Canopy temperature is one of promising signals for evaluating crop water status. The infrared thermal imager can provide real-time temperature distributions over larger areas with high spatial resolution. The main factors (the observation orientation, angle and distance) controlling the accuracy of measuring canopy temperature with the infrared thermal imaging were investigated in a cotton field. Moreover, the correlation relationships between the crop water stress index (CWSI), which was observed using different methods, and soil water content (SWC), leaf water potential (LWP), and stomatal conductance (g(s)) of cotton in different water treatments were analyzed. Results indicated that the CWSI, which was measured in the opposite direction of the sun with the observation angle of 45°, was in good correlation with LWP, g(s) and SWC, indicating it was a suitable observing method of canopy temperature. The canopy temperature gradually decreased with the increasing observation distance, so the calibration was necessary for long-distance measurement. By analyzing the relationship between the temperature at the dry/wet reference surface and the canopy temperature, we developed a suitable and simplified model of CWSI for cotton in the North China Plain.