Introduction: Vision is one of the most basic human necessitites and impairment of vision significantly hinders navigation and environmental cognition without any assistance. Accurate and reliable identification of signboards is essential in assistive technologies designed to meet the needs of individuals with visual impairment (VIPs). Methods: The article proposes an innovative Sign Board Detection Method for individuals with disabilities based on a Metaheuristic Optimization Algorithm (SBD -MOA). The first is Wiener Filtering (WF), which is used to pre-process images in order to remove noise and improve image quality, followed by the use of Mask R-CNN to detect objects and perform instance- segmentation. The feature extraction method is a Scale-Invariant Feature Transform (SIFT), and the classification method is a Graph Neural Network (GNN). The GNN hyperparameters are optimized with the Modified Mussel Length -based Eurasian Oystercatcher Optimization (MMLEOO) algorithm to enhance the performance of classification. Results: Experimental validation of a standard traffic sign image dataset shows that the proposed SBD-MOA framework can be effectively utilized with an average accuracy of 98.71 and a precision of 95.18 with an 80:20 training- testing split, and thus is a better model than several existing detection models. Discussion: The results reveal that a combination of segmentation-based detection, scaleinvariant feature extraction, relational learning using GNNs, and metaheuristic optimization is an effective approach for improving reliability in complex outdoor settings. The modular architecture is effectively useful in the field of assistive applications, where the need to accurately localize, interpret, and tolerate noise and illumination changes is more important than inference speed. Conclusion: The proposed SBD MOA model has high potential to enhance autonomous navigation and situational awareness of the visually impaired population, which would enhance safer and more robust assistive vision systems.
Batra et al. (Wed,) studied this question.