为了避免群死、群伤事故,必须对施工升降机轿厢内人数进行监控。结合升降机的工作特点,本文利用安全帽的颜色特征进行静态图像人数统计。考虑升降机轿厢空间小、安全帽图像容易产生粘连、因人货同梯导致图像背景复杂等问题,利用目标区域最小外接矩形的面积、质心位置以及占空比、长宽比等多种几何特征融合处理,解决了粘连图像的人数识别以及干扰背景的分离。现场测试结果表明,本算法对复杂背景粘连图像的人数识别准确率较高,实时性好,基本满足工程应用的需要。%In order to avoid the death or injury in group,it is necessary to supervise and count the number of people in the elevator.Combining with the elevators'working characteristics,this pa-per utilizes the color features of safety helmets to count the number of people.Considering that the space in the elevator is small and the images of safety helmets are easy to produce image over-lapping problems,and meanwhile the loading of people and groceries can lead to the image back-ground complicated overlapping problems,this paper utilizes the multi-source information fusion to solve such problems as the square of the minimum external rectangle,the location of the cen-troid,duty cycle,and the length-width ratio.The result of experiment on the site shows that this method improves accuracy,has an excellent real-time quality and can be applied to the engineer-ing applications well.
展开▼