Keywords

1 Introduction

Unmanned aerial vehicles (UAV) or commonly known as drones were initially invented for military purposed, then the uses of drone with additional functional by adding camera, the drone can be used for surveying high level areas or dangerous area (Madawalagama et al. 2016; Ajayi et al. 2018). The creation of structure from motion based method for digital photogrammetry from multiple converging images resulting to an outburst of 3D terrain reconstructions in high resolution (Dering et al. 2019). The UAV systems equipped with photogrammetric cameras have shown enough accuracy for many applications in civil engineering (Díaz-Vilariño et al. 2016). The advantages of UAVs are it require lesser time compared to other techniques for data acquisition and minimize the cost (Aber et al. 2010; Agüera-Vega et al. 2018). From the previoues study shows that, UAV photogrammetry is more accurate than Terrestrial Laser Scanning (TLS) when used in terrain mapping (Gruszczyński et al. 2017). Through the leveraging state of the art methodologies, the research mentioned in (Pierdicca 2018) paper shows how a precise reality-based model (compatible for documentation purposes) can be achieved with a robust pipeline, also beginning from unplanned acquisition and rough data sets. UAV photogrammetry monitoring of 3D areal displacements was revealed in (Kemal Özgür Hastaoğlu 2019) study.

The previous study has shown that to produce highly accurate reconstructions and models of land shapes and historic landscape structures, UAV technology can be used as a well-organized and dependable tool. This showed that it is necessary to use of high-resolution UAV-based data for the evaluation of the authentic storage capacity of historical landscape structures, that is essential information for the future application in water management and flood mitigation plans. Besides that, although the UAV imaging is a relatively low-cost technology, it is powerful enough to generate 3D models. The exactness obtained is compared to that obtainable from land survey approaches, Manned Aerial Vehicles and LiDAR data, all of which are more costly. It has been shown in (B. Kršák, 2016) that the surface model generated using a photogrammetric approach using low-cost UAV and low-cost cameras meets the desired accuracy requirements and can generally be considered as a convenient tool for data collection in surface mines. Thus, the objective of this paper is to review the application of UAV photogrammetry used for aerial mapping applications.

2 Reconstructed Point Cloud

Rather than using geo-referencing, UAV photogrammetry use the Ground Control Points (GCP) into a reconstructed point cloud based on UAV, or the camera exposure stations is used based on Real-Time Kinematic (RTK), for direct geo-referencing method. The studies carried out to compare these approaches show that the GCP measurements produce more reliable results in 3D positioning (Stöcker et al. 2017; Forlani et al. 2018; Rabah et al. 2018). In accordance with the same flight plan in the study of (Forlani et al. 2018), four different flights were conducted for each RTK mode over a field trial. Three block control configurations were used to produce the Digital Surface Model: GCP only, camera stations only, camera stations and one GCP. From the result, the first and third configurations offer the greatest accuracy in 3D positioning irrespective of the RTK mode. Moreover, (Stöcker et al. 2017) has reported that thorough investigations of the external orientation variables show that the addition of four GCPs will minimize systemic sensor misalignments and offsets of the image block. On this matter, the application of post-processing cinematic corrections decreases time-consuming field work to estimate the high quantities of GCPs and makes extensive scale UAV mapping a more practicable solution for cm-level practitioners. The research by (Rabah et al. 2018) showed that classical Aerial Triangulation is more reliable than the DG of UAV imagery. These researches show that while drones with on-board receivers capable of positioning RTK are being used, only few GCPs are needed to achieve more precise performance in 3D positions. Furthermore, the study by (Díaz-Vilariño et al. 2016) also show that the similarity of photogrammetry vs LiDAR as shown in Fig. 1.

Fig. 1.
figure 1

Digital elevation model (DEM) with resolution of 1 m per pixel from (a) LiDAR and (b) photogrammetric system.

Full size image

The overpriced cost associated with aircraft use and the consumption of time nature makes this strategy an unsuitable solution, particularly for small scale analysis (Al-Rawabdeh et al. 2016). Photogrammetry will create 3D and 2D full-colour terrain models in the different light spectrum that are easier to view. The key outputs of the photogrammetric surveys were raw photographs, ortho photograms, Virtual Surface Models and 3D point clouds created from the stitching and processing of hundreds or thousands of images (Agüera-Vega et al. 2018). Therefore, it is an optional method compare to Light Detection and Ranging (LiDAR), which is expensive but dominant in mid-air, to obtain high-resolution virtual surface models.

3 Method for Image Processing in Photogrammetry

End user will choose the software by of it graphical user interface, friendly user and easy to learn. In addition, basic of photogrammetric triangulation feature where it can process various types of imagery e.g. aerial & close range which also provides easy auto calibration. Software can create dense point clouds where its points classification to customize geometry reconstruction and elaborate model editing for accurate results. The final feature will be the generation and texturing of 3D model that creates photorealistic textures. The ambition is to unravel the figure in a compact image descriptor, which allows computing the gap between all images descriptors effectively. All the features is to match between candidate image pairs and continued to structure from motion where the input images, and infer the rigid scene structure (3D points) that provides the geometric relationship behind all pose observations (position and orientation) and internal configuration of all cameras. After that, depth map estimation is done and then followed by mesh where it creates a scene representing a dense geometric surface. After meshing there is texturing, then ended by localization based on the SFM results. Table 1 listed the method for image processing in UAV photogrammetry.

Table 1. List of image processing software based on previous study.
Full size table

4 Conclusion

It is very important for the method to produce accurate and consistent data in order to plan the measurements and to select ground sampling distance values according to the expected quantities of displacement. Parameters such as expanding the amount of deformation plates, reducing flight speed and altitude, and using a high-resolution camera would then enable smaller displacement quantities to be determined. With regard to the accuracy of the generated model, it is proven, due to the concept of digital photogrammetry and aerial imagery software processing, these models are much more precise compared to models created only from terrestrially determined detailed point.