A construction method of vehicle object multi-level knowledge dictionary for surveillance video compression

Abstract

The invention relates to a method for constructing a multi-level knowledge dictionary of vehicle objects oriented to surveillance video compression, which is used for expressing and compressing objects in surveillance videos, including texture layer knowledge extraction, and the texture layer knowledge includes different types of vehicles in multi-source surveillance videos common texture, use the texture dictionary to model the texture layer knowledge; structure layer knowledge extraction, the structure layer knowledge contains the common structure of the same type of vehicle, use the three-dimensional model to model the structure layer knowledge; personality layer knowledge extraction, the The above-mentioned individual level knowledge contains the transient knowledge of a vehicle individual that remains stable for a period of time, which is modeled using a residual dictionary. The method of the present invention establishes a multi-level knowledge dictionary of objects from three levels of texture layer, structure layer and personality layer, which can better express the appearance of moving objects in images in complex environments, and the information such as edge details is more abundant, which can improve the monitoring of vehicles. Tracking and identification efficiency can be widely used in intelligent transportation and other aspects.

Description

technical field [0001] The invention belongs to the field of surveillance video coding, and in particular relates to a method for constructing a vehicle object multi-level knowledge dictionary oriented to surveillance video compression, which is used to generate prediction references for moving objects in the surveillance video and improve the coding efficiency of the surveillance video. Background technique [0002] In recent years, the metropolitan surveillance network has become an indispensable urban infrastructure, playing an important role in the fields of intelligent transportation, smart city construction, fighting crime, and maintaining public safety. However, with the increase in the number of surveillance cameras and the high-definition trend of surveillance videos, the daily surge of massive surveillance video data poses a severe challenge to video compression efficiency. [0003] Existing surveillance video compression mainly uses a hybrid coding framework based...

AI Technical Summary

Problems solved by technology

However, 3D models usually do not consider physical factors such as artificial decoration and environmental factors such as lighting when modeling. This type of coarse-grained knowledge is quite different from objects in real videos. Only using 3D models to predict objects will lead to Larger prediction residuals reduce coding efficiency

Images