Performance Evaluation of Binary Robust Invariant Scalable Keypoints (BRISK) on Low Power Single Board Computer

Peneliti

Didit Andri Jatmiko, S.Kom.,MT. Taufiq Nuzwir Nizar, M.Kom, M.T

Deskripsi/Abstrak

In this paper, performance evaluation of Binary Robust Invariant Scalable Keypoints (BRISK) on low power Single Board Computer is presented. BRISK is a method that can be used as an interest point detection in image stitching applications, the BRISK method can be divided into three steps: keypoints detection, keypoints description, and descriptors matching. These stages are executed sequentially. The experimental results show that BRISK can run on low power Single Board Computers based on x86 or ARM architecture. Based on the tests performed, the BRISK performance on ARM Single Board Computer and x86 Single Board Computer is directly proportional to image resolution and the time required. Notable time acceleration has been reached. However, BRISK algorithm on x86 Single Board Computer BRISK is much faster than the ARM Single Board Computer for all test images. One of the test results shows in the test image that both 320x320 pixels require 1062810 ?s on the x86 Single Board Computer meanwhile, on the ARM Single Board Computer the time is 1572154?s. Future research direction will study another factor for BRISK performance. This initial analysis can provide an overview for developers or researchers for making decisions in implementing applications using the BRISK on Single Board Computer.