Abstract:
Road crack identification is a prerequisite for both road health monitoring and reduction in reconstruction outlay. Many techniques like edge detection algorithms have been used to identify cracks but most of them failed to detect cracks in highly textured images. This paper suggests a Bayesian approach for road crack detection as the method takes the initial probability together with some strong evidence to predict the output. The task of road crack identification is challenging because there is a small difference between a crack and noise. Besides the cracks have a strong variance in intensity throughout moreover the road images use to be highly textured. Other names for the devised algorithm are Sequential Monte Carlo, Bootstrap Filters, Condensation algorithms and survival of the fittest. Sequential Monte Carlo or particle filters are a collection of genetic algorithms which have vast applications in non-linear problems. Besides, the Gabor filter has been used for automatic seed point selection. For the algorithm to come to a termination point, a stopping criterion is used comprising of two stopping parameters. The formulated solution comprises of two main steps. A geometric model formation and application of a particle filter. The particle filter based algorithm has been optimized by considering unique geometric models, specific radii values. An important factor which determine the time of execution of algorithm is the type of geometric model used
