Abstract: The crayfish industry, centered on Procambarus clarkii, is rapidly expanding but faces challenges due to insufficient automation. Traditionally, manual visual inspection to assess the size and integrity of crayfish during breeding and processing is labor-intensive and susceptible to errors. This paper introduces an advanced algorithm using the YOLOv8n framework to intelligently recognize and grade crayfish by accurately identifying the body, tail, and claws of Procambarus clarkii. The proposed method innovates by replacing the conventional loss function with the CIOU (Complete Intersection over Union) and substituting it with the MPDIOU (Modified Perfect Dark Intersection over Union).A novel scale factor, 'ratio,' is integrated to regulate the size of the auxiliary bounding box in the loss calculation. This modification, when synergized with the MPDIoU loss function, significantly bolsters the precision and efficiency of bounding box regression. Consequently, this leads to the accurate identification of the crayfish's distinct body parts, which is a critical step towards automating the grading process. Empirical evaluation showed significant improvements in recognition rates. The integration of Inner-MPDIoU into the YOLOv8n model enhanced the mean Average Precision (mAP) from 83.7% to 90.8% at IOU thresholds from 0.5 to 0.95.This advancement not only streamlines the grading process but also paves the way for more nuanced and automated sorting systems in the crayfish industry. The study's findings underscore the efficacy of the proposed algorithmic model in accurately identifying key components of Procambarus clarkii. This research contributes to the broader objective of achieving intelligent and precise grading within the crayfish sector, potentially revolutionizing traditional methods and bolstering industry efficiency. The implications extend beyond mere automation, offering a foundation for future research into intelligent systems that can be tailored to the specific needs of the crayfish industry.

Key words: Procambarus clarkii, image recognition, YOLOv8, MPDIoU

摘要： 克氏原螯虾（Procambarus clarkii）产业发展迅速，然而仍面临智能化水平偏低的问题，在养殖过程及加工产业中主要通过人工肉眼观察克氏原螯虾规格及完整性并作出相关判断。为解决克氏原螯虾的智能识别问题，提出基于YOLOv8n识别克氏原螯虾的虾体、虾尾及螯足的算法。通过将原有损失函数CIOU替换为MPDIOU，并引入尺度因子ratio控制辅助边框的尺度大小用于计算损失，与MPDIoU损失函数相结合，提高边界框回归的准确性和效率，实现对克氏原螯虾虾体、虾尾及螯足的精准识别，为研究其分级智能化提供思路。结果显示，在YOLOv8n模型中加入Inner-MPDIoU的算法训练结果相比原有的CIoU损失函数识别率有所提高，mAP0.5-0.95从83.7%达到了90.8%。研究表明，该算法模型有助于对克氏原螯虾的主要部位实现精准识别，对研究克氏原螯虾的智能化精准分级具有重要推动作用。

关键词: 克氏原螯虾, 图像识别, YOLOv8, MPDIOU