Crowdsourcing Assessment Of Collapsed Buildings After The Earthquake

Aerospace remote sensing technology has the characteristics of large area synchronous observation and high time-effectiveness,it has been playing an important role in earthquake disaster response and disaster assessment.Due to the difficulty of the large-scale field investigation early after the earthquake,the observation data of aviation and space is an effective means to evaluate the disaster quickly.Building collapse is one of the most serious types of earthquake damage.Most casualties from earthquakes are associated with collapsing buildings.Meanwhile,the extent of buildings damage reflects seismic intensity,which is important information to assess the losses of life and property in an earthquake-hit area.And knowing the location and the damage extent of collapsed buildings is closely related to life-saving for emergency response.Due to the collapse of buildings after the earthquake,the original regular geometry and spectral features were lost,and the features of the collapsed buildings in remote sensing image are complex.Therefore,the accuracy of computer algorithm for automatic identification of building collapse is not high.The visual interpretation has shown the great reliability and is the main method adopted at present.The development of the Internet makes it possible for the general public to participate in scientific issues.Crowdsourcing geographic information for disaster response has become a research frontier.In this paper,Yushu earthquake is a case study,we use crowdsourcing for the assessment of post-disaster building collapse based on high-resolution aerial remote sensing image.This paper presents a new framework of disaster information acquisition and processing for building collapse assessment based on crowdsourcing and remote sensing images.The main research work and conclusions of this paper are as follows:1.The experiment platform of crowdsourcing data collection is built by using browser/server(B/S)structure.On the server-side,WebGIS server is used for publishing the high resolution aerial remote sensing image of study area.On the frontside,users access the remote sensing image concurrently through the browser,and the rapid sharing of spatial information among users is achieved.In addition,users not only acquire information through the browser,but also create information through the browser,to achieve interaction with the server.The assessment results of building collapse in the study area contributed by users based on remote sensing image,are stored in the database of the server-side,to achieve asynchronous interaction with the server.2.A probability model is established based on the relationship among users,buildings and damage types.And apply it to all user data we collected.Considering the differences in participant contributions,we use the EM algorithm to quantitatively estimate the individual error rate of each participant and infer the damage type of buildings on the ground.An experimental area of Yushu earthquake is provided to present the results contributed by participants,and then the results are discussed.3.The main conclusions of this paper are: the proportion of three building collapse types(“basically intact”,“partially collapsed” and “completely collapsed”)we set are 52.14%,34.64% and 13.22%,respectively,showing that there is no clear bias towards one or two damage types.The results of EM algorithm and majority algorithm are compared,finding that the former is more reliable,because the former includes the global data driven individual error rate of each participant and the posterior probability of the building collapse type,which is more in line with the actual situation.There is no clear agreement between participants on each building apart from one building that all participants label as the same type.That is to say,there are three collapse types on almost every building,but the proportion of each type is different.According to the crowdsourcing-derived classification of building collapse types,the features of buildings labeled as the same damage type are found highly consistent.This suggests that the building damage assessment contributed by crowdsourcing can be treated as reliable samples.This study shows potential for a rapid building collapse assessment through crowdsourcing and quantitatively inferring “ground truth” according to crowdsourcing data in the early time after the earthquake based on aerial remote sensing image.