一种基于投影变换的深度图像生成算法

A depth image generation algorithm based on projection transformation

深度图像的表示是实现三维目标识别、配准等工作的前提。针对地面三维激光扫描得到的点云数据,提出一种基于投影变换的深度图像生成算法。采用投影变换将三维点沿观察坐标系的法向量投影到观察平面上,使投影向量和观察平面垂直,这样观察点和观察平面间的距离就不会影响物体的投影大小和形状,有效避免了"失真"现象的发生。利用灰度图像生成深度图像,使用方案对地面激光扫描系统获取的建筑物实验数据进行具体处理。结果表明,文中基于投影变换的图像生成方法更能直观地反映建筑物表面的几何分布特征,为后续特征的提取奠定基础。

Depth representation of the image is a prerequisite to achieve a three-dimensional object recognition,registration and other work.For point cloud data obtained by terrestrial laser scanning,a depth image generation algorithm is proposed based on the principle of a central projection.The method uses the central projection vector along the observed three-dimensional coordinate system projected onto the observation plane,making the projection vector perpendicular to the plane and observed,so that the distance between the observation point and observation plane will not affect the projection size and shape of the object,avoiding the 'distortion'phenomenon generating a depth image from the grayscale image.Finally,with the present scheme,the experimental data terrestrial laser scanning systems are acquired.The results show that the image generated based on the projection transformation can directly reflect the distribution of the surface geometry of the building,which provides a subsequent feature extraction foundation.