A dead tree detection algorithm based on improved YOLOv4-tiny for UAV images
-
摘要: 针对目前枯死树木检测主要依赖人工实地勘察,容易受到森林地形限制、勘察效率低、易发生危险等问题,提出一种引进注意力机制及空间金字塔池化的YOLOv4-tiny枯死树木检测算法。首先,该方法在模型的Backbone部分后引入空间金字塔池化(spatial pyramid pooling,SPP)结构,融合局部和全局特征,丰富模型的特征表达能力; 其次,使用ELU替换模型中原激活函数LeakyReLU,使得激活函数单侧饱和,能够更好地收敛,提升模型鲁棒性; 最后,将注意力机制ECANet引入模型中,加强网络对图像中重要信息的学习,提升网络的性能。实验利用无人机采集辽南某风景区山林的树木影像,并使用不同模型对其中枯死树木进行检测。通过实验结果表明,改进算法检测精度达到93.25%,相比于YOLOv4-tiny,YOLOv4,SSD和文献[8]算法,精度分别提升9.58%,12.57%,10.54%和4.87%,能够较好地实现对于枯死树木的检测。
-
关键词:
- 枯死树木 /
- YOLOv4-tiny /
- 注意力机制 /
- SPP /
- ELU激活函数
Abstract: The current dead tree detection primarily relies on manual field surveys and, thus, is limited by forest topography, suffers a low detection efficiency, and is dangerous. Given these problems, this study proposed a YOLOv4-tiny dead tree detection algorithm based on the attention mechanism and spatial pyramid pooling (SPP) and improved the original detection model. First, the SPP structure was introduced after the Backbone part of the model to combine local and global features and enrich the feature representation capability of the model. Then, the original activation function LeakyReLU in the model was replaced with ELU, which made the activation function saturate unilaterally, thus improving the convergence and robustness of the model. Finally, the attention mechanism ECANet was introduced into the model to enhance the capacity of the network to learn important information in images, thus improving the performance of the network. The images of trees in a mountain forest of a scenic area in southern Liaoning were collected using an unmanned aerial vehicle (UAV). Then, dead trees in these images were detected using different models. The detection results show that the improved algorithm had a detection accuracy of 93.25%, which was improved by 9.58%, 12.57%, 10.54%, and 4.87% than that of the YOLOv4-tiny, YOLOv4, and SSD algorithms and an algorithm stated in literature [8], respectively, and achieved the effective detection of dead trees.-
Key words:
- dead tree /
- YOLOv4-tiny /
- attention mechanism /
- SPP /
- ELU activation function
-
-
[1] Kamińska A, Lisiewicz M, Stereńczak K, et al. Species-related single dead tree detection using multi-temporal ALS data and CIR imagery[J]. Remote Sensing of Environment, 2018, 219:31-43.
[2] 毕凯, 李英成, 丁晓波, 等. 轻小型无人机航摄技术现状及发展趋势[J]. 测绘通报, 2015(3):27-31,48.
[3] Bi K, Li Y C, Ding X B, et al. Aerial photogrammetric technology of light small UAV:Status and trend of development[J]. Bulletin of Surveying and Mapping, 2015(3):27-31,48.
[4] 汪沛, 罗锡文, 周志艳, 等. 基于微小型无人机的遥感信息获取关键技术综述[J]. 农业工程学报, 2014, 30(18):1-12.
[5] Wang P, Luo X W, Zhou Z Y, et al. Key technology for remote sensing information acquisition based on micro UAV[J]. Transactions of the Chinese Society of Agricultural Engineering, 2014, 30(18):1-12.
[6] Kamińska A, Lisiewicz M, K Stereńczak, et al. Species-related single dead tree detection using multi-temporal ALS data and CIR imagery[J]. Remote Sensing of Environment, 2018, 219:31-43.
[7] Manandhar A, Hoegner L, Stilla U. Palm tree detection using circular autocorrelation of polar shape matrix[J]. ISPRS Annals of Photogrammetry,Remote Sensing and Spatial Information Sciences, 2016, 3:465-472.
[8] 宋以宁, 刘文萍, 骆有庆, 等. 基于线性谱聚类的林地图像中枯死树监测[J]. 林业科学, 2019, 55(4):187-195.
[9] Song Y N, Liu W P, Luo Y Q, et al. Monitoring of dead trees in forest images based on linear spectral clustering[J]. Scientia Silvae Sinicae, 2019, 55(4):187-195.
[10] Culman M, Delalieux S, Van Tricht K. Individual palm tree detection using deep learning on RGB imagery to support tree inventory[J]. Remote Sensing, 2020, 12(21):3476.
[11] 王新彦, 吕峰, 易政洋. 基于深度学习的草坪树木检测方法研究[J]. 中国农机化学报, 2021, 42(7):136-141.
[12] Wang X Y, Lyu F, Yi Z Y. Research on lawn tree detection method based on deep learning[J]. Journal of Chinese Agricultural Mechanization, 2021, 42(7):136-141.
[13] Yu R, Luo Y, Zhou Q, et al. Early detection of pine wilt disease using deep learning algorithms and UAV-based multispectral imagery[J]. Forest Ecology and Management, 2021, 497:119493.
[14] 李海滨, 孙远, 张文明, 等. 基于YOLOv4-tiny的溜筒卸料煤尘检测方法[J]. 光电工程, 2021, 48(6):73-86.
[15] Li H B, Sun Y, Zhang W M, et al. The detection method for coal dust caused by chute discharge based on YOLOv4-tiny[J]. Opto-Electronic Engineering, 2021, 48(6):73-86.
[16] Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4:Optimal speed and accuracy of object detection[J]. Computer Vision and Pattern Recognition, 2020, 17(9):198-215.
[17] Lin T Y, Dollár P, Girshick R, et al. Feature pyramid networks for object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017:2117-2125.
[18] He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9):1904-1916.
[19] Clevert D A, Unterthiner T, Hochreiter S. Fast and accurate deep network learning by exponential linear units (ELUS)[C]// Proceedings of the 4th International Conference on Learning Representations.ICLR, 2015:375-387.
[20] Wang Q, Wu B, Zhu P, et al. ECA-Net:Efficient channel attention for deep convolutional neural networks[C]// CVF Conference on Computer Vision and Pattern Recognition.Seattle.IEEE, 2020:11531-11539.
[21] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018:7132-7141.
[22] 赵杰伦, 张兴忠, 董红月. 基于尺度不变特征金字塔的输电线路缺陷检测[J]. 计算机工程与应用, 2022, 58(8):289-296.
[23] Zhao J L, Zhang X Z, Dong H Y. Defect detection in transmission line based on scale-invariant feature pyramid networks[J]. Computer Engineering and Applications, 2022, 58(8):289-296.
[24] Everingham M, Van Gool L, Williams C K I, et al. The pascal visual object classes (voc) challenge[J]. International Journal of Computer Vision, 2010, 88(2):303-338.
[25] 郭晓征, 姚云军, 贾坤, 等. 基于U-Net深度学习方法火星沙丘提取研究[J]. 自然资源遥感, 2021, 33(4):130-135.doi:10.6046/zrzyyg.2020397.
[26] Guo X Z, Yao Y J, Jia K, et al. Information extraction of Mars dunes based on U-Net[J]. Remote Sensing for Natural Resources, 2021, 33(4):130-135.doi:10.6046/zrzyyg.2020397.
-
计量
- 文章访问数: 1363
- PDF下载数: 86
- 施引文献: 0
下载: