To localize a vehicle on a map, a reliable road selection system is essential. For this purpose, the article presents a credibilist multi-criteria association algorithm that performs data association between the infrastructure information (the map) and the noisy measurements of two sensors (DGPS, Odometer). The algorithm takes into account the inaccuracy, the uncertainty and the redundancy of the data. The multi-criteria fusion process is realized using Belief Theory and Dempster-Shafer's rule. A local strategy is developed to allot believes to two criteria. Experimental results show that the credibilist roads around an estimated position are well selected.