Research On Landslide Monitoring And Early Warning Based On Multi-Sensor Information Fusion

Landslides are one of the most common and dangerous natural disasters,posing a significant threat to human life and property while also wreaking havoc on the natural environment.Many new landslide monitoring methods and early warning models have been proposed by academics recently,which is vital for landslide monitoring and early warning.This research proposes a landslide monitoring and warning approach based on multi-sensor information fusion to address the drawbacks of present multi-sensor information fusion technology applied to landslides.The following are the key findings of the study:(1)Set up a comprehensive experiment system for landslide simulation.The landslide simulation experiment field is designed to closely resemble the emergence of a natural landslide.Rainfall,soil moisture content,surface displacement,soil stress,and other landslide parameters are monitored using many sensors.The landslide monitoring host computer and mobile phone APP based on the Aliyun Internet of Things platform are created to track changes in various data in real time throughout the landslide process.By evaluating experimental data and looking for correlations between the variables in the landslide process,we can learn more about the gestation and development of landslides.(2)Due to the inadequacies of the back propagation(BP)neural network,the seagull algorithm(SOA)is employed to optimize the weight and threshold of the BP neural network.The SOA-BP neural network improved using the gull method has superior performance and higher prediction accuracy when compared to the prediction effect of the neural network before and after optimization.(3)The correlation link between multi-source variables in the landslide process is utilized to determine the trustworthiness of monitoring data when combined with SOA-BP neural network and confidence interval approach.The confidence interval of15 different environmental quantities is provided for a specific monitoring attribute,and the minimal confidence interval is chosen as the foundation for judging the attribute data’s reliability.(4)The landslide risk is calculated using the TOPSIS approach,and the entire landslide process is separated into three hazardous states: low risk,medium risk,and high risk.To determine the dangerous state of a landslide,a feature-decision level information fusion model of SOA-BP-DS is constructed.The landslide monitoring attributes are separated into two categories: rock and soil attributes and rainfall attributes.The probability of three landslide danger states is calculated using the SOA-BP neural network model,and the landslide danger state is preliminarily judged.Then,to increase the accuracy of landslide hazard state judgment,the revised DS evidence theory is applied to fuse the two components of the result information.