Method Research On Landslide Change Monitoring Based On Long Time Series Multi-source Remote Sensing

The change monitoring of landslide recovery by remote sensing is an important part of landslide monitoring system facing the whole process of disasters,which can bring great ecological benefits.During the phase of landslide recovery,the landslide and its surrounding areas are ecologically fragile and prone to cause secondary disasters and ecological deterioration,which poses great challenges to the prevention and treatment of landslides.In order to obtain information on the dynamic changes of the phase of landslide recovery and assist in the prevention and treatment of landslides,this paper presents a method of change monitoring of landslides based on long time series multi-source remote sensing for long-range multi-source remote sensing data of medium-spatial resolution to improve the efficiency and accuracy of landslide change monitoring.Firstly,combined with texture feature analysis and NDVI optimization,NDVI is corrected to different degrees in areas of different soil complexity,and the improved index is obtained,which has better data consistency and is used as an indicator for landslide change.With the use of the improved index,the problem of poor consistency of multi-source data has been improved to some extent.Then,based on spatiotemporal correlation,remotely sensed time series data for change detection has been designed.With the sequence of improved index obtained,the Pearson correlation coefficient is calculated and used to express the spatiotemporal correlation of adjacent phases.If the correlation coefficient is higher than the given threshold,a method of estimation based on similarity measures is proposed to monitor the landslides in the state of recovery.Otherwise,a method of landslide extraction based on the improved index is proposed to discover and monitor the newly generated landslides during the recovery phase.During the process,an iteration is designed,which not only simplifies the process,but also further strengthens the role of spatiotemporal correlation,thus effectively solving the problem that lacks consideration of spatiotemporal correlation in the process of change monitoring of landslides.The Wenchuan County center and its surrounding area was selected as an example to conduct relevant experiments,aimed at the landslides after the Wenchuan 5·12 earthquake.To begin with,in order to test the effectiveness and accuracy of the method of landslide extraction based on the improved index,Landsat5 TM and Terra ASTER data from 2008 were used to conduct an experiment(Experiment 1)according to the method of landslide extraction based on the improved index.Compared with the method of landslide extraction based on NDVI and several other methods of supervised classification,the method of landslide extraction based on improved index has higher recognition accuracy for the newly generated landslides,and it is more adaptable to different data sources.Then,in order to obtain information on the changes of landslides in the study area in the past ten years,eight of multi-source data from 2008 to 2017 were used to conduct the experiment(Experiment 2)according to the method of change monitoring of landslides based on long time series multi-source remote sensing.Through the analysis and comparison of landslide recognition of different methods in 2013 and 2015,it is shown that the method proposed is better than the method of landslide extraction based on the improved index and the method of dual-phase change detection using pre-disaster data.At the same time,the statistical analysis of the monitoring results of the experimental area in the past 10 years has resulted in some change characteristics of landslide recovery: the area of landslides has been shrinking year by year while local landslides would still be generated.The overall recovery was rapid during the former 5 years,but the rate of recovery slowed down during the latter 5 years,with the rate of recovery peaking in 2~3 years.The recovery of the along-river district was generally faster than that non-along-river district.The rate of recovery was also related to elevation and slope while it had no significant correlation with slope aspect.The conclusions obtained are consistent with some previous research,again demonstrating the effectiveness of the proposed method of change monitoring of landslides based on long time series multi-source remote sensing.