Comparison of RGB-D processing for walkway detection in bionic eye

The Bionic Eye aims to restore vision for the blind, but limited hardware often leads to low-resolution images. Image processing is crucial for enhancing these images for the visually impaired. One promising approach is using RGB-D images to improve object detection and create a danger map for safe navigation. This study compares various RGB-D processing methods for generating a danger map in bionic eye walkway navigation. The methods include MSAC and four Convolutional Neural Network (CNN) models with semantic segmentation. Findings show that the Feature concatenated CNN models outperform other methods in terms of both accuracy and F1 score.