基于改进YOLOv8的柠檬果实识别方法

打开文本图片集

DOI： 10.13718/j. cnki. xdzk. 2025. 07.019

关键词：柠檬识别；YOLOv8；SPD卷积模块；Wise-IoU损失函数；注意力机制

中图分类号：TP391；S23 文献标识码：A

文章编号：1673-9868（2025）07-0219-12

A Lemon Fruit Recognition Method Based on Improved YOLOv8

LIU Yucheng， LIANG Xincheng， LI Falin， ZHANG Fengling， LI Yunwu College of Engineering and Technology，Southwest University，Chongqing 400715，China

Abstract：In order to address the challenges of high costs and low efficiency of manual picking lemon fruits，and achieve swift and precise identification of lemon fruits in intricate environments，a lemon fruit recognition method based on the improved YOLOv8 model was established. Firstly，the SPDConv module was introduced into the backbone network to enhance the accuracy of model's detection for low-resolution images and small targets. Then，the EMA attention mechanism was added to effectively extract the features of obscured fruits. Finally，the CIoU bounding box lossfunction was replaced with Wise-IoU to reduce the dependence on high-quality anchor boxes and improve the generalization ability of the model. Tested on a self-constructed dataset，the YOLOv8-SEW model exhibited precision，recalland mean average precision values of 94.5% ， 85.7% and 92.4% separately. Compared with before improvement，the precision，recall and mean average precision of the model was increased by 1.0% ， 4.2% and 2.9% ，respectively. The detection time for a single image was 44.8ms ， enabling rapid and accurate identification of lemon fruits，thus providing a technological foundation for automatic harvesting lemon fruits.

Key words： recognition of lemon fruits; YOLOv8; SPD convolution module; Wise-IoU loss function; attention mechanism

丘陵山区的柠檬因其皮厚气香、出汁率高而受到市场的广泛欢迎，在农业和食品行业中具有重要的经济价值和市场需求[-2]。（剩余13366字）