Recuperação de objetos 3D baseada em relações geométricas entre Surflets

Abstract

The fall in the cost of digital cameras over time made their access easier, which made their use more popular. This increased the number of images on the Internet, demanding the study of efficient techniques for information retrieval in images. Analogously, the same is happening with RGB-D cameras, since other features are now made available by these cameras, such as three-dimensional (3D) perception of the environment, and even the possibility of capturing in the dark. This brings new applications, these being from a 3D perception of the world, such as environment mapping, object scanning, 3D printing, among others. At this time, similar to what happened with the images, the number of 3D objects is growing, making algorithms that find patterns in such objects necessary for easy retrieval by similarity. From this, this work investigates techniques for extracting features of 3D objects based on geometric relations between their polygons, and also evaluates comparison metrics among their histograms. In addition, this work presents a proposal for extraction of features and comparison of 3D objects. The validation of the metrics investigated and proposed was based on four bases of objects selected and classified into groups. The results indicate that the intrinsic geometric relation to each object, proposed in this work, offers unique features that can be used to retrieve 3D objects by similarity.