Texture Image Classification Method and System Based on Local Binary Pattern and Zernike Moments

Abstract

The invention proposes a texture image classification method and system based on local binary patterns and Zernike moments, wherein the method includes the following steps: inputting texture images to be classified; feature and Zernike moment rotation invariant shape feature, among which, the local binary mode rotation invariant texture feature includes three types of signal, amplitude and direction; the local binary mode signal and amplitude rotation invariant texture feature and Zernike moment rotation The invariant shape features are combined; the chi-square distance between the image to be classified and the training image is calculated according to the combination result; and the above-mentioned chi-square distance is corrected according to the direction rotation invariant texture feature of the local binary pattern, and the training with the smallest distance is selected. The category to which the image belongs is the category of the image to be classified. According to the method of the embodiment of the present invention, the texture information of the texture image is made richer and more robust through the rotation invariant characteristic, and the classification accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a texture image classification method and system based on local binary patterns and Zernike moments. Background technique [0002] Texture classification is one of the hot topics in image processing and pattern recognition, and plays a vital role in the application fields of target tracking, recognition, remote sensing and similarity-based retrieval. [0003] Due to the non-uniformity of texture appearance, illumination changes and variability, it is very difficult to directly analyze the texture of objects. In the early stage, researchers mainly used statistical features to classify texture images, and the co-occurrence matrix statistical method was the earliest method used to describe texture features. [0004] In the 1990s, Gabor wavelet was used as one of the important tools for texture analysis. Although these methods have achieved excellent performance, they are ...

AI Technical Summary

Problems solved by technology

This tool uses the statistical information of the gray level difference between the central pixel and the adjacent pixels to describe the texture features of the image, but the existing LBP texture description only considers the phenomenon of this difference change, that is, the adjacent pixel is higher than the central pixel, so Set it to 1, otherwise, set it to 0, don’t pay attention to how much it changes, and don’t consider the direction information of the image

[0007] Although LBP in the existing methods has a good performance in texture image classification, it only describes the difference in local gray values ​​of the image, but lacks the overall shape and space representation of the image. Therefore, the description of the problem image is not comprehensive enough. And its classification results are not accurate enough

Images