Image Processing & OCR solution

License plate surveys are an important tool for transportation system planning for many years. These tools helps in origin-destination matrices, and provide basis for travel time estimations. The first generation of surveys involved human readers who would spend a lot of time in capturing data. The modern license plate surveys are based on video technology. License plate images are captured on videotape for data reduction at the analyst's office. If the tapes are manually transcribed, it's very expensive and labor intensive job.

This software developed for Indiana DOT was targeted to minimize the human role in the data reduction process and, in doing so, reduce the expenses involved. Image processing and artificial intelligence algorithms were implemented to extract license plate number and time stamp information from video data. A time stamp, that is present in all the frames, is also reported along with the license plate number. An open-source optical character recognition engine (OCR) was used to recognize individual characters in this software.

Video License Plate Data Reduction (VLPDR) software was developed and tested on a real dataset. The software can detect the presence of a vehicle and read the time stamp in approximately 85 percent of the cases, depending on the quality of the data based on lighting conditions, etc. Although the VLPDR software was not effective in recognizing most of the license plate numbers, VLPDR can reduced the labor time by upto 60 percent in manual transcription. The user is no longer required to work with difficult video players to identify and read frames with a vehicle present in the scene. The user only needs to use the software to identify and validate the results in a user friendly GUI.

This solution had been developed using core C/C++ technologies where image processing and artificial intelligence algorithms have been implemented.

Rajeev Gupta, CPIM
Principal
HyTech Professionals LLC
[The principals of HyTech Professionals led the development of this LPDR software while at Purdue University under Prof. Jon D. Fricker. This work has been published in ‘Transportation Research Record’ as Journal Title: Transportation Research Record, Issue 1804, pp 31-38, ISBN 0309077303.

Mr. Gupta has significant experience in image processing and command in artificial intelligence. He has been involved in utilizing some of the core software knowledge in Supply Chain Planning as well. He holds a graduate degree from Purdue University, West Lafayette, IN and is a graduate of Indian Institute of Technology (IIT), Kanpur India]