Crack detection on surfaces is a crucial task in various fields, e.g., civil engineering, non-destructive testing, and materials analysis. This article introduces a surface model designed to generate an infinite cluster using the nearest-neighbor clustering algorithm to identify cracks in two-dimensional surfaces. It also examines how performance varies based on the physical dimensions of the surfaces analyzed. Metrics are based on the generation of pores with uniform and randomly distributed radii, using the Monte Carlo method. The proposed algorithm evaluates the number of iterations needed to identify the crack that forms the infinite cluster, the computational time, and the total area of the detected crack. Results indicate that the size of the surface significantly impacts the porosity value, specifically the probability assigned in the simulated models. A proportional relationship was identified: for smaller surfaces, the critical probability is lower, while for larger surfaces, the critical probability increases significantly under both radius distributions.