Abstract -- The purpose of this study is to establish the detecting system for the occurrence of the debris flow applying on the road and bridge within the debris flow hazard areas to protect pedestrians and vehicles from damage. The structure of the detection system was constructed fundamentally and the overall assessment of the effectiveness and feasibility of the detection system were then examined. Many of detecting sensors were widely investigated and evaluated in this study. They can be categorized into two types. One is the traditional sensors which are mainly touchable with debris flow including wire sensor, rainfall warming gauge and geophone etc., the other is a non-contacts detecting device by using "machine vision detecting". Compared to the traditional sensors, the image process of machine vision for detecting the occurrence of debris flow has the advantages of re-use. Flume tests performing debris flow were also conducted to verify the feasibility of machine vision detecting. However, the behavior of the debris flow is complex, more detecting instruments and robust image processing methods developed to tackle different patterns of debris flow under various scenarios are inevitable. Thus, more efforts will be made in a long run.