Along with industrialization and urbanization, the exploitation of natural resources makes emissions of air pollutants increases. Fine Particulate Matter (PM2.5) is one of the pollutants which affect human health and has been attracting attention in recent years. Observations from the limited number of monitoring sites are not capable to depict the spatial concentration pattern of PM2.5. Diverse culture-specific emission contributors distributed in communities raise the difficulty of emission classification as well. This study selected Kaohsiung-Pingtung Air Quality Zone (Kao-Ping AQZ) as the study area. Annual average PM2.5 level was calculated based on the on-site observations from 15 EAP monitoring stations during 2006 to 2012. Land-use allocation surrounding the monitoring sites were obtained from GIS databases. A Land Use Regression (LUR) model then developed to estimate the spatial-temporal variability of PM2.5 concentrations of Kao-Ping AQZ. The overall model R2, adjusted R2, and RMSE was 0.82, 0.82, and 3.81µg/m3, respectively. Several land-use and environmental variables were selected as important predictor variables, including distance to the nearest industrial park, minimum of NDVI within 3000m buffer range, temperature, and area of farm within 1500m buffer range. With the partial R2 of 0.76, distance to the nearest industrial park was the dominant variable in the developed model. The results of model validation again confirmed the model performance while the R2 from cross validation and external data verification was 0.81 and 0.67, respectively. The predictions of PM2.5 variability based on the developed LUR model showed a slightly decreasing trend during the study period while coastal metropolitan areas were always the higher pollutant areas.

Nowadays, most satellite imagery vendors offer rational polynomial coefficients (RPCs) to users for processing geometric information. Although known RPCs in rational function model (RFM) would give explicit object-to-image correspondence, the physical meaning of the parameters is hard to be interpreted. Especially, when faced with underwater object points, the object-to-image correspondence cannot be directly realized by these RPCs due to the refraction effect. To cope with this situation, this study proposes an alternative way in the imaging rays under RFM and refraction effect to determine both the water surface and underwater object points. A generalized least-squares adjustment with constraints is developed to well handle functional and stochastic models. To its essence provided that the water surface is totally unknown, the refraction vectors under the water varying with estimated water surface through each iteration results in a weak geometry and leads to unstable solutions. In addition to revealing the characteristics of aforementioned object-to-image correspondence, this study explores how the underwater object point and water surface determination would benefit from the prior observations of water surface, vertical control point, full control point, and even water depth. The effect of using multiple points is also investigated in this study. In experiment part, this study uses not only simulation data but also real satellite imagery to verify the feasibility of the proposed model. And it can be summarized that the quality of water surface, either through priori information or derived from control information, is crucial to the positioning performance of underwater object points. The accuracy of water depth better than 2m, under employed satellite imagery, would supply water surface with 1-2 pixels positioning quality, and even bring about superior underwater object point coordinates as compared to when they are to be determined through perfect water surface.

Through analyzing consecutive image pairs, the robotic can localize itself autonomously and continuously in an unknown environment without any human input, and it is known as Visual Odometry. Localization can be achieved in photogrammetry and computer vision, but the methods and objectives are little different. Consequently, combinable methods are used to reconstruct the camera path in this study. Firstly, an appropriate matching strategy and an automatic method for solving relative orientation parameters (ROPs) are applied in this study. Secondly, coherent relative orientation of consecutive image pair can be determined by stacking individual ROPs. Since the localization of consecutive images is incremental, the errors are accumulated over time. In this study, a preliminary study of network adjustment of ROPs is proposed to suppress the accumulated error in the navigation. In the tests, results show the estimated camera path and reconstructed 3D object points are reasonable, and the features in reconstruction can be recognized clearly. Comparing estimated camera path before and after with actual camera path, the accumulated errors is reduced significantly. Furthermore, the deviation of object points derived from different image pairs are decreasing. In conclusion, the network adjustment in this study is effective to reduce the accumulated errors of the camera path.

With the development of Intelligent Transport System (ITS), the precise position is one of the primary cores to achieve the autonomous driving vehicles. However, Global Navigation Satellite System (GNSS) would be affected in poor accuracy by changes in surrounding obstacles, which is well-known as the multipath interference and Non-Line-of-Sight (NLOS) reception in urban environments. This study uses the integrated vehicle navigation system to establish High Definition maps (HD Maps) as reference information to detect the positioning errors due to multipath or NLOS reception. Moreover, the correction can feedback to the navigation process by height constraints system. The experiment applies commercial satellite receiver to observe single-frequency GPS code for Single-Point Positioning (SPP) calculation. The results show that with the HD Maps aided, the positioning accuracy improves by about 66%, which the height error is almost mitigated. This research is the prototype of HD Maps aided navigation application, which can make the autonomous driving positioning system more robust and feasible in urban area.

Conventional methods used to retrieve water depth for inland waterbodies rely on single-beam or multi-beam echo sounders, which sample depths along multiple survey lines and produce bathymetry through grid interpolation. However, for small or shallow waters where ships are not easy to navigate it would require other costly surveying strategies. Therefore, the maneuverability of sensor platform has been growing interest in related research fields. Reducing cost of commercial unmanned aerial vehicle (UAV) in recent years has facilitated its wide applications in geoinformatics. Owing to its high mobility, an innovative platform for point measurement is easily accessible. In order to understand water depths of small inland waterbodies and to calculate temporal changes of water volume, this study aims to combine an UAV, a smartphone, and a low-cost sonar to measure several ponds in Taoyuan. This integrated measuring system is operated over water surface to measure pointwise water depths for bathymetry mapping. The results show that an internal accuracy of water depth measurements (standard deviation of repeat observations) in four randomly chosen ponds of Taoyuan has achieved <0.13 meters. On the other hand, the root-mean-square of difference (RMSD) between depth measurements and historical data from Department of Water Resources in Taoyuan city (after appropriate water level adjustment) is 0.15 meters. Meanwhile, errors of area estimates are in a range of 3.2–8.9%. We conclude that in the absence of shipborne measurements, this method can sufficiently provide depth and volume estimation at low time/labor costs for small ponds whose depth is greater than 60 k.