Localizing hot spots in Poisson radiation data matrices: nonnegative tensor factorization and phase congruency

Abstract Detecting and delineating hot spots in data from radiation sensors is required applications ranging monitoring large geospatial areas to imaging small objects close proximity. This paper describes a computational method for localizing potential matrices of independent Poisson where, numerical terms, spot cluster locally higher sample mean values (higher intensity) embedded lower (lower background intensity). Two algorithms are computed sequentially 3D array 2D gross counts: (1) nonnegative tensor factorization the maximize likelihood (2) phase congruency pertinent matrices. The indicators closed contours these illustrated simulated datasets, including visualization contours. may be useful other which there counts, provided that distribution fits dataset.