Onboard real-time object detection in remote sensing images is a crucial but challenging task in this computation-constrained scenario. This task not only requires the algorithm to yield excellent performance but also requests limited time and space complexity of the algorithm. However, previous convolutional neural networks (CNN) based object detectors for remote sensing images suffer from heavy computational cost, which hinders them from being deployed on satellites. Moreover, an onboard detector is desired to detect objects at vastly different scales. To address these issues, we proposed a lightweight one-stage multi-scale feature fusion detector called MSF-SNET for onboard real-time object detection of remote sensing images. Using lightweight SNET as the backbone network reduces the number of parameters and computational complexity. To strengthen the detection performance of small objects, three low-level features are extracted from the three stages of SNET respectively. In the detection part, another three convolutional layers are designed to further extract deep features with rich semantic information for large-scale object detection. To improve detection accuracy, the deep features and low-level features are fused to enhance the feature representation. Extensive experiments and comprehensive evaluations on the openly available NWPU VHR-10 dataset and DIOR dataset are conducted to evaluate the proposed method. Compared with other state-of-art detectors, the proposed detection framework has fewer parameters and calculations, while maintaining consistent accuracy.

High security and effectiveness are critical performance metrics in the data transmission process for satellite remote sensing images, medical images, and so on. Previously, the receiver could gain a high-quality cover image (lossy) after decryption in a separable manner to balance embedding capacity (EC) and security. Completely separable, reversible data hiding in encrypted image (SRDH-EI) algorithms are proposed to address this issue. In this study, the cover image was preprocessed at the sender’s end. The pre-embedded pixels and most significant bits (MSB) were compressed via two coding methods to reserve space. Additionally, the header data were embedded for marking. Finally, auxiliary data and secret data were embedded in a forward “Z” and reverse “Z” shape before and after encryption, respectively. The receiver could extract secret data and decrypt the cover image separately using the keys and markers. The experimental results demonstrate that the algorithm reached a high EC for remote sensing images by utilizing pixel correlation at multiple positions within the groups. The cover image could maintain its entropy during the data embedding process, ensuring security. The decrypted image could be recovered without distortion, furthermore, the receiver could achieve complete separability, so it has good application prospects for remote sensing images.

Although convolutional neural networks (CNNs) have made significant progress, their deployment onboard is still challenging because of their complexity and high processing cost. Tensors provide a natural and compact representation of CNN weights via suitable low-rank approximations. A novel decomposed module called DecomResnet based on Tucker decomposition was proposed to deploy a CNN object detection model on a satellite. We proposed a remote sensing image object detection model compression framework based on low-rank decomposition which consisted of four steps, namely (1) model initialization, (2) initial training, (3) decomposition of the trained model and reconstruction of the decomposed model, and (4) fine-tuning. To validate the performance of the decomposed model in our real mission, we constructed a dataset containing only two classes of objects based on the DOTA and HRSC2016. The proposed method was comprehensively evaluated on the NWPU VHR-10 dataset and the CAST-RS2 dataset created in this work. The experimental results demonstrated that the proposed method, which was based on Resnet-50, could achieve up to 4.44 times the compression ratio and 5.71 times the speedup ratio with merely a 1.9% decrease in the mAP (mean average precision) of the CAST-RS2 dataset and a 5.3% decrease the mAP of the NWPU VHR-10 dataset.

With the increasing requirements for the security of medical data, military data, and other data transmission, data hiding technology has gradually developed from only protecting the security of secret data to all transmission data. As a necessary technical means, reversible data hiding in encrypted images (RDH-EIs) provides superior performance in terms of security. To simultaneously improve the effectiveness of RDH-EIs, this work proposes a mixed multi-bit layer embedding strategy in encrypted images. The cover image is processed into two categories: available hidden blocks (AHBs) and unavailable hidden blocks (UHBs) at the sender. Then, all data are embedded in the multi-bit layer of the encrypted pixels in AHBs through two embedding strategies to obtain the transmission image. At the receiver, the user can extract the needed data separably according to different keys to achieve error-free extraction of the secret data and lossless recovery of the cover image. The experimental results show that the proposed scheme has the advantages of superior embedding capacity and high decryption quality over the current state-of-the-art works.