Research On Detection Technology For Escalator Safety With Mechanical Vision

With the development of urbanization in twenty years,the number of escalators has risen sharply in China.More and more escalators are applied to the public places such as subways,airports,railway stations and shopping malls which bring great convenience for our daily life.However,the escalator accidents occur frequently at the same time which attracts public highly attention for their safety or not.The traditional escalator safety detecting system is mainly required to detect the components of broken-down escalator.Therefore,the escalator has always brought the irreversible personal injury once the accident occurred.In addition,the escalator in normal operation cannot prevent the potential accidental risk in ahead,such as the passenger body lean against balustrade,child play entrance section or climb the handrail etc.The problems above will hopefully be solved by image recognition function with the mechanical vision.We are exploring a new escalator safety detecting system technology depending on the effective neural network which base on mechanical vision system with the deep learning system that train the machine by Caffe(A frame of deep learning)and convolutional neural network as the algorithm.Trying to use VGG16 model to detect the danger pose instead of human pose estimation with skeleton model.On image data set,We will create special data set,which is classified base on danger pose and danger level,instead of general data for training and testing deep learning model.Detecting the risk and giving warning signal by the finished-train deep learning model with analyzing function of the operating escalator.