Automatic Vehicle License Plate Recognition Using Lightweight Deep Learning Approach

Intelligent transportation systems rely heavily on automatic vehicle license plate recognition (AVLPR). This system is required for traffic control, electronic toll collection, and access to parking lots. Due to the complexity and structure of license plate design for each country, especially in Arabic countries, license plate recognition is one of the most difficult recognition problems. Using image processing and deep learning (DL) techniques, the proposed AVLPR system is applied to new Iraqi license plates (LPs). We proposed a Lightweight Convolution Neural Network (LWCNN) in this study. It has four convolutional blocks, each with a batch of Max-pooling and Dropout layers, as well as a Flatten layer and two Dense layers. This architecture is simple, fast, and dependable, with promising classification accuracy that could propel the field forward. The proposed algorithm achieved 98% detection accuracy for the standard new PLs style and 85% training accuracy for digits dataset.