摘要
由于早腐区域在水果表面难以用肉眼直接观测,检测由真菌感染引起的柑橘早期腐烂成为柑橘产业面临的主要难题。为此开发一种基于可见LED结构光源的结构光反射成像系统,并提出结合螺旋相位变换(SPT)的解调方法。在解调原始图像过程中,使用3种不同的解调方法(三步相移法、二相SPT和一相SPT)。通过解调相移模式图像,可以得到直流分量(DC)和交流分量(AC)图像。与DC图像相比,AC图像和比值(RT)图像(AC/DC)可以清晰地显示早腐区域。在实际检测中,果梗会被误判为腐烂区域。针对柑橘早腐区域和果梗之间存在易误判的问题,提出一种基于YOLO v7的改进模型(YOLO v7-CA),即在YOLO v7模型骨干网络中引入坐标注意力机制(Coordinate attention,CA),从而提高模型对早腐区域的关注度。YOLO v7-CA模型结合二相螺旋相位变换(SPT)下的RT图像对柑橘早腐区域和果梗取得较好的检测结果。采用不同年份的早腐柑橘进行试验,检测精度达到98.5%。该研究为柑橘早期腐果检测提供了参考。
Early detection of decay caused by fungal infection in citrus fruit is a major challenge for the citrus industry,as the early decayed area is almost invisible on the surface of fruit.A detection system for structural illumination reflectance imaging was constructed based on light-emitting diode(LED)lamp and an effective methodology was proposed combining with spiral phase transform(SPT)algorithm for the early detection of decayed citrus.Three strategies(i.e.three-phase-shifting method,2-phase SPT and 1-phase SPT)were used to demodulate original patterned images.The direct component(DC)and alternating component(AC)images were recovered by demodulating phase-shifting patterned images.Compared with the DC image,the early decayed area can be clearly displayed in the AC image and the ratio(i.e.AC/DC)image.YOLO v7-CA model was proposed to address the issue of misjudgment between citrus decayed area and fruit stem.In actual testing,fruit stem can be misidentified as decayed area.The coordinate attention(CA)module was introduced in the backbone network of YOLO v7 model,in order to increase the attention to the decayed area.The improved YOLO v7-CA model combined with RT images under 2-phase SPT achieved good detection results for early decayed area and fruit stem in citrus.Early decayed mandarins from different years were tested with a detection accuracy of 98.5%.The research result can provide a reference for the detection of early decayed citrus.
作者
蔡仲磊
石瑞瑶
张君仪
蔡乐天
张译之
张亚伟
李江波
CAI Zhonglei;SHI Ruiyao;ZHANG Junyi;CAI Letian;ZHANG Yizhi;ZHANG Yawei;LI Jiangbo(Intelligent Equipment Research Center,Beijing Academy of Agricultural and Forestry Sciences,Beijing 100097,China;College of Mechanical and Electrical Engineering,Shihezi University,Shihezi 832003,China;College of Engineering,China Agricultural University,Beijing 100083,China)
出处
《农业机械学报》
北大核心
2025年第8期479-486,495,共9页
Transactions of the Chinese Society for Agricultural Machinery
基金
北京市农林科学院杰出科学家培育专项(JKZX202405)
北京市农林科学院改革与发展项目(装备中心-2024)
中央高校基本科研业务费专项资金项目(2024TC144)。
关键词
柑橘
早期腐烂
结构光反射成像
YOLO
v7
CA
citrus
early decay
structured-illumination reflecance imaging
YOLO v7
coordinate attention
