基于YOLOv8的鲁棒西红柿目标检测方法

Robust tomato target detection method based on YOLOv8

摘要

针对当前复杂环境下基于YOLO系列西红柿目标检测算法识别精确率不高、错检和漏检情况较多等问题,提出了一种基于YOLOv8单阶段西红柿目标检测识别方法.该方法以YOLOv8单阶段目标检测算法为主体,结合CSPDarknet结构改进、引入注意力机制、使用自适应锚框和混合损失函数等,提高算法检测精确率和效率.实验结果表明,YOLOv8模型在复杂环境下依然能够高效、准确地识别西红柿目标,展现出良好的鲁棒性,为农业自动化中的目标检测提供了可靠的技术支持.

Abstract

Aiming at the problems of low recognition precision,many false detection and missed detection of tomato target detection algorithms based on the YOLO series in the current complex environment,a single-stage tomato target detection and recognition method based on YOLOv8 was proposed.This method takes the single-stage target detection algorithm of YOLOv8 as the main body,combined with the improvement of the CSPDarknet structure,the introduction of attention mechanism,the use of adaptive anchor boxes and mixed loss functions,etc,to improve the detection precision and efficiency of the algorithm.Experimental results show that the YOLOv8 model can still efficiently and accurately identify tomato targets in complex environment with good robustness,which provides reliable technical support for target detection in agricultural automation.

关键词

西红柿 / YOLOv8 / CSPDarknet结构 / 注意力机制 / 混合损失函数

Key words

tomato / YOLOv8 / CSPDarknet structure / attention mechanism / mixed loss function

参考文献

[1] ZHAO Y S, GONG L, HUANG Y X, et al. Robust tomato recognition for robotic harvesting using feature images fusion[J]. Sensors, 2016, 16(2): 173.

[2] 王玉飞, 尹建军, 仲苏玉, 等. 基于特征分析的番茄识别及其软件的快速开发[J]. 计算机工程与设计, 2011, 32(11): 3824-3827.

[3] CHEN X M, YANG S X. A practical solution for ripe tomato recognition and localisation[J]. Journal of Real—Time Image Processing, 2013, 8(1): 35-51.

[4] IWASAKI F, IMAMURA H. A robust recognition method for occlusion of mini tomatoes based on hue information and the curvature[J]. International Journal of Image and Graphics, 2015, 15(2): 1540004.

[5] 赵源深, 贡亮, 周斌, 等. 番茄采摘机器人非颜色编码化目标识别算法研究[J]. 农业机械学报, 2016, 47(7): 1-7.

[6] REN S Q, HE K, GIRSHICK R, et al. Faster R—CNN:Towards real—time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(6): 1137-1149.

[7] LIU W, ANGUELOV D, ERHAN D, et al. SSD:Single shot multibox detector[C]// Computer Vision—ECCV 2016:14th European Conference,October 11—14,2016,Amsterdam,The Netherlands.Berlin:Springer, 2016: 21-37.

[8] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]// 2017 IEEE International Conference on Computer Vision(ICCV),October 22—29, 2017,Venice,Italy.IEEE, 2017: 2980-2988.

[9] REDMON J, FARHADI A. YOLO9000:Better,faster,stronger[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),July 21—26, 2017,Hawaii,United States.IEEE, 2017: 7263-7271.

[10] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4:Optimal speed and accuracy of object detection[J]. ArXiv, 2020: 1-27.