Commodity image feature description method based on a multi-scale visual word bag model

Abstract

The invention discloses a commodity image feature description method based on a multi-scale visual word bag model, solves the problem of poor commodity image classification effect of an existing visual word bag feature description method, and belongs to the field of image classification. The method comprises the following steps: carrying out wavelet multi-scale decomposition on a double up-sampled commodity image, extracting multi-scale feature points, carrying out edge filling, delimiting a local area of the image on the edge-filled image, describing the delimited local area by using an SURF local feature and a quantized and normalized color vector angle feature, obtaining a final descriptor of the local region features through serial fusion; clustering the local area features to form a word bag; carrying out distance measurement on each local region feature and a word feature vector in the word bag, coding the local region features, carrying out feature convergence by using a diagonal concentric moment space model to form an image descriptor, and carrying out dimensionality reduction to obtain a visual word bag feature vector; and classifying the obtained visual word bag feature vectors by using a classifier.

Description

Technical field [0001] The present invention relates to a merchandise image feature description method based on a multi-scale visual phrase model, belonging to an image classification. Background technique [0002] my country's e-commerce is growing, and has become an important industry that affects residents' lives. With the continuous advancement of the Internet promotes the rapid development of major e-commerce platforms such as Taobao, Jingdong and Vipshop, online shopping has become the main way of mass purchase products and is increasingly recognized. According to the statistics of China Internet Information Center, as of June 2020, my country's netizens is 904 million, and the Internet's penetration rate is 64.5%. my country's online shopping users reached 71 billion. The 2019 transaction size reached 10.63 trillion yuan, year-on-year growth 16.5 %. In order to provide users with fast browsing, search positioning, etc., a good product image feature description and classifi...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problem that the commodity image classification effect of the existing visual word bag feature description method is not good, the present invention provides a product image feature description method based on a multi-scale visual word bag model

Images