Identification of the hatching egg before the incubation based on hyperspectral imaging and GA-BP network
Zhihui Zhu1,2, Ting Liu1,2, Lirong Xiong1,2, Meihu Ma1,3
1National R&D Center for Egg Processing, Huazhong Agricultural University, Wuhan, Hubei, China, 430070
2College of Engineering, Huazhong Agricultural University, Wuhan, Hubei, China, 430070
3College of Food Science and Technology, Huazhong Agricultural University, Wuhan, Hubei, China, 430070
The removal of the unfertilized egg from the hatching egg before the incubation could improve the efficiency of incubation. The identification of the unfertilized and fertilized eggs by hyperspectral imaging technology combined with GA-BP algorithm was proposed. The comparative analysis for the unfertilized and fertilized eggs was implemented by different pretreatment and principal component. In order to improve the performance of BP neural network, GA algorithm was used to optimize the network parameters. The application of GA-BP network was established the qualitative detection model. The results of the study showed that the MSC + SD pretreatment method was the most suitable for the model. The determination coefficient was 0.95, which indicated the optimized network model had a good generalization ability and high prediction precision with unfertilized eggs accuracy being 93%, fertilized eggs accuracy being 94%, the overall accuracy being 93.5%. The results indicated that the method of non-destructive identification for fertilized and unfertilized eggs based on hyperspectral imaging technology combined with GA-BP algorithm was feasible.