基于压缩感知的光度立体视觉三维成像方法研究

发布时间：2018-03-26 18:17

本文选题：三维成像　切入点：光度立体视觉　出处：《遥感技术与应用》2017年03期

【摘要】：压缩感知是近年来发展迅速的一种新型信号处理理论,以此为基础发展的成像系统具有系统规模小、成本低等特点而受到广泛的关注。利用表征反射光强度信息的图像灰度值是物体表面属性及几何形状综合反映这一光度学基础,引入以随机采样及非线性重建为技术手段的压缩感知理论,提出了一种基于压缩感知的光度立体视觉三维成像方法。该方法通过调换探测器与光源的位置,采用单像素探测器代替传统CCD成像,从不同位置的单像素探测器采集经调制的测量数据,利用压缩感知算法重建二维目标的图像,而后利用图像中的阴影信息求解目标的表面法向量和三维表面模型。最后,利用该方法对人像和简单几何体进行三维成像实验,结果表明该成像方法能在数据欠采样的情况下,实现目标的高效三维重建;在实验基础上,进一步分析了图像数量、采样率、重构方法、探测器位置等方面对三维成像质量的影响。
[Abstract]:Compression sensing is a new signal processing theory developed rapidly in recent years. The imaging system developed on this basis has a small scale. Because of its low cost and so on, the image gray value which represents the reflected light intensity information is the basis of the photometry, which is reflected by the surface properties and geometric shapes of the object. This paper introduces the theory of compression perception based on random sampling and nonlinear reconstruction, and proposes a three-dimensional imaging method of photometric stereo vision based on compression perception, which changes the position of detector and light source. Instead of traditional CCD imaging, single pixel detector is used to collect modulated measurement data from single pixel detector in different positions, and compression sensing algorithm is used to reconstruct 2D target image. Then, the surface normal vector and 3D surface model of the target are solved by using the shadow information in the image. Finally, the 3D imaging experiments of human image and simple geometry are carried out by using this method. The results show that the imaging method can be used in the case of under-sampling data. On the basis of experiments, the effects of image quantity, sampling rate, reconstruction method and detector position on 3D imaging quality are analyzed.
【作者单位】：中国科学院光电研究院中国科学院定量遥感信息技术重点实验室;中国科学院大学;
【基金】：国家863计划项目“强度关联遥感成像外场综合测试技术”(2013AA122904) 中国科学院、国家外国专家局创新国际团队项目“基于强度关联的新型遥感机理与方法研究”(2013AA1229)
【分类号】：TP391.41

【相似文献】