03-15-2016 08:15 AM
hello. i am using ocr through vision assistant interface.i am recognizing license plate over there.but my effiecency of recognition is very poor.not even single number plate i can fully recognize.one or two letters are mismatching or not recognized.what can i do?
03-15-2016 08:28 AM
I suggest you post some of your sample images, and some details of the steps you are applying to the image in the vision assistant (upload the project/code?). Without information about what you're doing, it's impossible to offer any sensible suggestions. For all we know your source images might be too noisy to pick up the characters (dirty number plate? 🙂 ).
03-16-2016 07:11 AM - edited 03-16-2016 07:14 AM
hello sam_sharp
i have the number plate image "1" which is i extracted from whole image of car.then i applied some processing on it and after process image became "2".i am also putting image of this steps which i appplied before ocr.
so now my original image is image 1.then i appled niblack thresholding.then basic morpholgy used to dilate the image.then low pass filter to smooth it.then advance morphology to remove small objects.then i convert binary into gray scale image using "equalize".then it become image 2.which i give for ocr.
03-18-2016 10:49 AM
Each time you encounter some characters that are not recognized properly, you can edit the character set file and train the characters that are not recognized properly. This should improve the recognition rate.
Also, I don't see in your Vision Assistant script functions that can be used to reposition the region of interest (locating some fiducial and creating a coordinate system to be able to reposition the Region of Interest used in the OCR step.
When you go through your different images, is the license plate always at the same place? Is the region of interest of the OCR step position correctly for al the images?
To verify that, you can edit the OCR step, and click run once in the script toolbar (above the script icon representation) to go through your set of images one by one.
Hope this helps.
Christophe
03-18-2016 11:09 AM
hello.christopheC
can u explain more about script function?how it can implement and reposition the roi. every time for each image my license plate will be at different location.
03-18-2016 12:26 PM
Sagar,
Your messages are hard to read. Please use capital letters to start sentences like you were taught in school. Also, put a couple spaces after periods we can visually see where one sentence stops and another begins.
Also, be sure to use words such as "you" and not text message abbreviations.
Thank you!
03-18-2016 12:45 PM
Take a look at the following example that illustrates the concept of Region of Interest repositioning using a coordinate system.
From the main menu, select Help>>Solution Wizard..., then select Dental Floss Inspection from the Pharmaceuticals solutions in the tree.
Click Load Solution.
This example shows how you can use for example a pattern matching step to first locate the object, define what we call a coordinate system that will be used to reposition regions of interests in later steps using the position and angle found by the pattern matching step.
In this example, the ROIs defined in the Measure Area Intensity step are repositioned using the coordinate system and appear at the same location on each dental floss, allowing to accurately measure the intensity in specific areas to determine the presence/absence of toothpaste and dental floss wires.
Pattern matching is only one of the possible locating tools you can use to define locate an object, In some applications, edge detection is often used for a similar purpose. You can reposition ROIs based on horizontal, vertical and angular offsets.
Hope this helps.
Christophe
03-18-2016 01:56 PM
ok..got it. Thanks a lot.
04-07-2017 12:17 AM
Hi ,
i am working with "Hindustan times" newspaper located in New Delhi,India.
we are planning to buy the ocr software, our requirement is to convert the handwritten Hindi material into PDF doc format in hindi Devanagari script.so i
would request you to share some videos or ppt so that we can get an idea how it will work in real time.