The invention relates to a median filtering detection method based on a PCA network, and belongs to the information safety and digit image information processing technical field; the method comprises model building and target detection parts; the model building part comprises the following steps: 1, selecting an image and extracting materials; 2, filtering extracted material median so as to obtain the filtered material; 3, using the materials outputted by step 1and step 2 as a training set, and calculating a median filtering residual error; 4, building the PCA network; 5, training a support vector machine so as to obtain a trained model. The target detection part comprises the following steps: A, cutting a to-be-detected image into small size images; B, inputting small size images outputted by step A into the built model, determining whether median filtering treatment is finished or not according to the model output, and deciding whether to jump to step C; C, marking the image block with the median filtering treatment in A, and jumping to B. The method is simple in structure, short in training time, less in artificial setting parameters, suitable for various image sets, can be conveniently realized on a FPGA, thus greatly improving the operation speed.