That's not what the graph shows at all. The graph shows that there is no statistical correlation between income and votes, except that lower incomes slightly preferred Clinton. The problem with this data is that it doesn't appear to be weighted by electoral college result, so we don't get to actually see what the impact on the election result is. And, even worse, the survey appears to include the question "Did you vote for Clinton or Trump" yet doesn't include that answer in the graph, so we don't get to compare the relative results. I tried to extract that result by using the male/female data assuming 50/50 representation, but ended up in a tie, and we don't get to know the bias of participants vs non-participants.

Per that chart, Republicans in general tend to win every income bracket above $50k. Look at the gains and losses from 2012: Trump made massive gains with lower income brackets, and lost the Republican grip on the middle and upper middle classes.

So yes, it's reasonable to say that the cohort that Trump made the largest gains with (i.e., lower income brackets) is his "voter base."