Multi-class Scene Classification Method Based on Incremental Linear Discriminant Analysis

Abstract

The invention discloses a multi-label scene classification method based on incremental linear discriminant analysis, which mainly solves the problems of image dimensionality reduction and multi-label classification in the prior art. The implementation steps are as follows: firstly, decompose multi-class labels into multiple single-class labels; secondly, use the initial samples to calculate the transformation matrix of the initial linear discriminant analysis, use the newly added samples to update the transformation matrix of the linear discriminant analysis, and convert the high-dimensional The data is projected into a low-dimensional space; then randomly select training samples and test samples from the sample set after dimensionality reduction, use a single-instance multi-label K-nearest neighbor classifier to classify the feature samples after dimensionality reduction, and obtain the output value of the test sample; Finally, the label of the test sample is predicted and the classification result is obtained. The invention has the advantages of higher classification accuracy and shorter classification time, and can be used to quickly and accurately process massive, high-dimensional and multi-category multi-class data classification problems.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a multi-class image scene classification method, which can be used to quickly and accurately process complex scene detection with rich features and a large number of categories. Background technique [0002] In recent years, as the capacity of computers and other data collection equipment has increased, the size of the data has become larger and larger. A large amount of data emerges due to high dimensionality and data augmentation. It is widely expected that time and resource consumption can be reduced by reducing the dimensionality of the data. Dimensionality reduction can greatly reduce the amount of data, and facilitate subsequent processing by mapping data from a high-dimensional feature space to a low-dimensional feature space. Projections preserve as much information as possible in high-dimensional spaces. A widely used supervised dimensionality red...

AI Technical Summary

Problems solved by technology

However, the incremental linear discriminant analysis algorithm proposed by T.-K.Kim et al. and the GSVD-ILDA algorithm proposed by H.Zhao et al. encountered the same problem, that is, it is difficult to determine the trade-off level between efficiency and performance.

If you remove too many minor components, the performance will deteriorate, otherwise it will be very inefficient

In addition, the performance is very sensitive to parameter settings, and it is not easy to adjust the parameters, resulting in unstable classification results

Images