Enhanced pedestrian walkway object detection using deep learning and pelican optimization algorithm for assisting disabled persons.

Walking is a significant transportation method, but the convenience of pedestrian surroundings for individuals with blindness is highly challenging. Pedestrians with blindness familiarize themselves with guidelines in their surroundings, which might be artificial or natural. To overcome these troubles, it is highly significant for them to perceive the features of an environment. Currently, numerous methods like long white canes and GPS are deployed to improve pedestrian walkways for sightless people. So, they can utilize it as the primary assistive device for recognition and also the vital ecological features for persons with disability. Recently, a growing amount of success has been conveyed for vision navigation tasks depend upon deep learning (DL) and machine learning (ML) networks to aid visually impaired people. This study proposes an Enhanced Pedestrian Walkway Object Detection and Pelican Optimization Algorithm for Assisting Disabled Persons (EPWOD- POAADP) method. The main intention of the EPWOD-POAADP method is to enhance pedestrian walkways for blind people's navigation. At first, the image pre-processing stage applies median filtering (MF) to eliminate the noise in the input data. Furthermore, the Faster R-CNN model is employed for the object detection process to identify and locate objects within an image. The CapsNet model is used for the feature extraction process. In addition, the wavelet neural network (WNN) technique is implemented for the detection and classification process. Finally, the hyperparameter selection of the WNN model is performed using the pelican optimization algorithm (POA) technique. The experimental evaluation of the EPWOD-POAADP approach is examined under the UCSD anomaly detection dataset. The outcomes indicated the enhanced performance of the EPWOD-POAADP approach compared to recent approaches.