Vision-Based Location Recognition for Visually Impaired People

Vision-based location for place recognition is a method used in computer vision to identify and localise aspecific place or location based on visual data. This method entails employing a camera or sensor to take pictures or videos of the environment, which is then processed to extract features that can be applied to location recognition. The proposed system uses Fast Retina Keypoint (FREAK) descriptor and the Features from Accelerated Segment Test (FAST) algorithm as detectors to identify and extract features from the input image, enabling reliable feature matching. The system can determine the location of a person within the college environment . This information is relayed to the user through an audio interface making it easier and safer for visually impaired people to travel around indoor spaces. The system is trained using a dataset of images that represent the various locations within the college campus. To improve classification performance, the system incorporates K-means clustering and various classifiers, like Decision trees(DT), support vector machines (SVMs), and K nearest neighbour (KNN) algorithms. Comprehensive dataset of 2700 photos used for training a system and achieves high accuracy of 94% and robustness in detecting and recognizing different places within college campuses. The proposed technology has the potential to increase self-reliance and safety of visually impaired people.