In the field experiment of wheat breeding, an important measurement index is the plant height of the plot population. To solve the problem of low accuracy of wheat plant height measurement based on UAV remote sensing, two methods were proposed, including a nearest neighbor correction method (NNCM) and a spectral index correction method (SICM). NNCM based on the true value of manual measurement, the height information of the community group was obtained, the elevation correction was carried out in combination with the ridge, and then the accurate plant height of the community was obtained by sliding correction according to the true value of the neighbor. SICM of multi-spectral + RGB data fusion, by calculating vegetation index and performing index optimization, an accurate inversion model of plant height-vegetation index was constructed. The test results showed that the relative root mean square error (RMSE100) of the traditional UAV crop height measurement method in the six periods with ground truth were 11.15%, 59.44%, 11.76%, 12.31%, 8.05% and 59.76%; the RMSE100 of NNCM were 7.17%, 8.18%, 5.70%, 5.62%, 5.65% and 7.74%; the RMSE100 of SICM were 7.33%, 8.17%, 6.05%, 6.15%, 6.45% and 10.50%; the NNCM and SICM kernel density distribution curves were closer to the ground truth, and the median, quartile, maximum, and minimum deviations did not exceed 0.5%. These indicated that both the proposed methods can correct the plant height traits on the grain size of breeding plots measured by UAV. The two models proposed had high accuracy and strong robustness. NNCM was suitable for the scene of random sampling of ground truth on the ground, while SICM was used for plant height detection of largescale farmland, and different methods were selected according to the using conditions.