Advanced Real-Time License Plate Recognition Using Modified Lightweight Convolutional Greylag Goose Network

License plate recognition (LPR) involves reading and decoding car license plates based on optical character recognition captured from pictures or video. License plate recognition is used for monitoring traffic, parking, and security applications. This study presents a novel methodology for real-time LPR using an optical residual lightweight conventional Greylag goose (ORLCGG). The process begins with image acquisition, where input images are localized using a region-based linear layer (RLL) to effectively isolate license plates from their backgrounds, enhancing clarity. To improve image quality, the distortion blur preprocessing (DBP) addresses blurriness caused by optical distortions, particularly for side-mounted plates. It employs fuzzy logic for distortion estimation and calculates the point spread function (PSF) matrix, which normalizes the image to enhance clarity. A canny edge detection algorithm maximizes the features of the license plates and subsequently improves recognition accuracy. For detection with the system, segmentation and ResNeXt101 are combined for feature extraction, and then the appropriate lightweight convolutional neural network (LCNN) is tailored for its number plate recognition problem. Greylag goose optimization (GGO) is used to optimize the LCNN weight parameter. The combination increases efficiency for number recognition, with improved accuracy, and makes the system a viable solution for real-time LPR in a number of environments. The proposed method achieved a mean absolute error (MAE) of 0.11, a mean squared error (MSE) of 0.02 and an [Formula: see text] value of 0.93, demonstrating its effectiveness in improving detection accuracy across various datasets. Additionally, it attained a confidence level of 81% in real-time data, highlighting its superior feature extraction capabilities.