Design and Evaluation of Image Quality Enhancement to Augment Object Detection and Tracking Framework to Support Visually Impaired Persons

Abstract

Object Tracking after Prediction for the Visually Impaired was intricate, especially when identifying the right object with the right identity. The Image Quality Enhancement from captured images acquired from external devices significantly impacts the prediction of objects and tracking of their movement in Visual impairment devices. The foremost objective of the study was to design a framework for enhanced predictions in object tracking after prediction for visually impaired people. Initially, the captured image was applied with image quality enhancement measures in terms of filtering, segmentation and feature extraction to ascertain a quality image that could be used for object detection and tracking with enhanced prediction efficiency like accuracy, sensitivity, specificity, etc. The experimental analysis showed that the captured frame had been enhanced to a quality level that enhanced the prediction evaluation measures based on the confusion matrix assessment techniques. The research showed an enhanced accuracy level of 98.3% in comparison to benchmark models like Detectron2 (97.3%), YOLOv5 (84%), Faster RCNN (46%), HFYOLO (95.5%) to prove that when the image quality was augmented, Object Tracking after Prediction could be augmented being directly proportional to each other.