| Geological disasters,especially secondary geological disasters,are hidden from great disasters,and landslides is one of the most influential natural disasters in secondary disasters.According to statistics,there are thousands of geological disasters in China every year,and the losses are countless.Under the combined influence of various influencing factors,the probability of causing secondary disasters is very high.Therefore,the timely extraction of landslide information and risk assessment research has certain guiding significance for landslide warning,earthquake prevention and disaster mitigation.In 2008,an earthquake of magnitude 8.0 occurred in Wenchuan,causing heavy casualties and various geological disasters,which has had a serious impact on the local area.It has been more than ten years since the Wenchuan earthquake,although the disaster area has been reconstructed and repaired,the impact of secondary disasters is still there,and geological disasters continue to occur yet.This paper carried out the landslide information extraction and landslide risk assessment research in Weizhou Town,Wenchuan County,Sichuan Province.Weizhou Town is the county town of Wenchuan County,and it is also one of the towns that suffered severely from the 5.12 Wenchuan earthquake.In this paper,we use the 2015 GF-1 imagery and DEM as the data source.Using pixel-based methods such as Maximum Likelihood Classification(MLC),Support Vector Machine(SVM),Neural Network Classification(NNC),ISODATA and K-Means and object-oriented fuzzy classification Method to extract landslide information respectively.Based on the extracted landslide information and the selected evaluation factors,the 30m*30m grid is used as the evaluation unit to establish and verify the risk assessment result that based on the methods of the hierarchical information model and the analytic hierarchy process(AHP).The main achievements are:(1)In order to highlight the feature information about the image,and eliminate the influence of the atmosphere on the image information,we preprocess the GF-1 images data,including radiation correction,enhancement and cropping of the data.Finally,the basic data needed for the research is obtained.(2)Extracting the landslide information about Weizhou Town,Wenchuan County with the methods of pixel-based and object-oriented respectively.Firstly,according to the spectral characteristics of the GF-1 images,the classification of Weizhou Town is carried out by supervised classification and unsupervised classification.The confusion matrix is used to evaluate the accuracy of the classification results and the evaluation results show that the classification accuracy of the maximum likelihood method,neural network and support vector machine in the three supervised classification methods are 78.99%,80.07% and 80.11%,respectively.The accuracy of the ISODATA and K-Means methods are 49.50% and 41.22%,respectively;Secondly,on the eCognition software platform,combining the image with the elevation and slope data and using the multi-scale segmentation algorithm to classify the research area.Segmentation levels are established based on the segmentation scales 38 and 54 by the ESP2 optimal segmentation scale,combined with the object feature set,and established the rule of extracting the feature.Finally,we get the landslide information in the study area.Meanwhile,the extracted landslides were verified by the collected 339 landslide sample points and use the accuracy evaluation method provided with eCognition software for accuracy evaluation.The results show that the accuracy of the landslide extracted by the object-oriented method reaches 81.71%,which is in line with the requirements of remote sensing image information extraction.In addition,the accuracy of the extraction results is 92.60%.Comparing the accuracy of the two types of evaluations shows that the object-oriented results are better.(3)The analytic hierarchy model(AHP)and the hierarchical information model were combined with the extracted landslide data to evaluate the landslide hazard in Weizhou Town,Wenchuan County.Based on the actual situation of research area,this paper selects seven evaluation factors to establish a landslide hazard evaluation system,including elevation,slope,topographical moisture index(TWI),terrain roughness index(TRI),annual precipitation,enhanced normalized differential vegetation index(ENDVI)and distance to rivers.In the process of using analytic hierarchy process and hierarchical information models to evaluate the risk of landslides,the critical step is to apply the deterministic coefficient method combined with ArcGIS.The sensitivity of each evaluation factor to landslides is analyzed by the major range,thereby objectively ranks the evaluation factors.The evaluation factor ranking results are: annual precipitation > slope > terrain roughness index > elevation > distance to rivers > enhanced normalized differential vegetation index > topographical moisture index.Based on the ranks,we use AHP to establish an evaluation matrix and calculate the weight of the evaluation factor.By calculation,the weights of the evaluation factors are 0.336,0.221,0.143,0.099,0.062,0.035 and 0.025,respectively.Then,via the weight and evaluation layers,we obtained the risk assessment results of the analytic hierarchy model.Besides,in the evaluation of the hierarchical information model,combined with the calculated weight value,based on the established evaluation model,the information amount and weight are calculated,and finally the landslide risk assessment result map of the study area is obtained.(4)Comparing the accuracy of the two risk assessment results by prediction accuracy method.It has been found that the evaluation accuracy of the analytic hierarchy-information value model(81.6%)appears to be more accurate comparing to the analytic hierarchy process method(71.89%).Therefore,the hierarchical information model is more objective for risk assessment of the study area.An object-oriented approach can be used to extract landslide information more accurately and quickly.Based on the extracted landslide data,this paper uses two evaluation models to evaluate and analyze the landslide hazard in Weizhou Town,Wenchuan County,which can provide some reference for infrastructure planning,early warning and prevention of secondary disasters in the region. |
PDF Full Text Request