基于YOLOv10的玉米籽粒淀粉粒识别与粒径分析系统的开发

doi:10.13560/j.cnki.biotech.bull.1985.2025-0957

摘要/Abstract

摘要：

目的针对玉米成熟籽粒淀粉粒粒径检测传统方法（如激光衍射粒度分析等）存在的制样繁琐、效率低、准确性差等问题，开发一种高效、准确的玉米胚乳淀粉粒粒径自动分析工具，以满足玉米产量、品质改良及特用玉米淀粉生产中的淀粉粒粒径测定需求。方法基于YOLOv10目标检测算法，建立并训练玉米籽粒淀粉粒自动检测模型，并基于Django框架开发粒径分析系统。选取6个代表性玉米自交系（B73、B104、KN5585、MO17、RP125、W22）的成熟籽粒，获取其胚乳扫描电镜（SEM）图片，利用该系统对6个玉米自交系胚乳淀粉粒的粒径进行分析。结果模型评价表明，所构建的淀粉粒检测模型精确率为0.812，召回率为0.843，平均准确率（AP）达0.895。开发的系统支持用户批量上传籽粒胚乳SEM图片，利用训练好的模型自动识别图片中的淀粉粒，并分析计算每个识别出的淀粉粒的粒径数据，最终结果以Excel文件格式输出。对6个玉米自交系胚乳淀粉粒的分析结果显示，不同自交系淀粉粒平均粒径存在显著差异，MO17的淀粉粒平均直径最大（15.46 μm），RP125的最小（10.61 μm）。与扫描电镜图片的手工测量粒径法对比，该系统生成的粒径数据可靠，在省工节时方面具有巨大优势，工作效率大幅提升。结论开发了一个基于YOLOv10目标检测和Django框架的玉米籽粒淀粉粒粒径分析系统，该系统能够快速、批量、准确地自动识别玉米胚乳扫描电镜图片中的淀粉粒并分析其粒径，有效克服了传统方法的局限性，是一个便捷、高效、准确的粒径分析工具。

关键词: 玉米, 籽粒胚乳, 淀粉粒识别, 淀粉粒粒径检测, YOLOv10, 深度学习, 图像识别, Django框架

Abstract:

Objective To address the limitations of traditional methods for detecting starch granule size in mature maize kernels (such as laser diffraction particle size analysis), including cumbersome sample preparation, low efficiency, and poor accuracy, this study is aimed to develop an efficient and accurate automatic analysis tool. This tool is intended to meet the demand for starch granule size measurement in maize yield and quality improvement, as well as in the production of specialty maize starch. Method Based on the YOLOv10 object detection algorithm, an automatic detection model for maize kernel starch granules was established and trained. A granule size analysis system was developed using the Django framework. Mature kernels from six representative maize inbred lines (B73, B104, KN5585, MO17, RP125, W22) were selected. Scanning electron microscopy (SEM) images of their endosperms were acquired. Starch granule size of the six inbred lines endosperms was analyzed using the developed system. Result Model evaluation indicates that the constructed starch granule detection model has an accuracy rate of 0.812, a recall rate of 0.843, and an average precision (AP) of 0.895. The developed system supports batch uploading of kernel endosperm SEM images by users. It utilizes the trained model to automatically identify starch granules within the images, analyzes and calculates the size data for each detected granule, and ultimately outputs the results in an Excel file. Analysis of starch granules from the six maize inbred lines revealed significant differences in mean granule diameter among them. MO17 showed the largest mean starch granule diameter (15.46 µm), while RP125 presented the smallest (10.61 µm). Compared with manual starch granule size measurements from SEM images, the system-generated granule size data demonstrates reliability, offering significant advantages in labor-saving and time-efficiency, with a substantial improvement in work productivity. Conclusion This study developed a maize kernel starch granule size analysis system based on the YOLOv10 object detection algorithm and the Django framework. The system enables the rapid, batch, and accurate automatic identification of starch granules in maize endosperm SEM images and analysis of their size. It effectively overcomes the limitations of traditional methods, providing a convenient, efficient, and accurate granule size analysis tool.

Key words: maize, endosperm of kernel, starch granule recognition, starch granule size detection, YOLOv10, deep learning, image recognition, Django framework

万瑞恒, 李晓艳, 李莹莹, 郭扬, 刘应红, 张军杰, 胡育峰, 李炀平, 黄玉碧, 刘汉梅. 基于YOLOv10的玉米籽粒淀粉粒识别与粒径分析系统的开发[J]. 生物技术通报, doi: 10.13560/j.cnki.biotech.bull.1985.2025-0957.

WAN Rui-heng, LI Xiao-yan, LI Ying-ying, GUO Yang, LIU Ying-hong, ZHANG Jun-jie, HU Yu-feng, LI Yang-ping, HUANG Yu-bi, LIU Han-mei. Development of a YOLOv10-based System for the Starch Granules Recognition and Size Analysis of Maize Kernels[J]. Biotechnology Bulletin, doi: 10.13560/j.cnki.biotech.bull.1985.2025-0957.

图/表 11

图1 数据集原始图像概览A：放大2 000倍；B：放大1 000倍；C：放大500倍

Fig. 1 Overview of the original images of the datasetA: Enlarge 2 000 times. B: Enlarge 1 000 times. C: Enlarge 500 times

表1 训练环境

Table 1 Training environment

配置 Configuration	参数 Parameter
CPU	16vCPU AMD EPYC 9654 96-Core Processor
内存 Memory	60 GB
GPU	RTX 4090（24 GB） × 1
训练工具 Training tools	Cuda 12.1
操作系统 Operating system	Ubuntu22.04
开发环境 Development environment	PyTorch 2.1.0 Python 3.10

表2 系统开发软件配置

Table 2 Software configuration for system development

类型 Type	软件/版本 Software/Version
操作系统 Operating system	Windows10
开发语言 Development language	Python
后端开发框架 Backend development framework	Django
后端开发工具 Backend development tool	PyCharm
前端开发语言 Frontend development language	HTML、CSS、JS

图2 系统功能模块图

Fig. 2 Modules of system functions

图3 模型性能评估A：精确率‒召回率曲线；B：F1-置信度曲线；C：召回率变化曲线；D：精确率变化曲线

Fig. 3 Assessment on model performanceA: Precision-Recall curve. B: F1-confidence curve. C: Recall change curve. D: Precision change curve

图4 系统操作流程图

Fig. 4 The operation flowchart of the system

图5 淀粉粒粒径检测界面

Fig. 5 Detection interface of starch granules size

图6 六个玉米自交系胚乳淀粉粒的扫描电镜图像A： B73； B： B104； C： KN5585； D： MO17； E： RP125； F： W22

Fig. 6 Original SEM images of endosperm starch granules for six maize inbred lines

表3 六个玉米自交系粒径的人工和自动化检测

Table 3 The manual and automated measurement of starch granule diameter for six maize inbred lines

自交系 Inbred lines	人工检测			自动化检测			人工与自动化检测粒径的差异 Difference in granules diameter between manual and automated detection （μm）
自交系 Inbred lines	图片数量 Number of images	识别的淀粉粒总数 Total number of identified starch granules	淀粉粒平均直径 Average diameter of starch granules （μm）	图片数量 Number of images	识别的淀粉粒总数 Total number of identified starch granules	淀粉粒平均直径 Average diameter of starch granules （μm）
B73	1	30	16.12	5	297	15.31	0.81
B104	1	71	13.30	5	468	11.99	1.31
KN5585	1	64	12.24	5	561	10.80	1.44
MO17	1	63	14.89	5	263	15.46	0.57
RP125	1	55	10.92	5	565	10.61	0.31
W22	1	49	11.03	5	419	12.20	1.17

图7 不同玉米自交系的粒径变异系数

Fig. 7 Coefficient of variation for granule diameter in different maize inbred lines

图8 不同作物淀粉粒的模型识别结果A：小麦；B：大麦；C：水稻；D：马铃薯

Fig. 8 Model recognition results of starch granules in different cropsA: Triticum aestivum. B: Hordeum vulgare. C: Oryza sativa. D: Solanum tuberosum

参考文献 30

[1]	Bertoft E. Understanding starch structure: recent progress [J]. Agronomy, 2017, 7(3): 56.
[2]	Smith AM. The biosynthesis of starch granules [J]. Biomacromolecules, 2001, 2(2): 335-341.
[3]	余平, 石彦忠. 淀粉与淀粉制品工艺学 [M]. 北京: 中国轻工业出版社, 2011: 7-26.
	Yu P, Shi YZ. Technology of starch and derived products [M]. Beijing: China Light Industry Press, 2011: 7-26.
[4]	Zhu F. Starch based Pickering emulsions: Fabrication, properties, and applications [J]. Trends Food Sci Technol, 2019, 85: 129-137.
[5]	Myburgh JA, Finfer S, Bellomo R, et al. Hydroxyethyl starch or saline for fluid resuscitation in intensive care [J]. N Engl J Med, 2012, 367(20): 1901-1911.
[6]	Hebeish A, Aly AA, El-Shafei AM, et al. Innovative starch derivatives as textile auxiliaries for application in sizing, finishing and flocculation [J]. Starch Stärke, 2008, 60(2): 97-109.
[7]	Mohd Asharuddin S, Othman N, Ali Hamood Altowayti W, et al. Recent advancement in starch modification and its application as water treatment agent [J]. Environ Technol Innov, 2021, 23: 101637.
[8]	侯汉学, 董海洲, 宋晓庆, 等. 不同品种玉米淀粉的理化性质及相关性研究 [J]. 中国粮油学报, 2009, 24(1): 60-64.
	Hou HX, Dong HZ, Song XQ, et al. Correlations among physico-chemical properties of starches from different maize cultivars [J]. J Chin Cereals Oils Assoc, 2009, 24(1): 60-64.
[9]	张雅婷, 路梦莉, 余梦奇, 等. 种植密度对沿淮玉米籽粒淀粉粒度分布与糊化特性的影响 [J]. 粮油食品科技, 2024, 32(1): 66-76.
	Zhang YT, Lu ML, Yu MQ, et al. Effects of planting density on starch particle size distribution and pasting properties of maize kernels along Huaihe River [J]. Sci Technol Cereals Oils Foods, 2024, 32(1): 66-76.
[10]	李晓艳. 玉米淀粉合酶Ⅳ、Ⅴ（ZmSSⅣ、ZmSSⅤ）对淀粉粒形成的影响研究 [D]. 雅安: 四川农业大学, 2024.
	Li XY. Effect of maize starch synthase Ⅳ, Ⅴ (ZmSSⅣ, ZmSSⅤ) on the formation of starch granules [D]. Ya’an: Sichuan Agricultural University, 2024.
[11]	陈海花. 玉米淀粉合酶Ⅳ（ZmSSⅣ）、Ⅴ（ZmSSⅤ）过表达对淀粉合成的影响 [D]. 雅安: 四川农业大学, 2021.
	Chen HH. Effects of overexpression of starch synthase Ⅳ (ZmSSⅣ) and Ⅴ (ZmSSⅤ) on starch synthesis in maize [D]. Ya’an: Sichuan Agricultural University, 2021.
[12]	Seung D. Amylose in starch: towards an understanding of biosynthesis, structure and function [J]. New Phytol, 2020, 228(5): 1490-1504.
[13]	Chia T, Chirico M, King R, et al. A carbohydrate-binding protein, B-GRANULE CONTENT 1, influences starch granule size distribution in a dose-dependent manner in polyploid wheat [J]. J Exp Bot, 2020, 71(1): 105-115.
[14]	Lv XX, Hong Y, Gu ZB, et al. Effect of solution on starch structure: New separation approach of amylopectin fraction from gelatinized native corn starch [J]. Carbohydr Polym, 2024, 329: 121770.
[15]	Zhang J, Li JW, Fan LP. Effect of starch granule size on the properties of dough and the oil absorption of fried potato crisps [J]. Int J Biol Macromol, 2024, 268(Pt 2): 131844.
[16]	Cui LN, Dong ST, Zhang JW, et al. Starch granule size distribution and morphogenesis in maize (Zea mays L.) grains with different endosperm types [J]. Australian Journal of Crop Science, 2014, 8(11): 1560-1565.
[17]	Murphy KM, Ludwig E, Gutierrez J, et al. Deep learning in image-based plant phenotyping [J]. Annu Rev Plant Biol, 2024, 75(1): 771-795.
[18]	Huang YN, Qian YR, Wei HY, et al. A survey of deep learning-based object detection methods in crop counting [J]. Comput Electron Agric, 2023, 215: 108425.
[19]	Liu RY. Enhanced motorbike helmet detection using YOLOv10 with attention-based modifications [C]//2024 4th International Conference on Computer Science and Blockchain (CCSB). September 6-8, 2024, Shenzhen, China. Piscataway, NJ: IEEE, 2024: 114-118.
[20]	An JY, Dong SY, Wang XL, et al. Research on UAV aerial imagery detection algorithm for Mining-Induced surface cracks based on improved YOLOv10 [J]. Sci Rep, 2025, 15: 30101.
[21]	Bellout A, Zarboubi M, Elhoseny M, et al. LT-YOLOv10n: a lightweight IoT-integrated deep learning model for real-time tomato leaf disease detection and management [J]. Internet Things, 2025, 33: 101663.
[22]	Guan ST, Lin YM, Lin GY, et al. Real-time detection and counting of wheat spikes based on improved YOLOv10 [J]. Agronomy, 2024, 14(9): 1936.
[23]	Saito M, Tanaka T, Sato K, et al. A single nucleotide polymorphism in the “Fra” gene results in fractured starch granules in barley [J]. Theor Appl Genet, 2018, 131(2): 353-364.
[24]	Peng C, Wang YH, Liu F, et al. FLOURY ENDOSPERM6 encodes a CBM48 domain-containing protein involved in compound granule formation and starch synthesis in rice endosperm [J]. Plant J, 2014, 77(6): 917-930.
[25]	郎双静, 刘诗琳, 侯越, 等. 微细化马铃薯淀粉制备工艺优化及颗粒性质研究 [J]. 黑龙江八一农垦大学学报, 2020, 32(6): 61-69.
	Lang SJ, Liu SL, Hou Y, et al. Study on processing optimization and particle properties of micronized potato starch [J]. J Heilongjiang Bayi Agric Univ, 2020, 32(6): 61-69.
[26]	Sun Y, Li MQ, Liu ML, et al. A statistical method for high-throughput emergence rate calculation for soybean breeding plots based on field phenotypic characteristics [J]. Plant Methods, 2025, 21(1): 40.
[27]	李思进, 李金阳, 张伟. 基于改进YOLOv5s的大豆冠层叶片识别方法研究 [J]. 中国农机化学报, 2025, 46(5): 100-105, 140.
	Li SJ, Li JY, Zhang W. Research on soybean canopy leaf recognition method based on improved YOLOv5s [J]. J Chin Agric Mech, 2025, 46(5): 100-105, 140.
[28]	Feng J, Li ZT, Zhang SZ, et al. A microimage-processing-based technique for detecting qualitative and quantitative characteristics of plant cells [J]. Agriculture, 2023, 13(9): 1816.
[29]	朱怡航, 张小斌, 沈颖越, 等. 基于图像识别技术的金针菇表型高通量采集与分析 [J]. 菌物学报, 2021, 40(3): 626-640.
	Zhu YH, Zhang XB, Shen YY, et al. High-throughput phenotyping collection and analysis of Flammulina filiformis based on image recognition technology [J]. Mycosystema, 2021, 40(3): 626-640.
[30]	陈佳玮, 李庆, 谭巧行, 等. 结合轻量级麦穗检测模型和离线Android软件开发的田间小麦测产 [J]. 农业工程学报, 2021, 37(19): 156-164.
	Chen JW, Li Q, Tan QX, et al. Combining lightweight wheat spikes detecting model and offline Android software development for in-field wheat yield prediction [J]. Trans Chin Soc Agric Eng, 2021, 37(19): 156-164.