摘要
在激光三维测量系统中,应用人工神经网络技术实现图像坐标系到世界坐标系的映射,修正系统的非线性误差。在线状激光三维扫描技术中,用面阵 CCD 摄像机摄取图像,用前馈型人工神经网络对二值化后的图像进行变换,可直接得到物体轮廓线的二维坐标;在点状激光三维扫描技术中,用线阵 CCD采集深度坐标,用前馈型人工神经网络对深度坐标进行非线性修正。用这种方法无须测定仪器结构参数,能自动修正镜头的几何畸变,方法简单,实用性好。
In 3D laser scanning technique, artificial neural networks can be adopted for mapping coordinates from screen coordinates system to world coordinates system and for correcting non-liner errors of range sensor. In 3D scanning system based on line laser, a CCD video camera is used for taking pictures of an object, after black-white processing artificial neural networks transfer pixel coordinates that represent outline of the body to world coordinates. In 3D scanning system based on dot laser, artificial neural networks are used for calibrating the distortion of CCD range sensor. This method is simplicity, clear relationship between coordinates transformation, lens distortion is calibrated automatically and the stereo coordinates of 3D surface can be obtained without determining the structure parameters such as camera position and direction, laser direction, etc. 常用的激光三维扫描方法有线激光扫描法和点激光扫描
法,线激光扫描也称为三维面形的光刀测量法[1],由面阵 CCD
摄像机和线状激光源构成二维测量装置。点激光扫描法也称为
光针法,由线阵 CCD 摄像机和点状激光源构成一维(深度)测
量装置。线激光扫描法速度较快,点激光扫描法精度最高。在
应用光刀测量法时,若要获得较理想的效果,要解决如何实现
出处
《测控技术》
CSCD
2004年第11期52-54,共3页
Measurement & Control Technology
关键词
人工神经网络
激光
三维测量
坐标变换
artificial neural networks
laser
3-D measurement
coordinate transformation
