An automated segmentation method of goat trunk with machine vision and machine learning was proposed. 396 images of goat trunk were acquired randomly in a goat slaughtering plant by using two industrial cameras. 24 sets of feature parameters were extracted by image processing. The characteristic parameters included the center (O), the centroid (Q) and the four vertex coordinates (ABCD) of the minimum circumscribed rectangle corresponding to the four parts of the body. 16 groups of characters were selected by significance test. The data was preprocessed with abnormal value detection and normalization methods. The hierarchical sampling method is used to divide the training and test set to 7∶3. The models such as Lasso, Ridge, SVR and GBDT in auto-sklearn are selected as individual learning cell. The automatic integrated learning algorithm is designed and constructed with Bayesian optimization method. When Lasso, SVR and GBDT are integrated with a weight of 0.30∶0.25∶0.45, the model predicts best. Finally, the MSE and R² scored 7.93 and 0.961, on test set. The experiment was carried out on a multi-joint robot. The error is 3.4 mm and the theoretical reached 413 units per hour,which increased by 37.9%. The results indicate that the method is effective and feasible.