摘要
针对当前马铃薯种薯制备环节存在自动化水平过低问题,构建了基于YoloV8n的芽眼检测与种薯切割决策模型。使用工业级CCD摄像头从两个不同视角拍摄分级后的马铃薯种薯,获取待切割种薯的最大视场图像。经过训练后的YoloV8n目标检测算法,可准确识别种薯图像中的芽眼位置信息。通过透视变换将种薯图像和芽眼位置信息转换到鸟瞰图,使用图割算法提取种薯的轮廓。切割决策模型计算每个切割槽与轮廓交线形成的面积,制定切割种薯的最佳策略,以确保每个切块上均含有芽眼且各切块的投影面积尽可能相同,从而指引切割装置对种薯进行切割。试验结果表明:制备的种薯薯块合格率达到93.19%,较盲切提高5.41%,种薯切块重量的标准差降低6.87 g,满足了种薯制备的质量要求。
In response to the problem of low automation level in the current potato seed potato preparation process,a decision model for bud detection and seed potato cutting based on YoloV8n was constructed.Use an industrial grade CCD camera to capture the graded potato seeds from two different perspectives and obtain the maximum field of view image of the potato seeds to be cut.After training,the YoloV8n object detection algorithm can accurately identify the position information of bud eyes in seed potato images.Convert the image of the seed potato and the position information of the bud eye into a bird's-eye view through perspective transformation,and extract the outline of the seed potato using graph cutting algorithm.The cutting decision model calculates the area formed by the intersection of each cutting groove and the contour,and formulates the optimal strategy for cutting seed potatoes to ensure that each cutting block contains bud eyes and the projection area of each cutting block is as similar as possible,thereby guiding the cutting device to cut the seed potatoes.The experimental results show that the qualified rate of the prepared seed potato chunks reaches 93.19%,which is 5.41% higher than blind cutting.The standard deviation of the weight of the seed potato chunks is reduced by 6.87 g.Satisfied the quality requirements for seed potato preparation.
作者
赵文帅
冯全
孙步功
孙伟
ZHAO Wen-shuai;FENG Quan;SUN Bu-gong;SUN Wei(College of Mechanical and Electrical Engineering,Gansu Agricultural University,Lanzhou Gansu 730070,China)
出处
《林业机械与木工设备》
2024年第5期76-82,共7页
Forestry Machinery & Woodworking Equipment
基金
甘肃省科技重大专项(22ZD6NA046)
国家自然科学基金地区科学基金项目(32160421)。
