Elevator Operation Safety Monitoring System Based On MEMS And Machine Vision

The number of elevators in China has ranked first in the world,but the use of elevators presents the characteristics of high load,large transportation volume,and strong randomness,which cause elevator accidents to occur from time to time.Elevators,as a kind of special equipment that runs in a limited space and on tracked tracks,how to use fast-developing information technology to ensure the safe operation of elevators and promote the development of elevator safety monitoring is a hot spot for national and elevator industry researchers.This project is supported from the guiding project of Fujian Province.It aims to use Micro Electro Mechanical System(MEMS)nine axis inertial sensor to collect data and realize the estimation of the three-dimensional attitude of the elevator,and to use the face detection technology of the convolutional neural network to identify whether there are elevator passengers in the elevator car,and to obtain the number of people.Through the fusion of the two,a real-time monitoring system for detecting abnormal elevator operation status has been developed to provide novel ideas for improving the efficiency of elevator operation safety monitoring and ensuring the safety of elevator passengers.Firstly,in the elevator operation safety monitoring system,MEMS nine axis inertial sensor is applied to collect the multi-dimensional acceleration,angular velocity,magnetic field and other data information of elevator operation,and calculate the collected multi-dimensional information to obtain the elevator operation state data,such as three-dimensional attitude angle,operation speed and so on.When calculating the three-dimensional attitude angle of the elevator with MEMS nine axis data,according to the working characteristics of the elevator,the quaternion complementary filtering method is optimized to modify the gyroscope data,and the extended Kalman filter(EKF)method is combined to improve the accuracy of the attitude angle calculation;the three-dimensional attitude angle,three-dimensional acceleration and running speed of different elevators are compared and analyzed to provide key data support for elevator safety evaluation.Then,Multi-task Cascaded Convolutional Neural Network(MTCNN)face detection algorithm has the advantages of small model and fast operation,so it is applied to the elevator passenger recognition.Combined with the design idea of parallel operation(Inception)module of convolution kernels of different sizes,increase the depth and width of all levels of network and improve the feature extraction ability of network;at the same time,Batch Normalization(BN)layer is added to improve the model efficiency Convergence speed and strengthen the classification ability of the network.The experimental results show that the improved algorithm has higher face detection accuracy,the accuracy of elevator passenger face detection is improved by 2%,and the elevator passenger recognition with high accuracy is realized.Finally,based on the integration of MEMS elevator running state analysis and MTCNN elevator passenger detection,the abnormal state of elevator running can be judged;and the elevator running safety monitoring system based on the integration of MEMS sensor and MTCNN is designed at the mobile,and the performance and various functions of the system are tested.The developed elevator operation safety monitoring system is low-cost,easy to install,and can display the multi-dimensional data directly related to the elevator operation,the elevator passengers’ situation and the abnormal state of the elevator operation,which helps to ensure the safety of the elevator operation.