Design And Implementation Of A Mountain Landslide Concentrator Based On Multi-sensor Information Fusion

Landslide is a common natural disaster,which is a kind of landslide energy accumulation caused by changes in mountain structure and external environment.When the sliding force of the mountain is greater than the anti-sliding force,a landslide occurs.The accumulation of landslide energy is a slow process with certain regularity.The sensor can effectively monitor this process,and the monitoring data can determine the time range of the landslide through big data trend prediction,which can minimize the damage of person and property.Therefore,it is of practical significance to establish a landslide forewarning mechanism based on multi-sensor and cloud platform,and the concentrator is the key device in the landslide forewarning mechanism.The thesis uses embedded system processing technology,digital image processing technology,multi-sensor information fusion technology and wireless communication technology to complete the landslide concentrator system.The system mainly includes 433 RF communication module,digital camera module,main control module,liquid crystal display module and 4G wireless communication module.The sensor system completes the collection of mountain parameters such as mountain temperature,humidity,rainfall,pore water pressure,earth pressure,vibration intensity and displacement,and it uses the 433 RF module to realize the data transmission and communication with the concentrator system.The digital camera module mainly completes the acquisition of the mountain image and transmits the obtained digital image information to the main control panel.The main control module extracts the mean,variance,skewness and kurtosis of each component of the RGB and HSV images of the mountain as color features,and converts the RGB images into grayscale images to obtain the entropy,energy,inverse gap,moment of inertia and correlation of the gray level co-occurrence matrix as texture features,and the mean and standard deviation of the gradient are obtained by Sobel operator as edge features.By transplanting the dichotomous model based on the Support Vector Machine(SVM)method under the MATLAB platform of PC,the three image features are merged and used as model input variables to complete the classification and recognition of the mountain image at the concentrator.At the same time,the concentrator unifies the multi-source heterogeneous sensor data format,and extracts its anomalous characteristics based on the Pauta method.Besides,the autocorrelation function(ACF)method is used to extract its stationary characteristics,and the multi-sensor initial fusion is completed.The function of the liquid crystal display module is to display the image information of the mountain and themulti-sensor data information in real time.The 4G wireless module mainly completes the data transmission and communication between the concentrator and the server of the host computer.The landslide concentrator designed in the thesis can complete the real-time processing of multi-sensor data in the laboratory environment.It can calculate and fuse the relevant characteristics of the data,and complete wireless transmission and communication with the host computer.At the same time,it can realize the state recognition of the mountain images,upload only the abnormal images of the landslide,reduce the amount of data transmission,and improve the monitoring efficiency.Integrated the sensor system and host computer server to realize early warning and prediction of landslides.The experimental results show that the concentrator system developed in the thesis can basically meet the design requirements,and achieve the expected goals in terms of real-time,accuracy and effectiveness,and the experimental results are good.Compared with previous mountain image classification and recognition method based on the single feature,the accuracy rate of the system is improved.Also,compared with the traditional monitoring effect of landslides,the concentrator system has certain improvements,which provides characterization data research foundation for landslide prediction in decision-making layer,and has certain reference value for the study of the whole landslide monitoring system.