Research And System Implementation Of Intelligent Construction Safety Control Technology Based On LCB Theory

The construction safety issues in the building sector have attracted increasing attention from the state,governments at all levels,and the main body of the i ndustry.Despite this,the construction industry still has one of the highest rates of safety accidents among the pillar industries in the country,causing huge economic losses and negative social impact.In order to effectively reduce the occurrence of construction safety accidents,based on the LCB(Leadership-Culture-Behavior)theory of construction safety,this paper analyzes the possible causes of safety accidents from three dimensions,including safety leadership,safety culture,and safety behavior.Then combined with the dynamic identification and early warning technology of unsafe behaviors based on deep learning,this thesis constructs a construction safety management and control framework model based on LCB theory.Based on this model,this paper develops a set of construction safety intelligent management and control system combining "edge end + cloud end",realizing the visual display of construction safety hazards and real-time safety management at construction site,effectively improving the safety management level of construction sites and reducing the occurrence of construction safety accidents.The main tasks are as follows:(1)In order to put the LCB theory into practice,this paper builds a construction safety management and control framework model.The model first collects the evaluation data,including manual input and automatic generation,and then with the help of the LCB evaluation management module in the cloud system,the collected data is comprehensively analyzed.In the end,the visualization results are displayed,which provides a reference for the effective intervention of unsafe behaviors.(2)In order to realize real-time supervision and early warning at the construction site and automatic acquisition of worker behavior data,th is paper uses an improved deep learning algorithm to automatically identify unsafe behaviors in on-site surveillance videos,and then completes the statistics of unsafe behavior data by edge computing.At first,the YOLOX model is used as the basic detecto r,the original FPN is replaced with a simplified BiFPN structure,and the attention module CBAM(Convolutional Block Attention Module)is added to enhance the feature extraction capability of the network for small targets such as helmets and reflective clothing.Then statistics of the unsafe behavior data identified by the model are performed in the edge system.The improved YOLOX-L algorithm can achieve an average inference time of 23.7ms per image on NVIDIA GeForce GTX 1080Ti(62G),ensuring real-time detection speed,and the mAP trained on the self-made construction site safety behavior dataset is 89.15%,which is 1.8% higher than the accuracy of the original YOLOX-L detection model.(3)Combined with the construction safety management and control framew ork model and the unsafe behavior detection algorithm,this paper builds a construction safety intelligent management and control system based on LCB theory and applies the system constructed in this paper to a practical case.The rate of unsafe behavior i s analyzed,and the result shows that the system has played a significant safety control effect.At present,the system has been deployed in more than 20 project sites of a group company in Jiangsu Province,which has improved the safety management level of the construction sites.