黄花菜花蕾的精准识别与分级方法

doi:10.13304/j.nykjdb.2023.0915

摘要/Abstract

摘要：

针对大田环境下黄花菜花蕾识别背景复杂、个体过小及采摘后分级标准不统一的问题，提出了黄花菜花蕾的识别及采摘后分级方法。选取1 716幅不同光照、遮挡及模糊等复杂环境下的黄花菜花蕾图像建立数据库，在YOLOv5s模型的主干网络中引入Biformer自注意力机制对数据集进行训练，并与多种其他目标检测算法进行对比测试。在识别完成后，使用分级算法通过图像处理技术获取黄花菜花蕾的轮廓，并使用几何计算技术获得黄花菜花蕾的长度及直径，对其进行分级。结果显示，改进的YOLOv5s算法在黄花菜的大田识别中精确率、召回率、平均精确率（mAP）分别达到了94.80%、91.40%、96.60%，识别精确率显著提高，黄花菜分级算法准确率达到97.00%，满足生产实践中对黄花菜分级的要求，为黄花菜产业智能化提供可靠支持。

关键词: 黄花菜识别, Biformer注意力机制, 改进 YOLOv5s, 大田环境, 品质分级

Abstract:

Aiming at the problems of complex backgrounds， small individual size and inconsistent grading standards after harvesting of daylily buds in field environment， a recognition and post-harvest grading method for daylily buds was proposed. A database was established based on 1 716 images of daylily buds under various complex conditions such as different lighting， occlusion， and blurriness. The Biformer self-attention mechanism was introduced into the backbone network of the YOLOv5s model to train the dataset， and comparative tests were conducted against various other target detection algorithms. After recognition， a daylily bud grading algorithm was used to obtain the contours of the daylily buds through image processing technology， and geometric calculation techniques were used to measure the length and diameter of the daylily buds for grading. Experimental results showed that theimproved YOLOv5s algorithm significantly increased recognition precisionthe precision， recall rate， and mean average precision （mAP） of improved YOLOv5s algorithm of 94.80%， 91.40%， and 96.60%， respectively， in the open field recognition of daylily. The accuracy of the daylily grading algorithm reached to 97.00%， meeting the requirements for daylily grading in production practice and providing reliable support for the intelligent development of the daylily industry.

Key words: daylily recognition, biformer attention mechanism, improved YOLOv5s, open field environment, quality grading

中图分类号:

S126

袁嘉良, 连润楠, 张吴平. 黄花菜花蕾的精准识别与分级方法[J]. 中国农业科技导报, 2025, 27(5): 103-112.

Jialiang YUAN, Runnan LIAN, Wuping ZHANG. Accurate Identification and Grading Method for Daylily Flower Buds[J]. Journal of Agricultural Science and Technology, 2025, 27(5): 103-112.

图/表 13

参考文献 25

1	刘佩冶,李可昕,张超凡,等.黄花菜生物活性成分及功能研究进展[J].食品与发酵工业,2022,48(12):330-336.
	LIU P Y, LI K X, ZHANG C F, et al.. Research progress on bioactive components and functions of daylily [J]. Food Ferment. Industries, 2022,48(12): 330-336.
2	武珍珍,洪沙沙,吕虹瑞,等.黄花菜保鲜贮藏及后处理加工技术研究进展[J].食品与发酵工业, 2023,49(22):334-340.
	WU Z Z, HONG S S, LYU H R, et al.. Research progress in fresh-keeping,storage and post-processing technology of daylily [J]. Food and Ferment. Industries, 2023,49(22):334-340.
3	FU M, HE Z, ZHAO Y, et al.. Antioxidant properties and involved compounds of daylily flowers in relation to maturity [J]. Food Chem., 2009,114(4):1192-1197.
4	王卓,王健,王枭雄,等.基于改进YOLO v4的自然环境苹果轻量级检测方法[J].农业机械学报,2022,53(8):294-302.
	WANG Z, WANG J, WANG X X, et al.. Lightweight real-time apple detection method based on improved YOLO v4 [J]. Trans. Chin. Soc. Agric. Mach., 2022, 53(8): 294-302.
5	卢伟,邹明萱,施浩楠,等.基于YOLO v5-TL的褐菇采摘视觉识别-测量-定位技术[J].农业机械学报,2022,53(11):341-348.
	LU W, ZOU M X, SHI H N, et al.. Technology of visual identification-measuring-location for brown mushroom picking based on YOLO v5-TL [J]. Trans. Chin. Soc. Agric. Mach., 2022, 53(11): 341-348.
6	靳红杰,马顾彧,唐梦圆,等.复杂环境下黄花菜识别的YOLOv7-MOCA模型[J].农业工程学报,2023,39(15):181-188.
	JIN H J, MA G Y, TANG M Y, et al.. Identifying daylily in complex environment using YOLOv7-MOCA model [J]. Trans. Chin. Soc. Agric. Eng., 2023,39(15): 181-188.
7	张立杰,周舒骅,李娜,等.基于改进SSD卷积神经网络的苹果定位与分级方法[J].农业机械学报,2023,54(6):223-232.
	ZHANG L J, ZHOU S H, LI N, et al.. Apple location and classification based on improved SSD convolutional neural network [J]. Trans. Chin. Soc. Agric. Mach., 2023, 54(6): 223-232.
8	BHARGAVA A, BANSAL A. Classification and grading of multiple varieties of apple fruit [J]. Food Anal. Meth., 2021, 14(7): 1359-1368.
9	杨其晟,李文宽,杨晓峰,等.改进YOLOv5的苹果花生长状态检测方法[J].计算机工程与应用, 2022,58(4):237-246.
	YANG Q S, LI W K, YANG X F, et al.. Improved YOLOv5 method for detecting growth status of apple flowers [J]. Comput. Eng. Appl., 2022,58(4):237-246.
10	REDMON J, DIVVALA S, GIRSHIVK R, et al.. You only look once: unified, real-time object detection [EB/OL]. (2016-05-09) [2024-09-23]. .
11	FARHADI A, REDMON J.YOLOv3:Anincremental improvement [EB/OL]. (2018-04-08) [2024-09-23]..
12	ALBAHLI S, NIDA N, IRTAZA A, et al.. Melanoma lesion detection and segmentation using YOLOv4-DarkNet and active contour [J]. IEEE access, 2020,8:198403-198414.
13	YAN B, FAN P, LEI X, et al.. A real-time apple targets detection method for picking robot based on improved YOLOv5 [J/OL].Remote Sensing, 2021,13(9):1619 [2024-09-23]. .
14	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7:Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors [EB/OL].(2022-07-06) [2024-09-23]. .
15	REN S, HE K, GIRSHICK R, et al.. Faster R-CNN: towards real-time object detection with region proposal networks [J]. IEEE Trans. Pattern Anal. Mach. Intell., 2017, 39(6): 1137-1149.
16	ZHANG Q L, YANG Y B. Sa-net:shuffle attention for deep convolutional neural networks [C]//Proceedings of ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021:2235-2239.
17	HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design [C]// 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2021:13708-13717.
18	YANG L X, ZHANG R Y, LI L D, et al.. Simam: A simple, parameter-free attention module for convolutional neural networks [C]//Proceedings of International conference on machine learning. PMLR, 2021, 139:11863-11874.
19	VASWANI A, SHAZEER N, PARMAR N, et al.. Attention is all you need [EB/OL].(2017-06-19) [2024-09-23]. .
20	ZHU L, WANG X J, KE Z H, et al.. BiFormer: vision transformer with Bi-Level routing attention [EB/OL]. (2023-03-15) [2024-09-23]..
21	LAY Y M, STEAD A D, REID M S. Flower senescence in daylily (Hemerocallis) [J]. Physiol. Plant., 1992,86(2):308-314.
22	中国团体标准,大同黄花干制品质量分级标准:T/DTHH 001—2021 [S].大同:大同市云州区黄花菜合作协会,2021.
23	JOCHER G, STOKEN A, CHAURASIA A, et al.. ultralytics/yolov5: v6. 0-YOLOv5n'Nano'models, Roboflow integration, TensorFlow export, OpenCV DNN support [J/OL]. Zenodo, 2021: 5563715[2024-09-23]. .
24	王小荣,许燕,周建平,等.基于改进YOLOv7的复杂环境下红花采摘识别[J].农业工程学报,2023,39(6):169-176.
	WANG X R, XU Y, ZHOU J P, et al.. Safflower picking recognition in complex environments based on an improved YOLOv7 [J]. Trans. Chin. Soc. Agric. Eng., 2023, 39(6): 169-176.
25	龚惟新,杨珍,李凯,等.基于改进YOLOv5s的自然环境下猕猴桃花朵检测方法[J].农业工程学报,2023,39(6):177-185.
	GONG W X, YANG Z, LI K, et al.. Detecting kiwi flowers in natural environments using an improved YOLOv5s [J]. Trans. Chin. Soc. Agric. Eng., 2023, 39(6): 177-185.

模型 Model	P/%	R/%	mAP/%	每秒传输帧数/（帧·s^-1） Frames per second FPS/（frames·s^-1）	模型大小 Model size/Mb
YOLOv3	85.39	78.57	88.05	61.52	235
YOLOv4	87.67	79.20	88.00	63.40	244
YOLOv7	86.40	82.77	90.50	107.21	142
Faster-RCNN	61.75	86.13	84.86	136.69	108
YOLOv5s	84.63	81.54	89.00	118.49	13.7
YOLOv5m	84.70	79.20	88.80	37.85	239
YOLOv5l	85.40	85.30	91.90	26.89	88.5
YOLOv5x	84.70	85.70	90.70	17.57	165

模型 Model	P/%	R/%	mAP/%	每秒传输帧数/（帧·s^-1） Frames per second FPS/（frames·s^-1）	模型大小 Model size/Mb
YOLOv3	85.39	78.57	88.05	61.52	235
YOLOv4	87.67	79.20	88.00	63.40	244
YOLOv7	86.40	82.77	90.50	107.21	142
Faster-RCNN	61.75	86.13	84.86	136.69	108
YOLOv5s	84.63	81.54	89.00	118.49	13.7
YOLOv5m	84.70	79.20	88.80	37.85	239
YOLOv5l	85.40	85.30	91.90	26.89	88.5
YOLOv5x	84.70	85.70	90.70	17.57	165

模型 Model	P/%	R/%	mAP/%	检测速度/（帧·s^-1） Detection speed/（frames·s^-1）
YOLOv5s	84.63	81.54	89.00	118.49
YOLOv5s_CA	78.60	80.50	86.40	64.824
YOLOv5s_SA	91.90	87.20	93.00	118.28
YOLOv5s_SimAM	85.60	81.30	88.70	74.36
YOLOv5s_Biformer	94.80	91.40	96.60	149.75

模型 Model	P/%	R/%	mAP/%	检测速度/（帧·s^-1） Detection speed/（frames·s^-1）
YOLOv5s	84.63	81.54	89.00	118.49
YOLOv5s_CA	78.60	80.50	86.40	64.824
YOLOv5s_SA	91.90	87.20	93.00	118.28
YOLOv5s_SimAM	85.60	81.30	88.70	74.36
YOLOv5s_Biformer	94.80	91.40	96.60	149.75

模型 Model	第2层 2^nd layer	第4层4^th layer	第6层 6^th layer	P/%	R/%	mAP/%	检测速度/ （帧·s^-1）Detection speed/（Frames·s^-1）	模型大小 Model size/Mb
Yolov5s	×	×	×	84.63	81.54	89.00	118.49	13.7
Yolov5s_Biformer_1	×	×	√	81.90	84.60	89.90	177.89	13.6
Yolov5s_Biformer_2	×	√	√	86.80	81.60	88.70	151.47	14.1
Yolov5s_Biformer_3	√	√	√	94.80	91.40	96.60	149.75	16.3