| Summary: | 碩士 === 國立成功大學 === 水利及海洋工程學系碩博士班 === 95 === This paper presents a method of using circular Hough transform to identify the positions of particles in an image of granular flow or debris flow and then to track their moving trajectories. The images used in this paper includes the video image of granular flows in a vertically rotating annular flume experiments, and the video image of a debris flow.
The experiments of granular flows in a vertically rotating annular flume were conducted by using two types of spherical granular particles：polystyrene pellets (having uniform size) and glass marbles (having non-uniform size). The spherical particles moving in the rotating annular flume were recorded by a digital video camera. The positions of particles in the image were identified by the techniques of the digital image processing, the Hough transform and the particle tracking velocimetry (PTV). The result showed that as long as the particle size in the image is larger than eight pixels and the edge of particles in the image is clear, the proposed method can effectively track the moving particles for the two kinds of particle materials used in this study. The circular Hough transform not only can identify the position of particle but also can estimate the diameter of the particle, and this function would be very useful in analysis of the particle sizes and size distribution in the granular flows with non-uniform particle sizes.
The proposed method of particle tracking was also applied to analyze the movement of the large gravels on the surface of a debris flow. Theoretically, the circular Hough transform could not directly used to track the gravels in debris flows because gravels in the flows are not spherical in shape. Therefore, the non-circular shape technique of circular Hough transform was used track the gravels in debris flows. If taking the Hough transform on the entire image of a debris flow, it would get many unnecessary data, and the author had not found a suitable method to delete these unnecessary data yet. Therefore, the author take the Hough transform on the region of interest of the debris-flow image instead of the entire image in tracking the gravel movement in debris flow.
|
|---|
