Abstract: Traditional camera maintenance methods rely on periodic manual inspections, which are time consuming, costly, and inherently limited in that they only allow problems to be detected and addressed after they occur. In large-scale industrial sites operating hundreds of cameras, such methods frequently lead to human errors and delays in response, resulting in decreased safety and productivity. This study proposes an automated and intelligent predictive maintenance system by integrating artificial intelligence (AI)-based video analysis technology with industrial Internet of Things (IIoT) systems to overcome these limitations. The proposed system collects highresolution video data in real-time and operates through a series of processes including image preprocessing, feature extraction, anomaly detection, severity classification, and maintenance alert transmission. Utilizing CNN-based deep learning algorithms and OpenCV image processing techniques, the system can automatically detect issues such as lens contamination, focus blur, and image degradation. When anomalies are identified, they are immediately classified, and alerts are sent in real-time via a cloudbased notification system. Additionally, maintenance history is automatically logged and analyzed in a database, supporting the development of long-term asset management strategies. Experimental results in real industrial environments demonstrate that the proposed system improves detection accuracy by over 90–95% compared to manual inspection methods, reduces alert response time to within seconds, and lowers maintenance time and costs by more than 70% and 40%, respectively. This research validates the practical effectiveness of automated and predictive maintenance in camera systems as a core technology for smart factory implementation and is expected to contribute to the development of more scalable maintenance frameworks through integration of multi-sensor data and further advancements in predictive algorithms. Index Terms—Predictive Maintenance, Industrial IoT (IIoT), Automated Camera, AI Image Analysis, SmartFactory.

Keywords: Monocular Depth Estimation, RGB-D Fusion, Object Detection, Computer Vision, Deep Learning

Pages: 1955-1970