基于瞳孔定位的单目测距系统  

Monocular vision measurement system based on eye location

作　　者：李贤辉[1] 高盈[1] 钱恭斌[1] 

机构地区：[1]深圳大学信息工程学院,广东深圳518000

出　　处：《智能计算机与应用》2016年第2期70-73,共4页Intelligent Computer and Applications

摘　　要：针对人和机器人在协同工作的过程中,机器人需要对伙伴进行准确的目标定位这个问题,设计了一个以ARM Cotex-A9 i.MX6Q为微处理器处理模块,USB摄像头为采集模块的嵌入式视觉测距系统,实现了单目视觉测距定位。该系统设计大体分为人眼瞳孔定位和单目测距2部分。人眼瞳孔定位通过USB摄像头采集人脸图像,系统对人脸图像进行处理,获取人脸图像中瞳孔的精确位置,获取人脸图像中瞳孔的位置采用Adboost人脸检测算法和灰度投影人眼定位算法;测距采用小孔成像三角形相似原理,根据实际人眼瞳距和图像中人眼瞳距信息计算出目标人物到摄像机镜头间的距离,其中摄像机的内参数标定借助于Matlab工具箱来完成。实验结果表明,该测距定位系统测距精度较高,系统鲁棒性好,具有良好的应用前景。The target location problem of people-robot cooperation is discussed. A monocular vision measurement system with ARM CotexA9 i. MX6 Q microprocessor as the main processing module and the USB camera as the image capture module is presented. The system design includes two parts that eye location and measurement. The system capture the image of human face by USB camera,and then process the image of human face by microprocessor to get eye＇s location in the image. The system use the Adboost and eye location based on gray projection to achieve the eye location,and the distance measurement according to the principle of small hole imaging. The intrinsic camera parameter calibration is performed with Matlab toolbox. The experiments show that the system is high in precision and robust.

关 键 词：单目视觉 视觉定位 嵌入式图像处理 人眼定位 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]