This thesis explores the progress of creating a system for detecting license plates and navigating through a garage environment with onboard computing on the nano-drone CrazyFlie 2.1. The system uses the deep learning network MobileNetv2 to classify real-time images from the built-in camera to detect license plates in its immediate environment. The navigation system is built upon a wall-following strategy that passes by parked cars. To train the system, a new specialized dataset is created with the camera onboard the drone for classification. The dataset contains 747 images of license plates and 743 images of background. Eight test cases were set up to evaluate the system. The eight test cases cover cars that are parked in (a) one row, (b) two rows,(c) a straight lane, and (d) an uneven lane. Also, tests regarding three different speeds (e) 0.1 m/s, (f) 0.2 m/s, and (g) 0.3m/s.Lastly, a test with (h) a darker test environment. The first four test cases were tested three times, and the last four two times. The results showed that the system managed to navigate past all simulated vehicles in all cases. The averageF1 score for each test case is (a) 59%, (b) 76%, (c) 25%, (d) 42%,(e) 23%, (f) 71%, (g) 72% and (h) 49%.