Scholarlink

<p>From September 2017 to March 2021, the Insect Laser project was carried out. Three partners from stored product protection, micro-integration and laser research, as well as media technology and artificial intelligence (AI) worked on the task to provide image recognition for adult granary weevils (<i>Sitophilus granarius</i>; Coleoptera, Curculionidae) and adult Indianmeal moths (<i>Plodia interpunctella</i>; Lepidoptera, Pyralidae). Deep learning was facilitated by training the system with photos and short video clips. After training, the system was able to identify both granary weevils and Indianmeal moths with a confidence level of 93%. The camera for image recognition was mounted on a one-dimensional track equipped with an infrared LED lamp and a laser source at a distance of 500&#xa0;mm to insects on the floor of the test chamber. Digital image processing was carried out with YOLO v4-Tiny. Location coordinates of an insect detected in the scanned camera frame were forwarded to the laser beamer unit. Both, adult granary weevils and Indianmeal moths, were controlled with beams at wavelengths of 808&#xa0;nm (8&#xa0;W) or 1470&#xa0;nm (2.5&#xa0;W) in treatment times of 500 ms. The results suggest that using the system in a large warehouse environment is less feasible. However, its use in a closed trap could provide workers safety and be effective for both monitoring and control.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial intelligence for stored product insect detection and control with laser beams: the Insect Laser project

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月13日

查看全文
  • Cornel Adler,
  • Gunnar Böttger,
  • Kirko Große,
  • Christian Hentschel,
  • Dirk Höpfner,
  • Peter Kern

摘要

From September 2017 to March 2021, the Insect Laser project was carried out. Three partners from stored product protection, micro-integration and laser research, as well as media technology and artificial intelligence (AI) worked on the task to provide image recognition for adult granary weevils (Sitophilus granarius; Coleoptera, Curculionidae) and adult Indianmeal moths (Plodia interpunctella; Lepidoptera, Pyralidae). Deep learning was facilitated by training the system with photos and short video clips. After training, the system was able to identify both granary weevils and Indianmeal moths with a confidence level of 93%. The camera for image recognition was mounted on a one-dimensional track equipped with an infrared LED lamp and a laser source at a distance of 500 mm to insects on the floor of the test chamber. Digital image processing was carried out with YOLO v4-Tiny. Location coordinates of an insect detected in the scanned camera frame were forwarded to the laser beamer unit. Both, adult granary weevils and Indianmeal moths, were controlled with beams at wavelengths of 808 nm (8 W) or 1470 nm (2.5 W) in treatment times of 500 ms. The results suggest that using the system in a large warehouse environment is less feasible. However, its use in a closed trap could provide workers safety and be effective for both monitoring and control.