Baojun Lin

Baojun Lin


Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 April 2022 in Sensors
Verified
Journal Article
Open access
PDF file
FXFang Xie
Fang Xie
BLBaojun Lin
Baojun Lin
YLYingchun Liu
Yingchun Liu
The real-time performance of ship detection is an important index in the marine remote sensing detection task. Due to the computing resources on the satellite being limited by the solar array size and the radiation-resistant electronic components, information extraction tasks are usually implemented after the image is transmitted to the ground. However, in recent years, the one-stage based target detector such as the You Only Look Once Version 5 (YOLOv5) deep learning framework shows powerful performance while being lightweight, and it provides an implementation scheme for on-orbit reasoning to shorten the time delay of ship detention. Optimizing the lightweight model has important research significance for SAR image onboard processing. In this paper, we studied the fusion problem of two lightweight models which are the Coordinate Attention (CA) mechanism module and the YOLOv5 detector. We propose a novel lightweight end-to-end object detection framework fused with a CA module in the backbone of a suitable position: YOLO Coordinate Attention SAR Ship (YOLO-CASS), for the SAR ship target detection task. The experimental results on the SSDD synthetic aperture radar (SAR) remote sensing imagery indicate that our method shows significant gains in both efficiency and performance, and it has the potential to be developed into onboard processing in the SAR satellite platform. The techniques we explored provide a solution to improve the performance of the lightweight deep learning-based object detection framework.

30 January 2024 in Remote Sensing
Verified
Journal Article
Open access
PDF file
ZDZhong Dong
Zhong Dong
BLBaojun Lin
Baojun Lin
FXFang Xie
Fang Xie
In the realm of few-shot classification learning, the judicious application of data augmentation methods has a significantly positive impact on classification performance. In the context of few-shot classification tasks for remote sensing images, the augmentation of features and the efficient utilization of limited features are of paramount importance. To address the performance degradation caused by challenges such as high interclass overlap and large intraclass variance in remote sensing image features, we present a data augmentation-based classification optimization method for few-shot remote sensing image scene classification. First, we construct a distortion magnitude space using different types of features, and we perform distortion adjustments on the support set samples while introducing an optimal search for the distortion magnitude (ODS) method. Then, the augmented support set offers a wide array of feature distortions in terms of types and degrees, significantly enhancing the generalization of intrasample features. Subsequently, we devise a dual-path classification (DC) decision strategy, effectively leveraging the discriminative information provided by the postdistortion features to further reduce the likelihood of classification errors. Finally, we evaluate the proposed method using a widely used remote sensing dataset. Our experimental results demonstrate that our approach outperforms benchmark methods, achieving improved classification accuracy.

29 May 2024 in Remote Sensing
Verified
Journal Article
Open access
PDF file
WYWang Yinxiao
Wang Yinxiao
WYWei Yuan
Wei Yuan
FXFang Xie
Fang Xie
BLBaojun Lin
Baojun Lin
Super-resolution (SR) for satellite remote sensing images has been recognized as crucial and has found widespread applications across various scenarios. Previous SR methods were usually built upon Convolutional Neural Networks and Transformers, which suffer from either limited receptive fields or a lack of prior assumptions. To address these issues, we propose ESatSR, a novel SR method based on state space models. We utilize the 2D Selective Scan to obtain an enhanced capability in modeling long-range dependencies, which contributes to a wide receptive field. A Spatial Context Interaction Module (SCIM) and an Enhanced Image Reconstruction Module (EIRM) are introduced to combine image-related prior knowledge into our model, therefore guiding the process of feature extraction and reconstruction. Tailored for remote sensing images, the interaction of multi-scale spatial context and image features is leveraged to enhance the network’s capability in capturing features of small targets. Comprehensive experiments show that ESatSR demonstrates state-of-the-art performance on both OLI2MSI and RSSCN7 datasets, with the highest PSNRs of 42.11 dB and 31.42 dB, respectively. Extensive ablation studies illustrate the effectiveness of our module design.

25 August 2022 in Photonics
Verified
Journal Article
Open access
PDF file
LZLisha ZHOU
Lisha ZHOU
HHHongyu He
Hongyu He
JSJianfeng Sun
Jianfeng Sun
BLBaojun Lin
Baojun Lin
Compared to traditional phase laser ranging technology, phase-shift laser ranging technology based on carrier phase modulation has significant advantages, such as simple modulation and demodulation, as well as simultaneous ranging and velocity measurements with high precision. To expand the range and ensure the ranging precision in carrier phase modulation in phase-shift laser ranging technology, we propose a phase-shift laser ranging system based on multi-frequency carrier phase modulation. We analyzed the effects of the stagger coefficient, modulation depth, and signal-to-noise ratio on the ranging precision by simulation. Finally, we built an experimental system with a range of 300 m by choosing appropriate parameters and conducting linearity and range precision tests. The experimental results show that the system has good linearity, and the ranging precision can reach 1 cm when the signal-to-noise ratio is 20 dB.

28 July 2022 in Sustainability
Verified
Journal Article
Open access
PDF file
FXFang Xie
Fang Xie
HLHao Luo
Hao Luo
SLShaoqian Li
Shaoqian Li
YLYingchun Liu
Yingchun Liu
BLBaojun Lin
Baojun Lin
This paper studies the lightweight deep learning object detection algorithm to detect ship targets in SAR images that can be deployed on-orbit and accessed in the space-based IoT. Traditionally, remote sensing data must be transferred to the ground for processing. With the vigorous development of the commercial aerospace industry, computing, and high-speed laser inter-satellite link technologies, the interconnection of everything in the intelligent world has become an irreversible trend. Satellite remote sensing has entered the era of a big data link with IoT. On-orbit interpretation gives remote sensing images expanse application space. However, implementing on-orbit high-performance computing (HPC) is difficult; it is limited by the power and computer resource consumption of the satellite platform. Facing this challenge, building a processing algorithm with less computational complexity, less parameter quantity, high precision, and low computational power consumption is a key issue. In this paper, we propose a lightweight end-to-end SAR ship detector fused with the vision transformer encoder: YOLO−ViTSS. The experiment shows that YOLO−ViTSS has lightweight features, the model size is only 1.31 MB; it has anti-noise capability is suitable for processing SAR remote sensing images with native noise, and it also has high performance and low training energy consumption with 96.6 mAP on the SSDD dataset. These characteristics make YOLO−ViTSS suitable for porting to satellites for on-orbit processing and online learning. Furthermore, the ideas proposed in this paper help to build a cleaner and a more efficient new paradigm for remote sensing image interpretation. Migrating HPC tasks performed on the ground to on-orbit satellites and using solar energy to complete computing tasks is a more environmentally friendly option. This environmental advantage will gradually increase with the current construction of large-scale satellite constellations. The scheme proposed in this paper helps to build a novel real-time, eco-friendly, and sustainable SAR image interpretation mode.

23 April 2025 in Applied Sciences
Verified
Journal Article
Open access
PDF file
LXLicheng Li Xia
Licheng Li Xia
BLBaojun Lin
Baojun Lin
SZShuai Zhao
Shuai Zhao
YZYanChun Zhao
YanChun Zhao
Designing routing algorithms for Low Earth Orbit (LEO) satellite networks poses a significant challenge due to their high dynamics, frequent link failures, and unevenly distributed traffic. Existing studies predominantly focus on shortest-path solutions, which compute minimum-delay paths using global topology information but often neglect the impact of traffic load on routing performance and struggle to adapt to rapid link-state variations. In this regard, we propose a Multi-Agent Reinforcement Learning-Based Joint Routing (MARL-JR) algorithm, which integrates centralized and distributed routing algorithms. MARL-JR combines the accuracy of centralized methods with the responsiveness of distributed approaches in handling dynamic disruptions. In MARL-JR, ground stations initialize Q-tables and upload them to satellites, reducing onboard computational overhead while enhancing routing performance. Compared to traditional centralized algorithms, MARL-JR achieves faster link-state awareness and adaptation; compared to distributed algorithms, it delivers superior initial performance due to optimized pre-training. Experimental results demonstrate that MARL-JR outperforms both Q-Routing (QR) and DR-BM algorithms in average delay, packet loss rate, and load-balancing efficiency.