With the tightening of on-board diagnostics requirements, accuracy of sensors is essential to monitor the efficiency and ensure a proper control of the after-treatment systems. Temperature sensors are commonly used in the exhaust line at the diesel oxidation catalyst-inlet of turbocharged diesel engines for control and diagnosis of the after-treatment system. In particular, negative temperature constant sensors are used for this purpose. However, due to the necessary robustness that on-board sensors must fulfil, thermal inertia causes significant differences during engine transient operating conditions in temperature measurements. A Kalman filter is proposed in this paper for the on-line dynamic estimation of the catalyst-inlet temperature, which combines a slow but accurate measurement of the on-board temperature sensor with a fast but drifted temperature model. A fast research-grade thermocouple is used as reference of the actual exhaust gas temperature as well as a frequency analysis is performed in order to calibrate the model and analyse results of the signal reconstruction. Results of the algorithm are then successfully proved in experimental transient tests and typical European approval test cycles This research has been partially financed by the Spanish Ministerio de Economia y Competitividad, through project TRA2013-40853-R PRELIMIT.