基于光场成像的表面三维重构

Three dimensional surface reconstruction based on light field imaging

  • 摘要: 传统的成像方式单次曝光只能获取物空间二维横向分辨率信息,无法获取纵向深度信息,导致单次拍摄过程中物空间的深度信息丢失, 无法对物空间的目标物进行三维重构。光场相机内部采用光场传感器,不同于传统成像系统单次采集只能获取二维信息而造成的信息缺失,光场传感器可获取物空间的多维光场信息,同时其还具有便携等优点。采用光场相机进行拍摄,利用数字重聚焦以及散焦测距和相关计算的方法,实现密集深度图像的获取,基于matlab软件平台,对所获取的图像深度数据矩阵进行处理,最终实现物空间的三维重构。得到物空间的相对深度的归一化结果。本实验中,在深度范围为100 mm~1 500 mm范围内,实现平均误差为5.47%深度信息的表面三维重构,最大重构误差为8.30%。

     

    Abstract: Traditional imaging methods can not capture axial information through one shot, which lead to information deficiency and incapability of three dimentional surface reconstruction. Unlike traditional camera, lytro's first generation pocket-sized camera using light field sensor can capture multi-dimensional light field information through one single exposure, and at the same time it has advantages such as portable. By using light field camera, with digital refocusing method, defocusing and correspondence cues, acquirement of high quality depth map can be realized. By using matlab platform to process acquired depth map matrix, reconstruction of three dimensional surface can be realized. Normalized depth data of object space is obtained. In this experiment, three dimensional surface reconstruction is realized with average error of 5.47% in depth range of 100mm to 1 500 mm, and maximum reconstruction error is 8.30%.

     

/

返回文章
返回