Moisture regain of cotton has posed a great challenge on the quality of products in the whole processing links of cotton industry. Therefore, it is necessary to accurately measure the moisture regain of cotton. Most measurements of cotton moisture are contact detection in the field of online cotton processing, particularly requiring additional auxiliary devices. However, the current contact mode in the moisture measurement cannot realize fast online detection at a highly demanding speed. In this study, a non-contact measurement was proposed using near-infrared spectroscopy, and an experimental test was also conducted to investigate the influencing factors and feasibility in the measurement of cotton moisture regain. The specific sample collection was set to ensure that the moisture regain of cotton samples covered the moisture regain under natural conditions. Six cotton samples were prepared with the moisture regain levels of 6%, 8%, 10%, 12%, 14%, and 16%. Three measurements were selected to compare, including the most widely-used resistance-based, the newly proposed infrared-based, and the standard oven measurement. Two influencing factors were first explored in the infrared measurement, such as the detection distance and sample density. The reason was that different deformation capabilities were found in the cotton fibers with various moisture content when the infrared measurement was performed. The cotton samples under the same moisture regain level were regarded as a batch of cotton samples, where the dispersion degree of infrared measured values in the same batch of cotton samples was obtained under the various measuring distance and sample density. The experimental results show that there was great variation in the measuring distance, but the sample density had little effect on the measurement. The data range under different measurement distances was within 0.6%, with a standard deviation of 0.134%. The data range under different densities was about 0.5%, with the standard deviation of 0.15%, under the condition that there was no gap on the surface of visually inspected cotton sample, and no light leakage. The measuring error met the accuracy requirements of online processing for moisture regain. A feasibility verification was also performed for the infrared measurement. In correlation analysis, the data regression was proposed to achieve a more accurate measurement of cotton moisture regain, where the correlation coefficient of the calibration model was 0.978. A host computer software was also designed using Modbus communication protocol, where the calibration model was utilized to measure the moisture regain after calibration by oven data. It was found that the infrared measurement model using fitting calibration more accurately realized the online monitoring the cotton moisture regain. In cotton samples with the moisture regain of 5% to 15%, the measuring error less than 10%, with a minimum of 0 and a maximum of 0.12%, indicating a better prediction. Between 10% and 15%, the moisture regain increased slightly when the measurement error was low, where the minimum was 0.10%, and the maximum was 0.46%. The accuracy of calibration value in cotton moisture regain was reduced when the moisture regain level was close to 15%. The measurement error was expected to be less than 0.5%in the level of moisture regain. Therefore, the proposed near-infrared-based non-contact system was feasible for measuring cotton moisture regain, indicating high efficiency and sufficient real-time performance in cotton production.