In order to solve fish classification and identification problems based on acoustic scattering data, a data fusion method based on SVM posterior probability was deduced, and a multi azimuth acoustic scattering data cooperative fusion fish classification method based on support vector machine (SVM) was proposed. Firstly, the wavelet packets coefficients singular value feature, temporal centroid feature and discrete cosine transform coefficients feature using multi azimuth acoustic scattering data were extracted, which reflected acoustic scattering characteristics of fish from different aspects. Secondly, the SVM classifiers made the decisions for features of each azimuth and the results were expressed in the form of posterior probability, each azimuth decision probability was used to weight the decisions of other azimuth simultaneously. Finally, the classification results were the ultimate output. Three kinds of fish were selected as the research objects and the classification accuracy (more than 92%) was presented based on the cooperative fusion method under the conditions of different numbers of azimuth. The processing results of experimental data indicated that the overall classification accuracy showed an increasing trend with the increase of number of azimuth. To examine the performance of classification further, large carp samples and small carp samples were used as training and testing samples mutually. The classification accuracy showed a increasing trend with the increase of number of azimuth in both cases, which reached more than 90% ultimately. The multi azimuth acoustic scattering data cooperative fusion method based on SVM can improve the correct classification ratios effectively.