DispBDTs issueshttps://gitlab.desy.de/gernot.maier/dispBDTs/-/issues2021-04-14T08:16:35+02:00https://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/8Optimise Cherenkov image fitting with 2D-normal distribution2021-04-14T08:16:35+02:00Gernot MaierOptimise Cherenkov image fitting with 2D-normal distributionCurrent fit function yields parameter errors in x, y coordinates without covariance matrix.
For better usage in disp, change fit function to:
- fit in rotated coordinate system (provides error in cen_x, cen_y, image length, and phi wou...Current fit function yields parameter errors in x, y coordinates without covariance matrix.
For better usage in disp, change fit function to:
- fit in rotated coordinate system (provides error in cen_x, cen_y, image length, and phi would be most useful)
Should also allow for easier fitting of asymmetric Gaussian to large images.Gernot MaierGernot Maierhttps://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/7Loss of events in Image fitting with 2D-normal distribution2021-04-14T07:57:23+02:00Gernot MaierLoss of events in Image fitting with 2D-normal distributionDetermine number of events loss in image fitting with 2D-normal distribution (LL fitting) due to unsuccessful fits (look at FITSTAT as function of energy).
Related to #6
Add to https://gitlab.desy.de/gernot.maier/dispBDTs/-/blob/master...Determine number of events loss in image fitting with 2D-normal distribution (LL fitting) due to unsuccessful fits (look at FITSTAT as function of energy).
Related to #6
Add to https://gitlab.desy.de/gernot.maier/dispBDTs/-/blob/master/document/ImageFitting.mdGernot MaierGernot Maierhttps://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/6Image fitting with 2D-normal distribution2021-04-14T07:56:52+02:00Gernot MaierImage fitting with 2D-normal distributionInvestigate performance of image fitting with 2D-normal distribution.
Possible benefits are:
- calculation of errors on image parameters (e.g., on centroid position or image orientation)
- less dependent on camera pixelisation and gaps...Investigate performance of image fitting with 2D-normal distribution.
Possible benefits are:
- calculation of errors on image parameters (e.g., on centroid position or image orientation)
- less dependent on camera pixelisation and gaps between the sensitive area in the camera
Drawbacks:
- much slower than standard image calculation
- loss of events due to unsuccessful fits?
Development steps:
- [x] integrate time gradient into image fitting for sufficiently large images with time gradient measured
- [ ] use image parameter errors for stereo reconstruction
- [ ] geometrical method using image errors as weight
- [ ] dispBDT:
- [ ] add errors to improve prediction of disp
- [ ] add errors to improve prediction of dispError
- [ ] calculate in all cases the expected error on the stereo position from error propagationhttps://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/5Optimise and re-evaluate minimum angle between image lines2021-03-25T13:55:04+01:00Gernot MaierOptimise and re-evaluate minimum angle between image linesCurrently set to 15 degs (with '-minangle_stereo_reconstruction=15' when running *mscw_energy*.
- is this the optimum?
- implicit cut on core position and therefore on the effective area?Currently set to 15 degs (with '-minangle_stereo_reconstruction=15' when running *mscw_energy*.
- is this the optimum?
- implicit cut on core position and therefore on the effective area?https://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/4Determine true parameters for centroid and image lines2021-03-25T13:53:47+01:00Gernot MaierDetermine true parameters for centroid and image linesUse pe entries in simulations to determine best possible reconstruction and 'true' parameters of centroids and image lines.Use pe entries in simulations to determine best possible reconstruction and 'true' parameters of centroids and image lines.https://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/3Bias in dispBDT prediction2021-03-25T13:52:56+01:00Gernot MaierBias in dispBDT predictionSingle disp are often at a single side of the true position, see image examples in https://gitlab.desy.de/gernot.maier/dispBDTs/-/blob/master/document/StereoAnalysis.md#example-events-of-stereo-reconstruction
Why?Single disp are often at a single side of the true position, see image examples in https://gitlab.desy.de/gernot.maier/dispBDTs/-/blob/master/document/StereoAnalysis.md#example-events-of-stereo-reconstruction
Why?https://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/2Test dispBDTError prediction2021-03-25T13:52:11+01:00Gernot MaierTest dispBDTError predictiondispBDTError is used as weight for the stereo reconstruction. Test prediction.dispBDTError is used as weight for the stereo reconstruction. Test prediction.https://gitlab.desy.de/gernot.maier/dispBDTs/-/issues/1outliers of single dispBDT2021-03-25T13:51:23+01:00Gernot Maieroutliers of single dispBDTUnderstand many 'overflow' events for single DispBDT, see events images here: https://gitlab.desy.de/gernot.maier/dispBDTs/-/blob/master/document/StereoAnalysis.md#example-events-of-stereo-reconstructionUnderstand many 'overflow' events for single DispBDT, see events images here: https://gitlab.desy.de/gernot.maier/dispBDTs/-/blob/master/document/StereoAnalysis.md#example-events-of-stereo-reconstruction