Some of the applications of automatic license plate recognition algorithms are security, tracking, and toll collection. However, due to the lack of regulations on Peruvian car license plates, developing a robust system to detect them represents a challenge. The problem is not only related to different rules based on vehicle types or rule year inconsistencies but also that many vehicles on the road have degraded license plates or thick protectors that hinder the character recognition process. In this manuscript, we present an optimized method based on the k-nearest neighbors (k-NN) classification method, which is compared with regular k-NN and multiclass support vector machines to recognize Peruvian car license plates, with accuracy results bigger than 96% for the optimized k-NN method.