Abstract

This work explores the development of an elegant framework for LOGO Net-based Internet of Things (IoT)-enabled suspicious vehicle detection. The proposed design integrates a Convolutional Neural Network (CNN) to classify vehicles as suspicious or non-suspicious based on visual input. The process began with data collection and labeling, followed by preprocessing techniques such as image resizing and normalization. A custom CNN architecture was designed and trained using the prepared dataset, with data augmentation strategies applied to enhance model robustness. The training process was monitored using validation accuracy and loss metrics, with early stopping im-plemented to prevent overfitting. Detection results were evaluated using key performance indicators, including accuracy, precision, recall, and F1-score. The final model achieved a high overall accuracy of 96.7%, with 94.3% and 97.8% accuracy for suspicious and non-suspicious images, respectively. These results highlight the model's reliability and potential for enhancing re-al-time vehicle surveillance in IoT-based security systems through intelligent automation.