Research On Multi-Objective Optimization Model For Multi-Video Synopsis In Geographic Scenes

With the popularity of high-definition camera devices,the rise of the Internet of Things and the proposal of safe cities and smart cities,the expansion of surveillance cameras,complex scenes,numerous surveillance video objects and the difficulty of manually reviewing massive amounts of video,video enrichment technology has become a popular technology to solve this problem.The video synopsis optimization model is widely used in video synopsis optimization as a key model for comprehensive optimization of video concentration.Due to the transition from single-camera video enrichment to multi-camera video enrichment,some of the research ideas of multicamera video enrichment still follow the research ideas of single-camera video enrichment,which only extend the number of cameras and fail to effectively take into account the geospatial relationship between cameras and the spatio-temporal transformation of motion targets across cameras,resulting in the video enrichment optimization research neglecting the maintenance of spatio-temporal information of motion targets across cameras and motion target display across cameras.In response to the above problems,the research in this paper is as follows:(1)A multi-objective optimization model for geographic video synopsis is constructed to solve the problem that the traditional optimization model fails to take into account the geospatial relationship in multiple cameras and the spatio-temporal relationship of motion targets between cameras.The key steps of multi-video enrichment is analyzed from three aspects: motion target ontology construction,motion target trajectory arrangement and enrichment picture quality,and solves the problem that the global energy function of traditional optimization model fails to take into account the maintenance of spatio-temporal information of motion targets between cameras by establishing two optimization functions of motion coherence and motion target temporal consistency;solves the problem that the global energy function fails to take into account the maintenance of spatio-temporal information of motion targets between cameras by establishing two optimization functions of motion target display number,motion target By establishing two optimization functions of motion target display number and motion target display distance,we solve the problem that the global energy function fails to take into account the display of motion targets across cameras;by establishing the optimization function of compression ratio,which is used to balance the display arrangement of motion targets and prevent the synopsis time from being too long.(2)Based on the parameter method of geographic camera video synopsis multiobjective optimization model,the parameter method is proposed to improve and enhance the optimization effect of geographic camera video synopsis multi-objective optimization model.The proposed camera division scoring mechanism trajectory merging method optimizes the motion target trajectory coherence parameters and solves the missing of motion target cross-camera optimization based on camera division trajectory merging method;the proposed multi-video based temporal trajectory proximity fragment combination mechanism method optimizes the motion target intercamera and intra-camera timing,solves the missing of temporal trajectory proximity fragment combination mechanism on motion target inter-camera The proposed distance detection-based motion target trajectory calculation method sets a threshold value to detect motion target spacing,optimizes the motion target display distance,and solves the lack of optimization of the geospatial distribution of motion targets among multiple cameras by the distance constraint-based motion target trajectory calculation method;the proposed mechanism of displaying fixed number of motion targets in parallel optimizes the number of motion target displays,and solves the problem of The trajectory layout method fails to take into account the number of motion targets displayed,which makes it difficult for reviewers to track motion targets.(3)The study compares several multi-objective optimization algorithms and selects NSGA-Ⅲ as the calculation method of multi-objective optimization model for geographic video synopsis to solve the problem of difficulty in calculating multiobjective optimization by simulated annealing algorithm.Based on NSGA-Ⅲ algorithm,the local search mechanism is added to improve the efficiency of NSGA-Ⅲ algorithm to compute the model.