Cognitive load is a way to evaluate the physiological and psychological needs of pilots during flight missions, and it is an urgent scientific problem. This requires determining the ECG characteristics of pilots under different cognitive load conditions. In this research, a flight simulation experiment and NASA-TLX scale were used to induce and measure pilots’ cognitive load state respectively. These data were collected and recorded by PSYLAB software. The collected ECG data were preprocessed using threshold algorithm wavelet denoising. Then, using the person correlation coefficient method to analyze the characteristic variables of these data, and determined the difference in the pilot's ECG characteristics under different cognitive load states. Results show that ECG can identify the load status of pilots in controlling the aircraft. Time domain and frequency domain characteristics such as the average value of R-R interval, average heart rate, root mean square of the difference between adjacent R-R intervals, percent of NN intervals >50 ms (%), low frequency, and low/high frequency have significant differences under different load conditions, which could be utilized to identify the pilot's cognitive load status during aircraft control. This study provides a theoretical basis for the development of humanized flight autopilot and early warning systems.