Research And Development On Core Technology Of Intelligent Elevator Emergency Management

In view of the need for deep integration of artificial intelligence and real economy,this paper intends to give full play to the application of machine vision in the field of elevator emergency response service based on the elevator Internet of things platform.The traditional elevator black box uses infrared detection technology to identify whether someone is there in the elevator.Because the elevator belongs to the strong electric equipment,it is easily interfered by electromagnetic wave and other noise signals.In this paper,the video recognition technology can accurately judge the number of people in the elevator,which can provide more effective information for the rescue order and the rescue plan.At present,there are few personnel detection methods applied to elevators,most of which are applied to construction lifts.Traditional personnel detection methods include manual identification and electronic equipment counting.With the development of computer image processing technology,computer vision based method is applied to personnel detection of elevator.But deep learning algorithms generally require a lot of data training and computer configuration requirements.The existing elevator hardware configuration and computer vision technology can be more precise and convenient for personnel detection.The main contents of this paper are as follows:(1)Introduced the commonly used background extraction algorithms,and analyzed the various background extraction algorithms,summarized the advantages and disadvantages of each algorithm.(2)In view of the actual situation of elevator car video data,the most suitable video image preprocessing algorithm is studied,including histogram equalization,image binaryzation,morphological processing and image filtering.(3)In view of the simple and fixed elevator scene,an improved ViBe background modeling algorithm is proposed and combined with multi-frame differential method for personnel detection.Firstly,the concept of interframe processing is combined to increase the time of background modeling to avoid the static object being updated into the background,thus avoiding the introduction of the shadow area into the static object moving again in a short time.Then,the parameters of the algorithm are modified to achieve better target detection effect.Finally,the extracted non-background pixels are combined with different weights to detect personnel.(4)The implementation of embedded device algorithm is completed.A personnel detection system is designed and implemented.Based on Qt,a user graphical interface is developed independently.