Based on the simulation data of the edge radiation intensity of the oxygen A-band, the atmospheric temperature inversion of the adjacent space height (60~110 km) is studied and analyzed. Firstly, based on the forward model, the limb radiation intensity under noiseless and noise-added conditions are simulated respectively to derive the temperature profile, and the inversion results of all spectral lines were analyzed. The inversion results were analyzed, and the variation law of the weight function of each line in the oxygen A-band was determined as the basis for the judgment of temperature observation. Temperature influences the radiation intensity through line strength and self-absorption, and the influence of temperature on them is opposite. The weight function is used to express the influence of temperature on radiation intensity, and the difference of inversion results can be obtained from the weight function. In an ideal situation, when the effect of temperature on self-absorption is less than the influence on line strength, the weight function does not flip, the temperature inversion accuracy is higher, and the average inversion deviation is 4.1 K; when the effect of temperature on self-absorption is greater than the influence on line strength (mainly located at the height of 60~80 km), the weight function reverses because self-absorption reduces the sensitivity of radiation intensity to temperature, and the average inversion deviation reaches 34.9 K. In addition, the strong line has stronger anti-interference ability than the weak line in the presence of noise, and the inversion precision is higher, and it is more suitable for temperature inversion. In the practical observations, the line strength is also another line selection. Based on the weak line, this paper further analyzes the effect of radiation intensity on the inversion accuracy by improving the signal-to-noise ratio. The results show that the stronger the radiation, the larger the signal-to-noise ratio, the higher the accuracy of temperature inversion, and vice versa. When the line strength of the airglow spectrum reaches 10^-26, it can be used for temperature inversion above 80 km and obtain better inversion results, with the inversion accuracy <5 K.