The invention discloses a multi-target tracking model based on a composite recurrent neural network system. The method is based on a multi-target tracking network model of a composite recurrent neuralnetwork system formed by RNN and GRU, a boundary frame is adopted to label a tracked object, and the composite recurrent neural network system comprises a target detection unit based on RNN, a tracking management unit based on GRU and a data association unit based on GRU. The RNN and GRU-based multi-target tracking network model provided by the invention is an end-to-end multi-target tracking architecture; compared with a traditional method, the multi-target tracking model has the advantages that the parameter adjustment process and requirements are simplified, and the method has a good tracking effect under the complex environment conditions of different illumination conditions, frequent shielding, mutual influence between targets, similar appearances and the like.