基于YOLO的玉米全膜双垄沟种床苗带识别方法

Identification Method of Seeding Zone in Corn Whole Plastic-film Mulching on Double Ridges Based on YOLO

玉米苗期是玉米全膜双垄沟种床苗带中耕管理与追肥的关键时期,为进一步提升玉米全膜双垄沟智能机械化生产水平,基于深度学习的目标检测YOLOv3-SPP网络模型与YOLOx-Darknet53网络模型开展了玉米全膜双垄沟种床苗带识别方法研究。研究结果表明,YOLOx-Darknet53网络模型与YOLOv3-SPP网络模型的mAP分别为95.56%和94.72%,前者比后者高0.84%,使用100幅图像测得前者检测速度FPS为44.1,而后者为19.1,前者检测速度明显快于后者。研究方法与结论可为玉米全膜双垄沟种床苗带识别方法与玉米全膜双垄沟智能农机装备的有效结合提供理论支撑。

The maize seedling stage is the key period for the tillage management and fertilizer application in the planting belt of corn whole plastic-film mulching on double ridges.In order to further improve the intelligent mechanized production level of corn whole plastic-film mulching on double ridges, this paper carried out research on the identification method of seeding zone in corn whole plastic-film on double ridges based on deep learning target detection Yolov3-SPP network model and Yolox-Darknet53 network model.The research results show that the mAPs of YOLOx-Darknet53 network model and the YOLOv3-SPP network modelwere 95.56% and 94.72%,respectively, and the former was 0.84 percent point higher than the latter, and the detection speed FPS of the former was 44.1 and the latter was 19. 1 using 100 images,and the former was significantly faster than the latter in detection speed. The research method and conclusions can provide theoretical support for the effective combination of seed bed seeding zone identification method and intelligent agricultural machinery equipment in corn whole plastic-film mulching on double ridges.