Annual crop inventory information is important for many agriculture applications and government statistics. The synergistic use of multi-temporal polarimetric synthetic aperture radar (SAR) and available multispectral remote sensing data can reduce the temporal gaps and provide the spectral and polarimetric information of the crops, which is effective for crop classification in areas with frequent cloud interference. The main objectives of this study are to develop a deep learning model to map agricultural areas using multi-temporal full polarimetric SAR and multi-spectral remote sensing data, and to evaluate the influence of different input features on the performance of deep learning methods in crop classification. In this study, a one-dimensional convolutional neural network (Conv1D) was proposed and tested on multi-temporal RADARSAT-2 and VENµS data for crop classification. Compared with the Multi-Layer Perceptron (MLP), Recurrent Neural Network (RNN) and non-deep learning methods including XGBoost, Random Forest (RF), and Support Vector Machina (SVM), the Conv1D performed the best when the multi-temporal RADARSAT-2 data (Pauli decomposition or coherency matrix) and VENµS multispectral data were fused by the Minimum Noise Fraction (MNF) transformation. The Pauli decomposition and coherency matrix gave similar overall accuracy (OA) for Conv1D when fused with the VENµS data by the MNF transformation (OA = 96.65 ± 1.03% and 96.72 ± 0.77%). The MNF transformation improved the OA and F-score for most classes when Conv1D was used. The results reveal that the coherency matrix has a great potential in crop classification and the MNF transformation of multi-temporal RADARSAT-2 and VENµS data can enhance the performance of Conv1D.

Compared with a monoculture planting mode, the practice of crop rotations improves fertilizer efficiency and increases crop yield. Large-scale crop rotation monitoring relies on the results of crop classification using remote sensing technology. However, the limited crop classification accuracy cannot satisfy the accurate identification of crop rotation patterns. In this paper, a crop classification and rotation mapping scheme combining the random forest (RF) algorithm and new statistical features extracted from time-series ground range direction (GRD) Sentinel-1 images. First, the synthetic aperture radar (SAR) time-series stacks are established, including VH, VV, and VH/VV channels. Then, new statistical features named the objected generalized gamma distribution (OGΓD) features are introduced to compare with other object-based features for each polarization. The results showed that the OGΓD σVH achieved 96.66% of the overall accuracy (OA) and 95.34% of the Kappa, improving around 4% and 6% compared with the object-based backscatter in VH polarization, respectively. Finally, annual crop-type maps for five consecutive years (2017–2021) are generated using the OGΓD σVH and the RF. By analyzing the five-year crop sequences, the soybean-corn (corn-soybean) is the most representative rotation in the study region, and the soybean-corn-soybean-corn-soybean (together with corn-soybean-corn-soybean-corn) has the highest count with 100 occurrences (25.20% of the total area). This study offers new insights into crop rotation monitoring, giving the basic data for government food planning decision-making.

No abstract

Inspired by the tremendous success of deep learning (DL) and the increased availability of remote sensing data, DL-based image semantic segmentation has attracted growing interest in the remote sensing community. The ideal scenario of DL application requires a vast number of annotation data with the same feature distribution as the area of interest. However, obtaining such enormous training sets that suit the data distribution of the target area is highly time-consuming and costly. Consistency-regularization-based semi-supervised learning (SSL) methods have gained growing popularity thanks to their ease of implementation and remarkable performance. However, there have been limited applications of SSL in remote sensing. This study comprehensively analyzed several advanced SSL methods based on consistency regularization from the perspective of data- and model-level perturbation. Then, an end-to-end SSL approach based on a hybrid perturbation paradigm was introduced to improve the DL model’s performance with a limited number of labels. The proposed method integrates the semantic boundary information to generate more meaningful mixing images when performing data-level perturbation. Additionally, by using implicit pseudo-supervision based on model-level perturbation, it eliminates the need to set extra threshold parameters in training. Furthermore, it can be flexibly paired with the DL model in an end-to-end manner, as opposed to the separated training stages used in the traditional pseudo-labeling. Experimental results for five remote sensing benchmark datasets in the application of segmentation of roads, buildings, and land cover demonstrated the effectiveness and robustness of the proposed approach. It is particularly encouraging that the ratio of accuracy obtained using the proposed method with 5% labels to that using the purely supervised method with 100% labels was more than 89% on all benchmark datasets.

In recent years, global forest fires have occurred more frequently, seriously destroying the structural functions of forest ecosystem. Mapping the burn severity after forest fires is of great significance for quantifying fire’s effects on landscapes and establishing restoration measures. Generally, intensive field surveys across burned areas are required for the effective application of traditional methods. Unfortunately, this requirement could not be satisfied in most cases, since the field work demands a lot of personnel and funding. For mapping severity levels across burned areas without field survey data, a semi-supervised transfer component analysis-based support vector regression model (SSTCA-SVR) was proposed in this study to transfer knowledge trained from other burned areas with field survey data. Its performance was further evaluated in various eco-type regions of southwestern United States. Results show that SSTCA-SVR which was trained on source domain areas could effectively be transferred to a target domain area. Meanwhile, the SSTCA-SVR could maintain as much spectral information as possible to map burn severity. Its mapped results are more accurate (RMSE values were between 0.4833 and 0.6659) and finer, compared to those mapped by ∆NDVI-, ∆LST-, ∆NBR- (RMSE values ranged from 0.7362 to 1.1187) and SVR-based models (RMSE values varied from 1.7658 to 2.0055). This study has introduced a potentially efficient mechanism to map burn severity, which will speed up the response of post-fire management.