Continuous Multi-Angle Remote Sensing and Its Application in Urban Land Cover Classification

Yao, Yuan; Leung, Yee; Fung, Tung; Shao, Zhenfeng; Lu, Jie; Meng, Deyu; Ying, Hanchi; Zhou, Yu

doi:10.3390/rs13030413

Cited by 9 publications

(3 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Second, data sources were dimensionally limited; we only obtained 2D multispectral data. Further studies of multi-angle multispectral data may be helpful (Wang R. et al, 2021;Yuan et al, 2021), indicating that more information on the distribution of spectral values in the middle and lower parts of canopy would be learned by the model. Moreover, the spectral class was limited, fusion and inversion with hyperspectral images will provide richer spectral information (Xu, 2023), thereby extending the 3D spectral breadth of forests in terms of species.…”

Section: Future Improvementsmentioning

confidence: 99%

Forest stand spectrum reconstruction using spectrum spatial feature gathering and multilayer perceptron

Wang,

Song,

Liu

et al. 2023

Front. Plant Sci.

View full text Add to dashboard Cite

IntroductionThree-dimensional spectral distributions of forest stands can provide spatial information on the physiological and biochemical status of forests, which is vital for forest management. However, three-dimensional spectral studies of forest stands are limited.MethodsIn this study, LiDAR and multispectral data were collected from Masson pine stands in southern Fujian Province, China, and a method was proposed for inverting forest spectra using point clouds as a unit. First, multispectral values were mapped to a point cloud, and the isolated forest algorithm combined with K-means clustering was applied to characterize fusion data. Second, five deep learning algorithms were selected for semantic segmentation, and the overall accuracy (oAcc) and mean intersection ratio (mIoU) were used to evaluate the performance of various algorithms on the fusion data set. Third, the semantic segmentation model was used to reconfigure the class 3D spectral distribution, and the model inversion outcomes were evaluated by the peaks and valleys of the curve of the predicted values and distribution gaps.ResultsThe results show that the correlations between spectral attributes and between spatial attributes were both greater than 0.98, while the correlation between spectral and spatial attributes was 0.43. The most applicable method was PointMLP, highest oAcc was 0.84, highest mIoU was 0.75, peak interval of the prediction curve tended to be consistent with the true values, and maximum difference between the predicted value and the true value of the point cloud spectrum was 0.83.DiscussionExperimental data suggested that combining spatial fusion and semantic segmentation effectively inverts three-dimensional spectral information for forest stands. The model could meet the accuracy requirements of local spectral inversion, and the NIR values of stands in different regions were correlated with the vertical height of the canopy and the distance from the tree apex in the region. These findings improve our understanding of the precise three-dimensional spectral distribution of forests, providing a basis for near-earth remote sensing of forests and the estimation of forest stand health.

show abstract

Section: Future Improvementsmentioning

confidence: 99%

Forest stand spectrum reconstruction using spectrum spatial feature gathering and multilayer perceptron

Wang,

Song,

Liu

et al. 2023

Front. Plant Sci.

View full text Add to dashboard Cite

show abstract

“…It is simply defined as labeling an aerial image with a semantic category according to the image content. With the fast development of deep learning, most existing studies [19][20][21][22][23] have shown great success. These methods are all the singlelabel image classification.…”

Section: A Aerial Image Processing Tasksmentioning

confidence: 99%

A Confounder-Free Fusion Network for Aerial Image Scene Feature Representation

Xiong

Cui

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

“…Remote sensed data has become the pillar for all kinds of land observation tasks [2,3]. Given the availability of such data and thanks to deep learning methods, different kinds of implementations have emerged in domains such as agriculture, urban development and environmental purposes [3][4][5][6][7][8][9].…”

Section: Introductionmentioning

confidence: 99%

Deep learning for photovoltaic panels segmentation

Bouzaachane¹,

Darouichi²,

Guarmah³

2023

Math. Model. Comput.

View full text Add to dashboard Cite

Due to advanced sensor technology, satellites and unmanned aerial vehicles (UAV) are producing a huge amount of data allowing advancement in all different kinds of earth observation applications. Thanks to this source of information, and driven by climate change concerns, renewable energy assessment became an increasing necessity among researchers and companies. Solar power, going from household rooftops to utility-scale farms, is reshaping the energy markets around the globe. However, the automatic identification of photovoltaic (PV) panels and solar farms' status is still an open question that, if answered properly, will help gauge solar power development and fulfill energy demands. Recently deep learning (DL) methods proved to be suitable to deal with remotely sensed data, hence allowing many opportunities to push further research regarding solar energy assessment. The coordination between the availability of remotely sensed data and the computer vision capabilities of deep learning has enabled researchers to provide possible solutions to the global mapping of solar farms and residential photovoltaic panels. However, the scores obtained by previous studies are questionable when it comes to dealing with the scarcity of photovoltaic systems. In this paper, we closely highlight and investigate the potential of remote sensing-driven DL approaches to cope with solar energy assessment. Given that many works have been recently released addressing such a challenge, reviewing and discussing them, it is highly motivated to keep its sustainable progress in future contributions. Then, we present a quick study highlighting how semantic segmentation models can be biased and yield significantly higher scores when inference is not sufficient. We provide a simulation of a leading semantic segmentation architecture U-Net and achieve performance scores as high as 99.78%. Nevertheless, further improvements should be made to increase the model's capability to achieve real photovoltaic units.

show abstract

Continuous Multi-Angle Remote Sensing and Its Application in Urban Land Cover Classification

Cited by 9 publications

References 34 publications

Forest stand spectrum reconstruction using spectrum spatial feature gathering and multilayer perceptron

Forest stand spectrum reconstruction using spectrum spatial feature gathering and multilayer perceptron

A Confounder-Free Fusion Network for Aerial Image Scene Feature Representation

Deep learning for photovoltaic panels segmentation

Contact Info

Product

Resources

About