Image retrieval and recognition method and device based on binary semantic embedding

Abstract

The invention relates to a method and device for image retrieval and recognition based on binary semantic embedding. The method includes: determining the objective function, using the label information of the sample image in the training set, the pairwise similarity matrix and the original visual information, and learning from the original image space to the binary semantic space to obtain the binary code library corresponding to the deep neural network retrieval model and the sample image; use the deep neural network retrieval model to map the image to be retrieved into the binary semantic space to obtain the corresponding code library of the image to be retrieved The first binary code; calculate the Hamming distance between the first binary code corresponding to the image to be retrieved and each second binary code in the binary code library, and arrange in ascending order; according to the arrangement result of the Hamming distance Determine the approximate nearest neighbor retrieval result; identify the category of the image to be retrieved according to the image category in the approximate nearest neighbor retrieval result by majority voting. Through this technical solution, graph knowledge can be fully embedded, and the accuracy of node classification can be improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image retrieval and recognition method and device based on binary semantic embedding. Background technique [0002] With the progress of society and the rapid development of science and technology, urban congestion is becoming more and more serious, and the resulting frequent traffic accidents have gradually become an important issue that threatens people's lives. Therefore, the intelligent transportation system came into being, which is a comprehensive system integrating detection, communication, control and computer technology. Its core technology involves image processing, digital signal processing, pattern recognition, artificial intelligence, information technology, electronic technology , communication technology and system engineering technology, etc. Generally speaking, the intelligent transportation system mainly studies the following aspects: (1) collision ...

AI Technical Summary

Problems solved by technology

However, the huge amount of data will also bring high retrieval costs, and the time complexity and space complexity are too high to become an urgent problem to be solved

Images