Evaluation of Three Deep Learning Models for Early Crop Classification Using Sentinel-1A Imagery Time Series—A Case Study in Zhanjiang, China

Abstract

Timely and accurate estimation of the area and distribution of crops is vital for food security. Optical remote sensing has been a key technique for acquiring crop area and conditions on regional to global scales, but great challenges arise due to frequent cloudy days in southern China. This makes optical remote sensing images usually unavailable. Synthetic aperture radar (SAR) could bridge this gap since it is less affected by clouds. The recent availability of Sentinel-1A (S1A) SAR imagery with a 12-day revisit period at a high spatial resolution of about 10 m makes it possible to fully utilize phenological information to improve early crop classification. In deep learning methods, one-dimensional convolutional neural networks (1D CNNs), long short-term memory recurrent neural networks (LSTM RNNs), and gated recurrent unit RNNs (GRU RNNs) have been shown to efficiently extract temporal features for classification tasks. However, due to the complexity of training, these three deep learning methods have been less used in early crop classification. In this work, we attempted to combine them with an incremental classification method to avoid the need for training optimal architectures and hyper-parameters for data from each time series. First, we trained 1D CNNs, LSTM RNNs, and GRU RNNs based on the full images’ time series to attain three classifiers with optimal architectures and hyper-parameters. Then, starting at the first time point, we performed an incremental classification process to train each classifier using all of the previous data, and obtained a classification network with all parameter values (including the hyper-parameters) at each time point. Finally, test accuracies of each time point were assessed for each crop type to determine the optimal time series length. A case study was conducted in Suixi and Leizhou counties of Zhanjiang City, China. To verify the effectiveness of this method, we also implemented the classic random forest (RF) approach. The results were as follows: (i) 1D CNNs achieved the highest Kappa coefficient (0.942) of the four classifiers, and the highest value (0.934) in the GRU RNNs time series was attained earlier than with other classifiers; (ii) all three deep learning methods and the RF achieved F measures above 0.900 before the end of growth seasons of banana, eucalyptus, second-season paddy rice, and sugarcane; while, the 1D CNN classifier was the only one that could obtain an F-measure above 0.900 for pineapple before harvest. All results indicated the effectiveness of the solution combining the deep learning models with the incremental classification approach for early crop classification. This method is expected to provide new perspectives for early mapping of croplands in cloudy areas.

On the Joint Exploitation of Optical and Sar Satellite Imagery for Grassland Monitoring

“…A bi-directional Gated Recurrent Unit (GRU) network was chosen. It exhibits similar results to other types of RNNs for fewer parameters (Ndikumana et al, 2018, Zhao et al, 2019. Lastly, the third block decodes the outputs of the second block with an in-depth funnel-shaped MLP.…”

Mixture Optimization of Recycled Aggregate Concrete Using Hybrid Machine Learning Model

“…Nonetheless, GRU is considered a more efficient algorithm due to its simpler structure and formulation [ 27 ]. The formulation of GRU is summarized in the following: where and are the reset and update gate, respectively, is the candidate output, and is the corresponding output of the cell for the time step t. Accordingly, , , , , , and are the weight matrices that operate the input vector and the previous state , and ReLU is the rectified linear unit activation function [ 39 , 40 ].…”

Potential of Hybrid CNN-RF Model for Early Crop Mapping with Limited Input Data

“…When an incomplete time-series image set that cannot account for the full growth cycles of crops is used for the classification, distinguishing various crops with similar spectral responses is often difficult, thus achieving relatively poor classification performance. From a practical perspective, however, there is still a demand for generating crop type maps using only the images collected during the early crop growth period as decision-makers require such maps before the end of the crop seasons [11,12]. For crop type maps generated from an incomplete time-series set to be practically useful, achieving a classification performance comparable with that generated from a complete time-series set is necessary.…”

“…To overcome the limitation of insufficient temporal contextual information, convolutional neural network (CNN) that considers complex spatial patterns between neighboring pixels within a patch is efficient for classifying regions with similar spectral and spatial characteristics such as agricultural fields [24,25]. For example, the superiority of CNN in crop classification has been verified through comparisons with conventional ML and other DL models [9][10][11]31,32]. However, training the CNN model with limited information, in terms of the quantity and quality of training data as well as the length of the input time-series, is still challenging [33].…”