Bangladeshi License Plate Recognition Using Adaboost Classifier

License plate recognition (LPR) is a technology for the authentication of a vehicle by locating and recognizing the license plate number in an image through computer vision techniques and machine learning models. To develop intelligent traffic management such as vehicle monitoring, LPR is a key component. However, due to the diversity of layouts and characters of plates, universal solution is not possible. So, this research focuses on development of an algorithm for the recognition of license plate of Bangladesh by using image processing's and machine learning model. This algorithm executes in three steps: detection of the plate with shape verification, tilt correction and recognition of the number. For detection, RGB color space, median filtering, binarization, morphological analysis, region properties for filtering are applied. To discard noisy object, shape verification is done through robust distances to borders vectors. Before character segmentation, horizontal tilt correction is applied. Then, characters are extracted by using bounding box parameters from the extracted plate. Finally, the recognition is implemented by using the blending of Histogram Oriented Gradient (HOG) and Local Binary Pattern (LBP) features and adaptive boosting (Adaboost) classifier is used to categorize the characters. The proposed algorithm is simulated on the images which are captured from different roads of Bangladesh. The experimental result shows that the detection and recognition accuracy is noteworthy.