To solve problems in the existing JPEG steganalysis schemes, such as high redundancy in features and failure to make good use of the complementarity among them, this study proposes a JPEG steganalysis approach based on feature fusion by the principal component analysis (PCA) and analysis of the complementarity among features. The study fuses complementary features to reflect the statistical differences between cover and stego signals in the round, isolates redundant components by PCA, and finally achieves the goal of improving accuracy. Experimental results show that in various datasets and embedding rates, this scheme provides more accuracy than the main JPEG steganalysis schemes against steganographic methods of high concealment (e.g. F5, MME and PQ) and greatly reduces the time cost of the existing fusion methods on feature level.