The invention relates to a relative identification method based on a feature fusion neural network, and belongs to the field of image processing. The method comprises the following steps: step 1, preprocessing a face image in a data set to obtain a preprocessed face image; 2, constructing a positive sample pair and a negative sample pair; Step 3, generating corresponding positive sample pair tagsand negative sample pair tags; 4, respectively segmenting the positive sample pair, the negative sample pair, the positive sample pair label and the negative sample pair label to generate a training set and a test set; Step 5, generating training data and test data input by the network; 6, inputting the training data generated in the step 5 into a neural network in batches for T rounds of iterative training, storing the trained network parameters, and outputting the trained neural network; and step 7, testing is carried out through the trained neural network. According to the method, complementarity and difference between different features are learned, and the robustness of the features is improved.