Vision-Based Detection of Hospital and Police Station Scene

Abstract

This paper proposes an efficient method for scene detection and scene text recognition using computer vision and machine learning. The scene recognition of police stations and hospitals is implemented using an ORB feature detector and extractor. The extracted features are optimised with K-Means and PCA for dimensionality reduction. The police station and hospital scene is accurately recognized using Random Forest with an accuracy 94%. The recognized scene is then analysed for scene text extraction using localization and character segmentation techniques. The text recognition model is trained to accurately detect scene text using Random Forest with 98% accuracy. This novel method can be used for autonomous driving to assist drivers with information about the hospitals and police stations around, while driving.