Context :

Aims : We analyze near-infrared UKIDSS observations of a sample of 8325 objects taken from a catalog of intrinsically red sources in the Galactic plane selected in the Spitzer -GLIMPSE survey .
Given the differences in angular resolution ( factor > 2 better in UKIDSS ) , our aim is to investigate whether there are multiple UKIDSS sources that might all contribute to the GLIMPSE flux , or there is only one dominant UKIDSS counterpart .
We then study possible corrections to estimates of the star formation rate ( SFR ) based on counts of GLIMPSE young stellar objects ( YSOs ) .
This represents an exploratory work towards the construction of a hierarchical YSO catalog .

Methods : After performing PSF fitting photometry in the UKIDSS data , we implemented a technique to automatically recognize the dominant UKIDSS sources by evaluating their match with the spectral energy distribution ( SED ) of the associated GLIMPSE red sources .
This is a generic method which could be robustly applied for matching SEDs across gaps at other wavelengths .

Results : We found that most ( 87.0 \pm 1.6 \% ) of the candidate YSOs from the GLIMPSE red source catalog have only one dominant UKIDSS counterpart which matches the mid-infrared SED ( fainter associated UKIDSS sources might still be present ) .
Though at first sight this could seem surprising , given that YSOs are typically in clustered environments , we argue that within the mass range covered by the GLIMPSE YSO candidates ( intermediate to high masses ) , clustering with objects with comparable mass is unlikely at the GLIMPSE resolution .
Indeed , by performing simple clustering experiments based on a population synthesis model of Galactic YSOs , we found that although \sim 60 \% of the GLIMPSE YSO enclose at least two UKIDSS sources , in general only one dominates the flux .

Conclusions : No significant corrections are needed for estimates of the SFR of the Milky Way based on the assumption that the GLIMPSE YSOs are individual objects .
However , we found that unresolved binaries in GLIMPSE objects ( a few of them could be resolved at the UKIDSS resolution ) have a non-negligible effect , and would increase the SFR estimate by a factor \sim 1.2 –1.3 .