摘要
针对电力设施智能监控系统中目标人员及其穿戴特征,本文提出了一种高鲁棒性的检测方法。首先采集行人的目标正样本,以及所要监控的场景中无行人目标的负样本,接着基于HOG(Histogram of Oriented Gradients)特征以及线性SVM(Support Vector Machine)训练行人检测的分类器,检测进入场景中的行人目标;对检测到的行人的头部、腰部区域,采用HSV色彩模型判断人员是否进行安全着装,及时预警,避免危险发生。本文将目标识别及穿戴特征分析技术应用于电力设施作业现场监控中,可以更准确的识别入侵的人员,也可以在应急事故抢修时,加强工作人员规范操作,提前预警,避免危险。
In order to detect pedestrian and his wear features in intelligent surveillance video of power equipment, a highly ro- bust detection method was proposed. Firstly, collecting the positive samples and negative samples of pedestrians; secondly, HOG （Histogram of Oriented Gradients） features combined with linear SVM （Support Vector Machine） classifier were adopted for classifier training, and then the pedestrians who entered the scene were detected; thirdly, the head and middle region were scanned based on HSV color model for safety wearing. This paper not only achieves the pedestrian detection but also investi- gates the wear feature. It is useful for invader detection and keeping the staff safe in power equipment repairing.
出处
《电子设计工程》
2015年第10期68-71,共4页
Electronic Design Engineering
基金
国家自然科学基金资助项目(9092003)
关键词
电力设施
监控
检测
行人
安全帽
power equipment
surveillance
detection
pedestrian
safety helmet
