The automatic ship detection method for thermal infrared remote sensing images (TIRSIs) is of great significance due to its broad applicability in maritime security, port management, and target searching, especially at night. Most ship detection algorithms utilize manual features to detect visible image blocks which are accurately cut, and they are limited by illumination, clouds, and atmospheric strong waves in practical applications. In this paper, a complete YOLO-based ship detection method (CYSDM) for TIRSIs under complex backgrounds is proposed. In addition, thermal infrared ship datasets were made using the SDGSAT-1 thermal imaging system. First, in order to avoid the loss of texture characteristics during large-scale deep convolution, the TIRSIs with the resolution of 30 m were up-sampled to 10 m via bicubic interpolation method. Then, complete ships with similar characteristics were selected and marked in the middle of the river, the bay, and the sea. To enrich the datasets, the gray value stretching module was also added. Finally, the improved YOLOv5 s model was used to detect the ship candidate area quickly. To reduce intra-class variation, the 4.23–7.53 aspect ratios of ships were manually selected during labeling, and 8–10.5 μm ship datasets were constructed. Test results show that the precision of the CYSDM is 98.68%, which is 9.07% higher than that of the YOLOv5s algorithm. CYSDM provides an effective reference for large-scale, all-day ship detection.

/pdf/a-complete-yolo-based-ship-detection-method-for-thermal-26ha9byr.pdf

A Complete YOLO-Based Ship Detection Method for Thermal Infrared Remote Sensing Images under Complex Backgrounds

The thermal image segmentation has been widely concerned to alleviate the limitations of image segmentation caused by imaging of visible spectrum under challenge environmental conditions. The existing thermal image segmentation methods are not easy to obtain high quality segmented results, because thermal images are lack of color information, edges are unclear, details are not obvious. To this end, we proposed a multi-level pixel spatial attention network for thermal image segmentation. Specifically, we designed a pixel space attention module on each layer of the backbone network(MPAM), which recovers the more spatial detailed and maintains more semantic information. Then, we designed an edge extraction module (EEM) and a small target extraction module (STEM), which enhances the edge and small target features of the network by explicitly modeling the edge and small target features. Finally, the edge and small target features were fused with the output features of backbone, and the specialized loss functions was used to supervise them. Experimental results on SCUT-SEG, SODA and STI-Cityscpaes dataset demonstrate that our approach is slightly improved by 2.2% compared with other the state-of-art algorithms in the same scene. 

MPSA: A Multi-level Pixel Spatial Attention Network for Thermal Image Segmentation Based on Deeplabv3+ Architecture

Cloud detection for satellite cloud images is a challenging image processing task owing to the blurring of cloud boundaries, multiplicity, and complexity of cloud types. Currently, the commonly used cloud detection methods include original full convolutional neural network (original FCN), FCN with an 8-pixel stride (FCN- 8s), FCN with a 2-pixel stride (FCN-2s), and so on. However, the aforementioned methods exclusively rely on a single network layer − the final layer feature map; thus, shallow cloud image information, such as cloud profile information may not be captured. In this letter, a cloud detection method for satellite cloud images based on fused FCN features is proposed. The proposed method effectively fuses spatial and high-level semantic information, and a voting ensemble strategy is used to improve the accuracy and robustness of cloud detection. Finally, the experimental results demonstrate that the average overall accuracy (OA), average producer’ accuracy (PA), and average user’ accuracy (UA) of the proposed method for multiple training sample sizes and image sizes of the collected Fengyun satellite (FY-2 G) cloud image database increased by 7.15%, 9.04%, and 8.46%, respectively, relative to the average accuracies of the original FCN, FCN-8s, FCN-2s, SegNet, and DeepLabV3 methods. 

Cloud detection for satellite cloud images based on fused FCN features

ABSTRACT Cloud detection for satellite cloud images is a challenging image processing task owing to the blurring of cloud boundaries, multiplicity, and complexity of cloud types. Currently, the commonly used cloud detection methods include original full convolutional neural network (original FCN), FCN with an 8-pixel stride (FCN- 8s), FCN with a 2-pixel stride (FCN-2s), and so on. However, the aforementioned methods exclusively rely on a single network layer the final layer feature map; thus, shallow cloud image information, such as cloud profile information may not be captured. In this letter, a cloud detection method for satellite cloud images based on fused FCN features is proposed. The proposed method effectively fuses spatial and high-level semantic information, and a voting ensemble strategy is used to improve the accuracy and robustness of cloud detection. Finally, the experimental results demonstrate that the average overall accuracy (OA), average producer’ accuracy (PA), and average user’ accuracy (UA) of the proposed method for multiple training sample sizes and image sizes of the collected Fengyun satellite (FY-2 G) cloud image database increased by 7.15%, 9.04%, and 8.46%, respectively, relative to the average accuracies of the original FCN, FCN-8s, FCN-2s, SegNet, and DeepLabV3 methods.

Compared with low Earth orbit (LEO) satellites, the high altitude from geostationary Earth orbit (GEO) satellites leads to increased diffraction effects on hyperspectral infrared (HIR) sounders, which reduces the ensquared energy (EE) within the satellites’ field of view (FOV) and increases the pseudo noise of the measurements. To help understand how the instrument performance is affected by EE for the Geostationary Extended Observations Sounder (GXS), a point spread function (PSF) is used to simulate the contribution of each location within and outside of an FOV ( $4\times $ 4 km at nadir). The PSF is applied to the Moderate Resolution Imaging Spectroradiometer (MODIS) airborne simulator (MAS) data with a spatial resolution of 50 m, to determine an appropriate EE for GXS. Although wavenumber dependent, an EE of 70% is recommended, which ensures that all GXS channels have pseudo noise less than the instrument specifications. Regardless of the EE value, the pseudo noise reduces the precision of the temperature and moisture sounding retrievals in the troposphere. Even with an EE of 70%, the pseudo noise slightly increases the root-mean-square error (RMSE) by 3%–4% for temperature and by 1%–4% for relative humidity. If an EE of 70% is difficult to meet, due to cost for example, a lower EE can be a good tradeoff with only a slight degradation in the sounding retrieval quality, which may be overcome with spatial averaging using the inverted cone method. An EE of 50% would lead to an RMSE increase of about 6% for temperature and 3%–6% for relative humidity.

Studies Regarding the Ensquared Energy of a Geostationary Hyperspectral Infrared Sounder

Semantic segmentation (SS) has been widely applied for cloud detection (CD) in remote sensing images (RSIs) with high spatial and spectral resolution because of its effective pixel-level feature extraction structure. However, the typical model of lightweight SS, namely the fully convolutional network (FCN) with only seven layers, has difficulty in extracting high-level features, and the heavy pyramid scene parsing network (PSPNet) with complicated calculations is not practical in real-time CD, let alone on-orbit CD. So, in view of the problems above, we propose a compact attention mechanism cloud detection network (AM-CDN) based on the modified FCN to refine and fuse the multi-scale features for on-orbit CD. Specifically, taking the FCN as the baseline, our model increases the numbers of hidden layers and adds the residual connections between the input and output to eliminate the network degradation and extract the advanced context feature maps effectively. To expand the receptive field without losing the spatial information, the ordinary convolutions in FCN are replaced by the dilated convolution in AM-CDN. And inspired by the selective kernels of human vision, we introduce the convolutional attention mechanism (AM) into the encoder to adaptively adjust the receptive field to highlight the key texture features. According to experimental results using Landsat-8 infrared RSIs, the accuracy of the proposed CD method is 95.31%, which is 10.17% higher than that of FCN. And the calculation complexity of AM-CDN is only 7.63% of that of PSPNet.

Attention Mechanism Cloud Detection With Modified FCN for Infrared Remote Sensing Images

Atmospheric refraction is one of the most significant factors that affect the geolocation accuracy of high-resolution remote sensing images. However, most of the current atmospheric refraction correction methods based on empirical data neglect the spatiotemporal variation of pressure, temperature, and humidity of the atmosphere, inevitably resulting in poor geometric positioning accuracy. Therefore, in terms of the problems mentioned above, this study proposed a spatiotemporal atmospheric refraction correction method (SARCM) based on global measured data to avoid the uncertainty of traditional empirical models. Initially, the atmosphere was stratified into 42 layers according to their pressure property, and each layer was divided into 1,042,560 grid cells with intervals of 0.25 longitude and 0.25 latitude. Then, the atmospheric refractive index of each grid in the imaging region was accurately calculated using the high-precision Ciddor formula, and the result was interpolated using three splines. Subsequently, according to the rigorous geometric positioning model, the line-of-sight of each pixel and the viewing zenith angle outside the atmosphere in WGS84 were derived to provide input for atmospheric refraction correction. Finally, the coordinates of the ground control points were corrected with the calculated atmospheric refractive index and Snell’s law. The experimental results showed that the proposed SARCM could effectively improve the positioning accuracy of the image with a large viewing zenith angle, and especially, the improvement percentage for a viewing zenith angle of 34.2426° in the x-direction was 99.5%. Moreover, the atmospheric refraction correction result of the SARCM was better than that of the primary state-of-the-art methods.

/pdf/a-spatiotemporal-atmospheric-refraction-correction-method-1m7ntl8p.pdf

A Spatiotemporal Atmospheric Refraction Correction Method for Improving the Geolocation Accuracy of High-Resolution Remote Sensing Images

High-accuracy sea surface temperature (SST) retrieval near nuclear power plants (NPPs) is one of the most significant indicators for evaluating marine ecological environment quality, monitoring the real-time situation of thermal discharge, and supporting planning decisions. However, complex computations, the inaccessible real-time vertical profile of the atmosphere, and the uncertainty of atmospheric profile data increase the error of SST retrieval. Additionally, influenced by their low spatial resolution, the widely used AVHRR/MODIS remote sensing images (RSIs) are unable to retrieve the detailed distribution of SST in small scale regions such as coastal NPPs. In this paper, we propose a simplified split-window-based temperature retrieval method (the SW method) suitable for SDGSAT-1 30 m thermal infrared spectrometer (TIS) RSIs. Specially, this method only needs atmospheric transmittance and surface emissivity by counteracting the average atmospheric temperature to monitor the thermal discharge of offshore NPPs. First, the geometric and radiometric calibrated thermal infrared and multi-spectral cloudless data of the target regions are selected to obtain the corresponding apparent radiance of the RSIs. Second, in accordance with the red and near-infrared (NIR) bands of multi-spectral RSIs, the surface emissivity is calculated to distinguish water from land. Next, we determine the atmospheric profile parameters from the weather conditions of the target region at the imaging time. Finally, according to the theory of surface-atmosphere radiative transfer, the SST of target regions is retrieved with the proposed SW method, and the results are compared with those of the conventional radiative transfer equation (RTE), mono-window (MW), and the nonlinear sea surface temperature (NLSST) algorithms. The experimental results indicate that the SST retrieved from the split-window algorithms (i.e., SW and NLSST) are generally higher than those of the single-channel algorithms (i.e., RTE and MW). The SST difference between the SW algorithm and the NLSST algorithm is within 0.5 °C. In addition, SDGSAT-1 can monitor the seasonal detailed variation of the thermal discharge near coastal NPPs. This article is the first to attempt to quantitative small-scale SST retrieval based on thermal infrared and multi-spectral images obtained from the SDGSAT-1 TIS and a multispectral imager (MII), and therefore, provide an effective reference for marine environment monitoring.

/pdf/thermal-discharge-temperature-retrieval-and-monitoring-of-2quglqnm.pdf

Thermal Discharge Temperature Retrieval and Monitoring of NPPs Based on SDGSAT-1 Images

With the development of the infrared detection and imaging technologies, the infrared detection systems have played an important role in applications including military and civilian, which makes the requirements for the performance of infrared detectors increasingly high.However, due to the temperature drift, thermal shock and other factors, blind pixels and noise may change, which will greatly affect the imaging quality of the system, so it is necessary to accurately test and evaluate the relevant performance parameters of the infrared detection system such as the number and distribution of blind pixels and flickering pixels. But there are still lack of effective and accurate long-term test methods for these parameters, and the existing test methods have problems in automation, accuracy and precision. In view of the problems above, this research builds an automatic performance test and evaluation system for the infrared detectors, and additionally, except for response rate, noise, dynamic range, noise equivalent temperature difference and blind pixel, a special characterization of flickering pixel for the infrared focal plane arrays (IRFPA) is given in detail. Accordingly, experiment results based on two 512 × 512 IRFPAs demonstrate that the established system could automatically measure and assess the performance of the detector accurately, and meanwhile, the number variation and the spatial distribution of blind pixels and flickering pixels could be performed elaborately. This study can be used as a reference for the performance test and analysis of infrared detection system.

Wenwen Huang

Papers

Attention Mechanism Cloud Detection With Modified FCN for Infrared Remote Sensing Images

A Spatiotemporal Atmospheric Refraction Correction Method for Improving the Geolocation Accuracy of High-Resolution Remote Sensing Images

Thermal Discharge Temperature Retrieval and Monitoring of NPPs Based on SDGSAT-1 Images

An automatic performance test and evaluation system for IRFPA