Modular robots consist of multiple individual robotic modules that are connected to form various configurations. They are also equipped with locomotion capabilities so they can move to change their interconnections and self-reconfigure into different configurations. In this paper, we propose a distributed algorithm for configuration recognition designed for large-scale 2D lattice-based modular robots. The algorithm consists of searching the set of borders to be then transmitted to the modules in order to enable them to collectively discover an efficient global representation of their current configuration. To assess the performance of the proposed algorithm, we conducted simulations on three different configurations and compared it to a box-based approach. The results highlight significant reductions in communication complexity across all configurations. Additionally, our algorithm demonstrated improved memory efficiency in two out of three configurations.