Photometry omit large sources

CHANGES.md

@@ -1,3 +1,17 @@
+1.17.0 (2023-04-24)
+-------------------
+- We now omit sources in the photometry stage that have an area larger than 1000 pixels as they lead to long 
+  processing times and are almost invariably spurious.
+
+1.16.1 (2023-04-23)
+-------------------
+- Correction to aperture photometry. We were incorrectly using the radius instead of the diameter
+
+1.16.0 (2023-04-18)
+-------------------
+- Calibration frames are now associated with output data products rather than frames
+  so that we have more than one calibration data product produced per frame.
+
 1.15.2 (2023-04-12)
 -------------------
 - Fix to fpacking data when the image data array is None

banzai/lco.py

@@ -141,7 +141,7 @@ def write(self, runtime_context):
             if runtime_context.post_to_archive:
                 archived_image_info = file_utils.post_to_ingester(data_product.file_buffer, self,
                                                                   data_product.filename, meta=data_product.meta)
-                self.frame_id = archived_image_info.get('frameid')
+                data_product.frame_id = archived_image_info.get('frameid')
 
             if not runtime_context.no_file_cache:
                 os.makedirs(self.get_output_directory(runtime_context), exist_ok=True)
@@ -170,7 +170,7 @@ def to_db_record(self, output_product):
                              'instrument_id': self.instrument.id,
                              'is_master': self.is_master,
                              'is_bad': self.is_bad,
-                             'frameid': self.frame_id,
+                             'frameid': output_product.frame_id,
                              'attributes': {}}
         for attribute in self.grouping_criteria:
             record_attributes['attributes'][attribute] = str(getattr(self, attribute))

banzai/photometry.py

@@ -102,6 +102,10 @@ def do_stage(self, image):
             # Do an initial source detection
             segmentation_map = detect_sources(convolved_data, self.threshold, npixels=self.min_area)
 
+            # We now remove any sources with an area > 1000 pixels because they are almost invariably spurious
+            segmentation_map.remove_labels(segmentation_map.labels[segmentation_map.areas > 1000])
+            segmentation_map.relabel_consecutive(1)
+
             logger.info('Deblending sources', image=image)
             # Note that nlevels here is DEBLEND_NTHRESH in source extractor which is 32 by default
             deblended_seg_map = deblend_sources(convolved_data, segmentation_map,
@@ -125,9 +129,9 @@ def do_stage(self, image):
                              'xy': catalog.covar_sigxy.value,
                              'background': catalog.background_mean})
 
-            for r in range(1, 7):
-                radius_arcsec = r / image.pixel_scale
-                sources[f'fluxaper{r}'], sources[f'fluxerr{r}'] = catalog.circular_photometry(radius_arcsec)
+            for d in range(1, 7):
+                radius_arcsec = d / image.pixel_scale / 2.0
+                sources[f'fluxaper{d}'], sources[f'fluxerr{d}'] = catalog.circular_photometry(radius_arcsec)
 
             for r in [0.25, 0.5, 0.75]:
                 sources['fluxrad' + f'{r:.2f}'.lstrip("0.")] = catalog.fluxfrac_radius(r)

banzai/tests/test_frames.py

@@ -63,9 +63,8 @@ def test_frame_to_db_record():
                                   'CONFMODE': 'full_frame'}, name='SCI')]
     test_frame = LCOCalibrationFrame(hdu_list=hdu_list, file_path='/foo/bar')
     test_frame.is_bad = False
-    test_frame.frame_id = 1234
     test_frame.instrument = MagicMock(id=7)
-    mock_data_product = MagicMock(filename='test.fits.fz', filepath='/path/to/test/test.fits.fz')
+    mock_data_product = MagicMock(filename='test.fits.fz', filepath='/path/to/test/test.fits.fz', frame_id=1234)
     db_record = test_frame.to_db_record(mock_data_product)
 
     assert type(db_record) == CalibrationImage