基于智能布光的可移动文物原真采集系统

An authentic acquisition system for movable cultural relics based on intelligent lighting

文物摄影测量数字化采集需要布置均匀一致的光照环境,以满足多视图立体和纹理贴图的重建需求。但当前依赖经验的光源布置难以确保光照均匀性,从而导致原真采集图像数据量大与建模效率低下的问题。本文基于工程经验,提出了一套适用于文物摄影测量的智能布光原真采集系统。在不降低原真建模精度的前提下,通过减少采集图像数量,来提高三维建模效率。首先,将光源布置问题转化为球面点均匀分布问题,使用模拟退火算法计算不同光源数量下最优的光源分布位置集合;其次,提出在仿真计算环境下的光照均匀度计算方法,在三种不同球形构型下,计算不同光源数量的光照均匀度,并选择球形构型复杂度和光照均匀度较优的照明框架;最后,本文基于测地线球体构建了基于智能布光的文物原真采集硬件系统。仿真实验结果表明,与传统布光方式相比,本文方法在明显降低图像数量的同时,没有明显提高重投影误差,模型精度和完整度没有受到损失。

The digital collection of cultural relics photogrammetry needs to arrange a uniform lighting environment to meet the reconstruction requirements of multi-view stereo and texture maps.However,the current experience-based light source layout is difficult to ensure the uniformity of illumination,which leads to the problems of large amount of original image data and low modeling efficiency.Therefore,on the basis of engineering experience,this paper proposes a set of intelligent light distribution true acquisition system suitable for photogrammetry of cultural relics.Without reducing the accuracy of authentic modeling,the efficiency of 3D modeling can be improved by reducing the number of acquired images.First,the problem of light source arrangement is transformed into a problem of uniform distribution of spherical points,and the optimal light source distribution position set is calculated under different light source numbers using simulated annealing algorithm.Secondly,a calculation method of illumination uniformity under the environment of simulation calculation is proposed.Under three different spherical configurations,the illumination uniformity of different light source numbers is calculated,and the illumination framework with better spherical configuration complexity and illumination uniformity is selected.Finally,based on the geodesic sphere,this paper builds a hardware system for the original collection of cultural relics based on intelligent lighting.The simulation experiment results show that,compared with the traditional lighting method,the method in this paper not only significantly reduces the number of images,but also does not significantly increase the reprojection error,and the accuracy and integrity of the model are not lost.