The process of recognizing and identifying vehicle license plates is considered one of the important tasks in the monitoring of roads, institutions and the applications of intelligent transportation systems such as traffic control, autonomous driving, parking toll booths, etc. The current study’s objective is to compare how well local features perform in identifying and detecting car license plate numbers, estimate the detection time for each feature, and confirm it using the time constraints for detecting license plates on Iraqi inner and highways streets. The three models are Harr-like, Histogram of Oriented Gradients (HOG) and Linear Binary Patterns (LBP). The identifying and detecting process used a database created by recording multiple videos of various cars at parking in the city of Baghdad. The recognition performance results showed that all models managed to detect and recognize license plates within the specified time, except for the Histogram of Oriented Gradients (HOG) model in video recording. The latter model recognized the license plate beyond the specified time. The results also showed that all models managed to recognize license plates within the executed detected time per frame, which is within fractions of a second and falls between 0.00486 and 0.0225 s, thus attributing the success of the models and the goal of license plate recognition has finally been achieved.