A method for automatically screening two-dimensional codes in fashion image

Abstract

The invention relates to a method for automatically screening two-dimensional codes in fashion images. The method comprises the following steps: step S1, constructing a depth learning two-dimensionalcode detection frame; 2, designing that loss function of the target detection; 3, performing extraction preprocessing of the two-dimensional code target area; 4, decoding the two-dimensional code candidate region. The invention has the advantages of solving the problem that a large number of two-dimensional code pictures doped in a fashion database seriously affect the visual experience of customers and the performance of the algorithm, and a two-dimensional code recognition and screening solution with stronger robustness, higher accuracy and better automation is established.

Description

technical field [0001] The invention relates to the technical field of data identification, and relates to an automatic screening method for two-dimensional codes under large data sets of fashion images by using visual image processing and two-dimensional code decoding technologies. Background technique [0002] QR codes can store a large amount of information, and have been used in a wide range of fields, and have gradually become an indispensable communication tool in people's online life, which also means that a large number of pictures containing QR codes are flooding the network pictures. The fashion database comes from the Internet, which can provide customers with high-quality fashion style information and fashion element retrieval sources. It is inevitable to include a large number of pictures containing QR codes. In order to significantly improve the visual quality of the fashion database, it is necessary to The database performs automatic screening of QR codes. Th...

AI Technical Summary

Problems solved by technology

[0006] 1. The size of the two-dimensional code is different: the sizes of the pictures mixed in the hundreds of millions of fashion pictures are different, and the two-dimensional code is too small to be difficult to locate and decode

[0007] 2. The code system of the two-dimensional code is complex: the code system of the two-dimensional code is various, and there are also differences in the shape. A single two-dimensional code positioning and decoding method is difficult to meet the screening requirements of various two-dimensional codes

[0008] 3. Different angles of two-dimensional codes: in the fashion image database, a large number of two-dimensional code pictures have angle changes, which brings difficulties to the recognition of two-dimensional codes, and posture correction is often required to meet the recognition requirements

[0009] 4. The shape of the two-dimensional code is various: the conventional two-dimensional code appears in the form of a square, but there are some two-dimensional codes that appear in irregular shapes. Perform identification and screening, which affects the accuracy of screening to a certain extent

However, no corresponding disclosure has been made regarding the establishment of a technical solution for the identification and screening of two-dimensional codes with stronger robustness, higher accuracy, and better automation.

[0012] In summary, it is necessary to establish a set of more robust, more accurate and more automated QR code recognition and screening methods, but there is no report on this automatic screening method.

Images