Speaker
Description
We present a background-referenced method for localizing the spatial response associated with a wavelength-shifting (WLS) fiber using hierarchical DBSCAN clustering, without imposing a predefined fiber position.
The methodology was evaluated using experimental data acquired from a detector setup combining a WLS fiber embedded in a shielded scintillator bar, a pixelated photosensor, FPGA-based acquisition, and an independent hodoscope-based validation chain. A no-fiber background map is used as a pixel-wise instrumental baseline, while the fiber-coupled measurements are analyzed after background subtraction to identify the resulting spatial excess response. The datasets analyzed in this work were acquired under ambient laboratory conditions, without active temperature control.
The background-subtracted spatial excess map is processed in physical detector coordinates using the pixel pitch as the spatial metric. A hierarchical DBSCAN pipeline first identifies local microclusters and subsequently merges their centroids into larger superclusters. A parameter scan is used to identify configurations in which the reconstructed response remains spatially stable before independent structures merge artificially. The final localization is obtained by maximizing the spatial support enclosed within the physical fiber radius, yielding a reconstructed response centroid at (3.082 mm, 0.645 mm), corresponding to approximately (93.67, 8.83) in sensor column-row coordinates.
The reconstructed point is interpreted as the centroid of the measured spatial response rather than the geometrical center of the fiber, and the response footprint is not directly identified with the physical fiber diameter. An independent hodoscope-based acquisition provides temporal information on photon arrival and an additional cross-check of the reconstructed fiber-response position.
These results demonstrate that background subtraction combined with hierarchical spatial clustering can recover a localized WLS-fiber-associated response under non-uniform detector background conditions without imposing an external fiber position. The approach is detector-architecture agnostic and can be applied more broadly to pixelated photodetector systems where spatially varying background complicates direct localization.