The 3Dsurvey software enables the concept of videogrammetry, which allows the extraction of images from video for photogrammetric surveying. When performing video capture using a mobile phone, it is possible to attach a low-cost antenna GNSS-RTK, allowing direct georeferencing of images extracted from the video, without using the need for control point. The entire concept of videogrammetry was developed with the aim of faster spatial data acquisition compared to other geodetic methods. As part of the master's thesis, we evaluated the quality of videogrammetric products in two test areas, focusing on the accuracy of photogrammetric triangulation for the calculation of exterior orientation parameters of the images and the geometric quality of the photogrammetric point clouds generated through videogrammetry. Accuracy of photogrammetric triangulation was evaluated based on check points measured with a total station. The geometric quality of point clouds was assessed by directly comparing the videogrammetric point cloud with one obtained using a terrestrial laser scanner. We analyzed the impact of video capture configuration and the methods of calculating the external orientation of the images (indirectly, directly, or a combination of both). A total of 12 different photogrammetric processing of images were performed. The results showed that when using ground control points, the accuracy of photogrammetric triangulation ranged between 2 cm and 4 cm. In the case of direct georeferencing using GNSS-RTK antenna data combined with the phone's sensor data (enabling the measurement of image rotations), the triangulation accuracy ranged between 8 cm and 15 cm. The results of the comparative analysis between the photogrammetric point clouds and the TLS point cloud showed that the geometric quality varies depending on the method of video capture and the processing approach used. The best results were obtained using the indirect method of georeferencing with control points, where almost 80 % of the points in photogrammetric point cloud had a difference of less than 2 cm compared to TLS point cloud.
|