The camera traps is a common method to monitor wildlife. A large amount of data and complex background information make it difficult to label and detect. In order to solve the above problems, a wild animal detection method based on pseudo-labels and YOLOv4 was proposed. Firstly, a pseudo-labeling calibration method based on motion detection was proposed to realize automatic and fast calibration of video datasets by background difference method and morphological operation, which solved the problem of difficult automatic labeling caused by complex background in monitoring data. Then, a cross stage partial block was proposed to reduce the amount of computation required for the path aggregation network in YOLOv4. Finally, the Swish activation function was introduced in the dense convolution area, which improves the feature extraction ability of the model in the deep region. In this study, the monitoring videos of six kinds of wild animals in Zhejiang Jiangshan Xianxialing Provincial Nature Reserve were used as data sets for experiments. The results show that the method proposed in this study achieves 86.41% mean average precision (mAP) and a frame rate of 18.93 frames per second. These metrics represent improvements of 1.62 percentage points, 3.43 percentage points, and 7.11 percentage points over the YOLOv4, RFCN, and YOLOv8x algorithms, respectively. This demonstrates that the proposed algorithm effectively addresses the labeling and detection challenges faced by existing methods. Balanced improvement of detection mAP and frame rate contributes to automated and intelligent analysis of wildlife monitoring data.
早期主要使用数字图像处理方法来识别图像中的野生动物。如Steen et al.[3]提出了一种基于灰度分布直方图的动物识别方法;初未萌[4]基于边缘检测方法对野生动物图像边缘信息进行分析,并通过Hough变换的方式完成目标检测。图像中野生动物的特征信息会随着环境变化而变化,仅凭简单的特征很难准确识别野生动物。随后,传统机器学习方法被引入到野生动物识别中,如Matuska et al.[5]采用SVM分类器对全局区域算子提取的野生动物特征进行分类,在对5种野生动物的分类实验中,取得了86%的准确率;Kamencay et al.[6]利用主成分分析对野生动物图像中提取的核心特征降维后,输入到多种算法融合的分类模型中,达到了90%的识别准确率。但传统机器学习方法需要手动设计输入特征,无法保证特征的有效性,识别性能有待进一步提升。随着深度学习技术的发展,深度学习已经成为当前野生动物检测的主流方法。其中,目标检测方法利用深层次的卷积神经网络(convolutional neural network,CNN)能自动提取图像特征,并定位目标[7-10],大幅减少了对人工设计特征的依赖,提升了复杂环境下野生动物的检测能力[11-16],但这类方法在训练模型时需要大量的标记数据。目标检测模型的训练数据不仅需要标记类别,还需要给出野生动物的位置。人工标注足够的训练数据会耗费大量人力[17-18],亟需一种新的标注方法,提升标记数据的效率。
DELISLEZ J, HENRICHM, PALENCIAP, et al. Reducing bias in density estimates for unmarked populations that exhibit reactive behaviour towards camera traps [J]. Methods in Ecology and Evolution, 2023, 14(12): 3100-3111.
CHUW M. Research on animal image detection based on genera-lized Hough transform[D]. Harbin: Harbin Institute of Techno-logy, 2013.
[7]
MATUSKAS, HUDECR, KAMENCAYP, et al. Classification of wild animals based on SVM and local descriptors [J]. AASRI Procedia, 2014, 9: 25-30.
[8]
KAMENCAYP, TRNOVSZKYT, BENCOM, et al. Accurate wild animal recognition using PCA, LDA and LBPH [C]//2016 ELEKTRO, May 16-18, 2016. Strbske Pleso: IEEE, 2016: 62-67.
[9]
GIRSHICKR, DONAHUEJ, DARRELLT, et al. Rich feature hierarchies for accurate object detection and semantic segmentation [C]//2014 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), June 23-28, 2014. Columbus: IEEE, 2014: 580-587.
[10]
DAIJ F, LIY, HEK M, et al. R-FCN: object detection via region-based fully convolutional networks[C]//LEE D D, VON LUXBURG U, GARNETT R, et al. NIPS’16: Proceedings of the 30th International Conference on Neural Information Processing Systems. Barcelona: Curran Associates Inc., 2016: 379-387.
[11]
REDMONJ, DIVVALAS, GIRSHICKR, et al. You only look once: unified, real-time object detection [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 27-30, 2016. Las Vegas: IEEE, 2016: 779-788.
[12]
TANM X, PANGR M, LEQ V. EfficientDet: scalable and efficient object detection [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 13-19, 2020. Seattle: IEEE, 2020: 10778-10787.
ZHANGX Y, ZHANGH L, HANY Y, et al. Research pro-gress of the wildlife monitoring and identification based on deep learning [J]. Chinese Journal of Wildlife, 2022, 43(1): 251-258.
[15]
TUIAD, KELLENBERGERB, BEERYS, et al. Perspectives in machine learning for wildlife conservation [J]. Nature Communications, 2022, 13(1): 792.
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.
JIANGJ Z, XIEY H. Identification of wild animal species in airborne thermal imaging using improved youlu model algorithm [J]. Journal of Northeast Forestry University, 2022, 50(3): 109-112;124.
YUW H, XIEY H. Amur tiger detection based on lightweight YOLOX algorithm [J]. Chinese Journal of Wildlife, 2023,44(4):760-767.
[22]
XIEJ J, LIA Q, ZHANGJ G, et al. An integrated wildlife recognition model based on multi-branch aggregation and squeeze-and-excitation network [J]. Applied Sciences, 2019, 9(14): 2794.
[23]
SCHNEIDERS, GREENBERGS, TAYLORG W, et al. Three critical factors affecting automated image species recognition performance for camera traps [J]. Ecology and Evolution, 2020, 10(7): 3503-3517.
[24]
SCHAUSJ, UZALA, GENTLEL K, et al. Application of the random encounter model in citizen science projects to monitor animal densities [J]. Remote Sensing in Ecology and Conservation, 2020,6(4): 514-528.
CHOUY X, ZHANGA H, GUY, et al. Fast mathematical morphological filtering method and its application on pulse signal processing [J]. Chinese Journal of Scientific Instrument, 2020, 41(2): 253-262.
[27]
BOCHKOVSKIYA, WANGC Y, LIAOH M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. arXiv (2020-04-23) [2024-08-23].
[28]
LIUS, QIL, QINH F, et al. Path aggregation network for instance segmentation[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 18-23, 2018. Salt Lake City: IEEE, 2018: 8759-8768.
[29]
RAMACHANDRANP, ZOPHB, LEQ V. Swish: a self-gated activation function[EB/OL]. arXiv(2017-10-16)[2024-08-23].
[30]
HOWARDA, SANDLERM, CHENB, et al. Searching for MobileNetV3[C]//2019 IEEE/CVF International Conference on Computer Vision(ICCV), October 27-November 2, 2019. Seoul: IEEE, 2019: 1314-1324.
[31]
XUP F, ZHANGY Y, JIM H, et al. Advanced intelligent monitoring technologies for animals: a survey[J]. Neurocomputing, 2024, 585: 127640.