Integrating HOG-Based Vehicle Detection with CNN-Based Lane Detection for Autonomous Driving

Abstract

The advancement of autonomous driving systems hinges on accurate and reliable vehicle and lane
detection. This paper presents an integrated method to improve autonomous driving systems by
merging Histogram of Oriented Gradients (HOG)-based vehicle detection with Convolutional
Neural Network (CNN)-based lane detection. HOG effectively identifies vehicles by capturing
edge orientations and structural features, while CNNs excel in detecting intricate lane patterns
through deep learning. The combination of these techniques offers a robust solution for detecting
both vehicles and lanes, essential for autonomous navigation. Evaluated across a diverse dataset
featuring various driving conditions, the system's performance is measured using precision, recall,
F1 score (for vehicle detection), and accuracy (for lane detection). The results indicate significant
enhancements in detection capabilities, leading to improved situational awareness and safer
navigation. Future work will aim to refine the system further and tackle challenges in more
complex driving environments, marking this approach as a promising advancement in autonomous
driving technology.