Study Of A Counting Method For The Number Of People In The Elevator

With the increasing number of building elevators in construction,the accidents of elevator happen have a larger probability.The number of people in the elevator has been restricted measure to avoid group-death and group-wound accidents.In this paper,the lack of security awareness of workers and safety regulation,the automatic statistical method in the special elevator were studied taking into account.After analyzing the actual working conditions of construction elevator,firstly study statistical programs using the infrared detector.In order to solve the accumulated error in infrared detection scheme,this paper using motion detection method based on image processing to clear erroring elevator car based on the human movement characterized,and conducted a field experiment.Considering the characteristic that worker must wear safety helmet in construction site,helmet as personnel features used to the number of counts.Firstly LBP pyramid characteristics of helmet has been extracted from different scenarios,and training the SVM classifier to identify helmet.To solve those problems of multiscale target and sticking target in crowd in close image in helmet detection,this paper proposes a multi-scale method of ROI image processing,setting different ratios for different image scales and establishing a unified scale image.Finally,the algorithm is applied to test the recorded video in work-condition.According to the result,the recognition accuracy using this method for the umber counting in building elevator car has been improved compared to other methods,moreover,the best advantage is the ability to solve complex background,and the image vibration caused by construction elevator,and multiscale problems in close image and target adhesion sand has a good real-time.This method can meet the engineering application demand well.