PROJECT TITLE :

Polar Embedding for Aurora Image Retrieval - 2015

ABSTRACT:

Exploring the multimedia techniques to assist scientists for his or her research is an fascinating and meaningful topic. In this paper, we focus on the massive-scale aurora image retrieval by leveraging the bag-of-visual words (BoVW) framework. To refine the unsuitable illustration and improve the retrieval performance, the BoVW model is changed by embedding the polar info. The prevalence of the proposed polar embedding technique lies in two aspects. On the one hand, the polar meshing theme is conducted to see the interest points, that is more appropriate for images captured by circular fisheye lens. Especially for the aurora image, the extracted polar scale-invariant feature rework (polar-SIFT) feature will conjointly reflect the geomagnetic longitude and latitude, and therefore facilitates the additional data analysis. On the opposite hand, a binary polar deep local binary pattern (polar-DLBP) descriptor is proposed to enhance the discriminative power of visual words. Along with the 64-bit polar-SIFT code obtained via Hamming embedding, the multifeature index is performed to scale back the impact of false positive matches. Intensive experiments are conducted on the massive-scale aurora image knowledge set. The experimental result indicates that the proposed technique improves the retrieval accuracy considerably with acceptable efficiency and memory price. Moreover, the effectiveness of the polar-SIFT scheme and polar-DLBP integration are separately demonstrated.