Proposed Hybrid Color Histogram based Obstacle Detection Technique

Abstract

Assistive innovations for visually impaired persons are demonstrating a quick development, letting valuable devices to bolster their everyday exercises, therefore enhancing social consideration. This paper intends to propose a technique for helping blind persons in detecting obstacles in their path. Keypoint matching is an imperative feature of Computer vision obstacle detection. In this paper two techniques QC-LBP (Quantized Color based LBP) & QC-CSLBP (Quantized color based CS-LBP) are proposed based on hybrid features of LBP/CS-LBP, Gabor & HSV color Histograms. Then, these are compared to the already existing techniques such as SIFT, hybrid of SIFT with LBP & Gabor Filter. We grasp CS-LBP into our system due to its computational effectiveness & LBP due to its state-of-art execution in various issues. Gabor filter is coupled into our system due to its invariant nature. Color of each image is extracted using HSV, which on splitting undergoes different quantization levels & respective histograms are obtained. These obtained histograms are compared based on chi-square distance & matching object is obtained. In this paper, we present framework for detecting obstacles in the way of blind persons & also compare its efficiency with various existing algorithms. The output of the proposed system is the shape of obstacle or object in front of the blind user, which is intimated to the user in the form of sound. We show that our framework outperforms the other existing techniques.