| In modern society, eye-glasses have almost become necessities in human’s life.There are too many dazzling spectacles in traditional optical retail stores, which makes customers’ shopping experience not good─because of the time-consuming and boring in selecting eye-glasses. In order to improve the customer’s shopping experience, achieving virtual eye-glasses try-on system based on computer vision technique is very meaningful. This paper focuses on face detection technique and how to implement virtual eye-glasses try-on system based on computer vision, mainly including the following areas:1. Research on face detection and facial feature localization technology. MCT(Modified Census Transform) can be used to extract face contour information with illumination invariance. Therefore, we propose a face detection method based on MCT transform and HIK SVM classifier. The method uses two classifiers to detect.Firstly trained a linear SVM classifier as coarse classifier which based on local facial Haar features. Then use a HIK SVM classifier based on MCT features to confirm the results of coarse detection. The method achieve a good accuracy and is not sensitive to light.2. 3D Modeling technique on virtual eye-glasses. Based on 3D perspective theory, the obtained two-dimensional images of the front, left and right side images of eye-glasses were perspective projection transformed, to generate 3D eye-glasses images with dyeing effect. The generated 3D modeling perspective eye-glasses are vivid with good spatial scale and angle scalability.3. Achieve virtual eye-glasses try-on system based on Kinect. Somatosensory computer programming technique based on Kinect, facial feature positioning technique based on Kinect, virtual glasses 3D modeling technique are used in the development of virtual eye-glasses try-on system. The system realizes real-time virtual try-on, compare, recommend, photography and other functions, and successfully displayed in the 2014 retail trade exhibition in Dusseldorf, Germany,raising attention and praise by foreign customers.4. Initial implementation of virtual eye-glasses try-on system based on Android.This article combied Android programming technique and ASM facial featurelocalization technique in development of mobile phone version virtual eye-glasses try-on APP. The system realizes automatic extraction and positioning feature points on the face photoed by the mobil phone. The experiment has laid a good foundation for the next version of mobile Internet virtual eye-glasses try-on system in further research. |
