Research On The Estimation Method Of Catering Store Customer Flow Based On Video Analysis

Nowadays,intelligent video monitoring is an important application field of computer vision,and passenger flow statistics is one of the key technologies in this field in recent years.At present,passenger flow statistics have been widely used in the society,but because of the fast moving and dense target when the target enters the gate,there are still serious challenges to the passenger flow statistics.This paper takes the catering industry as the background,in order to realize the statistics of the passenger flow in each store,designs the strategies of target detection,target tracking and entrance behavior judgment,and puts forward the estimation method of the passenger flow in each store based on video analysis,the main research contents are as follows:The image enhancement pretreatment is carried out on the video image,and a comparison experiment is carried out on HE,AHE and CLAHE,and finally,a good processing effect is realized on the image under the catering background by using the CLAHE.In order to get the initial position information of pedestrian targets,the neural network model and detection algorithm of multiple convolution are analyzed,and the YOLOv1 detection model based on deep learning is constructed and trained.Through the experiment,the results show that the customer targets of different stores can be accurately detected.Aiming at the instability in target tracking caused by the scale change and the severe apparent change of the target tracking,a multi-channel scale adaptive target tracking algorithm based on a double correlation filter is proposed.Target apparent feature extraction is carried out by the fusion of grayscale,HOG and CN.The tracking robustness of the target apparent change is improved.The scale filter is established to realize the multi-scale judgment of the target,so that the tracking of the target can be stable when the scale changes.The behavior of the effective target is analyzed,firstly,a data association method of target detection and target tracking is designed,a detection algorithm and a tracking algorithm are effectively combined.The target occlusion behavior judgment method is designed so that the stability of the algorithm can be maintained when the passenger flow is dense.A method for judging the gate entry behavior of a target is designed,and the gate entry behavior of the target is judged through the overlapping rate,so that the accurate detection and counting of the gate entry behavior of the target are realized.Through experimental verification,the method proposed by text can achieve stable passenger traffic statistics under the condition of rapid moving and intensive targets,and obtain the average accuracy rate of passenger traffic statistics of 93.5% by using the statistics of 20 different catering stores.