Pedestrian Detection for Autonomous Driving within Cooperative Communication System Based on CNN-ELM Approach

The ability to perceive and understand the behaviors of other road users is crucial for autonomous vehicles to accurately plan their responses. Autonomous vehicles are able to complete a lot of tasks, like identifying pedestrians, thanks to computer vision systems that are mostly based on machine learning algorithms. Furthermore, in an entirely autonomous driving setting, all vehicles need to talk to each other and exchange perceived data in order for navigation to be safe. Training the model, segmentation, and preprocessing are the three primary components of the method. Image smoothing is a part of preprocessing; there are a number of filters available for this purpose, such as the Median, Gaussian, Histogram, Dilation and Erosion, mean, and so on. The best way to blur the noisy details is with the Gaussian filter. Negative sample decorrelation is used in segmentation. For the model's training, it utilized the CNN-ELM model. Compared to more recent techniques, such as CNN and ELM, this one seems antiquated. The data shows an astoundingly high accuracy percentage of 93.70 percent.