UCHIME and UCHIME2 are algorithms for detecting chimeric sequences. See the UCHIME and UCHIME2 papers for details. Note that the accuracy results reported in the original UCHIME paper are highly over-optimistic due to unrelaistic benchmark design. This is explained in the UCHIME2 paper.
Stand-alone UCHIME should not be used in OTU clustering or denoising pipelines because the cluster_otus and unoise3 commands have more effective chimera filters integrated into the algorithms. You will probably get much higher error rates, both false positives and false negatives, if you use UCHIME as a stand-alone filtering step.