Novel Vehicle Detection in Real Time Road Traffic Density Using Haar Cascade Comparing with KNN Algorithm based on Accuracy and Time Mean Speed
Aim: The main objective of the paper is to detect objects in iconic real time traffic density videos
from CCTVs and Cameras using Haar Cascade algorithm and to compare algorithms with
K-Nearest Neighbour algorithm (KNN). In this case we tried improving the rate of accuracy in
predicting the traffic density. Materials and methods: Haar Cascade algorithm is applied on 5
realistic videos and which consists of more than 250 frames. For the same we evaluated the
Accuracy and Precision values. Harr-like function displays the vehicle’s visual structure, and the
AdaBoost machine learning algorithm was used to create a classifier by combining individual
classifiers. The significance value achieved for finding the accuracy and precision was 0.445 and
0.754 respectively. Results and Discussions: Detection of vehicles in high speed videos is
performed by using Haar Cascade which has mean accuracy with 85.22% and mean precision with
90.63% and 60% of mean accuracy and 58.53% mean precision in KNN classifiers.
Conclusion: The performance of the Haar Cascade appears to be better than KNN in terms of both
Accuracy and Precision.
Author's Name: K. Pavani and P. Sriramya
Volume: Volume 11
Issues: Volume 11
Keywords: Haarcascade, KNN, Machine Learning, Novel Vehicle Detection, Real Time road Traffic Density, Digital Image Processing.