Debris flow detection using image processing techniques
This study is to detect the occurrence of debris flows and their characteristics by using machine vision techniques. Compared to conventional sensors for detecting debris flows, the image process of machine vision has the advantage of non-contacts and re-use. According to analyses of real time detection data, a decision can be made as to whether an alarm system should be announced to the public. The characteristics of debris flows determined by machine vision techniques can be used for fundamental research of debris flows. Prior to image detection or recognition, image preprocessing was performed on unclear images photographed in a harsh climate. Three preprocessing methods were developed to enhance image quality. They were noise removal, contrast stretch, and correction for non-uniform illumination. To consider various scenarios for detecting debris flows, three techniques for detecting debris flows were constructed: detection of specific moving targets; detection using textural features; and detection using filter banks. Videos of the historical debris-flow events and experimental flume tests were used to demonstrate the validity of the detection algorithms. Image preprocessing and detection were also integrated and tested at a field site to calibrate and verify the threshold values built into the models.
image preprocessing image recognition machine vision debris-flow detecting
S.Y. Chang C.P. Lin
Department of Environment and Resources Engineering, Diwan College of Management, Tainan, Chinese Ta Chung-shan Institute of Science and Technology, Taoyuan, Chinese Taipei
国际会议
成都
英文
549-560
2007-05-17（万方平台首次上网日期，不代表论文的发表时间）