Fluorescence targeted imaging in vivo has proven useful in tumor recognition and drug delivery. In the process of in vivo imaging, however, a high autofluorescence background could mask the signals from the fluorescent probes. Herein, a high contrast upconversion luminescence (UCL) imaging protocol was developed for targeted imaging of tumors based on RGD-labeled upconversion nanophosphors (UCNPs) as luminescent labels. Confocal Z-scan imaging of tissue slices revealed that UCL imaging showed no autofluorescence signal even at high penetration depth (approximately 600 microm). More importantly, region of interest (ROI) analysis of the UCL signal in vivo showed that UCL imaging achieved a high signal-to-noise ratio (approximately 24) between the tumor and the background. These results demonstrate that the UCL imaging technique appears particularly suited for applications in tracking and labeling components of complex biological systems.