Multiclass image classification method based on active learning and semi-supervised learning

Abstract

The invention relates to the technical field of image information processing, in particular to a multiclass image classification method based on active learning and semi-supervised learning. The method comprises five steps: initial sample selection and classifier model training, BvSB active learning sample selection, CST semi-supervised learning, training sample set and classifier model updating and assorting process iteration. Through the operations of BvSB active learning sample selection, CST semi-supervised learning and SVM classification, the invention has efficient image classification effect under the condition of less manual tagging, does not increase overmuch computation burden, can quickly provide classification effects and also can take consideration of the demand of a classification system on the computation complexity.

Description

technical field [0001] The invention relates to the technical field of image information processing, in particular to a multi-category image classification method based on active learning and semi-supervised learning. Background technique [0002] As an important application in image processing, image classification has always been a very active field in image processing. In recent years, many researchers have done a lot of research on image classification, and proposed many classification algorithms, such as Support Vector Machines (Support Vector Machines, SVMs) method, artificial neural network (Artificial Neural Network, ANN), genetic algorithm (Genetic Algorithm, GA), AdaBoost, Random Forest (Random Forest), etc. Most image supervised classification algorithms are based on statistical models. Users need to manually label a large number of image samples, and then obtain the model through training from training samples with category labels. [0003] In practical applica...

AI Technical Summary

Problems solved by technology

At present, some experts in this field have discussed the problem of classification uncertainty measurement in multi-class classification active learning. Aiming at the problem that entropy cannot describe the classification uncertainty of samples well in multi-class problems, a method based on the most Active Learning Sample Selection Criteria for Best vs Second-Best (BvSB) Labels

This method better solves the classification uncertainty measurement problem of samples in multi-class classification problems, but this method does not consider the information contained in the remaining unlabeled samples in the sample set, so that the useful information contained in a large number of unlabeled samples is wasted.

Images