Although efforts and progress have been made in crop classification using optical remote sensing images, it is still necessary to make full use of the high spatial, temporal, and spectral resolutions of remote sensing images. However, with the increasing volume of remote sensing data, a key emerging issue in the field of crop classification is how to find useful information from massive data to balance classification accuracy and processing time. To address this challenge, we developed a novel crop classification method, combining optimal feature selection (OFSM) with hybrid convolutional neural network-random forest (CNN-RF) networks for multi-temporal optical remote sensing images. This research used 234 features including spectral, segmentation, color, and texture features from three scenes of Sentinel-2 images to identify crop types in the Jilin province of northeast China. To effectively extract the effective features of remote sensing data with lower time requirements, the use of OFSM was proposed with the results compared with two traditional feature selection methods (TFSM): random forest feature importance selection (RF-FI) and random forest recursive feature elimination (RF-RFE). Although the time required for OFSM was 26.05 s, which was between RF-FI with 1.97 s and RF-RFE with 132.54 s, OFSM outperformed RF-FI and RF-RFE in terms of the overall accuracy (OA) of crop classification by 4% and 0.3%, respectively. On the basis of obtaining effective feature information, to further improve the accuracy of crop classification we designed two hybrid CNN-RF networks to leverage the advantages of one-dimensional convolution (Conv1D) and Visual Geometry Group (VGG) with random forest (RF), respectively. Based on the selected optimal features using OFSM, four networks were tested for comparison: Conv1D-RF, VGG-RF, Conv1D, and VGG. Conv1D-RF achieved the highest OA at 94.27% as compared with VGG-RF (93.23%), Conv1D (92.59%), and VGG (91.89%), indicating that the Conv1D-RF method with optimal feature input provides an effective and efficient method of time series representation for multi-temporal crop-type classification.

Farmland segmentation scenario from remote sensing images plays an important role in crop growth monitoring, precision agriculture, and intelligent agriculture. To achieve high precision segmentation of farmland boundary, a Multitask Deformable UNet combined Enhanced (MDE-UNet) network is proposed for farmland boundary segmentation. The network consists of two parts: a Multitask Deformable UNet (MD-UNet) segmentation module with Deformable UNet (D-UNet) as the basic network and an enhancement module with a lightweight UNet improved by residual attention. In the MD-UNet segmentation module, three branches are used for precise segmentation of deterministic, fuzzy, and raw boundary, respectively. In the enhancement module, an improved lightweight UNet is designed, which can enhance the feature extraction ability of the MD-UNet segmentation module and further improve the segmentation accuracy. The accuracy and mIoU in the GF-2 farmland segmentation test dataset can reach 96.41% and 91.29% using the proposed model, respectively. The MDE-UNet method outperforms other representative deep learning methods such as DeepLab v3+, FCN-8 s, SegFormer, and UTNet, and has potential for practical applications of farmland boundary segmentation.

Accurate building extraction from very high-resolution (VHR) remote sensing images plays an important role in urban dynamic monitoring, planning, and management. However, it is still a challenging task to achieve building extraction with high accuracy and integrity due to diverse building appearances and more complex ground background in VHR remote sensing images. Recently, Unity Networking (UNet) has been proven to be capable of feature extraction and semantic segmentation of remote sensing images. However, UNet cannot achieve sufficient multi-scale and multi-level features with larger receptive fields. To address these problems, an improved network based on UNet structure (Refine-UNet) is proposed for extracting buildings from the VHR images. The proposed Refine-UNet mainly consists of encoder module, decoder module, and refine skip connection scheme. The refine skip connection scheme is composed of an atrous spatial convolutional pyramid pooling (ASPP) module and several improved depthwise separable convolution (IDSC) modules. Experimental results on the Jilin-1 VHR datasets with a spatial resolution of 0.75 m demonstrate that compared with UNet, PSPNet, DeepLabV3+, and SegNet, the proposed Refine-UNet can obtain more accurate building extraction results and achieve the best Precision of 95.1% and intersection over union (IoU) of 87.0%, respectively, indicating the great practical potential.

Chlorophyll-a (Chl-a) is an important characterized parameter of lakes. Monitoring it accurately through remote sensing is thus of great significance for early warnings of water eutrophication. Sentinel Multispectral Imager (MSI) images from May to September between 2020 and 2021 were used along with in-situ measurements to estimate Chl-a in Lake Chagan, which is located in Jilin Province, Northeast China. In this study, the extreme gradient boosting (XGBoost) and Random Forest (RF) models, which had similar performances, were generated by six single bands and six band combinations. The RF model was then selected based on the assessments (R² = 0.79, RMSE = 2.51 μg L⁻¹, MAPE = 9.86%), since its learning of the input features in the model conformed to the bio-optical properties of Case 2 waters. The study considered Chl-a concentrations in Lake Chagan as a seasonal pattern according to the K-Nearest-Neighbors (KNN) classification. The RF model also showed relatively stable performance for three seasons (spring, summer and autumn) and it was applied to map Chl-a in the whole lake. The research presents a more reliable machine learning (ML) model with higher precision than previous empirical models, as shown by the effects of the input features linked with the biological mechanisms of Chl-a. Its robustness was revealed by the temporal and spatial distributions of Chl-a concentrations, which were consistent with in-situ measurements in the map. This research was capable of revealing the current ecological situation in Lake Chagan and can serve as a reference in remote sensing of inland lakes.

Snow cover plays an important role in climate, hydrology, and ecosystem. At present, passive microwave remote sensing is the most effective method for monitoring global and regional snow depth (SD). The traditional SD inversion algorithms use empirical or semiempirical methods to establish a fixed relationship between the SD and brightness temperature difference, given snow particle size and snow density. However, the snow characteristics present large temporal heterogeneity in Northeast China, and it leads to the inadaptability of the SD retrieval algorithm; using a fixed empirical coefficient will lead to large errors in SD inversion. In this study, a novel dynamic method was proposed to retrieve SD based on AMSR2 brightness temperature data. A snow survey experiment was designed to collect snow characteristics in different periods in Northeast China, and the microwave emission model of layered snowpacks was applied to simulate brightness temperature with varying snow characteristics to determine the dynamic coefficients in the SD retrieval algorithm. The validation results at 98 meteorological stations demonstrate that the novel dynamic SD inversion algorithm achieved better stability in the long-term sequence, its RMSE, bias, and R are 7.79 cm, 1.07 cm, and 0.61, respectively. Furthermore, compared with Che SD products, Chang algorithm, and AMSR2 SD products, the novel algorithm can obtain specific dynamic coefficients considering the snow metamorphism and has a higher accuracy of SD inversion in the whole winter. In conclusion, this novel SD inversion algorithm is more applicable and accurate than the existing SD inversion products in Northeast China.