A thermopile array and stereo vision fusion system for small-tube temperature sensing
Dexin Hou, Tiancheng Xie, Yiming Lv, Ji Mi, Jiong Ding, Dongfang HuThis paper presents a non-contact temperature measurement scheme for small-diameter tubes by fusing a low-resolution thermopile array (TPA) with visible-light stereo vision. Stereo vision locates the tube in three dimensions. Then, this geometry is projected onto the TPA imaging plane to constrain the radiation integration model. A spherical calibration target with an internal heat source is proposed, achieving a sub-pixel accuracy of 1/28 of a pixel via symmetry-based active aiming. Experiments on 3 and 2 mm heated tubes show that the proposed method reduces the mean absolute percentage error from 28% and 36% to 3.4% and 4.3%, respectively. Furthermore, the system resolves emissivity-induced circumferential temperature non-uniformities with a rotational resolution of 5°. These results demonstrate that multi-sensor fusion effectively overcomes the size-of-source effect in low-cost infrared sensing.