In response to the challenge posed by low recognition accuracy in rugged terrains with diverse topography as well as feature recognition agricultural settings,this paper presents an optimized version of the YOLOv5 alg...In response to the challenge posed by low recognition accuracy in rugged terrains with diverse topography as well as feature recognition agricultural settings,this paper presents an optimized version of the YOLOv5 algorithm alongside the development of a specialized laser weeding experimental platform designed for precise identification of corn seedlings and weeds.The enhanced YOLOv5 algorithm integrates the effective channel attention(CBAM)mechanism while incorporating the DeepSort tracking algorithm to reduce parameter count for seamless mobile deployment.Ablation tests validated this model’s achievement of 96.2%accuracy along with superior mAP values compared to standard YOLOv5 by margins of 3.1%and 0.7%,respectively.Additionally,three distinct datasets captured different scenarios,and their amalgamation resulted in an impressive recognition rate reaching up to 96.13%.Through comparative assessments against YOLOv8,the model demonstrated lightweight performance improvements,including a notable enhancement of 2.1%in recognition rate coupled with a marginal increase of 0.2%in mAP value,thus ensuring heightened precision and robustness during dynamic object detection within intricate backgrounds.展开更多
基金supported by Chongqing Science and Technology Bureau Key R&D Projects in Agriculture and Rural Areas(Grant No.cstc2021jscx-gksbX0003)Chongqing Municipal Education Commission Science and Technology Research Project(Grant No.KJZD-M202201302)+2 种基金Chongqing Municipal Science and Technology Bureau Excellence Programme Project(Grant No.20231102)Chongqing Municipal Science and Technology Bureau Innovation and Development Joint Fund Project(Grant No.CSTB2022NSCQ-LZX0024)the 2024 Chongqing Natural Science Foundation Joint Fund for Innovation and Development(Municipal Education Commission)Project(Grant No.CSTB2024NSCQ-LZX0091).
文摘In response to the challenge posed by low recognition accuracy in rugged terrains with diverse topography as well as feature recognition agricultural settings,this paper presents an optimized version of the YOLOv5 algorithm alongside the development of a specialized laser weeding experimental platform designed for precise identification of corn seedlings and weeds.The enhanced YOLOv5 algorithm integrates the effective channel attention(CBAM)mechanism while incorporating the DeepSort tracking algorithm to reduce parameter count for seamless mobile deployment.Ablation tests validated this model’s achievement of 96.2%accuracy along with superior mAP values compared to standard YOLOv5 by margins of 3.1%and 0.7%,respectively.Additionally,three distinct datasets captured different scenarios,and their amalgamation resulted in an impressive recognition rate reaching up to 96.13%.Through comparative assessments against YOLOv8,the model demonstrated lightweight performance improvements,including a notable enhancement of 2.1%in recognition rate coupled with a marginal increase of 0.2%in mAP value,thus ensuring heightened precision and robustness during dynamic object detection within intricate backgrounds.
