Fast acquisition of nutritional information of rapeseeds is important for rapeseed breeding programs, evaluation of soil nutrient conditions and even fertilization recommendations. Fourier transform mid-infrared photoacoustic spectroscopy (FTIR-PAS) was employed to determine nitrogen, phosphorus and potassium content in rapeseeds. Calibration models were developed using both partial least squares (PLS) and partial least squares combined with direct orthogonal signal correction (DOSC-PLS). According to the values of RPD (ratio of prediction to deviation), the PLS models for nitrogen and phosphorus were acceptable, while the PLS model for potassium needed to be improved. By contrast, DOSC-PLS models obtained the better predictive accuracy with RPD values of 2.54, 2.10 and 1.94 for nitrogen, phosphorus and potassium, respectively. This work demonstrates the good performance of FTIR-PAS for rapid and non-destructive quantification of nutritional information in rapeseeds.