A SIFT-GPR Multi-Class Indoor Object Recognition Method for Visually Impaired People

As indicated by renowned global organizations concerned with health statistics, vision disability and blindness have reached worrying numbers worldwide. Devoting all affordable means as to lessen daily difficulties being faced by blind people has therefore become a paramount necessity. Over the past decade, technology has been seeing light under what is known as blind rehabilitation through the proposal of several assistive designs, which have dealt with numerous issues, and among them comes the object recognition for blind people. In this concern, this paper explores the object recognition under a blind rehabilitation prospect. Unlike the literature witnessed so far, we introduce a coarse as well as simultaneous multi-class object recognition scheme based on Scale Invariant Feature Transform (SIFT) and Gaussian process regression (GPR) for indoor scenes acquired by a camera mounted on the chest of the blind individual. Experimental results performed on the basis of an indoor dataset have demonstrated the interesting potential exhibited by the considered method in terms of both recognition accuracy and processing time.