Research On The Number Of Small Target Tourists Based On Videos In Scenic Spot

In the const ruction of smart tourism,passenger flow of scenic spots,as an important indicator to measure the busy degree of a region,has been paid more and more attention by managers.As a research hotspot in the field of computer vision,pedestrian detection and tracking technology is widely used in intelligent video surveillance,passenger flow statistics and many other fields.In view of the low statistical accuracy of small target tourists in the existing statistical methods for the number of tourists based on videos in scenic spots,pedestrian detection and tracking methods are used to process tourist video to obtain passenger flow in this thesis.It improves the accuracy of statistical tourist targets and provides data support for scenic area managers to improve service quality.The main research work of this thesis includes :(1)The existing target detection model is not ideal for small target detection because of a large number of tourists in the videos.In this thesis,a series of improvements have been made on the target detection model,including expanding the detection scale and optimizing the loss function,so as to improve the detection accuracy of the model for small target tourists.On this basis,according to the real-time demand of passenger flow statistics in scenic spots,the detection speed of model is improved by using lightweight network design,and it has high detection accuracy.At the same time,aiming at the problem of serious occlusion between tourists in videos,Soft-NMS is used to process the detection results to reduce the problem of missing detection caused by occlusion.Through experiments on the pedestrian dataset of videos,it is verified that the model meets the real-time requirements and ensures high statistical accuracy.(2)A passenger flow statistics method based on deep learning is proposed for the statistics of tourists in the videos of scenic spots to obtain the passenger flow.This method is divided into three stages: pedestrian detection stage,multi-target tracking stage and passenger flow statistics stage.In the pedestrian detection stage,the target detection algorithm is used to detect and locate the tourists in the videos,and location information of all tourists are obtained.In the multi-target tracking stage,the multi-target tracking algorithm is used to correlate the detected target data and obtain the trajectory of each tourist.In the stage of passenger flow statistics,the trajectory of tourists is counted to obtain the passenger flow of scenic spots.(3)The method of crowd counting is used to solve the problem that the large error of target detection and tracking method in the dense crowd.When the crowd is too dense,the target detection algorithm cannot effectively detect all tourist targets,making statistical data error.The method of crowd counting can extract the head characteristics of tourists,generate a crowd density distribution map,realize the number of dense crowd statistics,and solve the problem of large statistical error of the number of tourists in dense crowds.