Radiometric calibration is an important step in the high-precision quantitative application of infrared remote sensing camera, and in order to obtain uniform images, infrared cameras need to perform relative radiometric calibration. To obtain the true radiation value of the target, it is necessary to carry out absolute radiometric calibration to establish the quantitative relationship between the radiance of the entrance pupil and the DN value of the detector output, and then invert the true radiance value of the detection signal. In recent years, the requirements for radiometric calibration are getting higher and higher, so it is very necessary to deeply analyze and suppress the main influencing factors of infrared camera radiometric calibration accuracy to improve the radiometric calibration accuracy.There are many factors affecting the calibration accuracy of infrared system, and it can be seen from the analysis of the radiometric calibration accuracy of infrared camera that in addition to the non-uniformity of blackbody temperature, the nonlinearity of system response is an important source of radiometric calibration error, and improving the response linearity is an important means to improve the radiometric calibration accuracy of infrared camera. Least squares is the most commonly used method of data fitting and considers that all data are equally weighted to find the best matching function by minimizing the sum of squares of the error. The weighted least squares used by the Changchun optical machine of the Chinese Academy of Sciences reduces the linearity deviation caused by the temperature fluctuation of the calibration blackbody. For systematic errors such as the nonlinearity of the low-end response of the detector, how to use a better fitting algorithm to compromise has not been publicly studied.The nonlinearity of response is an important source of radiometric calibration error of infrared camera, and a terrestrial and on-orbit radiometric calibration method based on weighted least squares fitting is proposed to improve the radiometric calibration accuracy. Firstly, the shortcomings of the existing linearity calculation methods are analyzed, and a new linearity evaluation method based on deviation/measured value is proposed, which uses the low-end temperature deviation of least squares fitting to be an order of magnitude larger than that of the high-end, and it is necessary to use weighted fitting to improve the response linearity of the system. Based on the analysis of the temperature and radiance fitting functions of different spectra, a ground and on-orbit radiometric calibration method based on the reciprocal power of the radiance weighted least squares linear fitting was proposed. The fitting effect was better when the weights were confirmed to be powers n=1 and 2, and then the weighted least squares fitting for outer blackbody linear calibration equation, inner blackbody linear calibration equation and inner to outer blackbody radiance conversion model with power n=1 were established. For the high-precision two-point calibration of the blackbody in orbit, a radiance conversion model based on weighted least squares fitting between the inner and outer blackbodies was established. The data processing results show that the low-end outer blackbody inversion deviation decreases from 1.63 K to 0.52 K, and the inner blackbody inversion temperature deviation decreases from 0.83 K to 0.39 K after weighted linear fitting, and the system response linearity is significantly improved.In view of the degradation of the in-orbit blackbody emissivity, a correction method for the in-orbit blackbody radiance calibration of cameras based on stellar calibration was proposed, and the conversion models of outer blackbody radiance, inner blackbody radiance and stellar radiance were established. After the camera is in orbit and stable, the blackbody calibration and stellar calibration can be carried out, and the relative relationship calibration equation can be established. Since most of the brightness measurements of stars are carried out in the bands of interest in astronomical research, these bands generally do not coincide with the earth observation bands of remote sensing satellites, and the radiance of stars is not equal to the blackbody on the star, so there may be inconsistencies in the calibration equation, but the stellar stability is very good, which can be used for long-term stability monitoring and correction of the blackbody on the satellite.