ACCURACY ASSESSMENT OF A COMPLEX BUILDING 3D MODELRECONSTRUCTED FROM IMAGES ACQUIRED WITH A LOW-COST UAS
ACCURACY ASSESSMENT OF A COMPLEX BUILDING 3D MODELRECONSTRUCTED FROM IMAGES ACQUIRED WITH A LOW-COST UAS
复制标题
根据低成本 UAS 获取的图像重建复杂建筑 3D 模型的精度评估
DOI：
10.5194/isprs-archives-xlii-2-w3-551-2017
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
F. Stătescu
中科院分区：
文献类型：
--
作者：
E. Oniga;C. Chirila;F. Stătescu
抽象的。如今，无人机系统（UAS）是一种广泛使用的采集技术，用于创建建筑物3D模型，在很短的时间内以非常高的分辨率或视频序列采集大量图像。由于低成本的UAS是首选，因此必须评估使用该平台创建的建筑物3D模型的准确性。为了取得成果，我们选择了罗马尼亚雅西“水利工程、大地测量和环境工程”学院的院长办公楼，这是一座形状复杂的建筑，屋顶由两个双曲抛物面组成。在建筑物周围的地面上放置了七个点，其中三个用作GCP，其余四个用作检查点（CP），用于准确性评估。此外，用Leica TCR 405全站仪测量了代表建筑物特征点的10个自然CP的坐标。建筑物3D模型被创建为点云，该点云是基于低成本UAS获取的数字图像自动生成的，使用图像匹配算法和不同的软件，如3DF Zephyr，Visual SfM，PhotoModeler Scanner和Drone2Map for ArcGIS。除PhotoModeler Scanner软件外，通过求解自校准光束法平差同时确定内部和外部定向参数。基于上述软件自动生成的UAS点云和GNSS数据，采用最小二乘法和统计粗差检测方法计算东侧双曲抛物面的参数。然后，为了评估建筑物3D模型的准确性，将立面和屋顶与参考数据进行了几次比较，考虑最小误差：立面的TLS网格和屋顶的GNSS网格。最后，使用PhotoModeler Scanner软件，根据建筑物的特征点，以3D形式创建建筑物的正面，从而生成CAD（计算机辅助设计）模型。结果表明，使用低成本的UAS进行建筑物3D模型创建的潜力很大，如果建筑物3D模型是基于其特征点创建的，则精度显着提高。
Abstract. Nowadays, Unmanned Aerial Systems (UASs) are a wide used technique for acquisition in order to create buildings 3D models, providing the acquisition of a high number of images at very high resolution or video sequences, in a very short time. Since low-cost UASs are preferred, the accuracy of a building 3D model created using this platforms must be evaluated. To achieve results, the dean's office building from the Faculty of “Hydrotechnical Engineering, Geodesy and Environmental Engineering” of Iasi, Romania, has been chosen, which is a complex shape building with the roof formed of two hyperbolic paraboloids. Seven points were placed on the ground around the building, three of them being used as GCPs, while the remaining four as Check points (CPs) for accuracy assessment. Additionally, the coordinates of 10 natural CPs representing the building characteristic points were measured with a Leica TCR 405 total station. The building 3D model was created as a point cloud which was automatically generated based on digital images acquired with the low-cost UASs, using the image matching algorithm and different software like 3DF Zephyr, Visual SfM, PhotoModeler Scanner and Drone2Map for ArcGIS. Except for the PhotoModeler Scanner software, the interior and exterior orientation parameters were determined simultaneously by solving a self-calibrating bundle adjustment. Based on the UAS point clouds, automatically generated by using the above mentioned software and GNSS data respectively, the parameters of the east side hyperbolic paraboloid were calculated using the least squares method and a statistical blunder detection. Then, in order to assess the accuracy of the building 3D model, several comparisons were made for the facades and the roof with reference data, considered with minimum errors: TLS mesh for the facades and GNSS mesh for the roof. Finally, the front facade of the building was created in 3D based on its characteristic points using the PhotoModeler Scanner software, resulting a CAD (Computer Aided Design) model. The results showed the high potential of using low-cost UASs for building 3D model creation and if the building 3D model is created based on its characteristic points the accuracy is significantly improved.