We propose an incremental null Foley-Sammon transform method.The method can upd ate the re-ID model rapidly which is computed previously when obtaining new samples.Firstly,the within-class divergence matrices and the overall divergence matrices are updated respectively in the null space Foley-Sa mmon transform by newly added samples.There is no need to recalculate the model by using the entire samples including old and new ones.Then carry out the orthogonal decomposition for updated divergence matrices to get projection matrix in the null space.Extensive experiments are carried out on two person re-ID benchmarks including VIPeR and PRID2011. Three different ways of increment learning are applied to the experiments and th e results show that the model obtained by incremental learning achieves comparab le accuracy to the batch learning counterparts.Moreover,the proposed m ethod can be applied to the real-time adaptive incremental updating under dynam ic scenes.It is not necessary to repeat the calculation of the original samples,the updating speed of model is fa st,and the precision grows steadily with the increase of the data.Moreover,the proposed method is expected to be applied for real-time surveillance systems.