Abstract: Foreign bodies in quick-frozen dumplings can seriously endanger consumers’ physical and mental health. However, the traditional metal detector can only detect metals and the results cannot be visualized intuitively. Motivated by this problem, in this study, we developed a method of recognizing foreign bodies dumplings with based on LeNet convolutional neural network (CNN) model, which can detect X-ray images of dumplings containing five foreign bodies, including metal balls, wires, screws, stones, and glasses. Firstly, X-ray images of dumplings without foreign bodies and those with foreign bodies were obtained and processed by denoising and contrast stretching transformation. Secondly, the CNN model was optimized, trained and validated using batch normalization (BN) method and Softmax linear regression classifier with ReLu as the activation function and Max-Pooling as the down-sampling method. The trained network model was applied to test 100 dumpling images without foreign bodies and 100 ones with foreign bodies. The experimental results showed that the recognition method could accurately identify the foreign body-containing dumplings with a recognition rate of up to 99.78%. Finally, the traditional texture features of local binary pattern (LBP), histogram of oriented gradient (HOG) and Gabor were extracted and used as the feature vectors for identifying dumpling images with and without foreign objects by using back propagation (BP) neural network, support vector machine (SVM), K-nearest neighbor (KNN) classifier, AdaBoost classifier, Naive Bayes classifier and decision tree classifier. The results were compared with those obtained with the network model, which verified the superiority of the algorithm and the effectiveness of the extracted features. This research provides a new approach for the detection of foreign bodies in food so as to ensure food safety.

Key words: quick-frozen dumplings, X-ray, foreign body recognition, convolutional neural networks, feature vector, food safety