Robust Pedestrian Detection and Tracking Using CNN and SORT Algorithms

In recent years, there has been a growing popularity of autonomous vehicles due to their significant impact on society. One of the key tasks of autonomous vehicles is accurate pedestrian detection, which plays a vital role in preventing accidents. However, accurately detecting and tracking pedestrians under various environmental circumstances poses a significant challenge. In this paper, an efficient model for pedestrian detection is proposed by integrating three modules: You Only Look Once version 8 (YOLOv8) for object segmentation, Histogram of Oriented Gradients (HOG) for feature extraction, and Custom Convolutional Neural Network (CNN) for classification and detection. For tracking purposes, a simple online and real-time tracking (SORT) algorithm is used to track pedestrians in consecutive frames. Extensive experiments were conducted using the EPFL dataset. By leveraging the strengths of these modules, the proposed model aims to improve the accuracy and performance of pedestrian detection and tracking. The experimental results demonstrate the remarkable capability of the suggested model to detect and track pedestrians, achieving an accuracy rate of 93.34% even in challenging weather scenes.