Remote sensing image analysis

With improvements in data availability, coverage, resolution, and diversity, remote sensing data can potentially provide a huge amount of information and knowledge. On one hand, machine learning, (particularly deep learning techniques), has drastically altered the methodological landscape of remote sensing and stimulated numerous novel applications. On the other hand, the challenges of properly using the data and methods, assessing the uncertainty in the modelling process, and understanding the knowledge we can/cannot not gain from the remote sensing data are becoming more and more prominent. We aim to better understand these questions in our research.

We are currently working on the research topics below:

Image classification

Deep learning and OBIA-based methods for coastal geomorphological classification.

This project focuses on innovating classic object-based image analysis with machine learning methods and understanding which and how spatial, temporal and spectral features contribute to the geomorphological classification.

This project serves as a high social impact application to advance techniques in remote sensing image analysis in general.

Urban element extraction and change detection.

We innovate and apply deep learning algorithms for a variety of remote sensing image analysis tasks for various applications such as building delineation, crop classification, image cloud detection and removal, waterbody delineation, and land cover change detection.

Water body segmentation.

Visualization of feature maps located at the last layer of the decoder of different network structures.

Coastal geomorphological classification from aerial imagery with optical and infrared bands and Lidar.

H1: hard substrate,
P1a1: sandy low dynamic flats ,
P1a2: silty low dynamic flats •
P2b: mega-ripples,
P2c: high dynamic shoal flats

Lu, M., Groeneveld, L., Karssenberg, D., Ji, S., Jentink, R., Paree, E., and Addink, E.:

GEOMORPHOLOGICAL MAPPING OF INTERTIDAL AREAS, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.,XLIII-B3-2021, 75–80,

https://doi.org/10.5194/isprs-archives-XLIII-B3-2021-75-2021, 2021.

Ji S, Shen Y, Lu M, Zhang Y.

Building Instance Change Detection from Large-Scale Aerial Images using Convolutional Neural Networks and Simulated Samples.

Remote Sensing. 2019; 11(11):1343.

https://doi.org/10.3390/rs11111343

Zhang Z, Lu M, Ji S, Yu H, Nie C. Rich

CNN Features for Water-Body Segmentation from Very High Resolution Aerial and Satellite Imagery.

Remote Sensing. 2021; 13(10):1912.

https://doi.org/10.3390/rs13101912

We focus on machine learning methods and specifically;

Transfer learning. Improving model generality for data from different sources.
Image fusion and downscaling: innovating current image fusion and downscaling methods using machine learning.
Spatiotemporal and multi-spectral information integration: developing algorithms that flexibly and explicitly model spectral and spatiotemporal information, as well as image pre-processing and postprocessing approaches.
Weakly and unsupervised learning to reduce the dependency on (perfect) labels and to include priors or additional information.
Generative models for feature simulation, transfer learning, and better image classification.

Change detection from remote sensing image time series and information integration from multispectral, spatiotemporal arrays

		Lu, M., Pebesma, E., Sanchez, A., & Verbesselt, J. (2016). Spatio-temporal change detection from multidimensional arrays: Detecting deforestation from MODIS time series. ISPRS Journal of Photogrammetry and Remote Sensing, 227-236. [117]. https://doi.org/10.1016/j.isprsjprs.2016.03.007

		Lu M, Hamunyela E, Verbesselt J, Pebesma E. Dimension Reduction of Multi-Spectral Satellite Image Time Series to Improve Deforestation Monitoring. Remote Sensing. 2017; 9(10):1025. https://doi.org/10.3390/rs9101025

Webmaster: Sebastian De La Serna

Faculty of Biology, Chemistry & Earth Sciences

Chair of Spatial Big Data - Juniorprofessor Dr. Meng Lu

Remote sensing image analysis