We developed a magnetic nanoparticle (MNP) imaging system that uses the harmonic signals of MNPs on the basis of their nonlinear magnetization curves. By measuring the harmonic signals, the interference of the excitation field can be significantly decreased, so the detection sensitivity can be improved. First, we obtained a contour map of the signal field from the MNPs, from which we estimated their positions. We could detect 100 μg of MNPs located at z=30 mm below the pickup coil with a signal-to-noise ratio of 4. Next, we analyzed the contour map using a mathematical technique called singular value decomposition (SVD) in order to improve the spatial resolution of the estimated position. Using the SVD method, we converted the field map to the MNP distribution. We could clearly distinguish two MNP samples separated by x=20 mm and located at Δz=20 mm. This simple detection system will be useful for medical applications such as sentinel lymph node detection.