Similar image matching method based on deep learning

Abstract

The invention discloses a similar image matching method based on deep learning, and the method is specifically implemented according to the following steps: outputting a detection frame correspondingto each object in a target image through a target detection network, and detecting precision; 2, analyzing and processing to obtain a relative position relationship between any two objects in the target image; 3, constructing a graph model of a target position relationship of the target image scene; 4, repeating the steps 1 to 3 to establish a graph model of a scene target position relationship for each sample image in the candidate data set, and performing scene matching on the graph model of the target position relationship of the sample image scene and the graph models of the target position relationships of all the target image scenes one by one, and determining and selecting three sample images with the highest similarity with the target image. According to the similar image matchingmethod based on deep learning, the spatial position relationship among the objects in the target image can be effectively analyzed, and the similar image of the target image can be accurately matched.

Description

technical field [0001] The invention belongs to the technical field of computer digital image processing, and in particular relates to a similar image matching method based on deep learning. Background technique [0002] As an emerging technology in machine learning algorithms, deep learning is motivated by the establishment and simulation of neural networks for human brain analysis and learning. Deep learning is essentially a hierarchical feature representation of observation data, so as to further abstract low-level features into high-level feature representations, and this process is usually implemented by neural networks. In recent years, with the development of deep learning, deep learning has attracted the attention of relevant researchers all over the world, and more and more researchers have integrated deep learning into their own research fields. The three most widely used deep learning The first areas are speech recognition, image processing and natural language p...

AI Technical Summary

Problems solved by technology

[0003] At present, when matching images similar to the target image, deep learning is often used to identify objects in the target image, but the existing deep learning target recognition algorithm can only detect the category and position of the object in the image, and cannot get the object in the image. The spatial relative positional relationship between each object makes it impossible to accurately match similar images of the target image

Images