Research On Virtual Fitting System Based On Kinect Somatosensory Device

Virtual fitting room is a new research field, it can bring us quick and convenient fittingexperiences, so it has a great research significance. The two core problems of virtual fitting systemare human-computer interaction and garment simulation. Interaction is an important factor todetermine the quality of the fitting experience. Using the traditional interaction method like themouse and keyboard, many previous virtual fitting products create an user Avatar by inputtingparameters to realize the virtual fitting function, but it is difficult to give the user a good immersionand interactivity. And the interaction method using multiple cameras to take photos of the user fromdifferent angles base on the gesture recognition, but its hardware environment is too complex, andthe accuracy of the recognition is not high enough. On the other hand, cloth simulation is the soul ofvirtual fitting system, in order to solve this problem, the predecessors had put forward manymethods, such as using simple2D texture or3D model for simulation, but these methods havedefects such as poor simulation effect, complex modeling and difficult extension method.This paper realizes a real time virtual fitting system based on the Microsoft kinectsomatosensory device, and gives some solutions for the problems mentioned above. The main workis as follows:1. This paper proposes a gesture decision algorithm based on the fingertips recognition, andencapsulates a human-computer interaction module based on hand gestures. This algorithm usescenter of gravity method to optimize the traditional Graham scan convex hull calculation method, itimproves the efficiency of the calculation, and on this basis realizes the decision to the "grab"gesture. The human-computer interaction interface based on this algorithm allows the user to usesimple and convenient gestures to operate the system’s GUI, with a good gesture recognitionaccuracy, the interaction effect is ideal.2. This paper improves the image transfer algorithm based on contour information matching,and proposes an image transfer algorithm based on skeleton information matching. This algorithmmodifies the matching keywords and uses the skeleton information as the matching keywords.Compared with taking the contour information as keywords, the method using the skeletoninformation has a higher accuracy and faster calculation efficiency, and its requirement of hardwaredevices is lower, so it has a better usability. In addition, this paper improves the matching algorithm,a continuity judgment unit is added to the energy formula in the original paper, and it makes thematching results more continuous. 3. Based on OpenNI development libraries and multithreading technology, this paperencapsulates a kinect somatosensory capture module, it can recognize the human body skeletonand the location of the palm center and guarantee the reliability of the gesture decision and theposture matching function.4. Based on the modules mentioned above, this paper implements a complete real-timevirtual fitting system with a good user experience.