|
Acadêmico(a): Danilo Conrado Eskelsen Junior |
Título: CORPUSVR: UM PROTÓTIPO PARA VISUALIZAÇÃO DA ANATOMIA DO CORPO HUMANO |
|
Abstract: |
This work presents the development of a prototype for visualizing the anatomy of the skeletal system of the hand in three dimensions. The prototype gets the images through the camera of a mobile device and from there, the images are processed in real time by a thread. During the execution of this thread, the obtained image is segmented by the color of the skin. The segmentation result forms a binary image containing the largest contour. Next, a list of convex points of the chosen contour is generated. This list of convex points is filtered and then sorted. Using the ordered list of convex points together with a Perspective-n-Point algorithm, the hand pose is estimated. Next, the rotation angles of the fingers are calculated. Then, the hand pose and the rotation angles of the fingers are updated on a shared object. This shared object is then used by another thread, the rendering of the 3D skeletal model of the hand thread. During the execution of this thread, the 3D model is drawn on top of the image obtained by the camera when the hand pose is identified. Finally, the prototype was able to identify and estimate the hand pose in controlled environments with good or moderate illumination, thus allowing real-time visualization of the skeletal system from different perspectives. All of this while maintaining a rate higher than 15 FPS in most cases. |
|
|