Even going out and counting is imprecise James - unless you are omnipotent you can not ensure that you did not miss people simply by not being in the same place at the same time.
For instance - the three homeless in that cafe at the corner - are they rough sleeping or customers eeking out a cup of tea?
What matters in terms of are things getting better or worse is period comaparisons using the same method. One can assume that the innaccuracies in the method can be cancelled out as a constant.
What are you suggesting, stand on a bridge and ask all rough sleepers to put their hands up and then have a head count?
Meanwhile, suffering the argument over method, more people are going to spend a night without a roof over their heads tonight - isn't that the real issue?