Comparison of Vehicle Detection Using Haar-like Feature, LBP and HOG Technique for Feature Extraction in Cascade Classifier

Abstract

Transportation continues to increase every year. Recorded in 2018, the number of vehicles registered in Indonesia is more than 111 million. Problems such as traffic congestion and traffic accidents need to be resolved. One of the solutions implements intelligent transportation systems (ITS). ITS plays a very important role in the suitability of the traffic conditions of the vehicle. Many researchers apply the Haar-like feature,  Histogram of Oriented Gradients (HOG) and Local Binary Pattern (LBP) to detect objects and vehicles. This paper describes the comparison of the applicability of the Haar-like Feature, the LBP Feature and the HOG Feature on vehicle detection. The results of the comparison of the three features are Haar-like features for vehicle detection system proves better than of using HOG features and LBP feature for vehicle detection. Its detection rate is higher than HOG and LBP where it detected 40 vehicles from the total of 42 vehicles rather by HOG and LBP with only 36 and 35 vehicles detected. In the execution process, the haar-like detection feature is faster at execution time of 14.56 s  rather by HOG and LBP with the execution time only 21.36 s and 19.41 s. Haar-like features faster by 46.7% times more than HOG feature detector and Haar-like features faster 33.3% times more than LBP. Haar-like feature based detector system is the best technique for vehicle detection using cascade classifier