基于多信息融合的疫苗制备中鸡蛋胚体分拣系统

Automatic Sorting System of Egg Embryo in Biological Vaccines Production Based on Multi-information Fusion

采用多信息融合方法研究疫苗制备中鸡蛋胚体状态的识别与分拣技术。在研究不同状态鸡蛋胚体图像特征、温度衰减和透光度变化情况的基础上,得出活胚图像中血管多、粗且呈放射状,弱胚图像中血管少、细且断裂,死胚图像内部均匀无血管,污染胚图像内部有明显黑块特征;鸡蛋胚体从37.8℃的孵化箱中取出置于25℃室温环境,活胚、弱胚、污染胚、死胚的温度衰减速度依次增大;活胚的透光度随孵化时间增加而逐渐降低,其他胚体透光度变化相对较小。将图像、温度、透光度信息特征融合,建立BP神经网络信息融合模型对鸡蛋胚体状态进行识别。最后,从37.8℃孵化箱中抽取80枚孵化6 d的鸡蛋胚体放置于室温10 min后,采集图像、温度和透光度信息,进行试验验证。结果表明多信息融合系统的识别准确率为96.25%,比单用图像、温度和透光度传感器进行识别的准确率分别提高了6.25%、13.75%和8.75%。

Identification and sorting technology of egg embryo in biological vaccines production with the method of multi-information fusion was researched. The regularity cognition was obtained by the researches on image features, degradation of temperature and variations of transmittance of different types of egg embryo, which indicated that there were many thick radial blood vessels in living egg embryo image and few thin fractured blood vessels in weak egg embryo image, no blood vessels in dead egg embryo image and obvious black blocks features in polluted egg embryo image were found. The rates of temperature decay of living embryo, weak embryo, polluted embryo and dead embryo increased successively, and the transmittance of living embryo decreased gradually with hatching time increase while the variation of the transmittance of other egg embryo were comparatively lesser. The BP neural network model was established to do the egg embryo condition identification based on the information fusion of images, temperature and transmittance. Finally, 80 eggs were randomly selected to collect the information of images, temperature and transmittance to do the verifying experiment, the eggs were hatched six days and taken out from the 37.8℃ incubator and placed under normal room temperature for ten minutes. The result indicated that the identification accuracy of the system was 96.25% , which was increased by 6.25%, 13.75% and 8.75%, respectively, compared with that only using image,temperature or transparency sensors.