A Novel Approach to Detect Parking Space Occupancy for Efficient Urban Management

Objectives: This study aims to develop A Novel Approach to Detect Parking Space Occupancy for Efficient Urban Management utilization and enhance user experience with real-time, accurate data. Methods/Analysis: The proposed system detects the parking space occupancy for efficient urban management by using a Multi-Component Attention Graph Convolutional Neural Network (DPSO-MCAGCNN) and processes data from the PKLot dataset. Pre-processing is performed using the Maximum Correntropy Quaternion Kalman Filter (MCQKF) for normalization. Key features like area, perimeter, and aspect ratio are extracted using the Second-Order Time-Reassigned Multi synchro squeezing Transform (SOTRMT) and analyzed through MCAGCNN. The Leaf-in-Wind Optimization (LWO) technique is incorporated to optimize the MCAGCNN for higher accuracy. Findings: The proposed system achieves significant improvements over existing methods, including 27.84%-29.27% higher accuracy, 25.87%-29.84% improved R², and 16.27%-19.84% reduced Mean Squared Error (MSE). Evaluation metrics such as RMSE, MAE, and MAPE confirm its robust performance. Novelty/Improvement: The integration of LWO into MCAGCNN enhances optimization and precision, surpassing the performance of state-of-the-art methods like EUPE-SVM, RTPM-YOLOv5, and MASP-LSTM, making it an innovative solution for intelligent parking management. Doi: 10.28991/CEJ-2025-011-02-015 Full Text: PDF