Research On Pedestrians Flow Measurement Method Based On Visual Dynamic Monitoring System Of Uav

In recent years,the four-rotor UAV has become a hot spot for people and business companies because of its unique structure,easy operation and other characteristics.The application fields of UAV have also been transformed from the military field slowly to the field of civilian and entertainment.The human flow statistics is the most significant part of the research in the field of safety monitor,and making use of unmanned aerial vehicles for human flow statistics on some occasions is another application area.This paper takes the UAV as a platform and combined with vision system and based on the computer vision technology,then the pedestrian detection and tracks counting method is studied.The main research work is as follows:Firstly,it summarized the importance of human flow statistics,introduced the development process and application of UAV and computer vision technology,and the present situation of pedestrian detection at the domestic and abroad and the present technology situation of human flow statistics.Secondly,for the four-rotor unmanned aerial vehicle,this paper introduces simply its structure and flight mode,designs the overall scheme of the visual tracking system,introduces its composition of hardware,introduces the visual module in detail,and summarizes workflows of the visual tracking system.Then,based on the noise pollution of the collected video images for pedestrian detection process,the preprocessing is carried out by using median filtering,gray level correction and histogram equalization.By comparing the advantages and disadvantages of frame difference method,background difference method and Optical flow method for the pedestrian extraction part,and finally proposes K-means to initialize the parameters of the Gaussian mixture model and estimates the number of mixed Gaussian model.And then use the background subtraction method to carry out the target extraction,and for the shortcoming exists in the object detection of pedestrian movement and by using morphological processing,according to the detection of the results to achieve the requirements of the late count.Finally,based on the problem of pedestrian target tracking,this paper introduces the algorithm based on region matching,model matching and feature matching,and all of the advantages and disadvantages of these algorithms.Finally,a classical and predictive Kalman filtering tracking algorithm is proposed.And the pedestrian tracking and counting program is developed.After the verification of the engineering experiment,the desired goal is achieved.The entire design requirements are completed.