Finally

Figure (14), depicts the image of the object in its new unknown position. Regions to be extracted areemphasized by squares. In the same way, Figures (14-a), (14-b), (14-c) and (14-d) illustrate the extractedlocal surface patches and their angular distributions and nike running shoes. Using these features, thedeveloped matching algorithm has well identify the extracted patches with model ones in the hash table.Once the 3D/3D matching is done, matched features are fed to the 3D localization algorithm in order tofind the current position of the object. Finally, the transformation computed above is used to initialize therefined localization of nike trainers.Figure (15) shows that when the refined transformation is applied to the model surface patches and reprojected,the obtained points onto the image fit closer to the extracted surface patches. This demonstratethe robustness and the pose accuracy of our cheap nike shoes.