基于YOLOv5s的蝴蝶种类检测
Butterfly Species Detection Based on YOLOv5s
DOI: 10.12677/MOS.2023.126529, PDF,    科研立项经费支持
作者: 赵凯旋*, 覃 林：贵州大学大数据与信息工程学院，贵州 贵阳；教育部功率半导体器件可靠性工程中心，贵州 贵阳；谢本亮：贵州大学大数据与信息工程学院，贵州 贵阳；教育部功率半导体器件可靠性工程中心，贵州 贵阳；贵州大学公共大数据国家重点实验室，贵州 贵阳
关键词: 蝴蝶识别；目标检测；YOLOv5s；CSandGlass；SE注意力；Butterfly Recognition； Object Detection； YOLOv5s； CSandGlass； SE Attention
摘要: 蝴蝶对周围环境敏感，能作为反应生态环境的指示物种，因此对其进行识别研究对研究生态稳定性具有重大意义。但蝴蝶分类细致，相似度高，传统识别方法效率低。为解决上述问题，本文以野外蝴蝶图像的种类自动识别为目标，提出了一种基于YOLOv5s的改进的目标检测方法。为了减少信息丢失，提高精度，在YOLOv5s的主干特征提取网络上设计了CSandGlass模块来代替残差模块；并加入了SE注意力机制和对损失函数进行改进。实验结果表明，改进后模型平均精度为92.6%，相比原模型平均精度提升2%，且具有较强的鲁棒性和稳定性，可满足自然环境下的蝴蝶种类识别需求。
Abstract: Butterflies are sensitive to their surroundings and can be used as indicator species responding to the ecological environment, so their identification studies are of great significance for studying eco-logical stability. However, butterfly classification is highly detailed with high similarity, and the tra-ditional recognition methods are inefficient. In order to solve the above problems, this paper pro-poses an improved object detection method based on YOLOv5s with the goal of automatic species identification of butterfly images. In order to reduce the information loss and improve the accuracy, the CSandGlass module is designed on the backbone feature extraction network of YOLOv5s to re-place the residual module; and the SE attention mechanism and the loss function are added and improved. The average accuracy of the improved model is 92.6%, compared with the original model, the average accuracy is improved by 2%, and has strong robustness and stability, which can meet the demand of butterfly species recognition in natural environment.
文章引用：赵凯旋, 覃林, 谢本亮. 基于YOLOv5s的蝴蝶种类检测[J]. 建模与仿真, 2023, 12(6): 5834-5842. https://doi.org/10.12677/MOS.2023.126529

参考文献

[1] Kumar, S., Simonson, S.E. and Stohlgren, T.J. (2009) Effects of Spatial Heterogeneity on Butterfly Species Richness in Rocky Mountain National Park, CO, USA. Biodiversity and Conservation, 18, 739-763. [Google Scholar] [CrossRef] 
[2] 尹晶, 施雯, 王灵敏, 等. 云南省元阳县蝴蝶群落结构与物种多样性研究[J/OL]. 云南农业大学学报(自然科学): 1-8.
http://kns.cnki.net/kcms/detail/53.1044.S.20231010.0927.002.html
[3] 武春生, 徐堉峰. 中国蝴蝶图鉴[M]. 福州: 海峡出版发行集团, 2017.
[4] Grajales-Múnera, J.E., and Restrepo-Martinez, A. (2013) Clasificación de Mariposas por Modelos de Color HSI y RGB Usando Redes Neuronales. Tecno Lógicas, 669. [Google Scholar] [CrossRef] 
[5] 谢娟英, 鲁银圆, 孔维轩, 等. 基于改进RetinaNet的自然环境中蝴蝶种类识别[J]. 计算机研究与发展, 2021, 58(8): 1686-1704.
[6] 李飞, 赵凯旋, 严春雨, 等. 基于残差网络的自然环境下蝴蝶种类识别[J]. 昆虫学报, 2023, 66(3): 409-418.
[7] Redmon, J., Divvala, S., Girshick, R., and Farhadi, A. (2016) You Only Look Once: Unified, Real-Time Object Detection. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, 27-30 June 2016, 779- 788. [Google Scholar] [CrossRef] 
[8] Redmon, J., and Farhadi, A. (2017) YOLO9000: Better, Faster, Stronger. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, 21-26 July 2017, 6517-6525. [Google Scholar] [CrossRef] 
[9] Redmon, J.F.A. (2018) Yolov3: An incremental improvement. arXiv: abs/1804.02767.
[10] Liang, B., Wu, S., Xu, K., and Hao, J. (2020) Butterfly Detection and Classification Based on Integrated YOLO Algorithm. Genetic and Evolutionary Computing, Vol. 1107, Springer, Singapore, 500-512. [Google Scholar] [CrossRef] 
[11] Zhou, D., Hou, Q., Chen, Y., Feng, J., and Yan, S. (2020) Rethink-ing Bottleneck Structure for Efficient Mobile Network Design. In: Vedaldi, A., Bischof, H., Brox, T., and Frahm, J.-M., Eds., Computer Vision—ECCV, Springer International Publishing, Cham, 680-697. [Google Scholar] [CrossRef] 
[12] Hu, J., Shen, L., Sun, G. and Li, Y. (2018) Squeeze-and-Excitation Networks. IEEE Conference on Computer Vision and Pattern Recognition, 71, 32-41. [Google Scholar] [CrossRef] 
[13] 郑远攀, 许博阳, 王振宇. 改进的YOLOv5烟雾检测模型[J]. 计算机工程与应用, 2023, 59(7): 214-221.
[14] 王红尧, 韩爽, 李勤怡. 改进YOLOv5的钢丝绳损伤图像识别实验方法研究[J]. 计算机工程与应用, 2023, 59(17): 99-106.
[15] 杨谢柳, 门国文, 梁文峰, 等. 水下图像增强与复原对深度学习目标检测精度的影响研究[J/OL]. 计算机工程: 1-10.[CrossRef] 

为你推荐