EUVIMEDEUVIMED entdecken
Crossref

A marigold corolla detection model based on the improved YOLOv7 lightweight

Yixuan Fan, Gulbahar Tohti, Mamtimin Geni, Guohui Zhang, Jiayu Yang

Signal, Image and Video Processing · 2024-07 · 18 (5) · 4703-4712

DOI: 10.1007/s11760-024-03107-2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Signal, Image and Video Processing, online erschienen am 26.03.2024. Verfasst von Yixuan Fan und 4 weiteren Autor:innen. Die Arbeit wurde bisher 15-mal zitiert (Crossref) und nennt 25 Literaturangaben. Gefördert von: State Key Laboratory for Manufacturing Systems Engineering of Xi'an Jiaotong University, The National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Marigold, as an important traditional Chinese medicinal plant renowned for its therapeutic attributes in heat dissipation, detoxification, liver protection, facial enhancement, and promotion of eye health, has witnessed a surge in demand. The demand for marigold is steadily rising, necessitating the inevitable shift toward mechanized harvesting in the industrialization of marigold cultivation. Consequently, this study endeavors to assemble a novel dataset in southern Xinjiang, China, crucial for marigold production. The focus is on improving the YOLOv7 model by lightweighting and proposing a set of detection models applicable to marigold. Through the elimination of superfluous object detection layers in the YOLOv7 model, substitution of conventional backbone network convolution with DSConv, replacement of the SPPCSPC module with Simplified SPPF, and subsequent model pruning and then retraining. This study aims to address challenges related to the deployment of mobile devices on the marigold-picking robot and the realization of high real-time detection. The experimental results demonstrate that the enhanced YOLOv7 model exhibited exceptional precision and mAP0.5 in marigold detection, achieving 93.9% and 97.7%, respectively, surpassing the performance of the original YOLOv7 model. Notably, the GFLOPs are a mere 2.3, representing only 2.2% of the computational load of the original model. Moreover, the parameter has been reduced to 15.04M, representing only 41.2% of the original model. The attained FPS, reaching 166.7, signifies a noteworthy enhancement of 26.7% compared to the original model. This showcases exceptional precision and speed in marigold detection, providing a strong technical foundation for the efficient harvesting of marigolds.

Förderung

  • State Key Laboratory for Manufacturing Systems Engineering of Xi'an Jiaotong University
  • The National Natural Science Foundation of China

Zitieren

APA

Fan, Y., Tohti, G., Geni, M., Zhang, G., & Yang, J. (2024). A marigold corolla detection model based on the improved YOLOv7 lightweight. Signal, Image and Video Processing, 18(5), 4703-4712. https://doi.org/10.1007/s11760-024-03107-2

Vancouver

Fan Y, Tohti G, Geni M, Zhang G, Yang J. A marigold corolla detection model based on the improved YOLOv7 lightweight. Signal, Image and Video Processing. 2024;18(5):4703-4712. doi:10.1007/s11760-024-03107-2

BibTeX
@article{Fan2024,
  author = {Fan, Yixuan and Tohti, Gulbahar and Geni, Mamtimin and Zhang, Guohui and Yang, Jiayu},
  title = {A marigold corolla detection model based on the improved YOLOv7 lightweight},
  journal = {Signal, Image and Video Processing},
  year = {2024},
  volume = {18},
  number = {5},
  pages = {4703--4712},
  doi = {10.1007/s11760-024-03107-2}
}
Kostenlos starten →