CAJ | 학술논문

为了能够检测自动售货机周围的行人数量，提出一种适用于自动售货机的人流量检测系统。该系统首先采用高斯混合模型的运动前景检测方法得到运动物体，并通过形态学处理进一步降低前景的噪声，然后提取运动物体的HOG特征并通过线性SVM分类器判断该物体是否为行人。在从现场采集到的视频上进行实验，结果表明该系统能够适应自动售货机前方的特殊环境，并能满足实时性的要求。
위료능구검측자동수화궤주위적행인수량，제출일충괄용우자동수화궤적인류량검측계통。해계통수선채용고사혼합모형적운동전경검측방법득도운동물체，병통과형태학처리진일보강저전경적조성，연후제취운동물체적HOG특정병통과선성SVM분류기판단해물체시부위행인。재종현장채집도적시빈상진행실험，결과표명해계통능구괄응자동수화궤전방적특수배경，병능만족실시성적요구。
In order to detect the number of pedestrians around the vending machine, puts forward a kind of traffic detection system applicable to the vending machine. Firstly this system gets the moving objects by using the mixture gaussian model and reduces the nosie of the foreground through the morphology processing. Then gain the HOG feature of the moving object through the linear SVM classifier to judge whether the moving object is pedestrians. Finally the results show that the system can adapt to the special environment in front of the vending machine ,and can satisfy the requirement of real-time after testing on the video captured from the vending machine.