CAJ | 학술논문

针对当前混合高斯背景模型运动目标检测中的更新率小、自适应学习率取值范围过窄、帧间差分引入噪声及背景模型更新没有选择性等缺点，提出了一种混合高斯背景模型运动目标检测的改进算法。在传统混合高斯的基础上，对更新率和方差值进行改进，以适应外界环境的变化；根据像素的连通区域，在线更新学习率和去除噪声；通过引用改进的三帧差分法，实现了对背景模型的选择性更新，减少减弱帧间差分引入噪声以及缓慢运动目标对背景的影响；利用色度信息消除阴影，得到较为精确的运动目标。实验结果表明，改进算法效果良好。
침대당전혼합고사배경모형운동목표검측중적경신솔소、자괄응학습솔취치범위과착、정간차분인입조성급배경모형경신몰유선택성등결점，제출료일충혼합고사배경모형운동목표검측적개진산법。재전통혼합고사적기출상，대경신솔화방차치진행개진，이괄응외계배경적변화；근거상소적련통구역，재선경신학습솔화거제조성；통과인용개진적삼정차분법，실현료대배경모형적선택성경신，감소감약정간차분인입조성이급완만운동목표대배경적영향；이용색도신식소제음영，득도교위정학적운동목표。실험결과표명，개진산법효과량호。
Aiming at the drawbacks of low updating rate, narrow scope of adaptive learning rate,an improved de-tection algorithm based on mixture Gaussian model was proposed. The update rate and the variance value were im-proved based on traditional method,so as to adapt to the change of the external environment. The leaning rate should be updated on-line and noises were suppressed according to the relationship between the pixel and its adjacent pix-els. The selectivity of background model update was realized, the noise caused by the frame differencing was sup-pressed and the influence of slow-moving objects on background update was removed by using the improved three frame difference method. Chromatic difference was employed to eliminate shadows and make the moving object region more accurate. The experimental results show that the improved algorithm has good effect.