Registration of 3D point clouds is one of the most fundamental phases during the process of reverse engineering and most challenging at the same time. This phase consists on matching two or more different point clouds into one data set in order to have them share the same global coordinate system. In this paper we present a new approach for automatic registration of 3D point clouds that uses the genetic algorithm (GA) as a global optimization method. We introduce a trips extraction technique for rough registration, which extracts important geometric information from a point cloud. Another contribution in this paper is the introduction of the Interpenetration Fraction Measure (IFM), which maximizes the number of points that overlap two different point clouds. The algorithm we present also takes advantage of the parallel computing power of today’s multi-core processors, and other techniques for further efficiency. Finally, we present some experimental data with comparisons for analysis and further discussion about the algorithm’s performance.