ALGORITHM FOR DETECTING WEAK INFRARED TARGETS ON A COMPLEX BACKGROUND USING THE YOLOv5 NEURAL NETWORK MODEL

Wang Ch, Afanasyev G. I., Afanasyev A. G.

The traditional algorithm for detecting infrared targets is based on the accurate selection and extraction of information from the surrounding background and does not always meet the requirements of detection in conditions of complex background and interference. The article discusses a modified algorithm for detecting weak infrared targets, which is based on the neural network model YOLOv5. This algorithm adds an attention mechanism to improve feature extraction ability and efficiency, and uses a modified loss function and a prediction frame filtering method to improve accuracy in detecting weak infrared targets.

Keywords:infrared weak targets, complex backgrounds, YOLOv5, attention mechanism, loss function