【作者】 孟盼盼；

【导师】 孔凡让；

【作者基本信息】 中国科学技术大学 ， 机械电子工程， 2011， 硕士

【摘要】 人脸肖像绘制机器人是一种可以自动绘制人脸肖像轮廓图的智能机械系统,它由图像采集模块、图像处理模块、机械控制绘图模块组成,能够自动拍摄人脸照片,提取肖像轮廓,并控制机械臂在画板上画出人脸线条画。人脸肖像绘制机器人是机器视觉的研究方向之一,可广泛用于科普展览,其中提出的基于机器视觉的研究技术在生产和生活等各个方面都有着广泛的应用。本文主要的研究内容和成果有:研究了图像采集系统的硬件和软件算法,通过对CCD摄像机、光学镜头、图像采集卡的选型和拍摄算法的设计,完成自动采集和保存人脸肖像照片的功能。在对基于肤色的图像分割技术研究的基础上,对比目前常用的色彩空间的原理和特点,采用YCbCr色彩空间进行图像分解,并提出了与迭代阈值法相结合的图像分割方法进行肤色分割,研究了图像轮廓提取的一般方法,并使用内部掏空法提取出人脸轮廓。研究了人脸五官特征提取的一般方法,提出一种基于搜索的投影法对眼睛和眉毛进行坐标定位,研究去除噪声提取眼睛和眉毛的搜索算法,优化处理人脸轮廓图,对改善人脸肖像图的整体效果达到较好效果。提出一种基于无干扰八领域搜索的矢量化算法,将人脸轮廓像素点转化为一系列相连的点,再检测每组相连的点中位于一条直线上的点,保存线段首位点坐标到数组中,为与机械臂控制器通信做好准备。研究绘图机械控制系统的硬件选型和控制算法,解决了外部坐标值与控制器内部变量通信的技术难点,在Visual C++ 6.0中实现了外部对机械臂绘图动作的自动控制,设计机械臂绘画动作流程,完成人脸轮廓图的自动绘制。更多还原

【Abstract】 Auto-portrait robot is a mechanical system which could draw face portrait automatically. Based on Ycbcr color space and iterative threshold method, we propose a face detection algorithm for extracting contour lines from color face images. By converting pixels of contour lines into vectorized points, the robot controller receives vectorization information and completes portrait drawing.Experimental results demonstrate successful contour extracting and portrait drawing over a wide range of facial variations.Face detection is the most crucial technology in an auto-portrait robot, and it plays an important role in applications such as video surveillance, human computer interface, face recognition, and face image database management. Although many different algorithms exist to perform face detection, each has its own weaknesses and strengths. These algorithms are computationally expensive. So in this paper we proposed a fast algorithm to increase the speed of face detection. The primary research contents and results are as follows:We researched on hardware of image acquisition system and algorithms of shooting human face images.We reviewed skin color method and iterative threshold method, and compare different color models with each other. We extracted external-contour of human body and clothes with an algorithm basing on Ycbcr color space and iterative threshold method.To detect eyes and eyebrows, we propose a shadow casting technique algorithm based on searching, it has a good effect to the whole image of human face boundary.We research on vectorization method based on 8 points neighborhood without interference in order to make robot controller understand contour information.In section 5 we propose the whole process that robot arm follows for drawing contour lines. We solved the most crucial problem of communication between coordinate outside and controller variables inside. In Visual C++ 6.0, we realized automatically control program of painting robot.更多还原