基于无人机测量的泥石流固体颗粒图像识别与泥石流预警

Recognition of Geo-Materials Composition by UAV Measurements and Its Application to Early Warning of Debris Flow

强震导致崩塌与滑坡碎屑物在泥石流流域内广泛堆积,挤压侵占天然沟道,完全改变了沟床物源级配构成,泥石流激发降雨临界值急剧降低。震后泥石流物源粒径和沟道宽度的变化直接改变了震后泥石流启动过程及机制。泥石流预警模型并未考虑震后泥石流物源组成的动态变化作为主控因素,需要进一步改进。本文现场调查了泸定9·5地震后泥石流沟的形成区及流通堆积区松散堆积物组成,采用无人机航测采集岩土粒径,结合PCAS软件进行泥石流粒径识别,分别通过系统误差、高度误差及坡度误差三个方面进行误差分析,探讨无人机航测泥石流形成区沟道粒径方法的准确性;对泸定大渡河流域得妥乡内的洛进沟和两岔河沟进行粒径测量,结合泥石流1 h预报模型和改进的精细化泥石流1 h预报模型,得到这两条沟震前震后的综合降雨临界值,发现强震后泥石流流域颗粒粒径明显减小,两条泥石流沟的激发临界降雨量都大幅降低;通过误差分析,得到无人机近地观测结合软件测量研究区域内泥石流沟道形成区物源粒径的方法与人工现场测量之间的误差较小,可以用于泥石流物源粒径识别,且由于强震后泥石流的激发降雨临界值是一个动态变化过程,采用无人机可以高效识别泥石流沟道物源的变化。

Strong earthquakes lead to extensive accumulation of debris from rockfalls and landslides in debris flow basins,followingly which squeezes and encroaches on natural debris flow gully,completely changing the composition of geo-materials on gully bed,and sharply reducing the threshold of debris flow initiation induced by rainfall.Changes in grain sizes and channel widths in post-earthquake gullies directly altered the process and mechanism of post-earthquake debris flow motivation;however,available debris flow warning models did not interpret the dynamic changes in the composition of geo-source of debris flow as a main control factor,which needs to be improved and justified.In this study,debris flow gullies in the post-shock areas of the Sichuan Luding Earthquake on September 5 of 2022 were surveyed in their section of forming,transporting and accumulating debris flow;unmanned aerial vehicles(UAVs)were used to photograph geo-materials,followed by PCAS software to be used to identify particle sizes by image recognition;the accuracy of the UAV approach to measurement of grain size in the forming section was addressed through three aspects of error analysis of systematic error,height error and slope error,respectively.The geo-materials in the Luojin gully and the Liangchahe gully at Detuo town in the Daduhe River basin of Luding county,Sichuan province of China was remotely photographed and then interpreted for image recognition for cases study;the comprehensive rainfall threshold for pre-shock debris flow as well as post-shock debris flow in the two gullies were estimated by combination of 1-hour prediction model of debris flow and a refined 1-hour prediction model.It was found that the particle size in the two gullies decreased evidently after the shock,along with significant decreases in rainfall thresholds.By error analysis,it was obtained that the UAV approach of near-earth observation combined with software to determine the particle sizes of geo-materials in the source areas of debris flow gullies had minor deviations as compared with in-situ measurement,justifying the applicability of the UAV approach;moreover,since the rainfall threshold of post-shock debris flow exhibited a dynamic change,using UAV approach in any field survey to collect geo-material information is a matter of time-saving and efficiency.