meas_extensions_trailedSources is a measurement plugin that characterizes fast-moving, trailed sources. Mainly, it aims to measure the length, angle, and flux. The current implementation involves two main algorithms:
- Naive: Utilizes previously measured second moments to estimate the length and angle of a trail, and compute the end points. This model is inexpensive to compute, but under-estimates length.
- Vereš: A chi-squared minimization using the analytic model from Vereš et al. 2012 – an axisymmetric Gaussian convolved with a line. Uses the Naive measurements and centroid as the initial parameter guess. This model assumes the flux in a given pixel is the value of the model at the center of that pixel. It estimates true length better, but is more computationally expensive than the Naive model. It also adds a flux, and centroid measurement.
Further development is ongoing. A correction to the Naive measurement is being investigated as well a third algorithm to provide a full forward model for high signal-to-noise trails. The overall goal is to provide a 'smart' plugin that runs each algorithm sequentially, base on signal-to-noise. Ie. For low signal-to-noise, the precision of the Naive measurement will be suitable. If the signal-to-noise is higher, then the next algorithm will be run, and so on.
The package is currently available at lsst/meas_extensions_trailedSources
lsst-dm/meas_extensions_trailedSources. It includes unit tests for each measurement algorithm and is able to be setup along side lsst_distrib. Only dependencies outside of lsst_distrib are scipy.optimize and numpy.