Object orientation estimation for high speed 3D object retrieval system
Vicky Sintunata, Terumasa Aoki · 2016
A 3D object retrieval system is one of the most difficult system to archive. If 3D model can be used as a query to the retrieval system, it will become less difficult. Unfortunately, creating 3D model is not a trivial task. Usually, users have only 2D images (photos, sketches, etc.) as query images for the retrieval system. In the past, the query images will then have to be compared with every image in the database. In this paper, we propose a new method by utilizing the skewness of the image, so that the system does not have to compare every image in the database. The skewness values will be used to generate contour maps of the object skewness data. By utilizing these skewness (contour) maps, the position of the image in the database can be extracted. Furthermore, the rotational degree (orientation) of the object can be approximated. This is done by simply using a simple thresholding method. The results of the proposed method are compared with the traditional SURF method. It shows that the proposed method is 25 times faster than the traditional SURF method when comparing the images.