Image Classification Recognition Judgment Method

Abstract

The present invention proposes a method for image classification, recognition and judgment, comprising the following steps: step 1, collecting images, preprocessing the images, removing image noise through the top-hat transformation method, and then using histogram equalization to enhance the images; step 2 , perform feature extraction on the enhanced image to obtain a joint histogram of N-dimensional vectors, perform dimensionality reduction operations on N-dimensional vectors, and the N is a positive integer; step 3, for the N-dimensional vectors of dimensionality reduction operations, pass the training set The classification set is summarized through training, and feature images are extracted through classification or screening.

Description

technical field [0001] The invention relates to the field of computer image recognition, in particular to an image classification recognition judgment method. Background technique [0002] HSI features are color features, based on the HSI color space established from the perspective of human visual perception. H is chroma, which is an attribute describing pure color; S is saturation, which gives a measure of the degree to which a pure color is diluted by white light; I is brightness, which is the brightness response of vision to light stimulation. [0003] The feature extraction method is to quantify the HSI color space according to the following formula, quantize the chroma H into 8 spaces, divide the saturation S into 3 spaces, and divide the brightness I into 3 spaces, that is, the color space is divided into 72 intervals. [0004] Construct a three-dimensional joint histogram of the three components of H, S, and I, but the HSI feature extraction method has no obvious ef...

AI Technical Summary

Problems solved by technology

[0004] Construct a three-dimensional joint histogram of the three components of H, S, and I, but the HSI feature extraction method has no obvious effect on feature image extraction in a complex image environment.

[0005] In addition, in the case of the SVM classifier, the effect of the complex image environment on the feature image extraction is not obvious, which requires those skilled in the art to continue to solve the corresponding technical problems

[0006] Also, the original image will always have some noise

For example, there are watermarks, uneven lighting, and insufficient lighting make the picture dark, etc.

Images