Abstract: Aiming at the difficulty in identifying defects caused by excessive noise and insufficient contrast in casting images, a method for identifying defects in castings based on ensemble learning is proposed in this paper. At first, grayscale transformation, bilateral filtering and adaptive image segmentation are employed to preprocess the casting image. Then, the HOG (Histogram of Oriented Gradients) feature, moment invariant feature, and LBP (Local Binary Pattern) texture feature are extracted to construct the full-information feature set. Meanwhile, the SVM-RFE (Support Vector Machine-Recursive Feature Elimination) algorithm is utilized to select sensitive features. In the end, the Adaboost-RF (Adaptive Boosting-Random Forest) method is used to recognize the casting defect. The results of the comparison experiment show that this method can effectively extract the sensitive features from the full-information feature set. Moreover, it has better classification performance and generalization ability than other classifiers.