Camellia oleifera Fruit Static and Dynamic Detection Counting Based on Improved COF-YOLO v8n

被引：0
|
作者：
Wang J. [1 ]
He M. [1 ]
Zhen Q. [1 ]
Zhou H. [1 ]
机构：
[1] College of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing
来源：
Nongye Jixie Xuebao/Transactions of the Chinese Society for Agricultural Machinery | 2024年 / 55卷 / 04期
关键词：
Camellia oleifera fruit; COF - YOLO v8n; count; machine vision; production estimates;
D O I：
10.6041/j.issn.1000-1298.2024.04.019
中图分类号：
学科分类号：
摘要：
Aiming at the problems of severe occlusion, close view and small target Camellia oleifera in Camellia oleifera fruit, the original YOLO v8n network was selected to improve the phenomenon of low detection accuracy and serious missed detection phenomenon by using the original YOLO network. MPDIOU was used as the loss function of YOLO v8n to effectively solve the problem of missed detection caused by fruit overlap. Adjusting the network and adding a small target detection layer to it, so that the network can pay attention to small target Camellia oleifera and Camellia oleifera obscured by leaves; SCConv was used to replace the C2f in the original YOLO v8n, so that the network can balance both detection accuracy and detection speed. The P, R and mAP of the improved COF - YOLO v8n network reached 97.7%, 97% and 99% respectively, which were 3.2 percentages, 4. 8 percentages and 2.4 percentages higher than P, R and mAP of the unimproved YOLO v8n, among which the P, R and mAP of Camellia oleifera reached 95. 9%, 95% and 98. 5% under severe occlusion, respectively, which was 4. 0 percentages, 9. 1 percentages and 4. 6 percentages higher than that of the original YOLO v8n. The COF - YOLO v8n network can significantly improve the recognition accuracy of Camellia oleifera under the conditions of severe occlusion, close vie, and small targets. In addition, the model can realize the counting of Camellia oleifera under dynamic and static input conditions. Dynamic counting draws on the multi-target tracking idea of DeepSORT algorithm, and took the recognition output of COF - YOLO v8n as the input of DeepSORT to realize the recognition and counting of Camellia oleifera fruits, and used the reduced resolution Camellia oleifera data to simulate the target situation in the field environment and restored the real picking environment. The resulting improved model had good robustness and simple model can be embedded in the robotic arm, which can not only be used to guide future automated harvesting, but also for yield estimation of orchards, providing reliable reference for orchard logistics distribution. © 2024 Chinese Society of Agricultural Machinery. All rights reserved.
引用
收藏
页码：193 / 203
页数：10
相关论文
共 37 条
  • [1] ZHOU Y, TANG Y, ZOU X, Et al., Adaptive active positioning of Camellia oleifera fruit picking points
  • [2] classical image processing and YOLO v7 fusion algorithm[J], Applied Siences, 12, 24, (2022)
  • [3] ZHOU Hongping, JIN Shouxiang, ZHOU Lei, Et al., Recognition of Camellia oleifera fruit in natural environment based on multimodal image [J], Transactions of the CSAE, 39, 10, pp. 175-182, (2023)
  • [4] YANG Y, KANG H., An enhanced detection method of PCB defect based on improved YOLO v7 [J], Electronics, 12, 9, (2023)
  • [5] CHEN J, CHEN J, ZHANG D, Et al., Using deep transfer learning for image-based plant disease identification, Computers and Electronics in Agriculture, 173, (2020)
  • [6] LEE C, YANG M, TSENG H, Et al., Single-plant broccoli growth monitoring using deep learning with UAY imagery [J], Computers and Electronics in Agriculture, 207, (2023)
  • [7] RAI N, ZHANG Y, RAM B G, Et al., Applications of deep learning in precision weed management: a review, Computers and Electronics in Agriculture, 206, (2023)
  • [8] JANANI M, JEBAKUMAR R., Detection and classification of groundnut leaf nutrient level extraction in RGB images [J], Advances in Engineering Software, 175, (2023)
  • [9] ALSHEHHI R, MARPU P R., Change detection using multi-scale convolutional feature maps of bi-temporal satellite high-resolution images, European Journal of Remote Sensing, 56, 1, (2023)
  • [10] QUAN Y, LI M, HAO Y, Et al., Tree species classification in a typical natural secondary forest using UAV-borne LiDAR and hyperspectral data[J], Giscience & Remote Sensing, 60, 1, (2023)
1234