To improve the success rate of target detection, a fusion method that comprises a polarimetric feature extractor and a deep convolutional neural network (CNN) with consecutive small (2 × 2) convolutions (or, CSC-CNN for simplicity) is proposed. First, we theoretically analyse the dispersion characteristics on a target surface based on the concept of information entropy, and it is concluded that the dispersion measure can be selected as a relevant feature for polarimetric images. Then, a polarimetric feature extractor is introduced for conveniently calculating dispersion measures with fewer prior parameters in outdoor measurements. Finally, a CSC-CNN is adopted for subsequent target detection with small scale training samples. The experimental results indicate that the proposed fusion method demonstrates great potential for reducing the detection error rate, which is less than that of the traditional method used for comparison.

An improved aerosol retrieval algorithm based on the Advanced Multi-angular Polarized Radiometer (AMPR) is presented to illustrate the utility of additional 1640-nm observations for measuring aerosol optical depth (AOD) over land using look-up table approaches. Spectral neutrality of the polarized surface reflectance over visible to short-wavelength infrared bands is verified, and the 1640-nm measurements corrected for atmospheric effects are used to estimate the polarized surface reflectance at shorter wavelengths. The AMPR measurements over the Beijing-Tianjin-Hebei region in north China reveal that the polarized surface reflectance of 670, 865 and 1640 nm are highly correlated with correlation slopes close to one (0.985 and 1.03) when the scattering angle is less than 145°. The 1640-nm measurements are then employed to estimate polarized surface reflectance at shorter wavelengths for each single viewing direction, which are then used to improve the retrieval of AOD over land. The comparison between AMPR retrievals and ground-based Sun-sky radiometer measurements during three experimental flights illustrates that this approach retrieves AOD at 865 nm with uncertainties ranging from 0.01 to 0.06, while AOD varies from 0.05 to 0.17.

Polarimetric imaging detection is a relatively new and largely undeveloped field. Although convolutional neural networks (CNNs) have achieved great success in two-dimensional (2D) normal intensity images in the field of target detection, traditional CNN methods have not been widely applied to optical polarimetric images, and they cannot take full advantage of the connection between different polarimetric images. To solve this problem, three-dimensional (3D) convolutions are adopted to consider the relationship between S0, S1, and S2 images as a third dimension. Based on the 3D convolutions, a CNN with 3D and 2D convolutional layers is introduced to further improve the success rate of target detection with limited polarimetric images. The evaluations in different natural backgrounds reveal that the proposed method achieves higher detection accuracy than that of two traditional methods for comparison.

Cloud detection is one of the critical tasks in remote sensing image preprocessing. Remote sensing images usually contain multi-dimensional information, which is not utilized entirely in existing deep learning methods. This paper proposes a novel cloud detection algorithm based on multi-scale input and dual-channel attention mechanisms. Firstly, we remodeled the original data to a multi-scale layout in terms of channels and bands. Then, we introduced the dual-channel attention mechanism into the existing semantic segmentation network, to focus on both band information and angle information based on the reconstructed multi-scale data. Finally, a multi-scale fusion strategy was introduced to combine band information and angle information simultaneously. Overall, in the experiments undertaken in this paper, the proposed method achieved a pixel accuracy of 92.66% and a category pixel accuracy of 92.51%. For cloud detection, the proposed method achieved a recall of 97.76% and an F1 of 95.06%. The intersection over union (IoU) of the proposed method was 89.63%. Both in terms of quantitative results and visual effects, the deep learning model we propose is superior to the existing semantic segmentation methods.

As the latest members of Chinese Environmental Protection and Disaster Monitoring Satellite Constellation, the first two of HuanjingJianzai-2 (HJ-2) series satellites were launched on 27 September 2020 by China and are usually abbreviated as HJ-2 A/B satellites. The polarized scanning atmospheric corrector (PSAC) is one of main sensors onboard HJ-2 A/B satellites, which is mainly used to monitor atmospheric components such as water vapor and aerosols. In this study, a columnar water vapor (CWV) retrieval algorithm using two bands (865 and 910 nm) is developed for PSAC. The validation results of PSAC CWV data based on ground-based CWV data derived from Aerosol Robotic Network (AERONET) show that PSAC CWV data has a high accuracy, and all statistical parameters of PSAC CWV data are better than those of Moderate-resolution Imaging Spectroradiometer (MODIS) CWV data released by NASA. Overall, there is no obvious overestimation or underestimation in PSAC CWV data. The root mean square error (RMSE), mean absolute error (MAE), relative error (RE), and percentage of CWV data with error within $\pm \left(0.05+0.10\ast CW{V}_{AERONET}\right)$ (PER10) of PSAC CWV data are 0.17 cm, 0.13 cm, 0.08, and 78.19%, respectively. The RMSE, MAE, RE, and PER10 of MODIS CWV data are 0.59 cm, 0.48 cm, 0.28, and 16.55%, respectively. Compared with MODIS CWV data, PSAC CWV data shows a 71% decrease in RMSE, a 73% decrease in MAE, a 71% decrease in RE, and a 372% increase in PER10. In addition, the results of day-to-day comparisons between PSAC CWV data and AERONET data show that PSAC CWV data can effectively characterize the change trend of CWV.

The DPC (Directional Polarization Camera) onboard the Chinese GaoFen-5 (GF-5) satellite is the first operational aerosol monitoring instrument capable of performing multi-angle polarized measurements in China. Compared with POLDER (Polarization and Directionality of Earth’s Reflectance) which ended its mission in December 2013, DPC has similar band design, with a maximum of 12 imaging angles and a relatively higher spatial resolution of 3.3 km. The global aerosol optical depth (AOD) over land from October to December in 2018 was retrieved with multi-angle polarization measurements of DPC. Comparisons with MODIS (Moderate Resolution Imaging Spectroradiometer) AOD products show relatively good agreement over fine-aerosol-particle-dominated areas such as northern China and Huanghuai areas in eastern China, the southern foothills of the Himalayas and India. AERONET (Aerosol Robotic Network) measurements over Beijing, Xianghe and Kanpur were used to evaluate the accuracy of DPC AOD retrievals. The correlation coefficients are greater than 0.9 and the RMSE are lower than 0.08 for Beijing and Xianghe stations. For Kanpur, a relatively lower correlation of 0.772 and larger RMSE of 0.082 are found.