The egg crack detection is one of the most important parts in the non-destructive detection of egg. The non-destructive testing parameters related the egg’s external conventional cracks, the external minimal cracks and internal cracks were extracted through image analysis, impact excitation frequency analysis and electronic nose pattern recognition respectively. A non-destructive testing criterion was designed according to the eggs cracks based on experimental research. The criterion was used to build the ‘one-to-many support vector discriminated model in eggs crack combined with experiment data. Performance parameters of the model (the model fitting degree was 0.9735, the convergence error was less than 0.0001) and the confirmatory test (the discriminated accuracy in the five cracks status was more than 90%) indicates that the model has credible structure and good discriminated ability.