A Garment Style Recognition Method Based on Hu Invariant Moment and Support Vector Machine

Abstract

The invention relates to a clothing style recognition method based on HU invariant moments and support vector machines. By preprocessing the clothing image, the outer contour of the clothing is obtained, and then the HU invariant moment features of the outer contour of the clothing are extracted, and then based on the support Garment Style Recognition with Vector Machines (SVM). The preprocessing of the clothing image refers to the clothing image segmentation process, finding the 8-connected region with the largest area is the clothing area, and filling the inner cavity of the clothing area; the acquisition of the outer contour of the clothing refers to the preprocessing of the clothing image. After processing, perform external edge detection to obtain the outline image of the clothing. The extracting the HU moment invariant feature of the outer contour of the garment refers to extracting the seventh order HU invariant moment feature vector of the shape feature of the garment outline. The SVM-based clothing style recognition uses SVM multi-classifiers to perform multi-category recognition of clothing styles. The invention can achieve a recognition accuracy rate of 83%, has better effect in recognizing styles with similar clothing outlines, has the characteristics of fastness and accuracy, and is applicable to the recognition of clothing styles in clothing images.

Description

technical field [0001] The invention belongs to the technical field of clothing style recognition, and relates to a clothing style recognition method based on HU invariant moments and support vector machines, in particular to a garment contour image obtained by edge detection after image segmentation processing and based on HU invariant moments and SVM approach to clothing style recognition. Background technique [0002] With the advent of the era of big data, merchants can use machine vision technology to analyze consumers' dressing styles, which will help merchants capture the consumption trends of various customer groups and formulate targeted product portfolios, marketing plans and business decisions. At the same time, with the popularization of face computer recognition technology, extracting face features and combining clothing style features will improve the accuracy of identity authentication. Clothing styles are composed of changes in the outer contour and inner de...

AI Technical Summary

Problems solved by technology

Since the similarity comparison method between WFD feature vectors is complex and depends on the complexity of the target object outline, WFD is not suitable for real-time classification of shapes

Although ELM can greatly improve the speed and generalization ability of network learning, it inevitably causes the hidden danger of over-fitting and reduces the classification effect.

At the same time, An recognizes the clothing design plan without the interference of color and texture, so the clothing outline is smoother and the recognition difficulty is slightly lower; its recognition method is not suitable for clothing with color and texture

Images