In order to increase the detecting accuracy and stability of the fertility of hatching eggs during early hatching period, information of vision and acoustic sensors were fused in the sensor level on the fifth day of incubation, and two different artificial neural networks were chosen to establish models for detecting fertility of hatching eggs. Results showed that the sensor fusion model by LVQ artificial neural network obtained a higher discriminating accuracy and stability than the sensor fusion model of BP artificial neural network. The discriminating accuracy of hatching eggs during the early hatching period was up to 92% and 88% by computer vision technique and impact excitation technique, respectively. However, the discriminating accuracy reached 98% by sensor fusion model, which implied that the sensor fusion was feasible for detecting fertility of hatching eggs during early hatching period.%为提高判别种蛋孵化前期受精的准确性和稳定性,将视觉和声学2种传感器信息在孵化第5天进行特征层融合,采用2种人工神经网络构建种蛋孵化前期受精性判断的融合模型.研究表明:采用LVQ神经网络判别模型的准确率和稳定性,优于BP神经网络.单独利用计算机视觉技术和敲击振动技术对鸭蛋孵化早期受精情况的判别准确率为92％和88％,而将2种传感器信息进行融合构建的模型的准确率可达98％,说明传感器信息融合技术在判断鸭蛋孵化前期受精性方面是可行的.
展开▼
