An improved YOLOv5-based approach to soybean phenotype information perception. (March 2023)

• Record Type: Journal Article
• Authors: Liu, Lichao; Liang, Jing; Wang, Jianqing; Hu, Peiyu; Wan, Ling; Zheng, Quan
• Is Part Of: Computers & electrical engineering. Volume 106(2023)
• Publication Date: 2023-03
• Subjects: Soybean phenotype -- Improved YOLOv5 algorithm -- Soybean leaves count -- Deep learning -- Targeted spraying; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2023.108582
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: A novel soybean phenotype information sensing method is proposed. An improved YOLOv5 model with ECANet and lightweight backbone network is established. The proposed method has good robustness and generalization ability to the soybean precision spraying. Abstract: Targeted spraying has been one of the hot issues in the field of plant protection robotics research, and whether the on-target application can be performed accurately requires first addressing the accurate identification of plant phenotypes. To solve the problem of targeted spraying of soybean, this paper proposes a method based on improved YOLOv5 for soybean phenotype information perception. First, the YOLOv5 backbone network was lightened by introducing MobileNetv2. Then, the robustness and generalization ability of the target detection model was improved by introducing attention mechanism and improving the loss function. Finally, the leaf identification experiments were conducted on the self-developed robot platform. The experimental results showed that the improved YOLOv5 had an mAP of 96.13%, an FPS of 79, and the number of model parameters was reduced by 39%, and the weights were reduced by 55.56%. The correlation coefficient between the acquired data and manual measurement of soybean phenotype information was 0.98384, with higher accuracy and better consistency.