Image Classification Method Based on Global Topological Representation of Visual Dictionary

Abstract

The invention discloses an image classification method based on visual dictionary global topological representations. The image classification method includes two steps of training and recognition, and further includes extracting features of target images with marked classes; performing global topological encode of the extracted features in the visual dictionary, and training the encoded results and modeling; and extracting features of images with unknown classes, performing global topological encode of the extracted features in the visual dictionary, inputting the encoded results to the trained model, and acquiring the classes of the target images. Since invariability of manifold representations of images of the global topological representations is provided, by the aid of the image classification method based on visual dictionary global topological representations, accuracy of image recognition is improved, and the image classification method plays a significant role in dynamic image understanding. By study of the visual dictionary of the global topological representations, the classes of images can be precisely recognized; and the image classification method can be widely adapted to the fields of security inspecting, web searching and digital entertainment and the like.

Description

technical field [0001] The invention relates to the fields of pattern recognition and computer vision, in particular to an image classification method based on the global topological expression of a visual dictionary. Background technique [0002] Image classification is a fundamental research problem in computer vision. Over the past few decades, researchers have attempted to study classification problems in terms of the physiology of human cognition. As an important basis, researchers believe that a basic element of image classification is the potential manifold expression of images. This manifold expression comes from neuron encoding, and the researchers proposed to use the topological relationship between neurons to obtain some topological invariance, and this invariance plays an important role in the manifold expression of images. Therefore, visual topology is a bridge connecting neuron encoding and manifold representation, and this relational mechanism has been widel...

AI Technical Summary

Problems solved by technology

Another limitation is that since the combination of visual words grows exponentially with the number of visual words, it is challenging to consider this high-order topology

Images