부산대학교 도서관

A novel robust package-height estimation and parcel-size classification scheme is proposed in this work. A pertinent prototype system consisting of a millimeter-wave radar, a shipment track, and a stepping motor has been built for proof of concept. Three new schemes using the range-profile data produced by the mmWave radar, namely, the detection based-on cumulative distribution (DBCD) scheme, the detection based-on peak-value clustering (DBPC) scheme, and the DBCD–DBPC hybrid scheme are introduced here. For benchmark evaluation of the proposed new system, two practical scenarios, namely, the undisturbed scenario (no people is close to the system) and the disturbed scenario (some people is close to the system), are tested. This new system can overcome the serious drawback of the existing camera-assisted systems, which have to rely on sufficient lighting conditions, and operate normally even in complete darkness. Meanwhile, the proposed new system can accommodate parcels/packages in arbitrary orientations, which cannot be allowed by the current camera-assisted systems, during the parcel-size measurements. According to the test experiments, the proposed DBCD–DBPC hybrid scheme can reach up to a very high average classification accuracy of 96.4% for both undisturbed and disturbed scenarios. The proposed novel package-height estimation and parcel-size classification technique in this work can be very useful for intelligent inventory management and smart logistics in the future.