Rear-Lamp Localization with Real-Coded Genetic Algorithm at Nighttime

We propose a nighttime rear-lamp localization method via a modified real-coded genetic algorithm (RCGA) with search history driven crossover (SHX). SHX generates offspring considering historical statistics provided by an archive consists of survivor individuals. SHX adjusts six geometric parameters of a rear-lamp model based on an aggregate objective function which is composed of four sub-objective functions. Moreover, the prior knowledge with respect to the red-light of rear-lamps is exploited to narrow down the solution space. Experimental results show that the proposed method can accurately localize rear-lamps under various scenarios, especially when clear redlight of rear-lamps are observable.