基于改进YOLOv8的无人机视角下青皮核桃目标检测

打开文本图片集

中图分类号：TP39 文献标志码：A DOI：10. 13705/j. issn.1671-6841. 2023256

文章编号：1671-6841（2025）05-0024-07

Abstract： Current methods of walnut yield measurement mainly relied on traditional statistical models， and the accuracy could hardly be guaranteed. Therefore，taking green walnuts as an example，an image dataset of walnuts from the perspective of drone aerial photography was established. The coordinate attention（CA） was innovatively applied to the YOLOv8 model for the first time. The improved YOLOv8-CA model algorithm was used for object detection of green walnuts. The experimental results showed that the newly improved model（YOLOv8-CA），improved the mAP value by O.004 and O.051，and the Recall value by 0. 019 and 0.089 compared with the original YOLOv8 and YOLOv5，respectively.

Key words： objection detection； drone perspective; computer vision； fruit yield measurement； walnutdetection

0 引言

核桃的果实发育周期较长，若将核桃采摘后再人工进行产量的测定，许多规划的制定以及政策的落实就会有一定的滞后性，这不利于核桃产业经济的发展，而如何准确地预测核桃产量，将对核桃产业的发展起到十分重要的作用。（剩余10959字）