Deep learning (DL)-based object detection algorithms have gained impressive achievements in natural images and have gradually matured in recent years. However, compared with natural images, remote sensing images are faced with severe challenges due to the complex backgrounds and difficult detection of small objects in dense scenes. To address these problems, a novel one-stage object detection model named MDCT is proposed based on a multi-kernel dilated convolution (MDC) block and transformer block. Firstly, a new feature enhancement module, MDC block, is developed in the one-stage object detection model to enhance small objects’ ontology and adjacent spatial features. Secondly, we integrate a transformer block into the neck network of the one-stage object detection model in order to prevent the loss of object information in complex backgrounds and dense scenes. Finally, a depthwise separable convolution is introduced to each MDC block to reduce the computational cost. We conduct experiments on three datasets: DIOR, DOTA, and NWPU VHR-10. Compared with the YOLOv5, our model improves the object detection accuracy by 2.3%, 0.9%, and 2.9% on the DIOR, DOTA, and NWPU VHR-10 datasets, respectively.

The on-board pose estimation of uncooperative target is an essential ability for close-proximity formation flying missions, on-orbit servicing, active debris removal and space exploration. However, the main issues of this research are: first, traditional pose determination algorithms result in a semantic gap and poor generalization abilities. Second, specific pose information cannot be accurately known in a complicated space target imaging environment. Deep learning methods can effectively solve these problems; thus, we propose a pose estimation algorithm that is based on deep learning. We use keypoints detection method to estimate the pose of space targets. For complicated space target imaging environment, we combined the high-resolution network with dilated convolution and online hard keypoint mining strategy. The improved network pays more attention to the obscured keypoints, has a larger receptive field, and improves the detection accuracy. Extensive experiments have been conducted and the results demonstrate that the proposed algorithms can effectively reduce the error rate of pose estimation and, compared with the related pose estimation methods, our proposed model has a higher detection accuracy and a lower pose determination error rate in the speed dataset.