Optimization of Canny Algorithm for Recognition of Female Clothing Styles from the Perspective of Features

Zhe Liu, Min Luo

Article
2024 / Volume 7 / Pages 292-302
Received 18 January 2024; Accepted 1 February 2024; Published 20 February 2024
https://doi.org/10.31881/TLR.2023.175

Abstract
The accurate and fast recognition of women’s clothing styles is conducive to the classification and recommendation of merchants, and it is also convenient for customers to choose from. This paper optimized the Canny algorithm used to extract the contour edge of the female clothing image and used a convolutional neural network (CNN) classifier to identify clothing styles based on the edges extracted by the Canny algorithm. Finally, the Canny algorithm and the CNN classifier were tested in the simulation experiment. The performance of the CNN classifier was compared with that of the template matching and SVM classifiers, the Resnet34-based recognition method, as well as the target detection network and genetic algorithm-back-propagation neural network combined recognition method. The results demonstrated that the optimized Canny algorithm extracted more distinct contour edges. The CNN classifier exhibited the best performance and the fastest recognition for female clothing styles.

Keywords
canny algorithm, female clothing, style recognition, convolutional neural network