Fine-grained image classification method and system based on saliency branch feature fusion

Abstract

The invention discloses a fine-grained image classification method and system based on saliency branch feature fusion, and the method comprises the steps: firstly carrying out the saliency detection of fine-grained images in M types of different saliency detection modes, and obtaining a plurality of saliency images with different saliency; constructing a multi-branch fine-grained image classification model, so that feature extraction can be performed on a fine-grained image and a plurality of saliency images at the same time, the efficiency of the method can be significantly improved; secondly, performing feature modulation on the fine-grained feature map by utilizing a plurality of saliency feature maps, so that the obtained modulation feature map can give higher attention to more different regions, and the problem of poor fine-grained image classification effect due to the fact that only one image feature is focused in the prior art is effectively solved; and finally, fusing the plurality of modulation feature maps and performing image classification according to the fused feature maps to obtain more comprehensive image features so as to effectively improve the classification precision.

Description

technical field [0001] The invention relates to the technical field of unmanned aerial vehicle autonomous landing, in particular to a fine-grained image classification method and system based on the fusion of salient branch features. Background technique [0002] Fine-grained image classification is also known as sub-category image classification. Its purpose is to divide images belonging to the same basic category (cars, dogs, flowers, birds, fruits and vegetables) into more detailed subcategories. Compared with ordinary image classification tasks, fine-grained image classification is more difficult. [0003] Fine-grained images are easily affected by many uncertain factors such as pose, illumination, occlusion, background interference, etc., making subcategories have the characteristics of large inter-class similarity and small intra-class similarity. The signal-to-noise ratio of fine-grained images is very small, and the information that contains sufficient discriminatio...

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a fine-grained image classification method and system based on the fusion of salient branch features, which is used to overcome the defects of low classification accuracy of fine-grained images in the prior art

Images