<p>This paper proposes a visual tracking system which utilizes images acquired from thermal and color cameras to robustly and continuously monitor a target body temperature in real time. In the proposed system, the thermal camera measures the human body temperature, and combining the color camera image, we can effectively monitors fever in a cluttered environment or crowded group of people. The general tracking results using color and thermal cameras only validates the target hypotheses. In this work, we design the sampling multiple importance resampling methodology with fuzzy inferences to efficiently generate and verify the hypotheses. The hypotheses, which are like the target visual and temperature distribution model, are selected and evaluated through the sparse appearance representations to achieve real-time recognition of heavily occluded objects. By means of this resampling methodology, target characteristics in the visual and temperature distribution are adaptively fusing with the fuzzy inferences. We also develop an update strategy for the target model to improve tracking robustness by fuzzy evaluation of occlusion rate and environment similarity. Finally, the proposed approaches are validated by experiments across several scenarios related to monitoring of human body temperature.</p>