基于高光谱图像的小麦种子纯度/含水量检测系统

Seed purity/moisture content detection system based on hyperspectral images

摘要:

小麦种子的纯度和含水量是反映小麦种子品质的两个重要指标；本文基于高光谱图像技术,开发了小麦种子纯度/含水量自动检测系统；检测系统集成了系统校准、高光谱图像采集、数据分析、模型更新、报告输出等功能模块,具有功能完善、操作简便、可扩展性强等优点；所开发系统利用STM32控制模块,配合SHT20温湿度检测模块、光源控制模块和步进电机驱动模块,实现操作环境信息获取、平台运动控制、光源功率和高光谱相机焦距调整；利用上位机实现高光谱图像的实时采集和STM32控制模块的信息交互；利用Python语言编写软件操作平台的后台逻辑和人机交互界面；所开发的系统和集成的数据分析模型(偏最小二乘回归模型、偏最小二乘鉴别模型),可实现小麦种子含水量(决定系数为0.943、均方根误差为0.0081g)和纯度(准确率为96.5%,精确率为93.0%,召回率为93.0%)的高精度检测；所设计的小麦种子纯度/含水量检测系统为促进高光谱图像技术在小麦种子质量检测领域的现场应用提供了可能。

Abstract:

The purity and moisture content of wheat seeds are two important indicators reflecting the quality of wheat seeds; Based on hyperspectral image technology, an automatic detection system for wheat seed purity/moisture content was developed. The detection system integrates system calibration, hyperspectral image acquisition, data analysis, model update, report output and other functional modules, which has the advantages of perfect function, easy operation and strong scalability; The developed system uses the STM32 control module, together with the SHT20 temperature and humidity detection module, light source control module and stepper motor drive module, to achieve operating environment information acquisition, platform motion control, light source power and hyperspectral camera focal length adjustment; Real-time acquisition of hyperspectral images and information interaction of STM32 control module are realized by using the host computer; Use python language to write the background logic and human-computer interaction interface of the software operation platform; The developed system and integrated data analysis model (partial least squares regression model, partial least squares identification model) can realize the high-precision detection of wheat seed moisture content (decision coefficient of 0.943, root mean square error of 0.0081g) and purity (accuracy of 96.5%, accuracy of 93.0%, recall rate of 93.0%); The designed wheat seed purity/moisture content detection system provides the possibility of promoting the field application of hyperspectral image technology in the field of wheat seed quality inspection.