Crowdsourcing Assessment Model Combined With Quality Control For Collapsed Buildings After The Earthquake

The aerospace remote sensing technology has the characteristics of strong timeliness.At the beginning of the earthquake,it can quickly take and record the damage situation in the affected areas after the earthquake,providing rescuers with timely disaster area information,and thus carrying out targeted rescue work.After the earthquake,the collapse of the building will bring a large number of casualties.Observing the degree of damage to the building can determine the intensity of the earthquake and determine areas where relief should be focused.Houses under aerial remote sensing photos often lose complex geometric features and spectral features in photos,and they have complex features on the photographs.Therefore,depending on the condition of buildings damaged by computers,the accuracy of classification may not be high.At present,artificial visual interpretation is still a reliable method for remote sensing classification.With the advent of the Web2.0 era,the general public on the Internet can also participate in scientific issues.Crowdsourcing can gather the intelligence of the public on the Internet to complete disaster assessment missions and provide reference for disaster relief.However,due to the different knowledge backgrounds and seriousness of the workers,there is a problem that the quality of workers' answers is not reliable enough in the real crowdsourcing platform.This paper takes the damaged houses in the affected area of Jiegu Town in the Qinghai Province Yushu earthquake on April 14,2010 as the research object,introduces the quality control link,and proposes a new type of evaluation model for collapsed crowdsourced houses after the earthquake.The main research work and conclusions of this paper are as follows:1.Set up a crowdsourcing website for the collapse of post-disaster buildings with the B/S framework.A total of 3,450 experimental data were collected from 30 workers.Gold data and consistency test methods were used to control the quality of volunteers' assessment results.Kappa test was used to filter out volunteers with lower evaluation results to ensure the accuracy of the final assessment results.2.Using the building collapse assessment model to calculate the volunteers' assessment results after quality control,establish a probability model,combine the maximum expectation algorithm to quantitatively obtain the damage situation of the building,and make a special map of the damage situation of the building.The experimental results show that before and after the quality control,the evaluation accuracy of the collapsed buildings in the experimental area is 70.59% and 85.29%,respectively,indicating that the use of the post-earthquake collapse assessment model proposed in this paper can effectively increase the accuracy of crowd-sourced interpretation of disasters in the early post-earthquake period.Provide reliable visual interpretation of remote sensing results,provide reference for disaster relief,and provide a new model for using remote sensing images to assess natural disasters.3.The main conclusions of this paper are:According to the three types of building collapse types("basically intact","partially collapsed" and "completely collapsed"),the proportions were 36.52%,39.13%,and 24.35%,respectively.The results did not obviously favor one or two types of damage.Type;compares the results of the EM algorithm with the results of most voting algorithms and finds that the EM algorithm can yield more reliable results because the former combines the global data-driven individual error rate for each participant and the type of building collapse The posterior probability is more in line with the actual situation.According to the results of crowd-driven building collapse type analysis,it is analyzed that under the same damage type,the image features of buildings are highly consistent,and can be used as a training sample for machine learning algorithms.Identification.This study confirms the feasibility of using crowdsourcing to assess the collapse of buildings after earthquakes.It also demonstrates the advantages and application prospects of the model.