Our goal was to identify the over-densities in the multi-colour matched database, where point-like sources had been previously removed. A possible approach is to determine the position of point-like source in the SExtractor parameter space, and isolate them.
To separate extended vs. point-like sources, a first selection was performed using a mag(core) vs. mag(tot) diagnostic plot. The ratio between these two quantities represents an index of a source concentration. In Figure 12 we plot the aperture magnitude (computed within a 5 pix area on the source center, i.e. the mag(core)), vs. the total magnitude given by mag(best) for the B, V, R images of the OACDF2. Galaxies appear as extended objects with a variable mag(tot)/mag(core) ratio (cloud-like distribution in the mag(tot) vs. mag(core) diagram), while stars, the point like objects, have a constant m(tot)/m(core) ratio (linear strip in the above diagram). At the bright end, the stellar locus is easily separated from the extended object cloud, while at faint magnitudes the PSF dominates onto the real spatial distribution and the star/galaxy separation becomes more difficult.
The stellarity index given by SExtractor, the CLASS_STAR parameter, shows a strong correlation with the classification outlined above: all objects with class_star>0.8 fall on linear strip (plotted in red), while all the other sources (class_star<0.8 plotted in blue) fall in the extended region. Both star/galaxy separation criteria do become less efficient at faint magnitudes, and a different selection consistent in all bands must be adopted. We first clean the catalogue from those objects which may have an erroneously determined photometry: from Figure 12 we can determine saturation in the three filters (in this case mag(core) cannot be larger than the CCD saturation and remains constant, while mag(tot) increases). The saturated stellar objects can then be deleted with the following magnitude constraints mag(best)_B> 17; mag(best)_V > 16; mag(best)_R> 17. All objects placed in an area equivalent to five times the isophotal area of each saturated object are also removed from the catalogues. These objects may in fact suffer of contamination from the ghost images of the saturated objects resulting from internal reflections in the field corrector.
To identify the stellar objects in a consistent way in all three filters, we plot in the (V-R) vs (B-V) diagram the locus of the stellar objects from our data and the calibration frames for the Landolt Fields. From Figure 13 we can see the locus identified by the Landolt standard stars over plotted onto a sample of bona bide point like objects selected from our multi colour catalogue, with class_star >0.95 and mag(best)_V<19). Both samples identify a strip-like region in this plane (a strip of 0.4 magnitude width near the line (B-V)=1.64*(V-R), which is situated along the main sequence).
The selection criteria for the star-galaxy separation, (partially based on the SExtractor stellarity index and on a selection in the colour-colour space) was used to clean the multi-colour catalogue. We also considered only those objects brighter than the limiting magnitude in R determined from the simulated images.