光传感器监控系统中行人识别模型的研究  

Research on pedestrian identification model in optical sensor monitoring system

作　　者：简远鸣[1] 嵇建波[1] JIAN Yuanming;JI Jianbo(Guilin University of Aerospace Technology,Guilin Guangxi 541004,China)

机构地区：[1]桂林航天工业学院,广西桂林541004

出　　处：《激光杂志》2020年第3期82-85,共4页Laser Journal

基　　金：广西自然科学基金项目(No.2018GXNSFAA281161)。

摘　　要：针对传统的行人识别模型行人识别速度慢的问题,设计了一种光传感器监控系统中行人识别模型。引入抛物线拟合,将三维场景中的点阵聚合成现实场景中的物体,对监控区域三维重建,在此基础上,采用野点去除和噪声去除的方法去除外界因素对行人目标造成的高频干扰,采集行人区域数据。采用红外图像分割方法,分别对行人目标与背景建模,最后将行人目标模型与背景模型结合生成行人识别模型。实验对比结果表明,此次设计的光传感器监控系统中行人识别模型比传统的图像阈值分割模型、激光雷达行人识别模型、模板匹配行人识别这三种传统模型行人识别速度快,能够满足行人识别需求。Aiming at the slow speed of traditional pedestrian recognition model,a pedestrian recognition model in optical sensor monitoring system is designed.The parabola fitting is introduced to aggregate the lattice in the three-dimensional scene into the object in the real scene,and the three-dimensional reconstruction of the monitoring area is carried out.On this basis,the high-frequency interference caused by the external factors on the pedestrian target is removed by the methods of wild point removal and noise removal,and the data of the pedestrian area is collected.The method of infrared image segmentation is used to model pedestrian target and background respectively.Finally,pedestrian recognition model is generated by combining pedestrian target model and background model.The experimental results show that the pedestrian recognition model of the designed optical sensor monitoring system is faster than the traditional image threshold segmentation model,lidar pedestrian recognition model and template matching pedestrian recognition model,which can meet the needs of pedestrian recognition.

关 键 词：光传感器 监控系统 行人识别 区域 

分 类 号：TN971[电子电信—信号与信息处理]