SmartEye Surveillance System using AI Detection

With the growing demand for intelligent surveillance systems, real-time abnormal activity detection has become an essential feature in security monitoring, aged care, and public safety applications. Traditional deep learning algorithms frequently necessitate big datasets, significant processing capacity, and complex training procedures. This study introduces the SmartEye Surveillance System, a lightweight, real-time anomalous activity detection platform based on MediaPipe pose estimation and absolute inter-frame distance analysis. To achieve translation and scale invariance, the system uses MediaPipe to extract 33 human body landmarks and then applies center and scale normalization. Abnormal activity is discovered by calculating the absolute Euclidean distance between consecutive frames’ landmarks. If the estimated motion intensity exceeds a predetermined threshold, the system considers the activity abnormal and issues an alarm. The suggested strategy reduces the requirement for massive training datasets while yet preserving real-time performance and computational economy. Under single-person observation, the experimental results show that abrupt falls, aggressive gestures, and strange posture changes can be reliably detected..