基于Yolov5s的蚕茧分类检测研究

An Analysis of Yolov5s-based Cocoon Categorization Detection

蚕桑是中国的传统文化。在现有的蚕茧加工过程中对于人工的依赖比例仍旧很高。随着人工智能不断发展,基于深度学习的蚕茧的分割、定位及分类应用水平不断提升。针对目前选茧技术人工依懒性强的特点,设计了一种基于深度学习的蚕茧检测系统。以Yolov5s模型为基础,对于上茧、双宫茧、黄斑茧、薄皮茧等进行有效识别。通过蚕茧模型匹配,将原先设定好的检测标签优先的结果显示出来,解决原先设备检测只能检测一面的弊端。可以更好地减少人工选茧的误判情况,提高检测的正确率和检测效率。

Sericulture is a traditional culture in China.The proportion of reliance on manual labor in the existing cocoon processing is still high.With the continuous development of artificial intelligence,the application level of segmentation,localization,and classification of silkworm cocoons based on deep learning is constantly improving.Aiming at the characteristics of the current cocoon selection technology with strong artificial laziness,a cocoon detection system based on deep learning is designed.Based on the Yolov5s model,effective identification is carried out for the upper cocoon,double palace cocoon,yellow spot cocoon,and thin skin cocoon.Through cocoon model matching,the results of the originally set detection label priority are displayed to solve the original equipment detection,which can only detect the disadvantages of one side.Can better reduce the manual selection of cocoon misjudgment,improve the detection of the correct rate,and increase detection efficiency.