This project is mainly implemented with C, Opencv 2.4.9 and MFC.
The program could be divided into twp parts:
- License plate localization
- Segment numbers and letters out and recognize them
(1) Use sobel calculator to get the X-direction derivative.
(2) Use dilate and erode to remove noise pieces and connect the neighboring regions.
(3) Find the contours based on (2) and get the bounding box. Select the most possible region as the license plate region.
(4) Segment ROI of the licnese plate and apply thresholding.
(5) Find contours of each letters and numbers and then segment them out as individuals.
(6) Apply template matching/feature matching to recognize each number and letter. Maybe in the future will add Convolution Network for classification.