Visual feature representing method based on autoencoder word bag

Abstract

The invention relates to a visual feature representing method based on an autoencoder word bag. The method includes the steps that training samples are input to form a training set; the training samples in the training set are preprocessed, and influences of illumination, noise and the like on the image representing accuracy are reduced; a visual dictionary is generated, an autoencoder is used for extracting random image block features, then the clustering method is used for clustering the random image block features into a plurality of visual words, and the visual dictionary is composed of the visual words; a sliding window mode is used for sequentially collecting image blocks of images in the training set, the collected image blocks serve as input of the autoencoder, and output of the autoencoder is local features of the images; the local features of the images are quantized into the visual words according to the visual dictionary; the frequency of the visual words is counted, a visual word column diagram is generated, and the visual word column diagram is overall visual feature representing of the images. By means of the visual feature representing method, feature representing characteristics are independently studied through the adoption of the autoencoder, and requirements for the quantity of the training samples are reduced through a BoVW framework.

Description

technical field [0001] The invention relates to a visual feature representation method in the fields of multimedia analysis, machine vision and the like. In particular, it involves a visual feature representation method based on autoencoder bag-of-words. Background technique [0002] Visual representation is one of the important issues in the fields of multimedia analysis, machine vision, etc. It profoundly affects many practical application problems such as multimedia retrieval, image classification, scene analysis, and object recognition. In recent years, with the rapid development of network technology, more and more digital images have flooded people's lives. How to accurately represent these visual data (especially in the environment of big data) has become an urgent problem to be solved. [0003] The bag of words (Bag of Words, BoW) model was originally applied in the field of document retrieval and classification. Its basic idea is: all the words in the statistical c...

AI Technical Summary

Problems solved by technology

Due to the inherent properties of the deep architecture (contains more hidden layers) and the use of the entire image as input, SAE is only suitable for situations where the image size is small and the number of training samples is large, while the image size is relatively large and the number of samples In rare cases, deep networks (not just SAE) are prone to overfitting, making it difficult to learn effective features

[0006] In the feature representation method based on deep network, because the deep network contains a large number of parameters, it is easy to overfit when there are few training samples; in the feature representation method based on visual bag of words, SIFT and other artificially designed characteristics, making it difficult to discover the inherent structure of the data

Images