This paper presents a P-RIF (Perceptually Robust Invariant Feature) based object recognition system. The P-RIF can encode visual
information by visual part decomposition and shape encoder. Especially, the P-RIF approximates arbitrary curves into line segments (1st
derivatives) and curvatures (2nd derivatives). These information is encoded in the form of histograms. Object recognition system can be
performed by nearest neighbor classifier-based voting and geometrical verification. The object recognition tests validate the feasibility of
the proposed object recognition system.