Abstract: In order to facilitate intelligent precise nitrogen fertilizer management, a model of lettuce leaves’ nitrogen content is constructed. In this article, the lettuce samples of several nitrogen levels were cultivated. At Rosette stage, color images of lettuce leaves with every nitrogen level were collected and preprocessed and the texture features and the color features were extracted. Through the correlation analysis, principal component characteristics were extracted and image feature vectors were constructed after being screened and optimized. The regression equations of image feature vector and lettuce leaf nitrogen content were constructed by the principal component regression analysis method and the multiple linear regression method respectively. Based on the same test samples, prediction error rates of two expression models were computed. Results showed that the average error ratio of the principal component regression expression model is 9.30% and the one of multiple linear regression expression is 12.66%. The root mean square errors (RMSEP) of PCR model was 0.4577 and the RMSEP of MLR model was 0.6284. It is also shown that the prediction result of the principal component regression expression model is better than the one of the latter and it can be applied to the nondestructive testing intuitive expression model of the nitrogen content of lettuce leaf. This study provides a basis or way to fertilize and manage nitrogen fertilizer precisely for lettuce or other crops.