Image set classification method and system based on representation learning reconstruction residual analysis

Abstract

The invention discloses an image set classification method and system based on representation learning reconstruction residual analysis, and the method comprises the steps: enabling the low-dimensional features of obtained image set data to be mapped into a high-dimensional space through a Gaussian kernel function, then learning a projection matrix through constructing a residual discriminant analysis model, the intra-class reconstruction residual error of the projected image set feature is minimum, and the inter-class reconstruction residual error is maximum. The method comprises the following steps: inputting image set data from a video sequence, a photograph album or a monitoring system; performing data preprocessing operation on the image set; inputting the obtained image data into a reconstruction residual analysis model for model training to obtain an optimal projection matrix; constructing a new image set classifier based on the optimal projection matrix; and inputting the test image set into the classifier for classification to obtain a corresponding category label.

Description

technical field [0001] The present invention relates to the technical fields of computer vision and machine learning, and specifically provides an image set classification method and system based on representation learning reconstruction residual analysis. Background technique [0002] With the continuous development of electronic technology, there are a large number of collection images in mobile phones, cameras or monitoring systems. Therefore, the classification of image collections has been studied in depth, and the classification of images based on collections has become more and more popular among researchers. their widespread attention. Different from the traditional single-image-based classification task, the set-based image classification task, that is, image set classification, can provide more abundant features for the object to be classified, and can effectively reduce the workload of data labeling. However, while the image set data provides rich discriminant in...

AI Technical Summary

Problems solved by technology

However, the deep learning model requires a large amount of data for model training, and it is a black box model with poor interpretability

[0013] At present, most of the deep learning methods are used to learn high-dimensional features, and then classify based on simple classifiers. Such methods rely too much on the parameters of the training features. If the parameters are suboptimal and the performance of the classification model is poor, it may lead to classification errors.

Images