Bangla number plate recognition from noisy video footage using deep learning

Abstract

The automatic number plate recognition of a vehicle from noisy footage has become an important issue in Bangladesh. In Bangladesh, the number of vehicles is increasing at a very high frequency. Moreover, unlike the English number plate, the Bengali number plate consists of two line inputs. As the number is increasing the difficulty of identifying a vehicle is also increasing. There are multiple problems where number plate recognition is very necessary such as, in a crime scene, finding out a lost vehicle, identifying the guilty vehicle in a road accident, etc. The main challenge in this system is to predict the Bangla numbers from very noisy images. Most of the methods of identifying vehicles from noisy data are not as accurate as they should be. In order to reduce such noise, we applied a total of ten filter algorithms but among those four filtration techniques gave us good results those are Conservative Filter, Wavelet denoising filter, Primal-Dual Algorithm, and Autoencoding. In order to detect the characters from the image files extracted from the camera footage, we used two detection algorithms. Those are Haar-Cascade and another one is Contour based Edge Detection algorithm. And to recognize the Bangla number characters from the bangla number plate we used ANN Backpropagation method.