Aiming at the problems of low object detection accuracy in remote sensing images caused by special angle of view，complex background information，dense objects，and diverse scales，an object detection algorithm for remote sensing images based on multi-scale feature fusion，EMD-YOLOv8 is proposed. Firstly，an enhanced backbone network，EnhancedDarkNet is introduced to realize shallow feature reuse and enhance the extraction capability of object details and texture features. Secondly，a Multi-scale Interactive Fusion Feature Pyramid Network (MIFFPN) is designed to reconstruct the neck structure and strengthen feature fusion in multi-scale spaces. Finally，a Dimension Interactive Polarization Attention (DIPA) mechanism is proposed to reduce background noise and interference of redundant information，enhancing the response of key object features. The proposed algorithm achieves an mAP@0.5 and mAP@0.5∶0.95 of 85.9% and 62.5% on the remote sensing dataset DIOR，which is 5.1 and 5.6 percentage points higher than that of the original YOLOv8n respectively，while the detection rate reaches 90.9 frames per second. Experimental results demonstrate that EMD-YOLOv8 can improve the accuracy of object detection in remote sensing images and meet the requirements of real-time detection performance.