A method for monitoring coastal water areas using video data analysis obtained from unmanned aerial vehicles (UAVs) to enhance water safety is presented. The proposed approach involves detecting water transport objects in images using a modern deep learning model and classifying their movement (moving or stationary) based on optical flow. A dataset of over 2500 aerial images was developed and annotated, including vessels and their wake traces. A detection model with high accuracy (F1-score up to 0.98) was trained to identify vessels on water. For movement classification, a combination of optical flow analysis and wake trace detection is used; a composite metric is proposed to accurately distinguish moving vessels from stationary ones. The results demonstrate the effectiveness of the approach for automated water area monitoring and early detection of potentially hazardous situations.