Research On Bus Passenger Detection Technology Based On MobileNet

Citizens living in cities often choose convenient and fast buses as the main means of transportation.Reasonable bus scheduling has an important impact on people’s travel flexibility and urban life experience.At present,the target detection algorithm has high detection accuracy,but the detection rate is low,so it is difficult to carry out real-time detection in the mobile terminal.This thesis is based on the detection algorithm designed by the lightweight convolutional neural network Mobile Net to detect the bus passenger dataset.The specific research content is as follows:(1)Design of bus passenger detection algorithm based on Mobile Net.In this thesis,the feature extraction part of SSD algorithm is replaced by the light network model Mobile Net based on deep separable convolution,and the bus passenger detection model is designed and trained on the bus passenger data set.Experimental results show that compared with SSD algorithm,the bus passenger detection algorithm based on mobilenet reduces the detection accuracy by about 3%,and improves the detection speed by 4 times.(2)Optimization of bus passenger detection algorithm based on Mobile Net.This thesis proposes an image compression algorithm based on the maximum value of regional pixels,which improves the problem that some common image scaling algorithms lose the original image feature information in the scaling process,retains the original image feature information to the greatest extent,and optimizes the bus detection model.The experimental results show that the detection accuracy of the optimized bus passenger detection model is improved by about 2% while keeping the detection rate unchanged.Based on the lightweight model Mobile Net,this thesis designs a bus passenger detection algorithm,and proposes an image compression algorithm based on the maximum value of regional pixels to optimize the bus passenger detection model.The experimental results show that the detection accuracy of the bus passenger detection algorithm based on Mobile Net is reduced by about 1% compared with SSD algorithm,and the detection speed is increased by 4times,which has practical significance for the implementation of bus detection algorithm..