CAJ | 학술논문

针对混合高斯模型存在的不足，提出一种改进的基于混合高斯模型的背景建模方法。为每一个背景像素建立多维混合高斯模型，融入三帧差分法实时判定背景区域和运动区域，为背景显露区中的像素点选择较大的更新频率，并通过实验进行验证分析。实验结果表明：该方法相比高斯背景建模具有更好的环境适应性和鲁棒性，能消除背景中的“鬼影”，适用于实时的目标检测和跟踪。
침대혼합고사모형존재적불족，제출일충개진적기우혼합고사모형적배경건모방법。위매일개배경상소건립다유혼합고사모형，융입삼정차분법실시판정배경구역화운동구역，위배경현로구중적상소점선택교대적경신빈솔，병통과실험진행험증분석。실험결과표명：해방법상비고사배경건모구유경호적배경괄응성화로봉성，능소제배경중적“귀영”，괄용우실시적목표검측화근종。
Aiming at the shortage of Gaussian mixture model, an improved background modeling method based on Gaussian mixture model is put forwards. It constructed a mixture Gaussian model for each background pixel, the three-frame differencing is used to judge background area and movement area in real time. Choose larger upgrade frequency for pixel point in background display area. Then verify and analyze it by test. The test results show that, compared with Gaussian background model, the method has better environment adaptability and robustness, it can eliminate“shadow”in background and can be use for object detection and tracing in real time.