Nowadays, in order to guarantee the security in passenger and goods railway transport, fixed systems located in rails are used to measure axles, wheels and brake discs temperatures during train circulation so that abnormally high temperatures as a result of a malfunction can be detected. Measurement systems in this kind of application may be affected by different uncertainty sources, characteristic of infrared temperature measurements, which limit the accuracy of the estimated measurement. Uncertainty sources are specially important in these applications due to: (1) Extremely variable emissivity as a result of stain or different paints used on the surfaces. (2) Difficult evaluation of the environment's radiation as measurements are made outdoors. (3) Alarm temperatures are only about 40 degrees Celsius to 80 degrees Celsius above the environment temperature. The paper analyses the effects of these uncertainties. The results show that, in order to get the minimum uncertainty peaks in the estimated temperature, the proper duty waveband is 3 - 5 micrometer. They also show that, with a proper choice of the wavelength, the uncertainty due to solar radiation remains masked by the uncertainty due to the lens emissivity.