Discriminant Analysis of Different Species of Meat and Bone Meal Based on Microscopic Image Processing

Abstract:

Discriminant analysis of different species of meat and bone meal (MBM) can provide necessary technical support for feed safety. A total of 20 MBM samples, including five porcine, five poultry, five bovine and five ovine samples, were chosen. All the samples were of reliable sources. Bone fragment samples were prepared using standard methods and their microscopic images were obtained by biological microscope imaging system. Five lacunas’ characterization data were extracted from each image by Matlab software including the area, perimeter, major axis and minor axis of each lacuna. Results showed that all the specific characterization data of lacunas fit the normal distribution and significant difference was found between different species. Partial least squares discriminant analysis (PLS-DA) results showed that it is feasible to classify mammalian MBM and poultry MBM based on the area and perimeter values of lacuna, while it was hard to discriminant different species of MBM samples based on the major axis and minor axis. Results of separate validation proved that successful discrimination of mammalian and poultry MBM could be performed based on the separate area, perimeter and pair combination values of bone fragments; the correct discriminant rate of established model were all 0.93, while result based on the pair combination values was better than that of separate parameter values. However, it was difficult to further discriminant ruminant and porcine MBM samples by this technique.