Bachelor and Master Theses Propositions in Remote Sensing

Focus Areas

  • Pattern recognition in large data sets
  • Classification methods, such as supervised learning, unsupervised learning, self-taught learning
  • Feature learning / representation learning
  • Interpretability and explainabilty for deep neural networks
  • Analysis techniques, such as clustering or archetypal analysis
Sparse Representation for root segmentation in  Hyperspectral images

German title: Sparse Representation für die Segmentierung von Wurzeln in Hyperspektralbildern

The segmentation of roots in images is challenging due to their similar color appearance and their thin structure. Sparse representation can be used for segmentation by representing each image patch by a weighted linear combination of basis patches (e.g., [2]). These basis patches are learned from annotated images and will be collected in discriminative dictionaries, which contain the most distinct structures of roots and background. The suitability of the approach will be evaluated by means of hyperspectral images of plants grown in flat rhizotrons [1].

[1] Steier, A., Cendrero-Mateo, P., Malolepszy, T., Gioia, T., Briese, C., & Rascher, U. (2013). HIDeR – Hyperspectral Imaging Device for Rhizotrons Simultaneously Scanning of Shoot and Root in Plant Phenotyping, DPPN.
[2] Berkels, B., Kotowski, M., Rumpf, M., & Schaller, C. (2011, May). Sulci detection in photos of the human cortex based on learned discriminative dictionaries. In International Conference on Scale Space and Variational Methods in Computer Vision, 326-337.

Supervision: Ribana Roscher, Uwe Rascher

Derivation of Sea Surface Heights from Altimetry Waveforms Using Deep Neural Networks

German title: Ableitung von Meereshöhen aus Altimeterwaveforms mit Deep Neural Networks

Satellite radar altimetry is one of the most powerful techniques for measuring sea surface height variations, with applications ranging from operational oceanography to climate research. Over open oceans, altimeter return waveforms generally correspond to the Brown model, and by inversion, estimated shape parameters provide mean surface height and wind speed. However, in coastal areas the parameter estimation is a challenging task as the observation signal is disturbed by land influences causing erroneous and imprecise height estimations [1].

The goal of this thesis is the development of an automated framework for the estimation of sea surface heights from radar altimetry waveforms. For this, regression networks will be utilized [1]. They state a powerful approach from the deep learning area, which are able to identify the parts of the signals which are relevant for the derivation of geophysical parameters.

[1] Roscher, R., Uebbing, B., & Kusche, J. (2017). STAR: Spatio-temporal altimeter waveform retracking using sparse representation and conditional random fields. Remote Sensing of Environment, 201, 148-164.
[2] Goodfellow, I., Bengio, Y., Courville, A., & Bengio, Y. (2016). Deep learning (Vol. 1). Cambridge: MIT press.

Supervision: Ribana Roscher, Jürgen Kusche

Leading Edge Detection with Convolutional Neural Networks

German title: Leading Edge Detektion mit Convolutional Neural Networks

Satellite radar altimetry is one of the most powerful techniques for measuring sea surface height variations, with applications ranging from operational oceanography to climate research. Over open oceans, altimeter return waveforms generally correspond to the Brown model, and by inversion, estimated shape parameters provide mean surface height and wind speed. However, in coastal areas the parameter estimation is a challenging task as the observation signal is disturbed by land influences causing erroneous and imprecise height estimations [1].

The goal of this thesis is the development of an automated framework for the detection of the leading edge from radar altimetry waveforms. For this, convolutional neural networks will be utilized [2]. They state a powerful approach from the deep learning area, which are able to capture spatial features from the signals.

[1] Roscher, R., Uebbing, B., & Kusche, J. (2017). STAR: Spatio-temporal altimeter waveform retracking using sparse representation and conditional random fields. Remote Sensing of Environment, 201, 148-164.
[2] Goodfellow, I., Bengio, Y., Courville, A., & Bengio, Y. (2016). Deep learning (Vol. 1). Cambridge: MIT press.

Supervision: Ribana Roscher, Jürgen Kusche

Statistical Analysis of Grapevine Clusters

German title: Statistische Analyse von Weintraubenparametern

Berry size, cluster density, and shape belong to the most important fruit traits in grapevine breeding. Non-invasive, image-based phenotyping promises a fast and precise method for the monitoring of these traits. However, even powerful machine learning methods can produce inaccurate results, making post-processing steps necessary.

The goal of this thesis is the development of an automated post-processing image analyzing framework based on given detection results of grapevine berries from images. The experiments will be conducted using RGB and NIR images acquired by a robot in a vineyard.

Supervision: Ribana Roscher, Laura Zabawa

Color classification of Grapevine Berries

German title: Farbklassifikation von Weinbeeren

The color of berries is one of the most important fruit traits in grapevine breeding since it indicates the state of ripeness. Non-invasive, image-based phenotyping promises a fast and precise method for the monitoring of this trait. However, the determine of the color is challenging due to different illumination conditions or effects such as reflections or shadows.

The goal of this thesis is the development of an automated image analyzing framework based on given detection results of grapevine berries from images. The experiments will be conducted using RGB images acquired by a robot in a vineyard.

Supervision: Ribana Roscher, Laura Zabawa