面向水电机组维护的增强现实云数据获取与虚实融合方法研究

Hydropower has an important strategic position in the national economic development of China. The maintenance of hydroelectric generating set which is the core equipment of hydropower units can effectively enhance the hydropower economic benefits. In order to further improve the level of hydropower generating units ,a more novel and perfect way should be established. In this context, we exploit augmented reality cloud system to develop the hydroelectric generating unit mantenance system. In this project, we try to start from the perspective of computer applications, put the improving the real-time and robustness of augmented reality cloud system as the entry point, focus on exploring cloud computing resource scheduling with QoS control and robust augmented reality registration method to effectively meet the need of enhanced reality of cloud systems in the application of unit maintenance. .Firstly,we propose a design an architecture of augmented reality cloud system to meet the demand of the actual maintenace of the hydropower generating units..Secondly, we propose a novell resource scheduling of cloud computing system to meet the requirements of real-time. In our research, we will apply the user QoS requirements to the scheduling policy, and build an intelligent QoS dynamic adjustment mechanism for the special users(unit parts).We compare and experimental study different distribute database systems and propose an optimization program to improve the abality of data analysis and processing of cloud system under different network conditions..Thirdly, we explore natural features tracking based 3D registration methods for augmented reality using the theory of multiple view geometry in computer vision. Our method can improves the robustnes and real-time of the augmented reality cloud..We build a prototype system and apply it to real projects to prove the effectiveness of the proposed method. We believe that the implementation of this project, which combine the universal thoeory and application research, have theoretical and practiical significance for the development of augmented reality cloud computing.

本项目以水电机组维护应用为基本背景，将增强现实和云计算相融合形成增强现实云，围绕如何从众多分布式数据库中获取维护信息并实现虚实信息融合这一核心问题，以提高增强现实云实时性和鲁棒性为切入点，重点开展以下研究：1）研究增强现实与云计算技术的融合策略，探索适用于水电机组维护的增强现实云模型；2）围绕数据获取问题，研究云资源调度策略，针对机组部件这一不具思维能力的"用户"，建立智能化的QoS动态调整机制，同时，探索分布式数据分析处理优化方法，从整体上提高资源调度的智能性和实时性；3）围绕虚实信息融合问题，以多视图几何理论为依据，探索基于自然特征的增强现实虚实配准方法，提高虚实信息融合的鲁棒性和实时性。通过研究，力争在增强现实云的几个关键问题上取得突破，项目将普适性的理论研究和适应性的应用基础研究相结合，在满足水电机组维护应用需求的同时，可以为云计算环境下的增强现实相关应用领域提供理论和应用支撑。

水电能源在国民经济发展中具有重要的战略地位，水电机组的维护是保障水电机组安全、可靠、高效完成发电任务的基本生产活动。本项目以水电机组维护应用为基本背景，将增强现实和云计算两种新技术相融合，围绕如何从众多分布式数据库中获取维护信息并实现虚实信息融合这一核心问题，以提高增强现实云系统实时性和鲁棒性为切入点开展研究，主要研究成果包括：(1)研究了MAR与云计算结合设计的关键技术问题，提出移动增强云系统中的体系结构和技术融合策略，设计并实现了基于分层的云计算系统中实现增强现实的体系结构。(2)围绕数据获取问题，提出了基于RBF-DE的改进蚁群算法和多QoS约束的云资源调度、基于PABFD-SA算法的虚拟机动态迁移、改进的基于感知能耗的虚拟机迁移整合过程优化方法等系列云计算资源调度策略和智能求算法。(3)针对水电机组维护的工作区光照变化较大的问题，从图像增强和图像分割两个基础性工作入手，建立了多模图像非局部联合去噪及加速、基于同源相似性的图像分割、基于共生直方图的显著性检测的改进和基于改进RKT的边缘检测等算法有效提升了系统抵抗噪声的能力。(4)针对机组在线维护过程中拟识别部件处于运动状态的情况，开展了运动目标跟踪技术研究，提出的结合方向纹理熵的Haar-like特征在线boosting跟踪、融合Haar-like特征和HOG特征的目标跟踪和重采样策略等算法能有效适应运动状态下的部件识别与实时跟踪。(5)提出了基于智能手机的图像文字块定位与识别算法和增强现实虚实配准系列方法，并建立了云环境下的移动增强现实系统。