Aiming at the problems of single feature, less data and low robustness of current detection methods for the fish feeding behavior, a detection method was proposed based on the multi-feature fusion and the machine learning, by using image processing technology to extract the color, shape and texture features of the fish feeding images, which were processed by normalization and feature fusion. The fish feeding behavior is checked by constructing a 3-layer BP neural network. The results show that compared with SVM and KNN, the BP neural network has the best effect, and the accuracy can reach 97.1%. Compared with the traditional method based on the single texture feature, the accuracy is improved by 4.1% while guaranteeing timeliness and enhancing robustness.