ABB Crane systems today apply a great number of cameras that capture projections of the three
dimensional surrounding to the container cranes under operation. LIDAR data is also being collected in
several cases. Fusing these two measurement modalities would possibly yield an aggregate data that
would benefit from both the strong point of the information rich camera image and the robust depth
estimate of the LIDAR. One way of fusing such data would be to formulate a model representation that
could be updated with the measurement of the image and lidar frame as they become available. The
suggested theme for this master thesis work is to investigate the details of such reconstruction.
Our understanding of the function of proteins, DNA, RNA and other biological
macromolecules, as well as the design of new drug molecules, rely strongly on the possibility
to obtain atomic-resolution structures by X-ray and neutron crystallography. Currently, almost
150 000 such structures are freely available in the protein databank. In Lund, data collection
for such structures can be performed at the Max IV laboratory and when ESS is running, it
will be possible to collect data for neutron structures at an unprecedented speed.
To start with, we will restrict the project to identify water molecules in X-ray crystallographic
maps. To train the model, we will employ a set of (~1000) curated maps where standard
methods clearly identify the presence or absence of a water molecule. Additional data can
easily be generated, both from existing crystal structures or from simulated molecular data.
The world leading group in laser remote sensing located at Lund University collects hundreds of thousands insect observation per day from airborne insects using entomological lidar measurements. The measurements, where the signals are of time-varying character and suffer from the influence of varying noise and other disturbances, are used for classification of species, sex and age groups. Exploitation relies on robust estimation of the parameters of the time-varying signals and especially the fundamental tone. In this master thesis we suggest an investigation of a novel technique, matched reassignment, which is a mathematical method that relies on the phase of the Fourier transform to reassign the power of the spectrogram. The method has been shown to outperform other time-frequency techniques and it gives accurate estimates of the time- as well as the frequency locations.
Supervisors: Maria Sandsten, sandsten@maths.lth.se; Mikkel Brydegaard and Samuel Jansson, Dept. Physics.
Prerequisites: FMSF10/MASC04, FMSN35/MASM26
Resistivity and time domain induced polarization
(DCIP) measurements is a great tool for mapping
properties of the subsurface. This master thesis project invloves analysis of
existiving DCIP data to find suitable machine learning
or signal processing apporaches to deal with non-stationary
disturbances from trams and metro.
Prerequisites: FMSF10/MASC04, (FMS051/MASM17), (FMSN35/MASM26)
Role Summary
Software development for an audio recognition and matching tool -- a chance to develop an application together with the end-user! You will be able to lead a project and develop a tool in close cooperation with the end user and the education team. The intent of the tool is to be able to allow the user to speak the correct number in Swedish.
Essential Responsibilities
The objective of this project is to develop an application for audio recognition and matching. The application needs to check "intended" target for speech versus the pronunciation in the area of numbers in Swedish. The purpose is to be able to detect which number the user pronounces in Swedish.
• Research on the state of the art systems in this area, Deep learning and similar approaches.
• Investigate how little user input the system must require in order to yield a good result.
• Evaluation of the system based on precision and ease of use.
Qualifications/Requirements
Qualifications regarding this opening:
• Experienced with programming
• Experienced with audio signal processing
• Experienced with machine learning