Research Of Construction Industry Behavior Abnormality Detection Based On Multi-Collaborative Vision Sensing Technology

Traditional construction industry behavior detection mainly depends on manual supervision through artificial supervision and inspection methods such as safety hazards and unsafe behavior,thereby taking corrective measures.It will be very human cost and time cost,and cannot eliminate the artificial process supervision personnel errors caused by subjective factors and personal lapses.With the development of computer vision technology and intelligent mobile devices,the analysis of many visual data can effectively assist traffic detection,engineering supervision,and agricultural monitoring,etc.,which provides a new opportunity for behavior detection in the construction industry.The collection of Visual data depends on the effective collaborative work of multiple Visual sensors.Visual Sensor Networks(VSN)can obtain large multimedia signals,such as images or video data,which can be used to assist in the detection of abnormal behavior in the construction industry.In the research of behavior anomaly detection in the construction industry based on VSN,opportunities are accompanied by challenges.It is the primary task of research that detects abnormal behavior in the construction area by the Effective monitoring of VSN.However,in VSN,the node’s deployment scheme directly determines the coverage quality of the Field of Interest(FOI),and the perceived coverage of the FOI of VSN is the most important indicator to measure the monitoring quality of the visual sensor network.Unlike traditional wireless sensor networks(WSN),visual sensors are much more sensitive to obstacles than other sensors,and obstacles directly affect the seffective sensing range of visual sensors.At present,a lot of research on the optimal deployment of visual sensor networks is mainly assumed within the non-obstacle area,ignoring the actual multi-obstacle situation,which is largely limited by the actual scene model simulation.Building Information Model(BIM)technology can automatically provide geometric and non-geometric attributes of actual scene elements and provide reliable scene Model input data for optimization algorithms.Therefore,to address the above problems,this paper studies how to effectively use VSN to conduct abnormal behavior detection in the construction industry using BIM technology and computer technology.The main work of this paper is as follows:1.This paper proposes a BIM-SAMDE framework,which integrates the advantages of BIM technology and optimization algorithm so that the model can automatically and accurately extract real scene data.This framework reduces the time consumption caused by manually enter the optimization algorithm input data.This paper also proposes an improved differential evolution algorithm called SAMDE,which uses adaptive segmented migration ideas to find the best solution efficiently and accurately.2.The problem of VSN node deployment optimization in a static multi-obstacle environment is deeply studied,and mathematical modeling is described with the optimal objective of maximizing the effective coverage of VSN nodes.A ray scan analysis algorithm is proposed to analyze the effective sensing coverage of nodes,and the SAMDE algorithm is used to find the optimal solution.A static obstacle scene experiment verifies the performance of SAMDE and the intelligence of the BIM-SAMDE framework.3.This paper discusses the optimal deployment of VSN in a dynamic multi-danger level environment.The regions with a high probability of abnormal behavior were classified into different levels.An optimization framework based on image technology is proposed,a data image is used to connect the BIM module,data modeling module,and optimization deployment and visualization module.The optimization results are visualized through data images,which makes the optimization deployment framework more intelligentized and more automated.A dynamic obstacle scene experiment verified the feasibility of the framework.The innovation point of the paper includes:1.This paper proposes a BIM-SAMDE framework integrating BIM technology and Segment-Adaptive-Migration Difference Evolution(SAMDE)optimization algorithm.It solves the difficulty of connecting the optimal deployment scheme of visual nodes with the actual scene model.2.An information fusion method based on image technology is proposed to transform the BIM extracted data into image data,which solves the problem that the discretization of scenes cannot directly evaluate the discrete results and visualization.3.This paper also proposes s multi-node effective coverage algorithm to assist the calculation of effective perception region.In addition,a nine-grid search algorithm is proposed to supplement it and improve the calculation accuracy.4.According to the probability of abnormal behavior occurrence,the author classifies the dangerous level of the detection area.In this way,the VSN can monitor the dangerous area more effectively.