Video-Based License Plate Detection Using Deep Learning Boundary Filtering Method for License Plate Detection from Surveillance Images

License plate detection from moving vehicles is useful in authenticating owners, detecting vehicle misbehaviors, etc. Roadside video outputs are analyzed using computer vision-based algorithms/ methods to improve the detection precision.This article thus introduces a Boundary Filtering Method (BFM) using Conditional Neural Learning (CNL).In this method, the conventional neural network with filtering conditions is used to identify the license plate boundary.The congruent textural features are filtered based on trained inputs from datasets.The similar boundary indices identified in the training images are used to shape the license plate region from the frame inputs.The conditions of maximum similarity and boundary displacement connectivity are verified throughout the training process until maximum precision is reached.The condition-failing features are filtered to reduce the false positives between different frame orientations.The proposed method is verified using accuracy, precision, similarity index, false positives, and time metrics.The proposed method improves precision by 9.57%and reduces false positives and analysis time by 10.43% and 6.28% respectively for the boundaries identified.