A lane detection system to assistant drivers for better driving using gradient decent algorithm comparison with KNN and SVM

Abstract

The aim of this study is to detect and identify the different lane signals using Gradient Descent algorithm and correlate with the KNN and Innovative SVM. A total of 370 samples were collected from different road signals available in kaggle. These samples were divided into a training dataset of 185 (50%) and a test dataset 185 (50%). Accuracy values were calculated to quantify the performance of the Gradient descent algorithm comparison with KNN and Innovative SVM. The study parameters are alpha=0.03, G power=0.8. On performing an independent samples T-test on the two groups considered. The Gradient descent algorithm has a 85% accuracy rate, KNN has an 88% accuracy rate and Innovative SVM has a 96% accuracy rate. Finally, Innovative SVM appears significantly better than Gradient descent and KNN. The statistical significance value is 0.001 (p<0.05). From the analysis of the experimental results, the Support Vector Machine algorithm (SVM) gives better results than the KNN (K Nearest Neighbour) and Gradient Descent for the detection of the various lane traffic signals.