90%。 A new method of crack detection in eggs was proposed with multi-level wavelet transform and texture analysis technology. First, G gray level image of all egg images were decomposed into approximation and detail sub-images at various levels by wavelet transform. Then, the feature vector which was composed of wavelet texture energy features, and the gray-level co-occurrence matrix features of the detail sub-images were analyzed and computed. Finally, with the most appropriate and effective eight parameters as inputs, the best BP neural network (8 input nodes, 20 hidden nodes, 2 output nodes)was employed to detect egg crack and classify eggs. The results of experiment proved that the correct discerning rate to detect eggs without crack and eggs with linear crack, meshy crack, point crack is respectively 95%, 90%, 95% and 80%, and the average correct rate to detect crack in eggs is 90%.