Advertisement

Using Aerial Platforms in Predicting Water Quality Parameters from Hyperspectral Imaging Data with Deep Neural Networks

  • Taina Hakala
  • Ilkka PölönenEmail author
  • Eija Honkavaara
  • Roope Näsi
  • Teemu Hakala
  • Antti Lindfors
Chapter
  • 53 Downloads
Part of the Computational Methods in Applied Sciences book series (COMPUTMETHODS, volume 54)

Abstract

In near future it is assumable that automated unmanned aerial platforms are coming more common. There are visions that transportation of different goods would be done with large planes, which can handle over 1000 kg payloads. While these planes are used for transportation they could similarly be used for remote sensing applications by adding sensors to the planes. Hyperspectral imagers are one this kind of sensor types. There is need for the efficient methods to interpret hyperspectral data to the wanted water quality parameters. In this work we survey the performance of neural networks in the prediction of water quality parameters from remotely sensed hyperspectral data in freshwater basins. The hyperspectral data consists of 36 bands in the wavelength range of 508–878 nm and the water quality parameters to be predicted are temperature, conductivity, turbidity, Secchi depth, blue-green algae, chlorophyll-a, total phosphorus, acidity and dissolved oxygen. The objective of this investigation was to study the behaviour of different types of neural networks with this kind of data. Study is a survey of the operation of neural networks on this problem, which can be used as a basis for the design of a more comprehensive study. The neural network types examined were multilayer perceptron and 1-, 2- and 3-dimensional convolutional neural networks with the effect of scaling the hyperspectral data with standard or min-max -scaler recorded. We also investigated investigated how the prediction of individual water quality parameter depends on whether the neural network model is done solely with respect to this one parameter or with several parameters predicted simultaneously with the same model. The results of the correspondence between the predicted and measured water quality parameters were presented with normalized root mean square error, Pearson correlation coefficient and coefficient of determination. The best models were obtained the 2-dimensional convolutional neural networks with standard scaling made separately for each parameter. The parameters showing good predictability were conductivity, turbidity, Secchi-depth, blue-green algae, chlorophyll-a and total phosphorus, for which the coefficient of determination was at least 0.96 (apart from Secchi-depth even 0.98).

Keywords

Remote sensing Hyperspectral Water quality Convolutional neural networks 

Notes

Acknowledgements

This research has been co-funded by Finnish Funding Agency for Innovation Tekes (grants 2208/31/2013 and 1711/31/2016).

References

  1. 1.
    Chebud Y, Naja GM, Rivero RG, Melesse AM (2012) Water quality monitoring using remote sensing and an artificial neural network. Water, Air, Soil Pollution 223(8):4875–4887CrossRefGoogle Scholar
  2. 2.
    Cybenko G (1989) Approximation by superpositions of a sigmoidal function. Math Control Signals Syst 2(4):303–314MathSciNetCrossRefGoogle Scholar
  3. 3.
    Goodfellow I, Bengio Y, Courville A, Bengio Y (2016) Deep learning, vol 1. MIT Press, CambridgeGoogle Scholar
  4. 4.
    Hakala T, Pölönen I, Honkavaara E, Näsi R, Hakala T, Lindfors A (2018) Variability of remote sensing spectral indices in boreal lake basins. International archives of the photogrammetry, remote sensing and spatial information sciences, vol 42. International Society for Photogrammetry and Remote SensingGoogle Scholar
  5. 5.
    Hakala TV (2018) Supplementary figures for the article ‘variability of remote sensing spectral indices in boreal lake basins’Google Scholar
  6. 6.
    Honkavaara E, Saari H, Kaivosoja J, Pölönen I, Hakala T, Litkey P, Mäkynen J, Pesonen L (2013) Processing and assessment of spectrometric, stereoscopic imagery collected using a lightweight uav spectral camera for precision agriculture. Remote Sens 5(10):5006–5039CrossRefGoogle Scholar
  7. 7.
    Krizhevsky A, Sutskever I, Hinton GE (2012) Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, pp 1097–1105Google Scholar
  8. 8.
    Markelin L, Honkavaara E, Näsi R, Nurminen K, Hakala T (2014) Geometric processing workflow for vertical and oblique hyperspectral frame images collected using UAV. International archives of the photogrammetry, remote sensing spatial information sciencesCrossRefGoogle Scholar
  9. 9.
    McCulloch WS, Pitts W (1943) A logical calculus of the ideas immanent in nervous activity. Bull Math Biophys 5(4):115–133MathSciNetCrossRefGoogle Scholar
  10. 10.
    Näsi R, Honkavaara E, Tuominen S, Saari H, Pölönen I, Hakala T, Viljanen N, Soukkamäki J, Näkki I, Ojanen H, et al (2016) UAS based tree species identification using the novel FPI based hyperspectral cameras in visible, NIR and SWIR spectral ranges. International archives of the photogrammetry, remote sensing and spatial information sciences, vol xli-b1Google Scholar
  11. 11.
    Nevalainen O, Honkavaara E, Tuominen S, Viljanen N, Hakala T, Yu X, Hyyppä J, Saari H, Pölönen I, Imai NN et al (2017) Individual tree detection and classification with uav-based photogrammetric point clouds and hyperspectral imaging. Remote Sens 9(3):185CrossRefGoogle Scholar
  12. 12.
    Pölönen I, Puupponen H-H, Honkavaara E, Lindfors A, Saari H, Markelin L, Hakala T, Nurminen K (2014) UAV-based hyperspectral monitoring of small freshwater area. Remote sensing for agriculture, ecosystems, and hydrology xvi, vol 9239. International Society for Optics and Photonics, pp 923912Google Scholar
  13. 13.
    Ranta E, Valtonen M, Ikonen E (2016) Hiidenveden alueen yhteistarkkailun yhteenveto vuodelta 2015. Länsi-Uudenmaan vesi ja ympäristö ry, KäsikirjoitusGoogle Scholar
  14. 14.
    Rumelhart DE, Hinton GE, Williams RJ (1986) Learning representations by back-propagating errors. Nature 323(6088):533CrossRefGoogle Scholar
  15. 15.
    Saari H, Pölönen I, Salo H, Honkavaara E, Hakala T, Holmlund C, Mäkynen J, Mannila R, Antila T, Akujärvi A (2013) Miniaturized hyperspectral imager calibration and UAV flight campaigns. Sensors, systems, and next-generation satellites xvii, vol 8889. International Society for Optics and Photonics, pp 88891OGoogle Scholar
  16. 16.
    Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R (2014) Dropout: a simple way to prevent neural networks from overfitting. J Mach Learn Res 15(1):1929–1958MathSciNetzbMATHGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  • Taina Hakala
    • 1
  • Ilkka Pölönen
    • 1
    Email author
  • Eija Honkavaara
    • 2
  • Roope Näsi
    • 2
  • Teemu Hakala
    • 2
  • Antti Lindfors
    • 3
  1. 1.Faculty of Information TechnologyUniversity of JyväskyläJyväskyläFinland
  2. 2.Finnish Geospatial Research Institute FGIKirkkonummiFinland
  3. 3.Luode Consulting LtdEspooFinland

Personalised recommendations