Abstract: Changes of leaf thickness can indicate the variations of plant growth state, in order to carry out real-time, live, non-destructive testing of leaf thickness, the study took cotton leaves as the research object. The correlation between plant leaf spectrum and thickness was studied on 84 couples of data of cotton, with DPS and Origin statistical softwares. Studies showed that the correlation between reflectance and leaf thickness showed significantly positive relationship in two visible light regions of 350-369 nm and 664-689 nm, and significantly negative relationship in two infrared regions of 917-1 884 nm and 2 048-2 380 nm. In general, the correlation degree between reflectance and leaf thickness in infrared light was higher than that in visible light. Red edge indices showed low correlation with leaf thickness, however 24 figure indices had significant correlation with leaf thickness, and the area of absorbtion with the center of 980nm had the highest correlation degree with correlation coefficient 0.848. Three models about leaf thickness were set and tested with reflectance, plant index and spectral figure index. Among these models, the highest relative error was 7.4%, and the RMSE was 0.051 mm. It is feasible to measure alive leaf thickness untouchably with hyper spectrum.