Temporal Convolutional Neural Network for the Classification of Satellite Image Time Series

By Charlotte Pelletier, Geoffrey I. Webb, Franccois Petitjean. Remote Sensing, 2018

585 citations recorded by the source corpus.

Latest remote sensing sensors are capable of acquiring high spatial and spectral Satellite Image Time Series (SITS) of the world. These image series are a key component of classification systems that aim at obtaining up-to-date and accurate land cover maps of the Earth’s surfaces. More specifically, current SITS combine high temporal, spectral and spatial resolutions, which makes it possible to closely monitor vegetation dynamics. Although traditional classification algorithms, such as Random Forest (RF), have been successfully applied to create land cover maps from SITS, these algorithms do not make the most of the temporal domain. This paper proposes a comprehensive study of Temporal Convolutional Neural Networks (TempCNNs), a deep learning approach which applies convolutions in the temporal dimension in order to automatically learn temporal (and spectral) features. The goal of this paper is to quantitatively and qualitatively evaluate the contribution of TempCNNs for SITS classification, as compared to RF and Recurrent Neural Networks (RNNs) —a standard deep learning approach that is particularly suited to temporal data. We carry out experiments on Formosat-2 scene with 46 images and one million labelled time series. The experimental results show that TempCNNs are more accurate than the current state of the art for SITS classification. We provide some general guidelines on the network architecture, common regularization mechanisms, and hyper-parameter values such as batch size; we also draw out some differences with standard results in computer vision (e.g., about pooling layers). Finally, we assess the visual quality of the land cover maps produced by TempCNNs.

Subfields: Ecology, Global and Planetary Change, Media Technology

View Paper: https://www.semanticscholar.org/paper/27a7170bf265352c1e9992eea641200bba58ce36

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecology

1. Yuan Yuan
2. Lei Lin
3. Marshall Burke
4. Marco Körner
5. David B. Lobell
6. Marc Rußwurm

In Global and Planetary Change

1. Yuan Yuan
2. Lei Lin
3. David B. Lobell
4. Marshall Burke
5. Marco Körner
6. Stefano Ermon
7. Marc Rußwurm

In Media Technology

1. Yuan Yuan
2. Lei Lin
3. Marco Körner
4. Stefano Ermon
5. Marc Rußwurm
6. Chenlin Meng

Citations in the corpus

• Transformers in Time-Series Analysis: A Tutorial (2022, methodology): “The performance of the Transformer architecture for vegetation classification, compared to LSTM-RNN [10], MS-ResNet [94], DuPLO [95], TempCNN [96], and random forest methods, is better for raw data.”
• Self-Attention for Raw Optical Satellite Time Series Classification (2019, methodology): “Recently, the light-weight time-convolutional feed-forward neural networks have shown to be a powerful alternative [53] and have achieved promising results in crop type identification [54].”
• SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagery (2022, methodology): “ML for SITS Deep learning has been used for many Satellite Image Time Series (SITS) supervisedlearning tasks such as crop-type mapping [29, 28, 37, 38], yield prediction [39, 40], understanding the economy [41, 42, 43, 44], precipitation forcasting [45], and land-cover classification [46, 47, 48, 27”
• Self-Supervised Pretraining of Transformers for Satellite Image Time Series Classification (2020, methodology): “3) CNN-1D: A three-layer CNN-1D network was used in the experiment [21], which was formed by three convolutional layers (128 units), one dense layer (256 units), and one softmax layer.”
• Review on remote sensing methods for landslide detection using machine and deep learning (2020, methodology): “According to Reference 82, it is standard and appropriate method for analyzing temporal data.”