CAJ | 학술논문

针对智能监控中运动目标检测的问题，提出了基于Davinci—DM6467的高斯混合模型像素级的背景分割策略。对彩色图像建立高斯混合模型，根据场景中象素点的稳定性来调整模型参数的更新速率；通过和马氏阈值进行对比来判断是不是要更新背景模型；通过和背景阈值进行对比来判断哪几个模型是属于背景区域。经验证性实验测试，结果表明，高斯混合模型在运动检测中实时性好，对环境有较强的鲁棒性。
침대지능감공중운동목표검측적문제，제출료기우Davinci—DM6467적고사혼합모형상소급적배경분할책략。대채색도상건립고사혼합모형，근거장경중상소점적은정성래조정모형삼수적경신속솔；통과화마씨역치진행대비래판단시불시요경신배경모형；통과화배경역치진행대비래판단나궤개모형시속우배경구역。경험증성실험측시，결과표명，고사혼합모형재운동검측중실시성호，대배경유교강적로봉성。
In the light of movement target detection during intelligence monitoring, the author put forward Gaussian mixture model of pixel level background segmentation strategy based on Davinei-DM6467. Firstly establishing Gaussian mixture model for colorful images and then adjusting the updating velocity of model parameters according to the stability of each pixels in frames ; secondly comparing them with Markov threshold to judge whether to update the background model; finally comparing them with back-ground threshold to judge which several models belong to background region. Experimental results show that Gaussian mixture model in motion detection is possessed of good real-time and strong robustness to environment.