GGG-PositioningSfMScale: Dataset for Testing Scaling and Uncertainty Propagation Through Camera Positioning Data

The proposed dataset can be used to test the precision of scaling SfM reconstructed objects using camera positioning data. In addition position uncertainty data from a DJI GPS RTK in X,Y and Z is added, so the propagation of uncertainty from the positioning sensor to the calculated scale can be calculated. The camera positions are also given with any uncertainty, so other types of position uncertainty can be used to test how the system will react with different uncertainty inputs.

Images from two objects are given - an angel statue and a wind turbine blade. Images are taken in a semi-circle pattern around the object and a total of 19 images are taken from each object. The positions of the images are given in a separate file, while all the images contain EXIF data with the used camera parameters. All images are captured using Canon 5Ds DSLR camera.

More information on how the scale and scale uncertainty can be calculated are present in the referenced paper.

本数据集可用于测试基于相机定位数据对运动恢复结构（Structure from Motion，SfM）重建对象的缩放精度。此外，数据集还附带了大疆（DJI）GPS RTK采集的X、Y、Z三轴位置不确定性数据，因此可计算定位传感器的不确定性向计算所得缩放比例的传播过程。相机位置数据同样包含对应的不确定性信息，故而可采用其他类型的位置不确定性数据，以测试系统在不同不确定性输入下的响应表现。

本数据集包含两类拍摄对象的图像：天使雕像与风力发电机叶片。图像沿环绕对象的半圆形轨迹采集，每个对象共计拍摄19张图像。图像拍摄位置存储于单独的文件中，所有图像均携带包含所用相机参数的EXIF元数据。所有图像均采用佳能（Canon）5Ds数码单反相机拍摄。

关于缩放比例与缩放不确定性的计算方法，详见所引用的学术论文。