Fast Thermal Infrared Image Ground Object Detection Method based on Deep Learning Algorithm

Nowadays, there is an increasing demand for wild searching and nature reserve monitoring in nighttime environment. Combined with the advantages of thermal infrared (TIR) imaging technology and the flexibility of unmanned aerial vehicle (UAV), the UAV thermal infrared remote sensing can provide sufficient image data as the information source of decision-making and judgment for multi scene object detection task at night. In order to process these image data quickly and accurately, and to realize the purpose of ground object detection, this paper investigates a fast thermal infrared image ground object detection method based on deep learning algorithm. The proposed method achieves good accuracy and speed performance on the dataset, which shows 64.35% mAP with 41.83 FPS. This method is superior to the other general detection algorithms in detection accuracy and detection speed.