Optical Character Recognition Based Intelligent Database Management System for Examination Process Control

Computer Vision techniques have recently opened several new avenues of research owing to advancement in hardware and software technologies. This technique relies on extracting relevant information from images to take intelligent decisions. In this project, we are using two subclasses of computer vision techniques namely optical character recognition (OCR) and object recognition to develop an intelligent database management system for post examination process control. Using MNIST database, in which we have 10,000 samples of test set and 60,000 samples of training set, we have applied a deep learning algorithm for handwritten digits recognition on an examination sheet. Images are adjusted using pre-processing techniques. After extraction of information from the first sheet of examination copy, we have classified the signatures from invigilator and examiner using feature extraction algorithm where signatures are treated as objects. The extracted information is stored in database and individual and collective scores are then computed. Our model also generates graphs of result for better understanding. This initial work reduces the workload of post examination data entry process and makes it time efficient. In future, it is expected to be integrated in an android based application for automatic post examination control expected to be used at author's university.