Abstract: A dual Kalman filter algorithm is proposed to overcome the defects of thermal imager efficiently, and information fusion data of multichannel thermal radiation of reagents, ambient temperature and heat source temperature are combined to estimate the real temperature of reagents surface by the proposed algorithm. Combining the iterated extended Kalman filter (IEKF) that has a fast convergence speed of nonlinear estimation and a linear Kalman filter (KF) that has the advantages of real-time capability, the algorithm can filter and control temperature in multichannel in real time. Test results show that the infrared thermal imager can reduce error from 2℃ to 0.3℃ after filtering. PCR (polymerase chain reaction) product test result of the system is more satisfactory and constant temperature setting is more reasonable.