The Establishment Of Sample Database Of Geological Disasters Based On UAV Images

Sichuan Province, located in the transitional zone between Tibetan Plateau and Sichuan Basin, is influenced by geographic factors like landforms, rainstorms and earthquakes, etc. Low altitude remote sensing technology by UAV is considered to be an effective technology, which is featured by good flexibility, high resolution and low price and widely applied into the tasks concerning the supervision of geological disasters and emergency rescue. Therefore, to establish the sample database of geological disasters based on UAV images can help rapidly extract and analyze the information on geological disasters in order to provide scientific support for the efficient investigation, scientific prevention and emergency rescue while in geological disasters.The research using the Object-oriented Classification Technology to extract the information on geological disasters in research areas and developing Python scripts in the ArcGIS environment established the sample database based on UAV images. Furthermore, the method of extracting information on geological disasters on the basis of transfer learning was put forward by using the established sample database on geological disasters and introducing the mechanisms of transfer learning. In addition, the information on geological disasters by using the methods of transfer learning and the extracted results employing the Object-oriented Classification Technology were comparatively analyzed. The main contents and innovative results were as follows.1. The Object-oriented Classification Technology was employed and extraction methods of information on geological disasters in UAV images in research areas were explained. The optimal segmentation scale was obtained by discussing Image segmentation and according to the combination of closest standard classification and degree of membership function classification the information on geological disasters was extracted by feature. Eventually, the categorized results were analyzed in accuracy and conformed with the norms practically.2. The UAV images were used to collected different categories of samples of geological disasters and based on the tool ArcPy in ArcGIS the research established a set of rapid tools for the establishment of sample database of geological disasters, which can to a larger extent improve the efficiency of establishing database.3. The established samples in geological disasters by introducing the mechanisms of transfer learning was employed to apply the feature extraction method from the course of CNN features into the extraction of information concerning geological disasters, which can realize the automatic extraction of information of geological disasters. And the extracted results and the information of object-oriented geological disasters were comparatively analyzed, leading to some insightful conclusions.