A multistage recognition approach is advanced for poorly classified clusters. According to this approach, if a sample is related to a cluster that is common for several samples, then further object recognition (within that cluster) is possible. Such two-stage recognition procedure is based, at each stage, on the fuzzy logic concept and enables one to perform practically complete recognition of all the samples under consideration.