TUPP01
HZDR, Dresden, Germany
TU Dresden, Dresden, Germany
| Paper | Title | Page |
|---|---|---|
TUPP01 |
The Applications of Machine Learning in Slit Scan Emittance Measurements | |
|
||
| The electron beam transverse projected emittance is one of the importance parameters to a photoinjector. Two traditional emittance measurement techniques are curried out widely in photoinjectors beamline, quadrupole scan and slit scan. Comparing the quadrupole scan, the slit scan can give the details of the electron beam phase space, which is useful for studying and understanding the beam dynamics. The ELBE SRF Gun diagnostics beamline was built science 2008 and both of two techniques were used to measure beam emittance. From 2019, to optimize beam parameters for users more effectively, a new fast slit scan system was built, which is based on a mask with one 100 um width slit, a fast motor, two cameras and a control and data analysis software. In one experiment, about one hundred images will be captured in 5 to 15 seconds. In traditional data analysis method, the image processing is time-consuming to reduce noise and make sure the emittance accuracy. Even so, the dark current noise still cannot be avoided in low bunch charge. The deep learning method is advanced to image processing and has potential application in this field to reduce the noise effectively and give a more accuracy emittance of the beam. The applications of deep learning in slit-scan images processing can be classified into two classes: images classification network and regression network. | ||
|
|
Poster TUPP01 [1.315 MB] | |
| Export • | reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml) | |