Performance Evaluation of AE-RBM Based Pedestrian Detectors in Autonomous Vehicles: Substantial Issues and Challenges

Autonomous vehicles' rising profile is a reflection of the widespread interest in them. Improved dependability, reduced fuel consumption, cost savings, and enhanced passenger convenience are all benefits of these cars. However, autonomous vehicles must first have their functionality problems resolved before they can be extensively used on major routes. Pedestrian detection is a crucial component in autonomous vehicles that helps to avoid accidents. Preprocessing, feature extraction, and model training all rely on proper sequencing. A number of filters, such as the Median, Histogram, Dilation and Erosion, Gaussian, mean, and others, can be used to smooth out images before lane marking recognition. Haar wavelet and LBP are the components that make up feature extraction. In order to train for AE-RBM, one must have precise control over the attributes. This method looks so much more modern than the RBM and AE algorithms of today. The results demonstrated a notable enhancement in accuracy, reaching 95.84%.