This article presents a method for real-time simultaneous measurements of the temperature and soot volume fraction distribution of volatile matter flames, forming during combustion of biomass pellets. This method uses flame radiation spectra, captured by a spectrometer, and images, captured by a digital camera, as inputs during combustion of three different types of biomass pellets (pine wood, rice straw, and corn straw). The pellets were ignited and burned in the post-combustion zone of methane gas at 1300 K. The radiation spectrum of the flame in the 500−800 nm wavelength band was selected for spectral analysis, combined with a spectral emissivity model based on polynomial fitting. The flame emissivity was measured at the response wavelengths of a dual band-pass filter (centered at 615 and 517 nm) fitted on the camera, and the resulting emissivity ratio showed the degree of departure from the gray radiation model. The measured emissivity ratio approached unity as the combustion intensity increased. The emissivity ratio of the spectrometer measurement was used to correct the flame temperature measurement obtained by the image method. Upon correction, the maximum relative error of the image temperature measurement was determined to be 2.7% by comparison to thermocouple readings. The experimental results showed that the maximum flame temperatures of pine wood, rice straw, and corn straw pellets were remarkably similar at 1876 ± 2 K. Moreover, the flame burnout durations and the peak soot volume fractions were both positively correlated with the volatile content of biomass pellets; the value of the latter was 1.19 ± 3 ppm. The experimental results measured in this article can provide data for models of the soot generation mechanism of biomass pellet combustion in industrial furnaces.