Remote Sensing Object Detection Enhancement Technology Based on Deformable Attention Convolution

Due to the specific imaging conditions and capture angles, remote sensing images often contain numerous small rotating objects, whose angular characteristics pose a challenge for traditional convolutions to accurately extract features. Addressing this challenge, this paper proposes an innovative remote sensing object detection method based on DeformAttnConv and KF-SIoU Loss, aiming to optimize the detection of small rotating objects. To enhance the model's ability to extract features from rotating small objects, we introduce the DeformAttnConv module. By integrating deformable convolutions with an attention mechanism, this module can dynamically adjust the size and shape of the receptive field, enabling precise capture of local object features. This innovation significantly improves the model's adaptability to the diversity and complexity of objects in remote sensing images. Furthermore, this paper also presents a novel loss function, KF-SIoU Loss, designed to optimize the model's bounding box regression performance. This loss function incorporates KFIoU Loss with an angular penalty term, comprehensively considering the positional, size, and angular differences between predicted and ground truth boxes. This effectively enhances the model's accuracy in boundary box regression, particularly when detecting small rotating objects. In our experiments, we employed R3Det as the baseline and integrated DeformAttnConv and KF-SIoU Loss, validating our approach on multiple remote sensing image datasets. The results demonstrate that our proposed method achieves significant performance improvements in the detection accuracy of rotating small objects compared to existing remote sensing detection algorithms, thus validating the effectiveness of DeformAttnConv and KF-SIoU Loss in remote sensing object detection tasks. Therefore, the remote sensing object detection method proposed in this paper provides a new perspective and tool for research in the field of remote sensing image processing, holding significant importance in enhancing the accuracy and efficiency of remote sensing object detection.