direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Theses

We offer Bachelor’s and Master’s thesis opportunities in the fields of processing and analysis of remote sensing images acquired by satellite systems for Earth observation. 

In the following you find currently available theses. If you are interested in one of the topics given below (or would like to propose a related topic), please contact the respective person or Prof. Demir.

Akshara Preethy Byju

akshara.preethybyju@campus.tu-berlin.de 

Topic: Generative Adversarial Network for Content Based Remote Sensing Image Retrieval in JPEG 2000 compressed domain

With the continuous advances in satellite technology, volume of remote sensing (RS) data is increasing massively. Accordingly, developing fast and efficient Content based Image Retrieval (CBIR) techniques has gained attention in the RS society. In order to reduce the storage size of the archive, RS images are preferred to be stored in compressed format. Recently JPEG 2000 (JP2) compression algorithm has become popular due to its multiresolution paradigm as well as increased compression ratio. Recent advances in deep learning shows that Generative Adversarial Networks (GAN) has the ability to model high dimensional distributions of data. In addition, GANs require the data to be decompressed before any retrieval operation in performed which is time-consuming and impractical in the case of large-scale RS archives. To address these limitations, the proposed research aims: 1) to develop a novel architecture that achieve efficient multi-scale feature representation in the compressed domain using GANs in a unified framework; 2) to address the possible solution to minimize the textural loss that are obtained from the partially decoded wavelet coefficients in the JP2 compression algorithm. In computer vision, there exist only a few studies that uses Convolutional Neural Network (CNNs) as well as GANs in the wavelet domain that has shown good performance for studies related with super-resolution as well as face hallucination [1, 2, 3]. In addition, it has to be noted that the performance of GAN for image retrieval has not yet been explored for the RS CBIR in compressed domain. Experiments for the proposed research can be performed in NWPU-RESISC45 dataset.


Nice to have: Prior knowledge on image processing


Prerequisite:
Prior knowledge on deep learning


We also encourage students to propose own topics. If you are interested in a topic in this area, please, provide me with some information about you, your interests, your programming skills, and your CV.


References:

[1] Huang, H., He, R., Sun, Z. and Tan, T. (2019). Wavelet Domain Generative Adversarial Network for Multi-scale Face Hallucination. International Journal of Computer Vision, 127(6-7), pp.763-784.
[2] Huang, H., He, R., Sun, Z. and Tan, T. (2017). Wavelet-srnet: A wavelet-based cnn for multi-scale face super resolution. In Proceedings of the IEEE International Conference on Computer Vision, pp. 1689-1697.
[3] Wang, T., Sun. W, Qi H. and Ren. P. (2018). Aerial image super resolution via wavelet multiscale convolutional neural networks. IEEE Geosci. Remote. Sens. Lett. 15, 769–773.

Bernhard Föllmer

bernhard.foellmer@tu-berlin.de

Topic: Change Detection and Change Classification in Remote Sensing with Attention based Deep Learning models

Lupe

Change detection and change classification is a challenging task in remote sensing, used to identify areas of change between two images acquired at different times for the same geographical area. Robust and accurate change detection methods are required in different fields like climate change, environmental monitoring or emergency management. To detect and classify different kind of changes, deep learning models have to extract significant spectral-spatial-temporal features from changed areas and suppress features from unchanged areas [1].
Attention mechanisms implicitly learn to suppress irrelevant regions in images while highlighting salient features. In computer vision, attention mechanisms are applied to a variety of problems, including image classification, segmentation, action recognition, image captioning and visual question answering [2].
In this study we want to 1) Analyze different attention mechanisms for change detection and change classification and 2) Analyze different attention mechanisms for change retrieval from EO data archives.


Nice to have: Prior knowledge on image processing


Prerequisite: Prior knowledge on deep learning


We also encourage students to propose their own topics. If you are interested in a topic in this area, please, provide me with some information about you, your interests, your programming skills, and your CV.


References:

[1] Wiratama, W., & Sim, D. (2019). Fusion Network for Change Detection of High-Resolution Panchromatic Imagery. Applied Sciences, 9(7), 1441.
[2] Schlemper, J., Oktay, O., Schaap, M., Heinrich, M., Kainz, B., Glocker, B., & Rueckert, D. (2019). Attention gated networks: Learning to leverage salient regions in medical images. Medical Image Analysis, 53, 197–207.

Gencer Sümbül

gencer.suembuel@tu-berlin.de

Topic: Multisource Multi-Label Remote Sensing Image Scene Classification

The increased number of recent Earth observation satellite missions has led to a significant growth of remote sensing (RS) image archives. Accordingly, associating one of the predefined categories to the most significant content of a RS image scene with deep neural network models, which is usually achieved by direct supervised classification of each image in the archive, has received increasing attention in RS. However, assigning different low-level land-cover class labels (i.e., multi-labels) to a RS image is not well studied in the literature. Since it is much more complex than the single label scene classification, joint use of different image sources together [1] in order to both model the co-occurrence of different land-cover classes and leverage the complementary spectral, spatial, and structural information embedded in different sources is crucial. This study requires to develop unified deep neural network framework that simultaneously i) learns the multi-label classification rules by accurately characterizing the information from different sources; and ii) overcomes the possible problems of using different sources together like alignment and registration of RS images from different sources. For this study, the BigEarthNet [2], which is a new large-scale Sentinel-2 benchmark archive, and the EU-DEM [3], which is digital surface model of whole Europe, will be used.


Nice to have: Prior knowledge on image processing


Prerequisite: Prior knowledge on deep learning


We also encourage students to propose own topics. If you are interested in a topic in this area, please, provide me with some information about you, your interests, your programming skills, and your CV.


References:

[1] X. Xu, W. Li, Q. Ran, Q. Du, L. Gao and B. Zhang, "Multisource Remote Sensing Data Classification Based on Convolutional Neural Network," in IEEE Transactions on Geoscience and Remote Sensing, vol. 56, no. 2, pp. 937-949, Feb. 2018.
[2] G. Sumbul, M. Charfuelan, B. Demir, V. Markl, BigEarthNet: A Large-Scale Benchmark Archive for Remote Sensing Image Understanding, International Conference on Geoscience and Remote Sensing Symposium (IGARSS), 2019.
[3] land.copernicus.eu/imagery-in-situ/eu-dem

Topic: Self-Supervised Feature Learning for Content Based Remote Sensing Image Retrieval

Supervised training of deep convolutional neural networks for the feature learning of content-based remote sensing (RS) image retrieval task requires massive amounts of manually labeled data in order to obtain high retrieval accuracy. However, this is both expensive and impractical to scale. Therefore, unsupervised semantic feature learning, i.e., learning without requiring manual annotation effort, is of crucial importance in order to successfully leverage the vast amount of freely available RS images. Recently, a novel paradigm for unsupervised learning called self-supervised learning is proposed in the computer vision literature [1], [2], [3], [4]. The main idea is to exploit different labeling (e.g. relative spatial co-location of image patches, different rotations of image patches etc.) that are available besides or within images, and to use them as intrinsic reward signals to learn general-purpose image features. The features obtained with existing self-supervised approaches have been successfully transferred to classification and detections tasks in computer vision literature, and their performance is encouraging when compared to fully-supervised training. However, self-supervision for content based RS image retrieval has not been investigated yet. This study requires 1) to define possible labeling that can be extracted only from RS images, 2) to create suitable neural network and training procedure for feature extraction with self-supervised learning and 3) to find a way for benefitting from extracted image features for content based retrieval. Experiments for this study will be conducted on the BigEarthNet, new large-scale Sentinel-2 benchmark archive [5].


Nice to have: Prior knowledge on image processing


Prerequisite: Prior knowledge on deep learning


We also encourage students to propose own topics. If you are interested in a topic in this area, please, provide me with some information about you, your interests, your programming skills, and your CV.


References:

[1] Z. Feng, C. Xu, D. Tao, “Self-Supervised Representation Learning by Rotation Feature Decoupling”, Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
[2] A. Kolesnikov, X. Zhai, L. Beyer, “Revisiting Self-Supervised Visual Representation Learning”, Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
[3] T. N. Mundhenk, D. Ho, B. Y. Chen, “Improvements to context based self-supervised learning”, Conference on Computer Vision and Pattern Recognition (CVPR), 2018.
[4] C. Doersch, A. Gupta, A. A. Efros, “Unsupervised Visual Representation Learning by Context Prediction”, International Conference on Computer Vision (ICCV), 2015.
[5] G. Sumbul, M. Charfuelan, B. Demir, V. Markl, BigEarthNet: A Large-Scale Benchmark Archive for Remote Sensing Image Understanding, International Conference on Geoscience and Remote Sensing Symposium (IGARSS), 2019.

Jian Kang

jian.kang@tu-berlin.de

Topic: Cross-modality deep hashing

With the rapid development of earth observation (EO) sensors, we step into the big EO data period. How to sufficiently and effectively retrieve semantic similarity preserved remote sensing (RS) images from large-scale data archives, e.g. BigEarthNet [1], plays a very important role in RS. One of the most popular way to solve that problem is indexing the dataset by exploiting binary codes, which is also dubbed as hashing. The keypoint of hashing is to learn an efficient hashing function, which can project the original images from the high-dimensional image space into a low-dimensional hamming space. To learn such hashing functions, label information is normally required. However, image labeling demands huge manual efforts, especially for large-scale dataset. One option is to exploit other data modalities, such as text description/tag information, to guide the hashing function learning. This study will investigate to develop deep learning based methods to learn a hashing function, which can project the features from text descriptions into hashing codes of the associated RS images. Specifically, 1) we will study the state-of-the-art cross-modality deep learning architecture, e.g. [2], for the RS image retrieval. 2) we will develop novel deep learning based cross-modality hashing methods, which can fully exploit spectral and spatial information of RS images and semantic information of the corresponding text descriptions. 3) Attention module for feature extraction from sequences of texts and the associated spatial-spectral RS images can also be an option for improving the quality of hashing code learning.

 

Nice to have: Prior knowledge on image processing

 

Prerequisite: Prior knowledge on deep learning

 

We also encourage students to propose their own topics. If you are interested in a topic in this area, please, provide me with some information about you, your interests, your programming skills, and your CV.

 

References:

[1] G. Sumbul, M. Charfuelan, B. Demir, V. Markl, BigEarthNet: A Large-Scale Benchmark Archive for Remote Sensing Image Understanding, International Conference on Geoscience and Remote Sensing Symposium (IGARSS), 2019.
[2] Jiang, Qing-Yuan, and Wu-Jun Li. "Deep cross-modal hashing." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.

Zusatzinformationen / Extras

Quick Access:

Schnellnavigation zur Seite über Nummerneingabe

Auxiliary Functions