A high-dimensional data model algorithm for image recognition

Abstract

The invention discloses a high-dimensional data model algorithm for image recognition, comprising the following steps: S1, compressing a source image to obtain a thumbnail image by a processing device, and identifying the thumbnail image according to a direct recognition method; If the identification is successful, the process goes to the end process; otherwise, the next procedure is executed; 2,carrying out image enhancement on that source image to obtain an enhancement image, and identifying the enhancement image according to a direct identification method; If the identification is successful, the process goes to the end process; otherwise, the next procedure is executed; S3, compressing the source image to obtain a compressed image, detecting an inclination angle of the compressed image according to an image region identification method, and then correcting the inclination angle by using the source image and / or the compressed image to obtain a target image. The invention combines the direct recognition method and the region recognition method to form a progressive image compression recognition algorithm, which is superior to the direct recognition method and the region recognition method in speed and recognition rate. It can complete image recognition in a short time and achieve a high recognition rate.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a high-dimensional data model algorithm for image recognition. Background technique [0002] In this era of big data, we often deal with data. Usually we can use a vector to represent a data sample, and the dimension of the data is the dimension of the vector. For example, our common two-dimensional data and three-dimensional data can be visualized intuitively. Some data dimensions are very high, such as data samples describing faces, voices, etc., and their dimensions are usually as high as hundreds. The classification of data samples is carried out by simple European clustering, and most of the low-latitude data samples have good classification results. But in the classification problem of high-dimensional data, the classification method based on Euclidean distance usually fails. [0003] Therefore, it is meaningful to propose a simple and effective method for the...

AI Technical Summary

Problems solved by technology

This model has strong generality and adaptability, and can better solve the subspace classification and multi-manifold classification problems of samples of different dimensions, but it takes a long time due to the complicated process and the correction of non-slanted images.

Images