Wind Turbine Detection and Analysis Based On Deep Learning Methods

In order to achieve China's "carbon peak", "carbon neutral" strategic development goals, the development and reuse of clean energy is very important, accurate and efficient extraction of wind turbines for wind energy utilization efficiency and environmental protection is of great significance, but the manual visual However, the manual visual interpretation method is too inefficient to meet the demand of large-scale accurate monitoring. Aiming at the variable scale of wind turbines in remote sensing images, a batch of 2m resolution samples are produced to train the model for accurate wind turbine extraction. Based on the deep learning network YOLOv9 (You Only Look Once), the SPD-Conv module is added to enhance the network's ability to detect small targets, and at the same time, adapt to the semantic features of wind turbines in different contexts. Finally, based on the domestic GF-1 satellite images, the extraction of wind turbines in North China is realized. The experimental results show that the extraction accuracy of the network can reach 91.2%.