An Improved YOLOv8 Algorithm for Aerial Image Detection

• Published in: 2024 7th International Conference on Computer Information Science and Application Technology (CISAT) pp. 56 - 61
• Main Authors: Su, Guolong, Zhu, Changming
• Format: Conference Proceeding
• Language: English
• Published: IEEE 12.07.2024
• Subjects: Accuracy; Attention mechanisms; Autonomous aerial vehicles; BiFPN module; Computational modeling; Feature extraction; GAM attention mechanism; improved YOLOv8 models; Object detection; Optimization; Satellite images; Target recognition; Urban planning; VisDrone dataset; YOLOv8
• DOI: 10.1109/CISAT62382.2024.10695323

In recent years, the rapid development of UAV technology has promoted the widespread application of UAV aerial imagery in urban planning, environmental monitoring, disaster assessment and other fields. However, the complexity and diversity of UAV aerial photography data pose challenges to traditional target detection methods, which are usually difficult to handle such data effectively. To address these challenges and improve the accuracy and robustness of target detection, this study proposes a method to improve the YOLOv8 model on the VisDrone dataset. Based on the YOLOv8 fast single-stage detection method, this study replaces the CIoU loss function with the WIoU loss function. This change improves the regression accuracy of the bounding box, resulting in more accurate target detection. The study also introduces the global attention mechanism GAM and integrates the bi-directional feature pyramid network BiFPN module into the YOLOv8 model. GAM helps the model better understand the semantic information of the whole image, thus improving the target recognition and localization ability, while BiFPN further strengthens the model's ability to perceive targets at different scales through multi-level feature fusion, feature optimization capability and detection performance. sensing ability of the model for targets at different scales through multi-level feature fusion, feature optimization capability and detection performance. These improvements further enhance the detection accuracy of the model and realize a better feature fusion and attention mechanism. The experimental results show that the improvements proposed in this study significantly improve the performance of the YOLOv8 model, resulting in a significant increase in the detection accuracy.