| Summary: | 碩士 === 國立中興大學 === 水土保持學系所 === 106 === Traditionally, the survey of grain-size distribution is mainly carried out by manual sampling. During the surveying-process, it needs to sample many analyzed areas and then analyzed for particle size. However, the process of sampling and grain-size analysis is time-consuming and labor-intensive. Some survey areas are dangerous, then we couldn’t sampled grain-size distributions of the overall river. This study presents a safety, high-mobility, and UAV-based image-processing system for the determination of grain size in gravel-bed rivers. The image characteristics of riverbed was photographed by the Unmanned Aerial Vehicles (UAV), and the automatic image-processing procedures were then used to recognize the outlines of the gravels. After that, the grain-size data were analyzed to obtain the grain-size distribution. In the image-processing procedure, the outlines of the gravels were recognized by using different methods, including the edge detection and the watershed segmentation and the improved watershed segmentation. The improved watershed segmentation was proposed by present study proposed to eliminate the problems of over-segmentation and under-segmentation existed in the previous studies. The indoor experiments were mainly used to check the accuracy of the gravel numbers and characteristic lengths identified by the system. Moreover, the outdoor experiments were used to evaluate the grain-size distribution produced by the system. The results of indoor and outdoor experiments showed that the improved watershed segmentation recognized the correct number of gravels. The average error percentage of the characteristic length is less than 9%. The average error percentage which improved watershed segmentation compared with the artificial methods is 1.38~4.99%. From the results of indoor and outdoor experiments, the improved watershed segmentation could accurately recognize the outlines and the correct number of the gravels, and acquired the distributions of feature-length similar to the results of artificial method. The UAV-based image-processing system proposed in this paper could effectively and correctly survey the grain-size distributions in gravel-bed rivers.
|
|---|
