基于单目摄像头的空中手写系统
In-Air Handwriting System Based on Monocular Camera
屈喜文 1吴响1
作者信息
- 1. 安徽工业大学计算机科学与技术学院,安徽 马鞍山 243000
- 折叠
摘要
目前基于视觉的空中手写系统主要是基于Leap Motion或Kinect等三维传感器.上述传感器体积太大且价格昂贵,不利于和手机、空调、电视等集成,于是开发了一种新的更利于应用推广的基于单目摄像头的空中手写系统.上述系统基于肤色特征从二维视频数据中检测书写手,然后利用书写手结构特点将书写手与手臂分离,通过寻找书写手轮廓到掌心的最远距离定位书写手指尖,记录连续视频帧中指尖位置形成空中手写字符轨迹,最后将字符送入高性能分类器进行识别.实验结果表明书写者可以利用上述系统自由地空中书写,实时前 5 识别率超过了 99%.
Abstract
At present,vision based in-air handwriting systems are mainly based on 3D sensors such as leap motion or Kinect.Such sensors are bulky and expensive,which is not conducive to integration with mobile phones,air conditioners,televisions,etc.Therefore this paper develops a new in-air handwritting system based on monocular camera,which is more conducive to application promotion.The system detects the writing hand from the two-dimen-sional video data based on the skin color feature,then separates the writing hand from the arm by using the structural characteristics of the writing hand,locates the fingertip by finding the farthest distance from the contour of the writing hand to the centre of the palm,and records the fingertip position in continuous video frames to form the trajectory of in-air handwritten characters.Finally,the formed characters are sent to a high-performance classifier for recognition.The experimental results show that users can use the system to write in the air freely,and the real-time recognition rate of the first five candidates is more than 99%.
关键词
空中手写/单目摄像头/字符识别Key words
In-air handwriting/Monocular camera/Character recognition引用本文复制引用
基金项目
国家自然科学基金青年基金(61906003)
安徽高校协同创新项目(GXXT-2021-004)
出版年
2024
NETL
NSTL
维普
万方数据