The Research Of Recommendation Model For Clothing Size Based On Image

Recently with the popularization of electronic commerce and the fast development of online shopping, the clothing shopping, as the most popular part of online shopping, its consumer demand has grown exponentially in recent years. Sometimes when consumers choose their clothing according to their body size, then after choosing the style which they like, they realized that the clothing they bought online did not fit them properly, and then they will choose whether return or continue to wear, this will make bad effects on the trust of online shopping. So how to help the consumers choose the right clothing to meet the need of them is a big problem. We propose that we should find a recommendation model for clothing size to help solve the problem.This paper will get the data information of clothing size recommendation from the image, and then combined with the data information of the clothing, by defining a garment fit assessment function to get the fit size of persons with different size, construct a size categorization database. And then we adopt the SVM model to make a recommendation for clothing size, compared with BP neural network that we will find a better way to make a recommendation model for clothing size.The main works of this thesis are as follows:This paper firstly collects the pictures from people and extracts the edge of the image, to get the edge contour line of people in the picture. According to the image of human body we make a data extraction of human body size. This article firstly adopts Roberts operator、Prewitt operator and Sobel operator and traditional Canny operator to make the edge extraction of images from human body. Then we adopt an optimized Canny operator to make the edge extraction, the algorithm is better than the way of other operator algorithm on the performance in the extraction. Through the application of the way we lay the foundation for the extraction of data from human body in the next part of this paper.And then we achieve the size data of human body and clothing, after the image processing of human body, firstly by using the Harris corner detection algorithm that we find the feature points of human body, then combined with definition of human size line we get the position of key part of body, and then we achieve the data of the height, width of neck, thick of neck, width of breast, thick of breast, width of shoulder, width of waist and the thick of waist. Then by adopting the way of BP neutral network, we make a transition from 2D size information to 3D size information to get the 3D size information of human body. For the achievement of data of clothing, this paper will use women blouses as the research object. Based on the rule of national garment size, we choose the data information of this type of clothing as the data information of clothing.Then we construct the model of garment fit assessment, based on the data of human body and clothing size, combined with the Emphasis of the comfortable sensation for the each part of the body. We build a garment fit assessment function as a way of recommending clothing size for people. Then based on the result of the evaluation, we can build a size categorization database.Then we adopt a way based on SVM model to construct a recommending model for clothing size. Based on the human size data and recommended size, we build a size categorization database. And then we adopt the SVM model based on genetic algorithm, particle swarm optimization and grid search algorithm to make a recommendation for clothing size and choose the most optimal model. At the same time, we compare the SVM model with BP neural network model, by the results of the recommending accuracy we try to find a better way for constructing the recommendation model for clothing size.At the end of the paper, some conclusions of the research work are made by us and some outlook of the next research direction or problems is discussed briefly.