Method for Hash image retrieval based on deep learning and local feature fusion

Abstract

The invention relates to a method for Hash image retrieval based on deep learning and local feature fusion. The method comprises a step (1) of preprocessing an image; a step (2) of using a convolutional neural network to train images containing category tags; a step (3) of using a binarization mode to generate Hash codes of the images and extract 1024-dimensional floating-point type local polymerization vectors; a step (4) of using the Hash codes to perform rough retrieval; and a step (5) of using the local polymerization vectors to perform fine retrieval. According to the method for Hash image retrieval based on deep learning and local feature fusion, an approximate nearest neighbor search strategy is utilized to perform image retrieval after two features are extracted, the retrieval accuracy is high, and the retrieval speed is quick.

Description

technical field [0001] The invention relates to the field of content-based image retrieval, in particular to a hash image retrieval method based on deep learning and local feature fusion. Background technique [0002] How to efficiently retrieve large-scale image data to meet the needs of users is an urgent problem to be solved. The traditional method is the image retrieval of the visual bag of words model, which is to first use the scale invariant feature transformation descriptor to extract the features of the image, and then use The hard clustering algorithm (K-Means) performs local feature clustering to obtain a visual dictionary, and finally counts the frequency of each visual word to generate a visual word histogram, and then matches and calculates image similarity. Since the initial feature extracted by the visual word bag model is Traditional manual descriptors, so the extracted features are relatively low-level, and cannot describe the high-level semantic informatio...

AI Technical Summary

Problems solved by technology

In addition, image retrieval based on deep learning generally extracts the features of the last convolutional layer or fully connected layer to directly perform similarity calculations, resulting in that although the final retrieved results are images of the same semantics, the local details between images are not similar. , because the high-level features have lost a lot of detailed information. For example, when searching for decorative clothing bags, industrial precision devices, and plant leaves in e-commerce, the images are similar in overall outline but have a large gap in details, resulting in retrieval results. Inconsistent with what the user expects

Images