Study On Detection Of People Flow Distribution And Lighting Energy-saving Control Method Based On Edge Computing

With the rapid development of emerging information technologies such as the Internet of Things,big data and artificial intelligence,the lighting field has entered a new era of intelligent lighting from a single lighting mode.By introducing artificial intelligence to improve the quality of lighting and reduce the energy consumption of lighting operation,it has become the key to improve the overall level of our lighting industry.At present,the traditional lighting system in public places has some problems,such as insufficient perception ability of lighting environment and simple control mode,which leads to serious waste of public energy consumption and can not meet the requirements of comfort lighting quality.Therefore,this paper proposes to use cameras for body perception and illuminance measurement.Based on the current monitoring system in public places,this paper implements a lightweight flow distribution detection algorithm in the edge equipment to realize fast and efficient flow distribution perception and provide information support for lighting control.At the same time,the optimization control model of multi-equipment cooperation is established to control the energy saving of lighting equipment,and realize the comfort lighting and green lighting in public places.The main study contents are as follows:I.The architecture of intelligent lighting system based on AIoT is designed.On the basis of investigating the lighting control requirements of subway stations,hospital halls,classrooms and other public places,the overall scheme of lighting environment perception and lighting optimization control based on digital image processing was designed,and the lighting control system architecture based on the Internet of Things and artificial intelligence was established.Through sensor data acquisition,processing and decision control in edge devices,problems of cloud computing such as large delay and high bandwidth are solved.II.YOLOv5 network is improved by light weight and its model is compressed.In view of the limited storage resources and computing resources of edge embedded devices,the YOLOv5 for the detection of human flow distribution was lightweight improved,and the calculation method of lightweight NMS was designed.By introducing the center distance of IoU and simplifying the calculation process of NMS,the efficiency of network forward reasoning was effectively improved.In the process of model quantization,a quantization parameter determination method based on MMSE was proposed,and the problem of serious precision loss caused by improper quantization parameter values was solved.Finally,the lightweight YOLOv5 was deployed to an embedded device for experimental testing.The experimental results show that for the image input of the same size,the volume of the lightweight model is reduced by 89.2M,the frame rate is increased by 12.1FPS,and the mean Euclianite distance and mean cosine distance after normalization before and after quantization are 0.08 and 0.99,respectively.The model precision loss is small,and the model compression method to minimize the accuracy loss is realized.Ⅲ.The optimized lighting model is established and the lighting optimization control is carried out.By analyzing the causes of high energy consumption of lighting in public places,the optimal control model of multi-lighting equipment is established.In order to solve the precocious convergence problem of sparrow search algorithm,a multi-strategy improved sparrow search algorithm is proposed.By introducing chaotic initialization,Cauchy variation and historical optimal factors to the sparrow search algorithm,the precocious convergence problem that the sparrow search algorithm is prone to fall into local optimal is solved.Experimental results show that the lighting control optimization method based on the improved sparrow search algorithm can find the combination of optimal dimming coefficients quickly and accurately.Compared with the full-open control,the power under four kinds of personnel distribution is reduced by 141.12w,186.48w,168.84w and 216w respectively.Achieve the maximum energy-saving requirements on the premise of meeting comfort.Finally,AIoT intelligent lighting control system has been built,and client Web management and control interface,cloud server and edge embedded devices have been designed and developed to meet the remote and unified management and control requirements of devices.Finally,the overall energy saving performance of the system is tested and analyzed.The experimental test results show that compared with direct switch control,the lighting energy consumption of this system can be reduced by 55.18%,realizing green lighting and intelligent lighting.