Research And Implementation Of Key Algorithms In Intelligent Video System Based On Object Detection

With the development of intelligent technology and the popularization of network monitoring devices,video system has become more important to guarantee the stable and orderly of modern society.However,limited by human resources,energy and many other factors,the traditional video system relying on artificiality has been difficult to meet the demand of multiple devices,multiple scenes and multiple tasks in daily life.Therefore,we focus on how to design and implement an intelligent video system to solve these complex problems effectively,taking the illegal parking detection as the key function of our system.We also design illegal parking detection algorithms in different ways based on computer vision and deep learning,and evaluate results with real data to ensure our intelligent video system reliable.Firstly,this thesis introduces the research background and current status of video system and computer vision,expounding the importance of computer vision in intelligent video system.We design and realize the key function of the intelligent video system,illegal parking detection,based on object detection,describing the sub modules in algorithm flow,and real illegal parking data are collected to evaluate its effect.Next,this thesis summarizes some typical relationships in illegal parking,and then designs an illegal parking detection algorithm based on relationship detection,which is also verified its feasibility in real data.After the implementation of the key function,we propose a general framework of the intelligent video system with 5 sub modules,describe the basic functions and connections of each module,and realized them in Python.We also test the system and the results prove its efficiency and stability.Finally,we summarize the main work of this thesis and look forward to the improvement of algorithm and system in the future research.