A new near-infrared monolithically integrated multichannel spectral sensor (MCSS) serves as a key component of Spectral-sensing Internet of Things (SSIOT). A wavelength calibration method for the MCSS was proposed. The method is based on the relative spectral response and includes specific principles of wavelength allocation. Several pixels comprised a channel with the calibrated wavelength. Experimental tests were conducted in the calibrated MCSS and other two commercial spectrometers. The compared spectrum results demonstrated calibration accuracy of 1.8 nm on average, with performance as good as a commercial spectrometer equipped with similar FWHM and sampling intervals, which can meet the application demand.