Thermal images have many applications in agronomy, including crop water stress status assessment. Nowadays, the miniaturization of thermal cameras allows installing them onboard the Unmanned Aerial Vehicles (UAV), but this miniaturization leads to some difficulties: the miniaturized thermal cameras have no temperature control system of their sensor. The instability of the miniaturized camera makes a high drift in the acquisition of temperature data so that acquired thermal images don't fit the real temperature of the studying object, so data have to be continuously corrected. We need to have stable reference on field in order to compute the actual temperature value. In this article we present a method for radiometric correction of UAV remote sensed thermal images. We have implemented a device in order to retrieve ground temperature measurements. This device is composed with four targets (cold, hot, dry soil, wet soil) which measured continuously the target temperature thanks to IR120 (Campbell ®) radio-thermometer. A meteorological station is included in this ground system and acquires air temperature and moisture, solar radiation, wind speed and direction every 10 seconds. The images are radiometrically corrected by linear regression from on ground thermal data collected. Corrected images have been compared with mean canopy surface temperature of a sample of 10 trees measured with radio-thermometers. The results showed a good link between data from on ground radio-thermometer and data from thermal camera after radiometric correction. We can conclude that images obtained by this method are of sufficient quality to be used in vegetation water stress studies.