In recent years many powerful computer vision algorithms have been proposed, making feasible automatic or semi-automatic solutions to many popular vision tasks, such as camera calibration and visual object recognition. In particular, the driving force was the development of new powerful algorithms, especially in the area of local features. On the other hand, embedded vision platforms and solutions, such as smart cameras, have successfully emerged. Smart cameras offer enough power for decentralized image processing for various kinds of tasks, especially in the field of surveillance, but are still strictly limited in their computational and memory resources. indent In this chapter, we investigate a set of robust local feature detectors and descriptors for application on embedded systems. We briefly describe the methods involved, that is, the DoG (difference of Gaussians) and MSER (maximally stable extremal regions) detector as well as the PCA-SIFT descriptor, and discuss their suitability for smart systems for camera calibration and object recognition tasks. The second contribution of this work is the experimental evaluation of these methods on two challenging tasks, namely, the task of robust camera calibration and fully embedded object recognition on a medium-sized database. Our approach is fortified by encouraging results which we present at length.
CITATION STYLE
Arth, C., Leistner, C., & Bischof, H. (2009). Using Robust Local Features on DSP-Based Embedded Systems (pp. 79–100). https://doi.org/10.1007/978-1-84800-304-0_4
Mendeley helps you to discover research relevant for your work.