Research On Dynamic Target Detection Algorithm For Intelligent Surveillance Video Image

The digital features of the intelligent video surveillance system play an important role in the military.With the improvement of people's safety awareness and the development of technology in the field of computer vision,the intelligent video surveillance system has been widely used in the ordinary civil scenes of life.Dynamic target detection is the core technology of intelligent video surveillance,and it is also a research hotspot in the field of computer vision.Dynamic target detection technology is the premise basis of machine vision,image matching,image detection,target tracking and pattern recognition.The quality of target detection directly affects the accuracy of subsequent target tracking and target positioning.This dissertation summarizes and analyzes the current research results of dynamic target detection in intelligent video surveillance,and proposes two new algorithms for dynamic target detection in video surveillance.The main work is as follows:(1)The experimental analysis of the two-frame difference method,three-frame difference method,SGM,and GMM is carried out.First,the GUI interface was designed,and four algorithms were tested on two video datasets highway and people with different lighting conditions,and then the results of the same video frame were compared and analyzed.The qualitative analysis and quantitative evaluation of the experimental results are provided to provide a basis for subsequent algorithm improvements.(2)Proposing an improved dynamic target detection algorithm based on GMM combined with shadow detection.First,the traditional GMM principle is studied to understand its detection characteristics.Then,the GMM proposes improvement measures in three aspects: model initialization,background model establishment and background model update,and combines with the Hotlin shadow suppression algorithm to complete the design of the overall detection algorithm.Finally,the effectiveness of the improved algorithm is proved by experiments and compared with other algorithms to prove its superiority.(3)An improved SACON algorithm is proposed.Firstly,the five-frame difference algorithm is improved,and the five-frame differential threshold selection is adjusted by the threshold selection method between the classes.Then the SACON algorithm is studied,and the improved method of threshold adaptive adjustment of SACON algorithm is proposed.Finally,the improved five-frame difference method is combined with the improved SACON to complete the dynamic target detection.The superiority of the proposed algorithm is proved by experiments.