key technology in infrared target searching via Gaussian mixture model to describe the background image sequences. In this paper, an adaptive Gaussian mixture model for background updating framework is proposed to solve the problem of the convergence and accuracy. Based on the framework, the adaptive Gaussian mixture model number and the relationship among learning factor, time and model state are built to improve the efficiency and robustness. This paper realized the design of the adaptive background updating model based on C6678DSP chip with video transmission, processing and other family functions. Experimental evaluation results for the real image sequences validate the upgraded estimation capability of the proposed method compared with other algorithms.