Infrared Small Target Detection Using Nested FPN

Infrared small-target detection is very important in surveillance, aerospace monitoring and defense systems. Traditional methods only look at one frame at a time. This leads to detections and a lot of false alarms. This project suggests a framework called Video-TNFPN. It helps in detecting targets by looking at both space and time. The system uses layers called ConvLSTM to understand movements over time. It also uses attention mechanisms to focus on features. A refinement module helps in making predictions stable across frames. The model looks at features in space. Combines them with features at different scales. This helps in detecting low-contrast targets in complex environments. Experimental results show that the system is more accurate. It has false positives and stable detection performance. The system can process information in real-time. This makes it suitable for surveillance, tracking and defense applications. The proposed approach works well in environmental conditions. It is also robust against noise and background clutter. This makes the system very reliable for use in world dynamic infrared scenarios. The Video-TNFPN framework is seful, for small-target detection. It improves detection accuracy and stability. The system can be used in practical applications.