Research On The Prediction Of Fashion Color Hue

Clothing industry occupies an important position in China's national economy,and clothing color is an important element to improve the competitiveness of commodity market.Stakeholders in the clothing industry need to participate in the design,sample,ordering,marketing and sales of clothing according to the predicted popular colors.At present,the samples used in the research are mainly the popular color determination information released by various authoritative institutions,without combining various cutting-edge fashion information in real life.In this background,this paper relies on the clothing designer space in the cloud design trading platform of dongwan humen clothing collaborative innovation center,this paper studies the method of fashion color prediction using fashion color determination information and fashion week show field picture information,and apply the results to the clothing designer "colour trend" in space module.This paper proposes a method for predicting the color trend of fashion colors based on the finalized information,and uses 288 finalized information of fashion colors released by INTERCORLOR from 2007 to 2013 published in recent studies as samples.This paper proposes to use FCM clustering algorithm to classify the hue and redefine the intervals of various colors,and calculate the proportion of various colors.The results are applied to the prediction model of color phase of fashion clothing constructed by GM(1,1)and BP neural network respectively.It was found that the color prediction model based on FCM had higher accuracy than the prediction model based on other classification methods.Among them,GM(1,1)model based on final information had the highest accuracy,and the MSE could reach 0.00028.On the basis of color hue trend prediction of fashion color based on finalized information,this paper proposes a color hue trend prediction process of fashion color based on picture information.In order to make a comparison with the research results based on finalized information,20,459 show pictures of 15 brands in the four fashion weeks of VOGUE from 2007 to 2013 were adopted.The foreground extraction algorithm based on GC-Ellipse(Grab Cut-Ellipse)proposed in this paper was used to extract the garment information in the image.The theme color extraction algorithm was used to obtain the four theme colors of the garment,and the HSV color space was used to quantify the colors and extract the hue value.The processed data was applied to the prediction method mentioned above for prediction.The color processingprocess of fashion color based on picture information designed in this paper broadens new ideas for the research of fashion color prediction and breaks the limitation of relying only on finalized information for prediction at present.Finally,the research results of this paper are integrated into the "color trend" module and applied to the space of fashion designers,which can provide guidance and help to stakeholders in the fashion ecosphere.For example,the prediction of fashion color trend can correctly guide and promote consumers' behavior awareness,and improve people's aesthetic taste and consumption taste.Assist stakeholders in the clothing industry to better grasp the trend of fashion color changes in the future,so as to develop more effective design,production plan.