The Ariana project has previously developped a road extraction algorithm based on Quality Candy model and relying on a marked point process using Reversible Jumps Markov Chains Monte Carlo (RJMCMC).

My research goal is to include a multiscale approach to this algorithm in order to efficiently process data from a great variety of remote sensing sources. The multiscale approach should also improve the detection itself since it will take each scale data into account.