红外与毫米波学报, 2019, 38 (6): 819, 网络出版: 2019-12-27
融合粗糙深度信息的低纹理物体偏振三维重建
摘要
对于表面光滑、低纹理的目标,传统基于多视几何重建的算法难以获得理想的结果,利用偏振信息来重建这类物体是便捷有效的方法之一.然而单纯利用偏振信息进行三维重建存在歧义性等问题,难以获得理想的结果.以粗糙深度图作为先验信息可解决歧义性问题.先对偏振相机与深度相机标定并配准图像,由粗糙深度图获得的法向量辅助纠正偏振方位角歧义,再利用纠正的法向量与粗糙深度图积分融合,从而获得较高精度的物体三维表面.
Abstract
For the smooth surface and low-texture target, the traditional multi-view geometry reconstruction algorithm was difficult to obtain the ideal result. However, it is difficult to obtain ideal results by simply using the polarization information for three-dimensional reconstruction with ambiguity problem. Using the rough depth map as a prior information can solve the azimuth ambiguity problem. Firstly, calibrated and aligned the polarized camera and the depth camera, and used the normal vector obtained by the rough depth map to correct the polarization azimuth ambiguity, then integrated the corrected normal vector with the rough depth map to obtain a high-precision three-dimensional surface of the object.
杨锦发, 晏磊, 赵红颖, 陈瑞, 张瑞华, 施柏鑫. 融合粗糙深度信息的低纹理物体偏振三维重建[J]. 红外与毫米波学报, 2019, 38(6): 819. Jin-Fa YANG, Lei YAN, Hong-Ying ZHAO, Rui CHEN, Rui-Hua ZHANG, Bo-Xin SHI.