Study On The Number Of Elevator Cabins Based On Convolution Neural Network

Elevator as commonly used tool in the high-rise building,its efficiency and safety must strictly guarantee,the elevator accident frequently in recent years,usually because of People's Daily misuse,cause long-term loss of elevator,so add the function of the number of detection in the elevator,and comply with the use of gravity detector,to avoid significant loss of personnel in the event of an elevator accident.In the past,the method of number counting was to use artificial extraction features,the accuracy of extraction was not high,and the degree of feature discrimination was not enough,convolutional neural networks could autonomously learn deeper image features,and improved the accuracy of detection,so this paper is proposed a method of counting people based on the neural network and experimenting in the elevator.This article is based on three-layer and multi-layer convolutional neural networks to achieve population statistics.Firstly,based on the three-layer convolutional neural network and ridge regression combined with the population statistics algorithm,the three-layer convolutional network is trained to learn the crowd density characteristics in the image,input regression model to forecast.an improved firefly algorithm of neighborhood renewal strategy is proposed to search for ridge parameters.Through the experimental comparison,the accuracy of the model is improved.followed by the multi-layer convolutional network Alex Net model,based on the Caffe open source framework,by studying the parameter influence on the performance of the model,adjusted the network layer and related parameters,combining the sliding window design,statistics on the number of people were achieved.Finally,based on Windows platform,a set of software of population statistics is built using MFC framework.