基于自由曲面棱镜的单目立体视觉测量关键技术研究

Prism-based single-lens stereovision measuring method has many advantages, such as small size, low cost, minimum parametric errors and synchronization error in image capturing when compared with multi-camera stereovision systems. It has significant research value; notably in systems requiring high density integration; and target measurement in confined space. However, the nonlinear light refraction using plane-surface prism will result in non-linear image distortion, the rectification of which is fastidious. Many applications of such prism based stereovision system are limited by the complex calibration process. Moreover, the accuracy of measurement was reduced as the accurate stereo disparity values are acquired at the expense of space resolution. These inherent problems mentioned above hinder further development, and hence the applicability of plane-surface prism based stereovision system. This project will focus on the practical necessity of optical measurement system. We will study the distortion rectification method based on freeform prism, develop the self-calibration algorithm of freeform prism based stereovision and the design of high precision three-dimensional reconstruction technique of freeform prism based 3 dimensional scene reconstruction. The principal objective of this research is to simplify the image pre-treatment process, to extend its field of applicable field, and to enhance the accuracy of the method. A prototype of a freeform prism based stereovision measuring system will be designed and implemented to verify the methods proposed in this study. The research results could be applied in other kinds of non-contact engineering measurement systems. The knowledge gained from this study would be valuable to enhance theoretical understanding of this field, which can hence be applied in many areas of applications especially for dynamic measurement.

基于棱镜的单目立体视觉测量方法具有体积小、成本低、无多相机参数误差和多相机同步误差等优点，对高集成度的系统需求和受限空间的目标测量具有重要研究意义。然而，因为该方法使用平面棱镜，光线折射的非线性导致图像产生畸变，矫正过程繁琐耗时；复杂的标定过程和对标定板的依赖，制约该方法的使用环境；以牺牲空间分辨率为代价获得视差图像，降低该方法的测量精度。本项目将针对视觉测量的实际需求，通过将棱镜的折射平面优化为自由曲面的手段，研究基于自由曲面棱镜的图像非线性畸变矫正方法、基于自由曲面棱镜的单目立体视觉自标定算法和基于自由曲面棱镜的单目立体视觉高精度三维重建技术，从提升成像系统性能角度简化图像预处理流程、扩大应用范围并提高测量精度。设计并实现一套基于自由曲面棱镜的立体视觉测量原型系统，验证提出方法的科学性和有效性。相关成果可应用到各类目标的非接触式测量，尤其对动态目标测量具有重要的理论意义和实际应用价值。

随着立体视觉技术的发展和应用领域的不断拓展，人们对立体视觉系统经济化、微型化、轻量化的使用要求也在不断提升，多目立体视觉系统的缺点愈加凸显，在很多领域的应用受到不同程度的限制。本项目依托基于棱镜的视觉测量技术优势，通过将棱镜的折射平面优化为自由曲面的方式，研究基于自由曲面棱镜的图像非线性畸变矫正方法、基于自由曲面棱镜的单目立体视觉自标定算法和基于自由曲面棱镜的单目立体视觉高精度三维重建技术，从提升成像系统性能角度简化图像预处理流程、扩大应用范围并提高测量精度。在项目执行期内，项目团队针对主要研究内容，从自由曲面棱镜建模与求解、基于棱镜的视觉测量算法全流程及参数优化和基于棱镜的视觉测量系统应用三个方面开展合作研究与探索，构建出自由曲面棱镜设计模型、基于棱镜的视觉测量系统极线几何模型、透视投影变换模型、三维重构算法等一整套系统运算流程，为该方法奠定了坚实的理论基础；此外，项目团队积极开展新领域的应用探索，分别在医用内镜、增强现实眼追踪、3D成像与显示以及拉曼成像等领域取得重要突破。依托该项目，项目负责人共发表相关SCI检索期刊论文8篇，其中JCR1区论文4篇，2区论文1篇，4区论文3篇；申请国家级发明专利7项，授权国家级发明专利2项；形成相关技术文档3篇；获2017年度沈阳市自然科学学术成果二等奖1项；培养硕士研究生6人。此外，依托本项目的资助，项目负责人职称也已由讲师晋升为副教授。