农机化研究2025，Vol.47Issue(11)：46-54,9.DOI:10.13427/j.issn.1003-188X.2025.11.006
基于多帧边缘信息融合的扁平种子高速穴盘播种检测算法
Flat Seed High-Speed Plug-tray Sowing Detection Algorithm Based on Multi-Frame Edge Information Fusion
摘要
Abstract
To address the issue of low accuracy in detecting double-seeding of flat solanaceous seeds during high-speed plug-tray sowing,a flat seed high-speed plug-tray sowing detection algorithm was proposed.Collected multiple frames of seed dropping images row by row through high-speed photography,with a roller scraper board as the background.An improved Gaussian mixture model with a luminance-aware learning rate was introduced to obtain binary seed location im-ages via background modeling of the image sequence,which was used as a mask to extract the original color area of the seeds.A multi-particle segmentation algorithm based on the Canny edge detection method was employed to separate the adhered connected components,and the number of connected components in the image was taken as the number of seeds.The results of grain number detection from all sowing image frames from one row were integrated via non-maxi-mum suppression,which served as the final detection outcome for the seeds from that line,finalizing the sowing detection process.Chile and tomatoes,two varieties of flat Solanaceae seeds,were chosen as experimental materials,and verifica-tion tests were carried out where the sowing speed was 600,750 and 900 trays per hour.The results indicated that for both types of test materials,the average accuracy of multi-seeding detection remained above 95%at different sowing speeds.The average detection time per tray was less than the theoretical sowing time per tray,which effectively improved the accuracy of multi-seeding detection under the conditions that allow for real-time monitoring at high sowing speeds.Additionally,the average miss-seeding detection accuracy rate for both types of test materials exceeded 99%,and the average single-seeding detection accuracy rate was above 98%,demonstrating that the algorithm can be applied to high-speed plug-tray sowing detection of flat solanaceous seeds.关键词
扁平种子/穴盘播种/重播检测/机器视觉/验证试验Key words
flat seed/plug-tray sowing/multi-seeding detection/machine vision/verification test分类
农业科技引用本文复制引用
何梓源,朱仕成,邓楚恒,陈杰庆,杨腾,黄润鑫,夏红梅..基于多帧边缘信息融合的扁平种子高速穴盘播种检测算法[J].农机化研究,2025,47(11):46-54,9.基金项目
广东省现代农业产业技术体系创新团队项目(2023KJ131) （2023KJ131）
