Abstract:
Using multi-head camera systems simultaneously with the nadir and oblique photographs from UAVs is being standardized in modern photogrammetry. Oblique photographs reveal side views of buildings and vegetation fields on the terrain. Thus, it induces more geometric detail information and also leads to precise interpretation and classification processes, compared with the view from nadir photographs as commonly practiced in traditional photogrammetry. From this research, multi-head camera systems increase the mapping coverage area significantly. The systems also provide efficient mapping missions, including flight plan designing, time-saving flight missions processing, and yielding a higher number of photos for multi-view geometry and point clouds. These advantages will further improve key point generating, which is crucial for processing photogrammetric computer vision software. The research objectives are to determine the camera rig parameters and numerical modeling of the 3DM V3 multi-head camera system. The multi-head camera system consists of one nadir camera and four 45-degree oblique cameras. All cameras are rigged up in an enclosure. The study also covers the improvement of the processing of 3D mapping data and point clouds from multi-view photographs. The final geometric model of the camera system should be consistent with the actual camera geometry in order to confirm the correctness of the model. The results of 3D mapping data and point clouds show the comparison of RMSE results between the nadir photos block and the five-camera combination block. The conclusion of parameterization and numerical modeling of 3DM-V3 multi-head camera systems are as follows. All 4 sets of relative translation parameters are called (Tx, Ty, Tz)reln when n has a value from 1 to 4, all offset values from four oblique cameras have to be constrained. The other 4 sets of relative rotation parameters at the reference camera referred to the other four oblique cameras (Rx, Ry, Rz)reln when n has a value from 1 to 4, the offset values must be set as an initial estimation and set free during numerical processing. Other conventional parameters that represent the value from all 5 internal camera models (f, xu, yu, R1, R2, R3, T1, T2) as free and will finally be numerical adjusted. The result of the final camera model from computation adjustment for triangulation affirms the precise geometry and the consistent rig parameters with the physical characteristics of the multi-head camera systems. This research triangulation block is tied-up with constraints and check points from 14 GCPs in the test field size 0.8 square kilometers. The results from final fit parameters for GCPs are RMSE 0.023 meters in horizontal and RMSE 0.019 meters for the vertical component. For point clouds production, in the case of combined oblique photos block, the amount of point clouds increases significantly 2.45 times for both density and volume, compared to the nadir photo block case. The positioning accuracy in 3D mapping and orthophoto from oblique photos block processing gives RMSE at 0.105 meters, which are considered from 30 sample points in the open field within the study area.