Paper Title
Improved Fast (Feature from Accelerated Segment Test) Algorithm Based on Histogram Oriented Gradient (HOG) Descriptor

Abstract
In computer vision and robotics, every application must need accurate and fast feature detection and extraction algorithms. There are many traditional feature detection and extraction algorithms. However, being influenced by different factors, these traditional algorithms are troublesome to get an adequate digital image processing result. The traditional Feature from accelerated segment test (FAST) algorithm is based on identifying interest point detector in an image to get real-time frame rate application. However, in some situations, the FAST algorithm returns false matching results by adding some inconsistent mismatching points with less key points detection. These types of drawbacks of this algorithm can hamper the overall performance of the system. To overcome these types of drawbacks, this paper proposes an advanced Histogram oriented gradient (HOG) based modified FAST algorithm that combines traditional FAST algorithm with HOG feature detectors followed by random sample consensus (RANSAC) algorithm. The proposed method applies the FAST algorithm to detect corner points to find features. After that, it uses a HOG feature extractor to extract HOG features from a grayscale image and returns feature with encoded local shape information within the image. This method eliminates inconsistent matching points through the RANSAC algorithm and after that proper matching is performed with correct matching points. The experimental analysis of this study confirms that our proposed method provides higher matching accuracy and low time consumptions over traditional FAST, SURF, BRISK, HARRIS algorithm. Keywords - SURF, FAST, SIFT, HARIIS, RANSAC, HOG, Feature