Semi-supervised Hash image search device based on Group Lasso

Abstract

The invention provides a semi-supervised Hash image search device based on Group Lasso, and belongs to the field of image search. The device comprises a preprocessing module, a training learning module and a calculation module, wherein the preprocessing module is used for identifying a tag image and a non-tag image in an image database, and preprocessing an input image, the tag image and the non-tag image; the training learning module is used for carrying out semi-supervised Hash learning based on the Group Lasso according to the preprocessed input image, tag image and non-tag image to obtain a binary Hash code corresponding to each image; and the calculation module is used for calculating a Hamming distance between the input image and each image in the image database according to the binary Hash code, and returning the image corresponding to a minimum Hamming distance as an image search result. In the device, an existing image data situation can be combined to effectively carry out modeling on an image data structure, required images can be quickly and accurately searched, the image does not need to be stored, and a storage space is greatly saved.

Description

technical field [0001] The invention relates to the field of image search, in particular to a semi-supervised hash image search device based on Group Lasso. Background technique [0002] The advent of the era of big data, the rapid development of Internet technology and the increasing popularity of imaging devices such as smartphones and cameras have made data collection of media resources such as images more and more convenient. In the era of Web 2.0, people are no longer satisfied with just using words to convey information, especially with the popularity of social software such as Facebook, Twitter, WeChat, Weibo, etc., people have become very proficient in using chat "emoticons" in daily life package”, small videos in Moments, voice messages, etc. These images, videos, audios and other massive unstructured data are growing at an alarming rate every day. According to a survey report by the market research company IDC, 80% of the data in the world are unstructured data, ...

AI Technical Summary

Problems solved by technology

[0004] However, considering the status quo of image data labeling in real life, especially in the fields of image search and recognition, the largest amount of image data that is most convenient to obtain is unlabeled, and only a small part is obtained with manpower and material resources. labeled image data

Images