Research On Risk Video And Early Warning Technology Of Deep Foundation Pit Construction

Construction industry is in full swing,in order to meet the fast development of our country's social economy and the urbanization process.With the rise of construction industry,as a special project in road and bridge engineering and highrise buildings construction industry,the deep foundation pit has become greater in number and even towards to a wide range and depth.Compared with other construction projects,deep foundation pit has more security risks in the construction process.According to statistics,the construction of deep foundation pit causes nearly one hundred accidents and nearly one thousand casualties every year.Therefore,the construction safety of deep foundation pit has become an increasing attention to the industry.However,the traditional detection methods are relatively outdated and cannot be found and managed in time.This paper is based on the third street planning project of China Railway Construction Engineering Group in Lin Fen city,Shanxi Province(project funding no.: 193040449-j).Aiming at the deep foundation pit project at the project site,an automatic real-time detection and early warning system is designed for the hazard sources that may cause casualties in the deep foundation pit.On the basis of literature survey and field investigation,hazard sources at the construction site of deep foundation pit project are classified,which are mainly carried out in two aspects: construction hazard area and unsafe behavior of construction personnel.In this paper,the moving object identification algorithm is adopted in the dangerous construction area.According to the mobile camera used in the actual scene,the existing moving target detection algorithm is improved to be applicable to this subject.There is a background pixel displacement because the camera is moving.Therefore,the background modeling method in the difference method is adopted,which is improved on the basis of the D-GSM background modeling method,and the motion compensation technology is used to transform the moving background into the moving target recognition under the static background.In terms of motion compensation,the original corner detection is improved for the purpose of this paper.The improved corner detection can effectively eliminate the interference of foreground pixels and conduct background registration for two consecutive frames.Finally,RANSAC is used to estimate the parameters of the monotonic transformation model to improve its robustness.In this paper,an initial motion recognition method based on critical bone Angle is adopted to identify dangerous behaviors.In this method,the identification of dangerous behaviors of construction personnel is mainly transformed into the identification of initial actions,and the key bone Angle of pre-action of dangerous behaviors is extracted and compared with the pre-set database of initial actions of dangerous behaviors to identify the construction dangerous behaviors Compared with the general machine learning method,this method has the advantages of less computation and faster recognition,and can meet the real-time requirement of this project.At last,the algorithm of hazard identification is transplanted,and an early warning platform for hazard identification is designed,which is verified and analyzed in the construction site of deep foundation pit.Finally,it is proved that the danger source warning platform has good practical value and opens up a new way of thinking for construction safety management.