Flower image retrieval method based on convolutional neural network

Abstract

The invention discloses a flower image retrieval method based on a convolutional neural network. The flower image retrieval method comprises the following steps that: constructing a flower image dataset with category information, and using a random image transformation method to process a flower image to enhance the flower image dataset; constructing a flower classification deep convolutional neural network model VGG-F based on VGG-16, wherein the node numbers of FC1 (First Full Connection), FC2 and a classification layer are independently 1024, 256 and a flower category sum; through the modelVGG-F and the FC2, independently extracting the feature vector of each flower image in an image to be inquired and the flower image dataset, and carrying out similarity comparison to obtain the mostsimilar 2N pieces of flower images; and independently calculating the image to be inquired, obtaining the HSV (Hue, Saturation and Value) color distribution feature vector of each flower image in the2N pieces of flower images, and carrying out comparison to obtain the most similar N pieces of flower images. By use of the method, on the basis of the convolutional neural network, and flower retrieval accuracy is effectively improved.

Description

technical field [0001] The invention belongs to the intersection field of image processing, computer vision technology and pattern recognition, and more specifically relates to a flower image retrieval method based on a convolutional neural network. Background technique [0002] The traditional flower image retrieval is based on text, and the user searches by inputting the name of the flower. However, due to the wide variety of flower images and changing colors and shapes, it is difficult for ordinary people to accurately understand and describe flower images, which makes it difficult for the text description entered by the user to meet the actual query needs of the user. [0003] Content-based image retrieval can effectively use the image's own characteristics and reduce the user's subjective description. In recent years, the convolutional neural network model represented by VGG has achieved good results in the field of computer vision, such as image classification and tar...

AI Technical Summary

Problems solved by technology

However, the existing flower image retrieval methods based on the VGG convolutional neural network model only use simple image low-level features such as shape, color, texture, and key points, and the accuracy of retrieval results is low.

Based on the features described by key points such as SIFT (ScaleInvariant Feature), it has certain robustness to transformations such as illumination, viewing angle, and scale, but its computational complexity is large, and its ability to describe flower images is limited.

For color features, the traditional color histogram statistics can only reflect the overall characteristics of the image, and can only express image color information, such as image global color information, but cannot describe other information, so the accuracy rate is relatively low when directly used for flower retrieval

Images