guest ::
| Home > Publications database > Evaluation of Line Detection Methods for the Analysis of Charge Stability Diagrams |
| Home > Publications database > Evaluation of Line Detection Methods for the Analysis of Charge Stability Diagrams |
| Book/Report | FZJ-2023-02705 |
2023
Forschungszentrum Jülich GmbH Zentralbibliothek, Verlag
Jülich
This record in other databases:
Please use a persistent id in citations: doi:10.34734/FZJ-2023-02705
Report No.: 4441
Abstract: The various approaches for the implementation of qubits are objects of ongoing research. One approach uses the spin states of electrons that are confined in double quantum dots. For the calibration of such a qubit, charge stability diagrams (CSDs) are measured and analyzed. They constitute the changes in dependence on the voltages of two electrodes and, thus, can be represented as grayscale images. The voltages at which electrons tunnel in or out of the quantum dots correspond to line structures in the CSD images. Therefore, the automatic detection of lines in CSD images plays an important role in the qubit calibration process. Various line detection methods are explored in this thesis. Since edge detection is a substep employed by several line detectors, a selection of edge detection methods is explored as well. Additionally, postprocessing methods for the validation, grouping, and merging of lines and the segmentation of infinite lines into finite line segments are investigated. This thesis focuses on traditional approaches as opposed to approaches based on machine learning. Two methods are proposed for the automatic evaluation of the detection results. Both methods function by comparing the detections against a manually labeled ground truth data set. The first method evaluates edge maps by comparing the individual edge pixels with the pixels corresponding to the ground truth lines. The second method benchmarks the detected lines against the ground truth lines. Using these evaluation methods, the free parameters of the edge detectors, line detectors, and postprocessing methods are optimized. The detection results of the methods are evaluated and compared for four data setsof varying quality. Additionally, the implications of the detection quality for the analysis of CSDs are explained.
; 
|
The record appears in these collections: |
PDF