0

I work currently on my project of end of study: the title of my project is "the detection in real time of the Moroccan license plates of vehicles ( ALPR )", I tried to post this my issue in group ALPR unfortunately ,i received no reply . i decide to contact you when i see your video in youtube,I want to inform you that I have many problem with Moroccan license plates. this is an example of Moroccan license plates: Morrocan alpr

i tried to use OPENALPR : When I run the command line " alpr lprmr.jpg" it give me this result :

plate0: 8 results - 4316 confidence: 83.3094 - 43I6 confidence: 76.1957 - 431B confidence: 68.5094 - 431G confidence: 67.8764 - 431S confidence: 64.9272 - 43IB confidence: 61.3957 - 43IG confidence: 60.7626 - 43IS confidence: 57.8135

Like you see tesseract fail because of the vertical lines and Arabic characters. However, i can use OpenCV's connected components functions to split the image into 3 sections. The first and third sections will work fine with tesseract. For the Arabic characters, i can build a neural network to classify them if you've enough dataset. To get the full number plate, and I will have to concatenate the 3 sections. for do that i think i should modify the code : /openalpr/src/openalpr/ocr/segmentation/charactersegmenter.cpp

The Question is how can be modify this code and detect the full Moroccan Plate ??

Edit: The Arabic letters we use in Morocco are : أ ب ت ج ح د هـ The expected result is : 77143 د 6 دThe vertical lines are irrelevant, I have to use them to separate the image and read data separately. Thanks in Advance

user13134819
  • 1
  • 1
  • There are a lot of flags to add to alpr, country flag for example. – Surt Mar 27 '20 at 11:34
  • what do you mean by flag : – user13134819 Mar 27 '20 at 11:36
  • like this command line : alpr -c us image.jpg – user13134819 Mar 27 '20 at 11:37
  • Look at the documentaion of the alpr for more options, and yes -c is one of them. – Surt Mar 27 '20 at 11:37
  • yes , used already : ./alpr -c us -p mr --config openalpr.conf.defaults images.jpg --debug it give me the same result , i think the problem is about segmentation. – user13134819 Mar 27 '20 at 11:40
  • I don't think `-c us` is correct for Moroccan plates? – Surt Mar 27 '20 at 11:43
  • i tried all country i used ** -c eu us br au ...** , the us and eu is closer of morrocan plate. – user13134819 Mar 27 '20 at 11:46
  • I have work on the same project, and trust me , moroccan plate aren't easy , morocco plate have several form (rectangles , square,...) And colors , and specifications , you cannot do that with tessrcat or any pretrained model, you need to train your own model and you need to find alot of data ( which is illegal by moroccan authority) – MoxGeek Nov 15 '20 at 00:56

0 Answers0