After the revival of deep learning in computer vision in 2012, SAR ship detection comes into the deep learning era too. The deep learning-based computer vision algorithms can work in an end-to-end pipeline, without the need of designing features manually, and they have amazing performance. As a result, it is also used to detect ships in SAR images. The beginning of this direction is the paper we published in 2017BIGSARDATA, in which the first dataset SSDD was used and shared with peers. Since then, lots of researchers focus their attention on this field. In this paper, we analyze the past, present, and future of the deep learning-based ship detection algorithms in SAR images. In the past section, we analyze the difference between traditional CFAR (constant false alarm rate) based and deep learning-based detectors through theory and experiment. The traditional method is unsupervised while the deep learning is strongly supervised, and their performance varies several times. In the present part, we analyze the 177 published papers about SAR ship detection. We highlight the dataset, algorithm, performance, deep learning framework, country, timeline, etc. After that, we introduce the use of single-stage, two-stage, anchor-free, train from scratch, oriented bounding box, multi-scale, and real-time detectors in detail in the 177 papers. The advantages and disadvantages of speed and accuracy are also analyzed. In the future part, we list the problem and direction of this field. We can find that, in the past five years, the AP₅₀ has boosted from 78.8% in 2017 to 97.8 % in 2022 on SSDD. Additionally, we think that researchers should design algorithms according to the specific characteristics of SAR images. What we should do next is to bridge the gap between SAR ship detection and computer vision by merging the small datasets into a large one and formulating corresponding standards and benchmarks. We expect that this survey of 177 papers can make people better understand these algorithms and stimulate more research in this field.

Through the use of video technology, satellites can detect dynamic targets and analyze their motion characteristics. Target tracking can extract dynamic information about key ground targets for target monitoring and trajectory prediction by satellite video. Tracking algorithms are affected by target motion characteristics, such as velocity and direction, as well as background characteristics, such as illumination changes, occlusion, and background similarities with the target. However, these problems are seldom studied with satellite video cameras. Current algorithms are unsuitable for satellite video because of the poor texture and color features of the target in satellite video. Therefore, in this article, we enhance target tracking for satellite video technology using two aspects: 1) sample training strategy and 2) sample characterization. We establish a filter training mechanism for the target and background to improve the discrimination ability of the tracking algorithm. We then build a target feature model using a Gabor filter to enhance the contrast between the target and background. Moreover, we propose a tracking state evaluation index to avoid tracking drift. Tracking experiments using nine sets of Jilin-1 satellite videos show that the proposed approach can accurately locate a target under weak feature attributes. Therefore, this article contributes to more robust tracking using satellite video technology.

Video synthetic aperture radar (Video SAR) has sparked lots of research attention since it provides the ability of continuously observing and tracking the object of interest. Instead of tracking a target directly, in Video SAR, it is preferred to track its shadow due to the target's instable backscattering characteristic and defocused patterns. However, the current tracking algorithms do not work well since they always suffer from the interference caused by surrounding clutters and cannot locate the target precisely. To address these issues, we propose a robust shadow tracking algorithm for Video SAR in this letter. We equip our tracker with the spatial-temporal (ST) information and saliency-based detection mechanism (SD) against the distractions and background clutters. The proposed optimization formula could be solved efficiently using the alternating direction method of multipliers (ADMM) technique and fully carried in the Fourier domain at a low computational burden. Experiments have demonstrated that the proposed algorithm performs favorably against other state-of-the-art methods.

The second batch of Zhuhai-1 microsatellites was successfully launched on 26 April 2018. The batch included four Orbita hyperspectral satellites (referred to as OHS-A, OHS-B, OHS-C, and OHS-D) and one video satellite (OVS-2A), which have excellent hyperspectral data acquisition abilities. For the first time in China, a number of hyperspectral satellite networks have been realized. To ensure the application of hyperspectral remote sensing data, a series of on-orbit geometry processing and accuracy verification studies has been carried out on the “Zhuhai-1” hyperspectral camera since the satellite was launched. This paper presents the geometric processing methods involved in the production of Zhuhai-1 hyperspectral satellite basic products, including geometric calibration and basic product production algorithms. The OHS images were used to perform on-orbit geometric calibration, and the calibration accuracy was better than 0.5 pixels. The registration accuracy of the image spectrum of the basic product after calibration, the single orientation accuracy, and the accuracy of the regional network adjustment were evaluated. The spectral registration accuracy of the OHS basic products is 0.3–0.5 pixels, which is equivalent to the spectral band calibration accuracy. The single orientation accuracy is better than 1.5 pixels and the regional network adjustment accuracy is better than 1.2 pixels. The generated area orthoimages meet the seamless edge requirements, which verifies that the OHS basic product image has good regional mapping capabilities and can meet the application requirements.

The GaoFen-3 (GF-3) satellite is the first C-band multi-polarization synthetic aperture radar (SAR) imaging satellite with a resolution up to 1 m in China. It is also the only SAR satellite of the High-Resolution Earth Observation System designed for civilian use. There are 12 different imaging models to meet the needs of different industry users. However, to use SAR satellite images for related applications, they must possess high geometric accuracy. In order to verify the geometric accuracy achieved by the different modes of GF-3 images, we analyze the SAR geometric error source and perform geometric correction tests based on the RPC model with and without ground control points (GCPs) for five imaging modes. These include the spotlight (SL), ultra-fine strip (UFS), Fine Strip I (FSI), Full polarized Strip I (QPSI), and standard strip (SS) modes. Experimental results show that the check point residuals are large and consistent without GCPs, but the root mean square error of the independent checkpoints for the case of four corner control points is better than 1.5 pixels, achieving a similar level of geometric positioning accuracy to that of international satellites. We conclude that the GF-3 satellite can be used for high-accuracy geometric processing and related industry applications.

Deep learning-based aircraft detection methods have been increasingly implemented in recent years. However, due to the multi-resolution imaging modes, aircrafts in different images show very wide diversity on size, view and other visual features, which brings great challenges to detection. Although standard deep convolution neural networks (DCNN) can extract rich semantic features, they destroy the bottom-level location information. The features of small targets may also be submerged by redundant top-level features, resulting in poor detection. To address these problems, we proposed a compact multi-scale dense convolutional neural network (MS-DenseNet) for aircraft detection in remote sensing images. Herein, DenseNet was utilized for feature extraction, which enhances the propagation and reuse of the bottom-level high-resolution features. Subsequently, we combined feature pyramid network (FPN) with DenseNet to form a MS-DenseNet for learning multi-scale features, especially features of small objects. Finally, by compressing some of the unnecessary convolution layers of each dense block, we designed three new compact architectures: MS-DenseNet-41, MS-DenseNet-65, and MS-DenseNet-77. Comparative experiments showed that the compact MS-DenseNet-65 obtained a noticeable improvement in detecting small aircrafts and achieved state-of-the-art performance with a recall of 94% and an F1-score of 92.7% and cost less computational time. Furthermore, the experimental results on robustness of UCAS-AOD and RSOD datasets also indicate the good transferability of our method.