A spatial image retrieval system and retrieval method based on multi-feature fusion

Abstract

The invention belongs to the technical field of image retrieval and discloses a spatial image retrieval system and a retrieval method based on multi-feature fusion. The spatial image retrieval systembased on multi-feature fusion comprises an input module, a main control module, an image detection module, a feature extraction module, a similarity measurement module, a feature fusion module, a matching module and a display module. A large gradient operation matrix is adopt by that feature extraction module, more gradient intervals are divide, the calculation of the HOG cell movement is omitted,the calculation amount is greatly reduced, the overall speed is increased by about 4 times, and the method is very suitable for application occasions which require high real-time performance. At thesame time, the matching module classifies the fused feature images of heterogeneous image blocks instead of the cascaded feature vectors, which is helpful to improve the performance of the network. The heterogeneous image matching method based on depth learning provided by the invention is not only superior to other methods in performance, but also superior to other methods in training efficiency.

Description

technical field [0001] The invention belongs to the technical field of image retrieval, and in particular relates to a spatial image retrieval system and retrieval method based on multi-feature fusion. Background technique [0002] The development of image retrieval is a process from simple to complex, from low-level to high-level, from the initial text information query to content-based image retrieval. At the same time, with the deepening of people's research on image understanding and image recognition, a retrieval based on image semantics is proposed, which makes full use of the semantic information of images and improves the ability of image retrieval systems. In addition, in order to solve the problem of the semantic gap, people have proposed information retrieval technology based on feedback, which uses human-computer interaction behavior to improve the system's ability and improve the accuracy of retrieval results. Finally, with the development of artificial intelli...

AI Technical Summary

Problems solved by technology

Although a certain degree of efficiency has been achieved, there are still some shortcomings: the matching between the implementation process of the algorithm and the image resolution is not intuitive; the algorithm is complex, resulting in a large amount of data processing and high operating costs; the average analysis of the algorithm leads to important geometric features in the graph. The impact of changes in the overall similarity, resulting in a certain deviation in accuracy and stability

In the prior art, the quality measurement effect of the whole image is poorly predicted by the local quality score

Images