Target-Adaptive CNN-Based Pansharpening

Scarpa, Giuseppe; Vitale, Sergio; Cozzolino, Davide

doi:10.1109/tgrs.2018.2817393

Cited by 329 publications

(181 citation statements)

References 64 publications

(104 reference statements)

Supporting

Mentioning

179

Contrasting

Unclassified

Order By: Relevance

“…To what extent we can use a machine learning model trained elsewhere? This is a key problem in machine learning and is very relevant for a number of remote sensing applications, such as coregistration [62] or pansharpening [24]. In [62], the importance of selecting training data which are homogeneous with the target has been underlined.…”

Section: Discussion and Future Perspectivementioning

confidence: 99%

“…In the last few years, CNNs have been successfully applied to many classical image processing problems, such as denoising [50], super-resolution [51], pansharpening [8,24], segmentation [52], object detection [53,54], change detection [27] and classification [17,[55][56][57]. The main strengths of CNNs are (i) an extreme versatility that allows them to approximate any sort of linear or non-linear transformation, including scaling or hard thresholding; (ii) no need to design handcrafted filters, replaced by machine learning; (iii) high-speed processing, thanks to parallel computing.…”

Section: Convolutional Neural Networkmentioning

confidence: 99%

“…This relatively shallow CNN is characterized by a rather small number of weights (as CNNs go), counted in Table 1, and hence can be trained with a small amount of data. Moreover, slightly different architectures have proven to achieve state-of-the-art performance in closely-related applications, such as super-resolution [51] and data fusion [8,24].…”

Section: Proposed Prediction Architecturesmentioning

confidence: 99%

“…The choice depends on the domain of the output and affects the convergence properties of the networks [58]. Our experiments have shown the L 1 -norm (Equation (4)) to be more effective than other options for training; therefore, we keep this choice, which proved effective also in other generative problems [24]:…”

Section: Input Bands Model Namementioning

confidence: 99%

“…One of the most frequent applications is pansharpening [6,23,24], although many other tasks can be solved under a multi-resolution paradigm, such as segmentation [25] or feature extraction [26], to mention a few. -multi-temporal: is one of the most investigated forms of fusion in remote sensing due to the rich information content hidden in the temporal dimension.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A CNN-Based Fusion Method for Feature Extraction from Sentinel Data

et al. 2018

Self Cite

View full text Add to dashboard Cite

Sensitivity to weather conditions, and specially to clouds, is a severe limiting factor to the use of optical remote sensing for Earth monitoring applications. A possible alternative is to benefit from weather-insensitive synthetic aperture radar (SAR) images. In many real-world applications, critical decisions are made based on some informative optical or radar features related to items such as water, vegetation or soil. Under cloudy conditions, however, optical-based features are not available, and they are commonly reconstructed through linear interpolation between data available at temporally-close time instants. In this work, we propose to estimate missing optical features through data fusion and deep-learning. Several sources of information are taken into account-optical sequences, SAR sequences, digital elevation model-so as to exploit both temporal and cross-sensor dependencies. Based on these data and a tiny cloud-free fraction of the target image, a compact convolutional neural network (CNN) is trained to perform the desired estimation. To validate the proposed approach, we focus on the estimation of the normalized difference vegetation index (NDVI), using coupled Sentinel-1 and Sentinel-2 time-series acquired over an agricultural region of Burkina Faso from May-November 2016. Several fusion schemes are considered, causal and non-causal, single-sensor or joint-sensor, corresponding to different operating conditions. Experimental results are very promising, showing a significant gain over baseline methods according to all performance indicators.

show abstract

Section: Discussion and Future Perspectivementioning

confidence: 99%

Section: Convolutional Neural Networkmentioning

confidence: 99%

Section: Proposed Prediction Architecturesmentioning

confidence: 99%

Section: Input Bands Model Namementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A CNN-Based Fusion Method for Feature Extraction from Sentinel Data

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

Bibliography

2021

Deep Learning for the Earth Sciences

View full text Add to dashboard Cite

Anatomically Designed Triboelectric Wristbands with Adaptive Accelerated Learning for Human–Machine Interfaces

Fang

Wang

et al. 2023

Advanced Science

View full text Add to dashboard Cite

Recent advances in flexible wearable devices have boosted the remarkable development of devices for human-machine interfaces, which are of great value to emerging cybernetics, robotics, and Metaverse systems. However, the effectiveness of existing approaches is limited by the quality of sensor data and classification models with high computational costs. Here, a novel gesture recognition system with triboelectric smart wristbands and an adaptive accelerated learning (AAL) model is proposed. The sensor array is well deployed according to the wrist anatomy and retrieves hand motions from a distance, exhibiting highly sensitive and high-quality sensing capabilities beyond existing methods. Importantly, the anatomical design leads to the close correspondence between the actions of dominant muscle/tendon groups and gestures, and the resulting distinctive features in sensor signals are very valuable for differentiating gestures with data from 7 sensors. The AAL model realizes a 97.56% identification accuracy in training 21 classes with only one-third operands of the original neural network. The applications of the system are further exploited in real-time somatosensory teleoperations with a low latency of <1 s, revealing a new possibility for endowing cyber-human interactions with disruptive innovation and immersive experience.

show abstract

Target-Adaptive CNN-Based Pansharpening

Cited by 329 publications

References 64 publications

A CNN-Based Fusion Method for Feature Extraction from Sentinel Data

A CNN-Based Fusion Method for Feature Extraction from Sentinel Data

Bibliography

Anatomically Designed Triboelectric Wristbands with Adaptive Accelerated Learning for Human–Machine Interfaces

Contact Info

Product

Resources

About