Research on Cotton Top Bud Target Detection Algorithm Based on Improved RetinaNet
- 1 College of Mechanical and Electrical Engineering, Tarim University, Alar, Xinjiang, China
- 2 Department of Xinjiang Uygur Autonomous Region, Modern Agricultural Engineering Key Laboratory at Universities of Education, Tarim University, Alar, Xinjiang, China
- 3 Key Laboratory of Tarim Oasis Agriculture (Tarim University), Ministry of Education, Tarim University, Alar, Xinjiang, China
- 4 Department of Xinjiang Uygur Autonomous Region, Modern Agricultural Engineering Key Laboratory at Universities of Education, Tarim University, Alar, Xinjiang, China
Abstract
To solve the problems of low accuracy and high miss rate in the recognition of cotton apical buds during mechanical topping, an enhanced method based on the RetinaNet network is proposed for the accurate identification of cotton apical buds under natural light. The traditional RetinaNet algorithm is validated to improve the recall rate and average accuracy of cotton apical bud recognition ([email protected]) at 83.61% and 77.64% respectively. Due to the shallow nature of the network, there is still overfitting and the RetinaNet algorithm is improved. This algorithm incorporates R-CBAM and ShuffleViT Block network modules and uses Atrous Spatial Pyramid Pooling (ASPP) to connect the cross-domain feature layer to the feature fusion layer. The results indicate thatcompared with the traditional RetinaNet algorithm, theimprovedRetinaNet algorithm has an average accuracy ([email protected]) of 96.25% and a recall rate of 91.10% for cotton apical bud recognition. This indicates that the improved RetinaNet algorithm has optimal recognition performance and high recognition accuracy for cotton apical buds, laying a solid foundation for precise topping operations in cotton cultivation.
DOI: https://doi.org/10.3844/ajbbsp.2024.94.104
Copyright: © 2024 Jikui Zhu, Shijie Lin, Fengkui Zhang, Ting Zhang, Shijie Zhao and Ping Li. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
- 1,192 Views
- 598 Downloads
- 0 Citations
Download
Keywords
- Cotton Apical Bud
- RetinaNet
- Target Detection