The improved model of YOLOV8 integrating LSKNet

Due to the characteristics of satellite remote sensing targets such as small size, dense distribution, variable angles, and complex backgrounds, traditional de-tec-tion methods are prone to false detections and missed detections. To im-prove the detection performance of the model, an improved Yolov8 model for remote sensing small target detection is proposed. LSKNet is adopted as the backbone network, which dynamically adjusts its large spatial receptive field to better model the ranging scenarios of objects in remote sensing scenes. The ver-ifica-tion through the DOTA-v1.0 dataset shows that the average precision (mAP) value of the improved model reaches 60.2%, which is 2.9 percentage points higher than that of the traditional model. The improvement measures can sig-nificantly enhance the accuracy and efficiency of detection, providing a more reliable and efficient solution for the field of satellite remote sensing.