南京农业大学学报2025，Vol.48Issue(3)：714-723,10.DOI:10.7685/jnau.202403012
基于改进YOLOv7的田间环境下食用玫瑰检测方法
Method for detecting edible roses in the field environment based on improved YOLOv7
摘要
Abstract
[Objectives]In order to accurately detect the maturity of edible roses in the field environment,in view of the poor recognition accuracy caused by factors such as illumination and occlusion,an improved model based on YOLOv7 was proposed to detect three growth states of edible roses.[Methods]SCConv was combined in the YOLOv7 backbone network to improve the efficient aggregation module ELAN,compressed redundant features for better performance and enhanced detection performance.CARAFE lightweight up-sampling operator was introduced into the model to optimize conventional up-sampling methods,improved feature recombination quality,and better aggregated contextual feature information.SimAM attention mechanism was integrated into the model to better focus on the detection target,and improve the phenomenon of missing or false detection.[Results]The average precision of the improved model reached 91.7%,3.6%higher than that of YOLOv7 algorithm,the detection precision of the bud stage,picking stage and abortive stage reached 96.1%,96.0%and 83.1%,3.7%,2.0%and 5.3%higher than the original model,and the improved model had higher overall accuracy.[Conclusions]This study provides a more accurate method for the detection of the flowering period of edible roses in the unstructured environment.关键词
成熟度检测/YOLOv7/CARAFE/注意力机制/食用玫瑰Key words
maturity detection/YOLOv7/CARAFE/attention mechanism/edible roses分类
信息技术与安全科学引用本文复制引用
化春键,黄宇峰,蒋毅,俞建峰,陈莹..基于改进YOLOv7的田间环境下食用玫瑰检测方法[J].南京农业大学学报,2025,48(3):714-723,10.基金项目
国家自然科学基金项目(62173160) （62173160）
