High-performance multi-layer dictionary learning feature image processing method for visual combat scene

Abstract

The invention belongs to the field of combat scene intelligence, and particularly relates to a high-performance multilayer dictionary learning feature image processing method for a visual combat scene. The method specifically comprises the steps of forming an initial dictionary; performing dictionary dimension reduction on the initial dictionary, and a preliminary feature library of the image target is formed; performing simplified sparse representation on the preliminary feature library to obtain a sparse dictionary; minimizing the sparse dictionary, and updating the initial dictionary to obtain a minimized sparse dictionary; generating a model from top to bottom, wherein the generation model belongs to a generation template model of dictionary learning; reconstructing a model from bottomto top, wherein the reconstructed model belongs to a reconstructed template model for dictionary learning; and forming effective identification and classification of targets by generating a templatemodel and reconstructing the template model. Aiming at visual image features of different combat scenes, target identification and classification are carried out by establishing a deep multilayer network, and the precision is improved; realizing reconstruction optimization through error iteration; and forming an optimal approximation direction by initializing the dictionary matrix.

Description

Technical field: [0001] The invention belongs to the field of intelligent combat scenarios, in particular to a high-performance multi-layer dictionary learning feature image processing method for visualized combat scenarios. Background technique: [0002] The visualization-oriented target display provides accurate target display features, and through the image processing architecture based on dictionary learning, and through dictionary feature extraction, it combines basic abstract functions such as texture enhancement. A large-scale dictionary data training is performed offline to form a template in a large-scale typical simulation scene, and a template model can be added for autonomous control. At the same time, it has the texture enhancement function of the scene, such as the shadow of the scene, texture and other special enhancement features. [0003] At present, it is difficult to achieve the goal of reducing the dimensionality of the dictionary by updating the diction...

AI Technical Summary

Problems solved by technology

[0003] At present, it is difficult to update the dictionary atoms through dictionary learning itself to reduce the dimensionality of the dictionary, and the image target feature selection process results in a large amount of computation.

Images