3D Convolutional Neural Networks for Crop Classification with Multi-Temporal Remote Sensing Images

Ji, Shunping; Zhang, Chi; Xu, Aiqiang; Shi, Yun; Duan, Yulin

doi:10.3390/rs10010075

Cited by 282 publications

(152 citation statements)

References 34 publications

Supporting

Mentioning

150

Contrasting

Order By: Relevance

“…No particular adaptation of the architecture for the temporal aspect of the data was proposed. Ji et al [24] adapted the patch methods to better take into account the temporal dynamics by using 3D 3 × 3 × 3 convolutions (2 spatial dimensions and 1 time dimension). They show results using 4 m and 15 m resolution RGB+NIR tiles using only 4 dates with significant improvements with respect to other approaches.…”

Section: Semantic Segmentation Vs Patch Based Methodsmentioning

confidence: 99%

Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems

et al. 2019

View full text Add to dashboard Cite

The Sentinel-2 satellite mission offers high resolution multispectral time-series image data, enabling the production of detailed land cover maps globally. When mapping large territories, the trade-off between processing time and result quality is a central design decision. Currently, this machine learning task is usually performed using pixel-wise classification methods. However, the radical shift of the computer vision field away from hand-engineered image features and towards more automation by representation learning comes with many promises, including higher quality results and less engineering effort. In particular, convolutional neural networks learn features which take into account the context of the pixels and, therefore, a better representation of the data can be obtained. In this paper, we assess fully convolutional neural network architectures as replacements for a Random Forest classifier in an operational context for the production of high resolution land cover maps with Sentinel-2 time-series at the country scale. Our contributions include a framework for working with Sentinel-2 L2A time-series image data, an adaptation of the U-Net model (a fully convolutional neural network) for dealing with sparse annotation data while maintaining high resolution output, and an analysis of those results in the context of operational production of land cover maps. We conclude that fully convolutional neural networks can yield improved results with respect to pixel-wise Random Forest classifiers for classes where texture and context are pertinent. However, this new approach shows higher variability in quality across different landscapes and comes with a computational cost which could be to high for operational systems.

show abstract

Section: Semantic Segmentation Vs Patch Based Methodsmentioning

confidence: 99%

Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Recently, these methods have been applied to traditional remote sensing tasks, largely by adapting methods initially developed in the computer vision community. Notable examples include applications to land cover segmentation [21,26], land degradation estimation [20], and crop classi cation [1,15].…”

Section: Related Workmentioning

confidence: 99%

A weakly supervised approach for estimating spatial density functions from high-resolution satellite imagery

Jacobs

Kraft

Rafique

et al. 2018

Proceedings of the 26th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems

View full text Add to dashboard Cite

We propose a neural network component, the regional aggregation layer, that makes it possible to train a pixel-level density estimator using only coarse-grained density aggregates, which re ect the number of objects in an image region. Our approach is simple to use and does not require domain-speci c assumptions about the nature of the density function. We evaluate our approach on several synthetic datasets. In addition, we use this approach to learn to estimate high-resolution population and housing density from satellite imagery. In all cases, we nd that our approach results in be er density estimates than a commonly used baseline. We also show how our housing density estimator can be used to classify buildings as residential or non-residential.

show abstract

“…Convolutional neural networks have become increasingly popular in image analysis due to their ability to automatically learn relevant contextual features. Initially devised for natural images, these networks have been revisited and adapted to tackle remote-sensing problems, such as road extraction (Cheng et al, 2017), cloud detection (Chai et al, 2019), crop identification (Ji et al, 2018), river and water body extraction (Chen et al, 2018b;Isikdogan et al, 2018), and urban mapping (Diakogiannis et al, 2019). As such, they seem particularly wellsuited to extract field boundaries but this has yet to be empirically proven.…”

Section: Introductionmentioning

confidence: 99%

Extracting field boundaries from satellite imagery with a convolutional neural network to enable smart farming at scale

Waldner¹,

Diakogiannis²

2020

Preprint

View full text Add to dashboard Cite

Applications of digital agricultural services often require either farmers or their advisers to provide digital records of their field boundaries. Automatic extraction of field boundaries from satellite imagery would reduce the reliance on manual input of these records which is time consuming and error-prone, and would underpin the provision of remote products and services. The lack of current field boundary data sets seems to indicate low uptake of existing methods, presumably because of expensive image preprocessing requirements and local, often arbitrary, tuning. In this paper, we address the problem of field boundary extraction from satellite images as a multi-task semantic segmentation problem. We used ResUNet-a, a deep convolutional neural network with a fully connected UNet backbone that features dilated convolutions and conditioned inference, to assign three labels to each pixel: 1) the probability of belonging to a field; 2) the probability of being part of a boundary; and 3) the distance to the closest boundary. These labels can then be combined to obtain closed field boundaries. Using a single composite image from Sentinel-2, the model was highly accurate in mapping field extent, field boundaries, and, consequently, individual fields. Replacing the monthly composite with a single-date image close to the compositing period only marginally decreased accuracy. We then showed in a series of experiments that our model generalised well across resolutions, sensors, space and time without recalibration. Building consensus by averaging model predictions from at least four images acquired across the season is the key to coping with the temporal variations of accuracy. By minimising image preprocessing requirements and replacing local arbitrary decisions by datadriven ones, our approach is expected to facilitate the extraction of individual crop fields at scale.

show abstract

3D Convolutional Neural Networks for Crop Classification with Multi-Temporal Remote Sensing Images

Cited by 282 publications

References 34 publications

Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems

Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems

A weakly supervised approach for estimating spatial density functions from high-resolution satellite imagery

Extracting field boundaries from satellite imagery with a convolutional neural network to enable smart farming at scale

Contact Info

Product

Resources

About