1.College of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing 210037,China
2.State Key Laboratory of Integrated Management of Pest Insects and Rodents in Agriculture,Institute of Zoology,Chinese Academy of Sciences,Beijing 100101,China
As an important component of the ecosystem, the dynamic monitoring of wildlife is of great significance for maintaining ecological balance, understanding species interactions, and assessing the health status of the ecosystem. Wildlife monitoring mainly relies on unmanned aerial vehicle (UAV) onboard cameras and fixed infrared cameras to capture the natural behavior of animals. However, due to the unpredictability of wildlife behavior, there are issues with small targets, multi-scale variations, and animal body occlusion in the actual tracking process. To address these challenges, this paper proposed an animal target tracking method based on an improved Siamese network, which transforms the tracking problem into a similarity learning problem. Introducing multi head attention mechanism in the feature extraction stage of Siamese relation network (SiamRN), including concatenated window self-attention operation and sliding window self-attention operation, the precise tracking ability of small targets is enhanced. At the same time, the introduction of multi head attention mechanism reduces the number and complexity of network parameters and improves computational efficiency. The experiment was conducted on both public and self-made datasets, and the results showed that the success rate and accuracy of the wild animal tracking method used in this paper are 0.698 and 0.928, respectively, which was superior to mainstream Siamese network tracking methods. The method proposed in this paper can accurately track and locate wildlife targets, achieving wildlife monitoring.
早期野生动物追踪采用射频识别(radio frequency identification,RFID)技术,RFID技术通过给动物植入耳标、佩戴项圈或腿带等装置,实现了对动物个体的识别和追踪[3]。GPS定位装置也被用于捕获、传输动物的位置数据[4]。Ullmann et al.[5]还利用加速度计记录了动物运动数据,进而分析动物的行为模式。此外,结合北斗与GPS定位技术的卫星追踪器也被应用于鸟类的救助与放飞过程[6-7]。这些方法丰富了野生动物监测方式,为野生动物保护提供了重要的数据。但值得注意的是,这些方法需要动物携带RFID或GPS装置,这些携带的装置可能会对动物造成一定的压力,进而影响其自然行为。
计算机视觉和深度学习技术在人工智能领域发展迅速,能够在数据收集的同时,最大限度地减少对野生动物的干扰,因此,其展现出了巨大的潜力[8]。张雪莹等[9]展示了深度学习技术在野生动物监测中的显著进展。在野生动物的检测和识别方面,杨帆[10]通过引入注意力机制改进YOLOv3算法,有效提升了对野生动物的检测和识别能力。Kim et al.[11]利用Faster R-CNN模型对26种亚马逊鹦鹉进行了分类。黄志静等[12]提出了一种基于深度残差收缩网络的野生动物识别模型,该模型通过减少图像中的噪声干扰来提高野生动物图像识别的准确性。这些研究成果不仅推动了野生动物保护工作的发展,也为计算机视觉技术在生态保护领域的应用提供了新的思路和方法。
LIX Y. Research on the current situation and application of technical methods of wildlife monitoring[J]. Journal of Agricultural Catastrophology, 2023, 13(10): 16-18.
SIMANUNGKALITG, BARWICKJ, COWLEYF, et al. Use of an ear-tag accelerometer and a radio-frequency identification(RFID) system for monitoring the licking behaviour in grazing cattle[J]. Applied Animal Behaviour Science, 2021, 244: 105491.
[6]
ARAVINDD, ANUPRIYAS, AARTHIA T, et al. An automatic wildlife tracking system using GPS and wireless sensor networks[R/OL]. Pune: National Conference on Electronics, Communication and Computing (NCECC-2017), 2017 [2024-10-28].
[7]
ULLMANNW, FISCHERC, KRAMER-SCHADTS, et al. The secret life of wild animals revealed by accelerometer data: how landscape diversity and seasonality influence the behavioural types of European hares[J]. Landscape Ecology, 2023, 38(12): 3081-3095.
LIX M, LIX M, SONGY B, et al. Analysis of movements of a juvenal great egret on the Sanjiang Plain by satellite tracking[J]. Chinese Journal of Wildlife, 2015, 36(4): 395-401.
HAOY, CHENJ W, YAND, et al. Wild bird rescue-release and satellite tracking in Changling Longfeng Lake wetland[J]. Chinese Journal of Wildlife, 2019, 40(4): 964-968.
[12]
KAYSR, CROFOOTM C, JETZW, et al. Terrestrial animal tracking as an eye on life and planet[J]. Science, 2015, 348(6240): aaa2478.
ZHANGX Y, ZHANGH L, HANY Y, et al. Research progress of the wildlife monitoring and identification based on deep learning[J]. Chinese Journal of Wildlife, 2022, 43(1): 251-258.
YANGF. Research and application of wild animal recognition under artificial intelligence[J]. China High and New Techno-logy, 2024(16): 72-74.
[17]
KIMJ I, BAEKJ W, KIMC B. Image classification of Amazon parrots by deep learning: a potentially useful tool for wildlife conservation[J]. Biology, 2022, 11(9): 1303.
HUANGZ J, SHAOM Y, ZHANGT R, et al. Wild life recognition based on deep learning[J]. Electronic Test, 2022(22): 69-71; 10.
[20]
MANOHARN, SHARATH KUMARY H, KUMARG H. An approach for the development of animal tracking system[J]. International Journal of Computer Vision and Image Processing, 2018, 8(1): 15-31.
YANJ Y, XIEY H. Real time tracking of wild animals using unmanned aerial vehicles based on YOLOv7-Tiny algorithm[J]. Chinese Journal of Wildlife, 2024, 45(2): 251-261.
MAG K, WUW, LIUD, et al. Automatic tracking method of group activities of captive Panthera tigrisaltaica cubs[J]. Chinese Journal of Wildlife, 2022, 43(3): 605-613.
XUQ, XIEY H. Automatic individual tracking method of Amur tiger based on attention mechanism fusion of multiple features[J]. Biodiversity Science, 2024, 32(3): 141-154.
[27]
BERTINETTOL, VALMADREJ, HENRIQUESJ F, et al. Fully-convolutional Siamese networks for object tracking[C]//HUA G, JÉGOU H. Computer vision: ECCV 2016 workshops. Cham: Springer International Publishing, 2016: 850-865.
[28]
LIB, YANJ J, WUW, et al. High performance visual tracking with Siamese region proposal network[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 18-23, 2018. Salt Lake City: IEEE, 2018: 8971-8980.
[29]
CHENZ D, ZHONGB N, LIG R, et al. Siamese box adaptive network for visual tracking[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), June 13-19, 2020. Seattle: IEEE, 2020: 6668-6677.
[30]
CHENGS Y, ZHONGB N, LIG R, et al. Learning to filter: Siamese relation network for robust tracking[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), June 20-25, 2021.Nashville: IEEE, 2021: 4421-4431.
[31]
LIUZ, LINY T, CAOY, et al. Swin Transformer: hierarchical vision transformer using shifted windows[C]//2021 IEEE/CVF International Conference on Computer Vision(ICCV), October 10-17, 2021. Montreal: IEEE, 2021: 10012-10022.
[32]
SVANSTRÖMF, ALONSO-FERNANDEZF, ENGLUNDC. A dataset for multi-sensor drone detection[J]. Data in Brief, 2021, 39: 107521.
WANGJ X, PANX. An auxiliary labeling method based on labelimg[J]. Technology Innovation and Application, 2023, 13(29): 145-148.
[35]
CHICCOD. Siamese neural networks: an overview[M]//CARTWRIGHT H. Methods in molecular biology: Vol.2190: artificial neural networks. 3rd ed. New York: Humana Press, 2020: 73-94.
[36]
THRUNS. Is learning the n-th thing any easier than learning the first?[C]//TOURETZKY D S, MOZER M C, HASSELMO M E. Proceedings of the 9th international conference on neural information processing systems. Cambridge: MIT Press, 1996: 640-646.
YANGM L, ZHANGX, GUOY, et al. Recognition of wild animals using infrared camera images based on YOLOv5[J]. Laser & Optoelectronics Progress, 2022, 59(12): 1215015.
[41]
FLEMINGC H, DRESCHER-LEHMANJ, NOONANM J, et al. A comprehensive framework for handling location error in animal tracking data[EB/OL]. BioRxiv (2021-07-24) [2024-10-08].