Towards a comprehensive and efficient close range photogrammetry solution for low cost, high precision surveying

使用简单和低成本的技术解决方案，同时提供空间准确和精确的3D对象信息，对复杂城市和景观地标的3D获取和建模的需求日益增加。地面物体和场景的三维建模是图形学、视觉学和摄影测量学领域一个深入而持久的研究问题。为许多应用指定的要求，包括基础设施检查和测量，涉及高几何精度和完整细节的建模，以及建模技术的自动化、低成本、可移植性和灵活性。人们提出了不同的基于图像的方法来产生准确和详细的模型。然而，由于目前大多数自动化方法尚未在实际应用中得到验证，因此它们仍然具有高度交互性。因此，结果的质量取决于用户的专业知识。因此，目前还没有专门针对不断扩大的要求低成本和高精度系统的非摄影测量师用户社区的摄影测量解决方案。提出的研究计划的长期目标包括设计一种全面的近景摄影测量方法，用于复杂物体和场景的三维采集和建模。目标方法将在低成本、高精度、自动化和耗尽性约束下进行设计。在地面摄影测量中，三维采集和建模是一个分两个阶段的过程。首先（第一阶段），获取图像和相关定位数据（即GPS和INS数据）；第二（阶段2），它包括利用这些图像，以恢复3D对象的细节。第二阶段的表现在很大程度上取决于第一阶段的执行情况。因此，为了实现我们的长期目标，我们建议解决以下短期目标：1)开发创新的策略和方法来有效地指导图像采集过程；2)设计3D建模方法的新概念，提高摄影测量过程的自动化程度，并使其能够被非测绘专家使用。这项研究将产生新的通用概念，可在各种计算机和工业视觉领域中重复使用。

There is an increasing need for 3D acquisition and modeling of complex urban and landscape landmarks using simple and low cost technological solutions while offering spatially accurate and precise 3D object information. Terrestrial three-dimensional modeling of objects and scenes is an intensive and long-lasting research problem in the graphics, vision and photogrammetric communities. The requirements specified for many applications, including infrastructure inspection and surveying, involve high geometric accuracy and the modeling of the complete details, as well as the automation, low costs, portability and flexibility of the modeling technique. Different image-based approaches have been proposed to produce accurate and detailed models. However, they remain highly interactive since most of the current automated methods are still unproven in real applications. Consequently, the quality of the result depends on the user expertise. As a result, there is currently no photogrammetric solution dedicated to the expanding non-photogrammetrists user community requesting low cost and high accuracy systems.The long-term objective of the proposed research program consists in designing a comprehensive close range photogrammetry method for the 3D acquisition and modeling of complex objects and scenes. The targeted method will be design under low cost, high accuracy, automation and exhaustiveness constraints. In terrestrial photogrammetry, 3D acquisition and modeling is a two stages process. First (stage 1), it consists in acquiring the images and related positioning data (i.e. GPS and INS data); second (stage 2), it consists in exploiting these images in order to recover 3D object details. The performances of stage 2 are strongly relying on how stage 1 is carried out. Therefore, to meet our long-term objective, we propose to address the following short-term objectives: 1) to develop innovative strategies and methods to efficiently guide the images acquisition process; 2) to design new concepts for 3D modeling methods enhancing the automation of photogrammetric processes and their use by non specialists in geomatics. This research will yield to novel generic concepts, reusable in various computer and industrial vision domains.