The Segmentation And Recognition Of The Mannequins Pictures Based On Image Processing

With the rapid development of the Internet in recent years, the online shopping has become part of people's lives, people's clothing online shopping has become a hot spot. So how to crop the mannequins images automatically according to the show rules on the online shopping website, how to identify the garment types from the mannequins images automatically and create the corresponding label marked the closes and how to split the model from the picture replaced by a new background become a hot spot of a study. The following works is mainly studied in the design.1.Image preprocessing. By fractional differential operator for image enhancement,median filtering and a series of image preprocessing methods to mannequins photo edge enhancement, with canny operator edge extract information mannequins. Use face detection locate the position of the face, the extracted core area model.2.The image classification. Through various different mannequin picture clothes type, the corresponding characteristics are different, extract the most favorable distinguishing feature,this paper extracted the number of edge point distribution of a particular segment binary direction corner feature and image edge features, and then classified by the minimum distance classifier method.3.The image recognition. By extracting the color characteristics of the mannequins pictures from the color histogram information under the YCr Cb color space. By extracting the texture features from the image gray co-occurrence matrix. By extracting the shape features from the corner detection and so on to identify the different characteristics of different closes and generate the corresponding label automatically.4.The image segmentation. By determining the adaptive threshold to complete the binarization, through the Canny operator for the edge detection, to combine them, andto improve the noise make the target area more accurately by inflation and corrosion two times, to separate the model from the background, and can choose other background to replace.This paper puts forward the following innovations and improvements during the research process:1. For the edge recognition of the object, proposed the method that first to classify then to recognize.2. For net purchase clothing made in accordance with image recognition, to extract the corresponding feature then generate labels automatically.3. For image segmentation,proposed the adaptive threshold segmentation in combination with the edge detection method.In this paper, the mannequins pictures at regular press clipping transforming into a garment type classification problem, and then find the appropriate clothing hem position according to the characteristics of the type of finish cutting. This article identifies the clothes according to color, texture and other characteristics of the picture corresponding features clothing and automatically generate labels saved. Also he completed the characters separated from the background and replace the background.