Automatic License Plate Recognition for Indonesia Plate Number
- Preprocess
- Plate detection and characters segmentation on vehicle images are using contours.
- For images/characters classification, I follow Tensorflow tutorial
- 58/60 plates detected
- Sometimes get wrong classification result for similar shaped characters, examples: 1 <--> T <--> I, 2 <--> Z
- For better detection result, you can do some experiments with preprocessing and contours.
- For better classification result, you can separate the plate number characters into three parts, the first letter, the second number, and the third letter;
- You can add more training datasets.
- Python 3.8.6
- OpenCV 4.5.1
- NumPy 1.19.4
- Tensorflow 2.4.0
- Matplotlib
- Resolution 2560 x 1920 pixels
- Daylight
- The vehicle is +- 1 meter from the camera
- I give you 5 test images
- Open mainProgram.py
- Change the test image path with the same image requirements
- Unzip dataset.zip
- Open Training.py
- Change train dataset folder path
- You can add another character dataset to the dataset folder for better result
https://www.kangghani.com/2021/02/deteksi-dan-membaca-plat-kendaraan-otomatis-python.html