Detection And Tracking Of Moving Objects Based On OpenCV

Intelligent monitoring includes video source acquisition,image preprocessing,target detection and tracking,behavior analysis and judgment warning.It combines advanced technologies in many fields,including automation control,pattern recognition,etc..In the community,hospitals,schools,supermarkets and other occasions?In areas,hospitals,schools,supermarkets and other occasions,we often see the figure of surveillance.Not only is it closely related to people's livelihood,but also industrial control,military guidance and so on.With the rapid development of the information society,the traditional manual model cannot meet the demand and has begun to change to automation.As a sign of intelligent monitoring,such as behavior analysis,early warning and so on often rely on the early evidence provided by the system-detection and tracking.It is very important for the system to find the target which people need or are interested in accurately and quickly.Therefore,after studying the literature background,this thesis mainly studies the detection and tracking of moving objects in video.In the implementation of detection and tracking issues,it depends on the VS platform and OpenCV open source library.The open source library OpenCV provides a good encapsulation interface API.For various situations in image processing,not only basic functions such as image reading and gray processing are processed.For video processing,There are also mature encapsulation libraries such as ML and video.In addition,there are many standard image algorithms,such as Gaosi filtering,wavelet algorithms and so on.Using the established platform and tools,the image and processing technology are discussed under the condition of static background,the factors affecting video processing in the early stages are analyzed,and the relevant comparative experiments on object detection and tracking are emphasized,and improvements are put forward.And give an analysis of the final results:(1)In object detection,the algorithms and flow of Gaussian background modeling,mean background difference modeling and frame difference modeling,and improved Gaussian modeling with changed LBP characteristics are presented.The advantages and disadvantages of each algorithm are explained.The performance of several algorithms is compared in the experimental results.(2)In the tracking of objects,several classical algorithms are compared and their connotation and ideas are analyzed.This paper mainly introduces the knowledge of MeanShift,including its mathematical principle,kernel function,feature sub and so on.The traditional description is improved in light and template updating,and the performance of traditional and improved algorithms is compared.In general,in terms of detection,the improved algorithm presented in the paper has good performance and modeling time.However,the improved tracking description features can be kept from losing in comparison with the traditional tracking methods under the same conditions and achieve the expected results.