Remote sensing image target detection algorithm based on improved YOLOv8

Abstract:

In this field of target detection, remote sensing targets are characterised by small size, complex background and imprecise target positioning, which can reduce the performance of the detection network and lead to misdetection and omission. On the basis of YOLOv8 backbone detection network, a module named LSKB_ECA is designed, which dynamically adjusts the spatial sensing field and the positioning of key points by combining the large selective kernel LSK Block with the ECA attention mechanism to enhance the model's attention to the target features; in order to improve the model's detection ability in complex scenes, a module named C2f_C-DCN is designed to replace the C2f_C-DCN in YOLOv8. module is designed to replace the C2f module in YOLOv8. The experimental results show that the mean average accuracy of the improved YOLOv8 algorithm is improved by 2.7% and 2.1% on DOTA and DIOR datasets, respectively, and the performance of target detection is improved significantly.

Key words:Remote sensing image target detection; YOLOv8; LSKB_ECA; C2f_C-DCN