Brazil-based Mercosur Plate Safety Enhancement
Budget: $30 – $250 USD
Requirement Description
The project involves capturing frames from video streams, with a focus on detecting and recognizing Mercosur License Plates. An algorithm for image cropping (after detection), grayscale conversion, thresholding (using Otsu's method and binary inversion), deskewing and filtering characters of the image has already been developed with OpenCV.
It is necessary to train a YOLO model for detecting these plates to generate weight, cfg and names files that will be used by the cv2 DNN. Furthermore, training of Tesseract for recognizing mercosur plate characters from the same plates is required, with training data consisting of images pre-processed according to the aforementioned sequence.
Tesseract must recognize the license plate characters with this single command:
tesseract -l <trained-font-for-OCR> --psm 6 --oem 3 -c tessedit_char_whitelist=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ-
*Tools and Versions*
Python 3.11.7
numpy==1.26.3, pytesseract==0.3.13, imutils==0.5.4, deskew==1.5.1, opencv-python==4.7.0
OpenCV 4.7.0
Tesseract 5.3.3.20231005
*Links*
https://en.wikipedia.org/wiki/Vehicle_registration_plates_of_the_Mercosur
https://en.wikipedia.org/wiki/FE-Schrift
The project involves capturing frames from video streams, with a focus on detecting and recognizing Mercosur License Plates. An algorithm for image cropping (after detection), grayscale conversion, thresholding (using Otsu's method and binary inversion), deskewing and filtering characters of the image has already been developed with OpenCV.
It is necessary to train a YOLO model for detecting these plates to generate weight, cfg and names files that will be used by the cv2 DNN. Furthermore, training of Tesseract for recognizing mercosur plate characters from the same plates is required, with training data consisting of images pre-processed according to the aforementioned sequence.
Tesseract must recognize the license plate characters with this single command:
tesseract -l <trained-font-for-OCR> --psm 6 --oem 3 -c tessedit_char_whitelist=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ-
*Tools and Versions*
Python 3.11.7
numpy==1.26.3, pytesseract==0.3.13, imutils==0.5.4, deskew==1.5.1, opencv-python==4.7.0
OpenCV 4.7.0
Tesseract 5.3.3.20231005
*Links*
https://en.wikipedia.org/wiki/Vehicle_registration_plates_of_the_Mercosur
https://en.wikipedia.org/wiki/FE-Schrift