基于YOLO v4的松材线虫病变色木自动检测

Automatic detection of pine wood nematode disease discolored trees based on YOLO v4

【目的】对受松材线虫病影响的树木进行快速、高效和精确的检测。【方法】利用深度学习技术中的YOLO v4(you only look once version 4)目标检测模型,对高分辨率影像中的松材线虫病变色木进行检测,并与SSD(single shot multibox detector)模型进行对比。【结果】YOLO v4模型的检测精度较高,精确度(P)为0.961 3,召回率(R)为0.764 9,F1分数为0.851 9。【结论】YOLO v4可准确地识别和定位松材线虫病变色木,且精确度比SSD高。

【Objective】 The rapid,efficient and accurate detection of pine wood nematode disease was realized.【Method】 The object detection model such as YOLO v4(you only look once version 4) in deep learning technology was used to effectively detect pine wood nematode disease in high-resolution images.It was compared with the model of SSD(single shot multibox detector).【Result】 YOLO v4 model had higher detection precision,the detection precision(P) of the model was 0.961 3,the recall rate(R) was 0.764 9,and the F1 score was 0.851 9.【Conclusion】YOLO v4 could accurately identify and locate pine nematode discolored wood with higher detection precision than SSD.