Iss388 cvt cosmics #421

zieglerv · 2025-01-08T22:51:03Z

No description provided.

…here for BST clusters.

… surface.

…p to bottom. In the case of a shallow angle ray there can be 2 intersection with a passive surface in the same hemisphere, hence the need to sort by y.

…he track if the next iteration chi^2 is worse.

… intersection point when there are 2 intersection points of the ray with the cylinder.

Use cluster information to determine y-intersection with the surface and hemisphere for active surfaces. Compute the y-intersection with the surface to be used for sorting in the kalman fitter.

raffaelladevita

it looks to me as if the main fix is in the calculation of the trajectory point that doesn't assume anymore the cylinders representing the detector surfaces are centered at 0,0, but I'm unsure about some of the changes to the measurement class

raffaelladevita · 2025-01-09T15:09:28Z

common-tools/clas-tracking/src/main/java/org/jlab/clas/tracking/kalmanfilter/Surface.java

@@ -42,6 +42,7 @@ public class Surface implements Comparable<Surface> {
    public double swimAccuracy;
    public boolean passive = false;
    public double hemisphere = 1;
+    public double rayYinterc = 9999;


let's make it inf or nan

Also, to be more general, it could make sense to add a Point3D to store the full information of the trajectory point instead of only y

raffaelladevita · 2025-01-09T15:13:51Z

...tools/clas-tracking/src/main/java/org/jlab/clas/tracking/kalmanfilter/helical/StateVecs.java

-                sv.z = inters.z();
-                sv.path = inters.distance(st);
-            }
+//                mv.surface.toLocal().apply(st);


let's remove the commented-out lines...they will still be in the history

raffaelladevita · 2025-01-09T15:14:46Z

...tools/clas-tracking/src/main/java/org/jlab/clas/tracking/kalmanfilter/helical/StateVecs.java

+                int ints = mv.surface.cylinder.intersection(toPln, inters);
+                if(ints<1) return false;
+
+                    sv.x = inters.get(0).x()  ;


is the first intersection always the right one?

raffaelladevita · 2025-01-09T15:18:18Z

...-tools/clas-tracking/src/main/java/org/jlab/clas/tracking/kalmanfilter/straight/KFitter.java

                this.setTrajectory(sv, mv);
-                setFitFailed = false;
+                //System.out.println(finalStateVec.toString());
+                //setFitFailed = false;


I assume setFitFailed is not really used elsewhere but it may be better to still set it to false
(and remove the commented text)

raffaelladevita · 2025-01-09T15:23:03Z

...ools/clas-tracking/src/main/java/org/jlab/clas/tracking/kalmanfilter/straight/StateVecs.java

+                } else {
+                    int index =0;
+                    if(ints>1) {
+                        double dh0 = Math.abs(inters.get(0).y()-mv.surface.rayYinterc);


Shouldn't the right intersection be the one with the largest y among the ones with y<rayInterc?

raffaelladevita · 2025-01-09T15:29:59Z

reconstruction/cvt/src/main/java/org/jlab/rec/cvt/measurement/Measurements.java

                }
-                surfaces.add(surf);
+
+                sMap.put(surf.getIndex(), surf);


is this changing something?

raffaelladevita · 2025-01-09T15:36:17Z

reconstruction/cvt/src/main/java/org/jlab/rec/cvt/services/CVTEngine.java

these changes are not really necessary

raffaelladevita · 2025-01-09T17:52:16Z

reconstruction/cvt/src/main/java/org/jlab/rec/cvt/measurement/Measurements.java

+                return hemisphereSign == (int) Math.signum(intersections.get(0).y());
+            case 2:
+                // Check if either of the two intersections match the hemisphere
+                return hemisphereSign == (int) Math.signum(intersections.get(0).y()) ||


for small y, isn't possible the sign doesn't match because of resolution?

raffaelladevita · 2025-01-09T18:08:58Z

reconstruction/cvt/src/main/java/org/jlab/rec/cvt/measurement/Measurements.java

-        if(surface.cylinder==null)
-            return 0;
-        List<Point3D> trajs = new ArrayList<>();
+    private double handleCylinderIntersection(int h, Cylindrical3D cylinder, Line3D line) {


h could be renamed as hemisphere for clarity

raffaelladevita · 2025-01-09T18:12:05Z

reconstruction/cvt/src/main/java/org/jlab/rec/cvt/measurement/Measurements.java

+        double Y = this.getY(cluster, ray, surface);
+        if(Y == Double.POSITIVE_INFINITY) return 0;
+        surface.rayYinterc=Y;
+        return (int) Math.signum(Y);


if misalignments are large in y, the sign of the y coordinate of the intersection point doesn't really correspond to the hemisphere

Change isCrossed method to use the value of the hemisphere which is calculated as the sign of the y value of the intersection of the ray with the surface if it crosses it or 0 if it does not.

Fix intersection of ray with cylinder for tracks from the target. Intersection initialized as positive infinity.

raffaelladevita · 2025-01-10T20:10:53Z

reconstruction/cvt/src/main/java/org/jlab/rec/cvt/measurement/Measurements.java

@@ -283,11 +297,14 @@ private void addClusters(DetectorType type, List<Surface> clusters) {
 //    }

    private List<Surface> getClusterSurfaces(DetectorType type, List<Cluster> clusters, Ray ray) {
-
-        List<Surface> surfaces = this.getClusterSurfaces(type, clusters, ray);        


it seems the initial source of the problem is this line, a call to the method from the same method. This was introduced with #351.

Fixing this and reverting to assigning the hemisphere based on the cluster.y instead of with the ray intersection seems to solve the problem.

The ray intersection with the surface doesn't seem to work consistently because when two intersection points are found (for cylindrical surfaces) the first one in the List is not necessarily the closest to the ray origin. This is the case also with intersectionRay

ziegler added 10 commits December 12, 2024 10:41

Bug fix in setting the measurements for cosmics. Added missing hemisp…

56cd809

…here for BST clusters.

fix in propagation to surface for shifted cylinders.

f1f1f02

Fix measurements sorting by Y intersection of the cosmic ray with the…

6bf24ee

… surface.

removed unused import

a1d3d37

Initialized chi^2

6aaf945

Added y intersect for rays. This is used to sort the surfaces from to…

861caf3

…p to bottom. In the case of a shallow angle ray there can be 2 intersection with a passive surface in the same hemisphere, hence the need to sort by y.

For cosmic fitting, save the output for the best chi^2; do not kill t…

4fb1e22

…he track if the next iteration chi^2 is worse.

Use global fit y-intersection with the surface to assign fitted track…

4528930

… intersection point when there are 2 intersection points of the ray with the cylinder.

Assign correct hemisphere for cosmics.

92952b4

Use cluster information to determine y-intersection with the surface and hemisphere for active surfaces. Compute the y-intersection with the surface to be used for sorting in the kalman fitter.

Cosmic debug statement

bb51011

zieglerv requested a review from raffaelladevita January 8, 2025 22:51

zieglerv linked an issue Jan 8, 2025 that may be closed by this pull request

Lots of log messages when reco CVT cosmic using 11.0.4 #388

Open

raffaelladevita requested changes Jan 9, 2025

View reviewed changes

ziegler added 2 commits January 9, 2025 19:48

Change intersection of ray with surface from y-coordinate to Point3D.

508ee56

Change isCrossed method to use the value of the hemisphere which is calculated as the sign of the y value of the intersection of the ray with the surface if it crosses it or 0 if it does not.

Remove commented out old code.

d569978

Fix intersection of ray with cylinder for tracks from the target. Intersection initialized as positive infinity.

raffaelladevita reviewed Jan 10, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Iss388 cvt cosmics #421

Iss388 cvt cosmics #421

zieglerv commented Jan 8, 2025

raffaelladevita left a comment

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 9, 2025

raffaelladevita Jan 10, 2025

Iss388 cvt cosmics #421

Are you sure you want to change the base?

Iss388 cvt cosmics #421

Conversation

zieglerv commented Jan 8, 2025

raffaelladevita left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment