考虑夜间视觉图像误差的人行道识别技术  

Sidewalks Recognition Technology Considering the Night Vision Image Error

作　　者：孟俊仙[1] 陈燕雷 

机构地区：[1]首钢工学院教务处,北京100144 [2]中国移动通信集团设计院有限公司研究所,北京100080

出　　处：《科技通报》2014年第5期164-167,共4页Bulletin of Science and Technology

摘　　要：传统视觉测量方法在夜间环境下,存在误差因素繁多、对比度差、结构相似性差等缺陷,无法准确识别人行道中的行人目标,提出了考虑夜间视觉图像误差的人行道识别方法,塑造目标模型更新机制,根据目标颜色直方图相似度和形状变化信息运算目标变化程度,及时更新夜间人行道上行人目标模型,利用夜间视觉图像误差达到识别目的,将待识别行人图像当成期望图像,对同类运动物体进行注视跟踪,确保实时采样图像收敛于期望图像,由图像反馈以及物体运动自适应补偿构成控制律,完成夜间环境下行人目标的准确识别。仿真实验说明,所提方法可对夜间环境下人行道中行人目标进行准确识别,具有较高的识别效率和鲁棒性。Put forward the consideration on the sidewalks of night vision image error identification method, shape target model update mechanism, based on color histogram similarity and shape change information operation target, update the sidewalk pedestrians at night target model, using the night vision image error for identification purposes, will be to identify the pedestrian image as expected, to look at the similar moving object tracking, to ensure the real-time sampling image converge to expect, made up of image motion adaptive compensation feedback and object control law, the accurate identification of pedestrian targets in night environment. Simulation experiment shows that the proposed method can be to accurately night pedestrians in the pavement under the environment of the target recognition and has high recognition efficiency and robustness.

关 键 词：夜间 视觉图像 误差 人行道识别 

分 类 号：U491.25[交通运输工程—交通运输规划与管理]