The purpose of this study is to collect ortho-images and point clouds acquired from UAVs in February and May to comprehensively assess whether they are suitable for time series offshore monitoring, such as volume change calculation and shoreline extraction. In February and May, UAV photogrammetry was performed at an altitude of 100 m using Zenmuse7 of Inspire-2 for the research area, and 245 chapters and 240 chapters were collected in ground sample resistance (GSD) 1.59 cm and 1.62 cm respectively. We obtained 40 and 21 ground control points (GCPs) that will be used for UAV photogrammetry and TLS surveying by using RTK-GNSS. The collected UAV images were treated as Pix4D mapper software. As a result, we deployed each point cloud and ortho-images in February and May. Image processing showed that the root mean square error (RMSE) in February was 0.015, 0.017, 0.040 m (x, y, z), and in May was 0.018, 0.015, and 0.035 (x, y, z).To verify accuracy, point clouds data collected with TLS surveying were collected. Using TLS point clouds and UAV point clouds (Feb, May), each DEM was deployed and the volume was calculated. In addition, a physical crosssection analysis was performed using 2 lines at the deployed TLS DEM, UAV DEM (2, 5 month). Finally, the coastline for the Imlang beach was extracted by applying the object based image segmentation technique obtained from UAV.
In this study, a seismic vulnerability of Gyeongju city, where the 9.12 Gyeongju earthquakes occurred, was analyzed and compared the prediction accuracy using frequency ratio (FR) and logistic regression (LR) models. The buildings damaged by the 9.12 Gyeongju earthquakes were used as dependent variables, of which the buildings were randomly selected data for training (70%) and validation (30%). The total eighteen seismic-related factors were used as independent variables as slope, elevation, groundwater level, distance to epicenter, distance to faults, peak ground acceleration (PGA), age of children, age of elderly, population density, building density, construction materials, number of floors, age of buildings, distance to police stations, distance to fire stations, distance to hospitals, distance to gas stations, and distance to road network. The spatial relationship between damaged buildings and seismic-related factors was analyzed using FR and LR models. The produced seismic vulnerability maps were classified into five zones, i.e., very high, high, moderate, low, and very low. The two maps validated and compared prediction accuracy using relative operating characteristic (ROC) curve and the areas under the curves (AUC). The validation results indicated that AUC value of FR seismic vulnerability map (73.1%) was about 3% higher than LR map (71.4%). The seismic vulnerability maps produced in this study could possibly be used to minimize damage caused by earthquakes and could be used as a reference when establishing policies.
Information on land cover is very important variable not only affecting on human activities but also studying the functional and morpho-functional changes occurring in the earth. The goal of this study is an assessment of support vector machine (SVM) for land cover classification over South Korea using normalized difference vegetation index (NDVI) of geostationary ocean color imager (GOCI). We collected level-2 land cover maps in South Korea and defined the seven most common land cover types (urban, croplands, forest, grasslands, wetlands, barren, and water) in South Korea to assess SVM model and produce land cover map. To train SVM model, we decided 1,000 training samples per classes. In addition, We repeated 50 times random selection of training samples. In order to evaluate accuracy of SVM`s kernels, we selected four kernels; linear, polynomial, sigmoid, and radial basis function (RBF). The parameters of each kernel were determined by the grid-search method using cross validation approach. The best accuracy of four kernel is linear kernel, the overall accuarcy was calculated 71.592%.
Land cover relates to the biophysical characteristics of the Earth’s surface, identifying vegetation, water, bare soil or artificial infrastructure. Land cover mapping is essential for planning and managing natural resource, and for understanding distribution of habitats. Land cover classification for land cover mapping has been developed in a variety of ways. Among them, there are many attempts to classification land cover using deep learning techniques such as Convolutional Neural Network(CNN). CNN has been developed in many models, and semantic segmentation techniques that combining segmentation are also being announced. Among the Semantic Segmentation models developed until recently, SegNet has high accuracy and learning efficiency. We analyzed the availability of SegNet in the Land Cover classification. The study area was conducted in parts of South Chungcheongnam-do in South Korea. For the learning of the model, 2,000 data were constructed with the same size using the aerial image, and the constructed data was divided into training and validation data by 8 to 2. To solve the problem of class imbalance, which causes problems such as overfitting due to the difference in area per class, the weight value of each class was calculated using medium frequency balancing method. In order to calculate the hyper parameter optimization, the batch size was changed from 1 to 5 and the iteration was changed from 0 to 100,000 times Our experiments show that an overall accuracy (OA) of up to 85%, which confirmed the positive possibility of the semantic segmentation technique in the study of land cover classification.
This study aims to produce landslide susceptibility map (LSM) using landslide conditioning attributes selected by different feature selection methods and compare predictive capability. Among the total 140 landslide locations, 98 locations (70%) were selected randomly for model training and remaining 42 locations (30%) were used to validate. Fourteen landslide conditioning attributes related to topography, hydrology, and forestry factors were considered. These factors were analyzed importance using four feature selection methods, such as information gain, gain ratio, Chi-squared, and filtered subset evaluator. From the results, the top seven attributes were selected and the LSMs were produced by random forest model. The results showed that the all LSMs had a prediction rate of more than 0.80 that yielded higher accuracy than the LSMs produced using all attributes. In addition, the LSM produced using attributes selected by gain ratio performed slightly better than another LSMs. These results indicate that the produced LSMs had good performance for predicting the spatial landslide distribution in the study area. In addition, selection of input attributes using feature selection methods was contributed to improve model performance. The produced LSMs could be helpful for establishing mitigation strategies and for land use planning in the study area.