With the ever-changing photogrammetry technologies and different types of sensing instruments, a wide range of methods and alternatives can be applied for mapping underwater geospatial information. Among them, ray tracing of optical imagery is targeted to determine the underwater object points in this study. To simplify the task, it is assumed that all involved media are homogeneous and the water surface is still. Under the ideal physical and optical phenomenon with the known interior/exterior orientation parameters, the object-to-image correspondence of whole imaging paths of visible points can be retrieved by considering both collinearity condition and refraction effect. With that, the position of the water surface can be an observation or unknown parameter being solved together with underwater object points in accordance with practical requirements. Current study only considers imagery acquired by frame cameras and the generalized least-squares adjustment model with constraints is employed to carry out an effective parameter estimation. To characterize the core issues of this study and qualitatively verify the positioning quality varied with the influential factors, simulation data and experiments are well designed and conducted. The results of geometric conditions, solvability analysis of parameters and positioning quality tuned by influential factors are fully addressed and concluded, implementing a cornerstone for the air-to-water photogrammetric intersection solving for water surface and underwater object points.

The purpose of this study is to explore the benefits and problems by using UAV images in the crop classification of slope areas in Taiwan. We first made use of UAV to obtain the mosaic orthoimage in the hillside of Taizhong Taiping District. The eCognition multi-scale segmentation was then utilized to create image blocks. The crop types in the research area were then interpreted in an object-oriented manner. Because the category of “Longan (Dimocarpus longan) and Litchi (Litchi chinensis)” in the research area was difficult to distinguish effectively with the forest, which led to the difficulty of automatic classification, this study adopted artificial interpretation. The classification accuracies indicated that “Longan (Dimocarpus longan) and Litchi (Litchi chinensis)" was 88.2%, forest was 96.5%, nursery was 6.8%, and others was 99.7%. Finally, this study suggests to differentiate Longan (Dimocarpus longan) and Litchi (Litchi chinensis) using different flowering stages and to improve the efficiency of field data collection using segmentation blocks for subsequent research.

Traditionally, 3D spatial information can only be projected on the 2D maps for visualization due to the limitation of equipment and technology. Benefit from the rapid development of computer hardware and software, 3D visualization of the 3D spatial information can be achieved either by virtual reality (VR) or by augmented reality (AR). The 3D visualization is not only much more ocular for human, but also much more detailed and accurate. Due to the close environment and the spatial limitation, the tunnel survey is usually carried out by the plan table or the theodolite. The 3D tunnel is projected on the 2D map. Although close-range photogrammetry can be used to reconstruct the 3D model and to survey 3D coordinates of the tunnel, it requires many artificial targets as control points. Additional light sources are also necessary for shooting clear photographs. These requirements leads to huge labor work and consumes time. The terrestrial light detecting and ranging (LiDAR) integrated with a camera is capable of acquiring 3D point cloud in an efficient way, and therefore, is very suitable for tunnel survey. However, the terrestrial LiDAR is too expensive to be adopt by non-surveyors. In addition, due to the winding design of the military tunnel, it will require much more station setups to scan the whole tunnel. The connection of every scanning scene requires a powerful computer and professional human interaction. Recently, the handheld 3D scanners are developed based on the structured light approach, which can rapidly scan 3D point cloud within a short distance (within 4 meters). In this study, we apply a DotProduct DPI-8 handheld 3D scanner and a FARO Focus3D terrestrial LiDAR in the tunnel survey under the 48 Highland in Ci Sin Tan, Hualien. We compare the requirements, the operating procedures, and the results of the handheld 3D scanner and of the terrestrial LiDAR. The results show that the handheld 3D scanner is very suitable for the tunnel survey.

Three-dimensional point clouds can be generated using multi-frame images or multi-panorama images taken by a multi-camera system. Panorama image is a 360-degree image covering in horizontal and vertical directions. The multi-camera system produces different types of panorama images, such as spherical projection and cylindrical projection. The objective of this study is to compare the 3-D point clouds generated from frame images and panorama images using different projection modes. Five GoPro Hero4 cameras were integrated into a multicamera system and applied to acquire three type of images (i.e., original frame images, spherical panorama images and cylindrical panorama images) in a video mode. Because of the slight time lag problem in synchronous images taking, each camera implements time synchronous before stitching panorama images. The comparison includes quantitative and qualitative analysis. The experimental results indicated that the accuracies of frame images and panorama images were 4 cm and 7 cm in an indoor space. The frame images provided better geometric constraint and offered higher efficiency than panorama images. To compare spherical and cylindrical panorama images, the number of points from cylindrical images was larger than spherical images. Consequently, the computational time of cylindrical images was also slightly longer than spherical images.

Accompanied by the advancement of technology, multi-rotors unmanned aerial vehicle (MUAV) becomes mature gradually, and is applied to agriculture, energy, and media. There are many publications in the application of aerial photogrammetry field. However, the small and light platform with consumer camera is easily influenced by environmental factors during the process of aerial photogrammetry so to cause the inconsistent of flight height, tilt, and offset. Also, the low flight height results in narrow frame, lens distortion, and a large number of photos. In order to solve the problems, the research uses Structure from Motion (SFM), and dense matching by Patchbased Multi-view stereo (PMVS) to process the images. This process can calculate the interior and exterior orientation of image, and produce the dense point cloud. In this study, we obtained MUAV imagery with 1 cm resolution, and executes the flight planning and image processing according to the resolution requirements. The study area is the old wireless telecommunications station of Fengshan, Kaohsiung City. There are houses, roads and other buildings in the area, which belong to flat terrain, and the highest building is 9 meters. Finally, by examining the results by the standard of topographic map, the study proves that proceeding aerial photogrammetry with UAV and consumer camera can provide high-accuracy topographic maps.