The work aims to realize automatic sorting of crankshaft bearing caps packaging production line, improve production efficiency and reduce production cost of enterprises. A classification method of crankshaft bearing caps contour based on machine vision was proposed. Firstly, the rows and columns of pre-processed crankshaft bearing cap image were extracted at equal intervals, and the number of target pixels in each row and column was calculated. Then, the number of target pixels in two columns with symmetrical image center was summed. The above extracted features were sequentially composed into feature vectors that were invariant to the positive and negative placement of bearing caps. Then the normalized feature vectors were reduced by principal component analysis. Finally, the support vector machine was used to classify the feature vectors. The experimental results showed that the classification accuracy of contour of parts can reach 99.8% by extracting the first five principal components from the feature vector of the sample set. The method described in this paper can realize the accurate classification of bearing cap parts.