Image classification method through fast and locality-constrained low-rank coding process

Abstract

The invention discloses an image classification method through the fast and locality-constrained low-rank coding process. The method comprises a dictionary learning part, a training part and a testing part. During the dictionary learning part, a corresponding dictionary is obtained based on SIFT features extracted out of all data sets according to the k-means clustering method or other dictionary learning methods. During the training part, one part of images in each data set are selected as training images, and then the corresponding low-rank representation corresponding to each data set is learned based on the tags of the images. Through the SPM quantization process and the SVM classifier-based training process, category labels and a kernel function corresponding to a feature representation matrix are obtained. During the testing part, all remaining images in each data set are adopted as to-be-classified test images. For any unclassified image, the SIFT features of the image are extracted firstly, and then a low-rank coding matrix thereof in the dictionary is found out. Through the SPM quantization process, a final feature representation matrix is obtained. The final feature representation matrix is input into an SVM classifier obtained during the training part, so that the category of the image can be known.

Description

technical field [0001] The invention relates to a method for classifying images by using a fast local constraint low-rank coding method, which is fast and has high precision, and is a data-driven classification method based on image content. Background technique [0002] Image classification is an important research direction of computer vision, which includes image preprocessing, feature expression and classifier design, where feature expression includes image feature extraction, feature dimensionality reduction, and feature encoding. Feature extraction is mainly to describe the image through attributes such as color, brightness, texture, shape and spatial distribution of pixels displayed by various contents in the image. Feature expression refers to statistics (vector quantization), coding or other methods based on the most basic features to form the final features of an image, usually with better performance than the original basic features; good feature expression It ca...

AI Technical Summary

Problems solved by technology

Fast Low-rank Representation based Spatial Pyramid Matching (LrrSPM) image classification method based on spatial pyramid, the classification speed of this method is 5 to 16 times that of ScSPM, but the classification accuracy is lower than ScSPM and LLC methods

[0005] At present, most image classification methods use sparse coding and low-rank representation coding methods, but since sparse coding is sensitive to image changes and noise, it consumes a lot of memory; low-rank representation coding methods achieve a balance between accuracy and computational complexity

Images