LBP-Haar Multi-Feature Pedestrian Detection for Auto-Braking and Steering Control System

Fatality due to road accidents are increasing with the increase in population and number of vehicles. Intelligent systems are developed to counter act the loss due to road accidents. The paper proposes one such method to counter the accidents by the implementation of pedestrian detection by the use of LBP histogram and HAAR-like features. LBP histogram are used for cross checking the HAAR-like features where the upper body, lower body, face are detected using the Haar like features and LBP classifiers. The twin stage algorithm is used because LBP classifiers are 20% percent faster than Haar based algorithm and Haar features are more accurate than LBP classifiers. Thus this can be considered as a 4 layer cross check using two different algorithms in a cost efficient way to improve the accuracy. The use of Beagle Bone Black (cortex A8 sitara) is made here for the simulation of vehicle control which includes auto-braking and steering locking. The implementation is a 2-phase system, which includes city-drive and high-way drive for speed control.