Research And Implementation On Infrared Pedestrian Traffic Intelligent Statistical Method Under Vertical View

The development of social mobility is convenient for people to travel,but also brings traffic congestion,frequent safety incidents and other issues,and pedestrians are one of the key factors of these problems.At present,some of the key scenes of human traffic statistics methods have exerted certain effects in practical applications,but there are still some shortcomings,such as low accuracy of identification statistics,high labor cost,and poor system reliability.With the continuous development and application of emerging technologies such as artificial intelligence,new pedestrian detection,tracking and statistical methods have emerged and have been applied in practical engineering.In order to solve the problems of inaccurate pedestrian identification and high labor cost in common methods,this paper proposes an intelligent statistical method for infrared pedestrian traffic under the vertical view.The method uses pedestrian detecting and tracking technology based on deep learning to identify and track pedestrians respectively,and then obtains the required traffic data by analyzing pedestrian movement trajectories.The experimental results show that the method has strong accuracy and real-time performance,it can effectively complete the human traffic statistics task under the monitoring scene.The main work of this paper is as follows:1.Aiming at the influence of different postures on the detection effect of pedestrians in infrared images,an optimization method based on pedestrian size ratio is used.This method optimizes the size and proportion of anchor detection window in Faster-RCNN algorithm,which not only improves the quality of suggest area also reduces the number of inspection windows.The experimental results also show that the optimization method improves the accuracy and speed of detection.2.Aiming at the influence of posture change on tracking accuracy duringpedestrian movement,an optimization method for indirect updating target template is proposed.When the tracking accuracy is lower than the threshold,the tracking result of the previous frame is added to the tracking of the current frame.Improved the ability of the SiamFC approach to respond to changes in pedestrian posture.The experimental results show that the optimization method improves the accuracy of tracking,but slightly reduces the tracking speed.3.Based on the above work,using the optimized pedestrian detection and tracking method,an intelligent statistical method for infrared pedestrian traffic under the vertical view is proposed.The method firstly performs pedestrian detection on the infrared image,acquires the pedestrian information in the image,and then tracks the detected pedestrians one by one,and finally obtains the human flow data by analyzing the pedestrian movement track.Based on this method and the actual application needs,the infrared pedestrian traffic statistics system from the vertical perspective is designed and implemented.The experimental results show that the statistics method can obtain the traffic data of the monitoring area for a period of time,which has certain practical application value.