A Trajectory Optimal Combination Method for Two-Layer Video Summarization

Abstract

Disclosed is a trajectory optimal combination method for double-layer video abstractions. The method comprises the steps of (S1) adding trajectories of all targets in target trajectory sequence sets of basic paragraphs of the abstractions into target sets to be added, and calculating a target activity energy function Eam and a target time energy function Etm; (S2) calculating correlation energy Ecm and target total energy EOm of each target to be selected, and selecting the targets with the highest total energy to serve as optimal hitting targets; (S3) generating abstraction frame matching solution sets of the optimal hitting targets; (S4) calculating matching solution time energy EFtk and total energy EFk, and selecting abstraction frame matching solutions with the lowest total energy to serve as optimal matching solutions; (S5) moving the optimal hitting targets from the target sets to be added to added target sets; (S6) repeating S2-S5 until the target sets to be added are null sets, and obtaining the fusion sequence of each target in abstraction frames.

Description

technical field [0001] The invention relates to a trajectory optimal combination method for a two-layer video summary. Background technique [0002] The two-layer video summary based on object content is oriented to surveillance video, with "object" as the basic unit, through the analysis of the structure and content of the video, meaningful "objects" are extracted from the original video, and they are passed through the object The optimal combination of trajectories generates a compact video enrichment that can fully express the semantic content of the video, and condenses tens of hours of video into tens of minutes of summary video. [0003] The optimal combination of trajectories is to recombine M segments of target trajectories into N frames of video summary frames. The current mainstream trajectory optimal combination method relies on the genetic algorithm and the method of traversing all solutions to solve the optimal combination of trajectories. However, these two Th...

AI Technical Summary

Problems solved by technology

[0003] The optimal combination of trajectories is to recombine M segments of target trajectories into N frames of video summary frames. The current mainstream trajectory optimal combination method relies on the genetic algorithm and the method of traversing all solutions to solve the optimal combination of trajectories. However, these two There are two problems in various methods: First, the definition of the optimal solution is vague. It only considers the less collision between the target trajectories after combination, and does not consider the correlation between the target trajectory and the video summary frame, and there will be problems for some video summary results. The problem of uneven density; second, using the optimal solution method of the genetic algorithm will cause falling into a local optimal solution; third, the algorithm complexity is too high, the method of the genetic algorithm is related to the number of traversal times of the genetic operator, and traversing all the solution methods there will be The optimal solution can be basically reached after only one operation

Images