基于双层路由注意力和自校准卷积的豹个体识别

doi:10.3969/j.issn.0255-8297.2025.02.012

应用科学学报 ›› 2025, Vol. 43 ›› Issue (2): 348-360.doi: 10.3969/j.issn.0255-8297.2025.02.012

• 计算机科学与应用 • 上一篇

基于双层路由注意力和自校准卷积的豹个体识别

杨婉¹, 陈爱斌¹, 赵莹², 武阅², 甑鑫², 肖治术³

1. 中南林业科技大学人工智能应用研究所, 湖南长沙 410004;
2. 中国猫科动物保护联盟, 北京 100875;
3. 中国科学院动物研究所, 北京 100101

收稿日期:2024-07-11 出版日期:2025-03-30 发布日期:2025-04-03
通信作者: 陈爱斌，教授，研究方向为深度学习、音频学习、机器学习。E-mail:hotaibin@163.com
基金资助:
国家自然科学基金（No.62276276）；湖南省自然科学基金（No.2024JJ5647）资助

Leopards Individual Recognition Based on Bi-level Routing Attention and Self-Calibrated Convolution

YANG Wan¹, CHEN Aibin¹, ZHAO Ying², WU Yue², ZHEN Xin², XIAO Zhishu³

1. Institute of Artificial Intelligence Application, Central South University of Forestry and Technology, Changsha 410004, Hunan, China;
2. Chinese Felid Conservation Alliance, Beijing 100875, China;
3. Institute of Zoology, Chinese Academy of Sciences, Beijing 100101, China

Received:2024-07-11 Online:2025-03-30 Published:2025-04-03

摘要/Abstract

摘要： 自然环境中豹的图像在用于个体识别任务时，个体与环境融合度高、类间相似性高这两个因素会导致识别困难，为此结合自校准卷积和双层路由注意力，提出了一种改进的EfficientNet模型。自校准卷积能够自适应地在每个空间位置周围构建远程空间和通道间的依赖关系，并显式地结合更丰富的信息来增强对细节特征的识别能力，解决了类间相似性高带来的识别难题。双层路由注意力结合自顶向下的全局注意力和自底向上的局部注意力，解决了个体与环境融合度高的问题。实验结果显示，改进后的模型在豹个体识别任务上的准确率达到了95.56%，显著高于原始的EfficientNet模型，证明了所提出的模型在处理豹个体识别任务上的有效性和先进性。

关键词: 个体识别, 自校准卷积, 双层路由注意力, 深度学习, 自建数据集

Abstract: Infrared camera images of leopards in natural environments pose significant challenges for individual recognition due to issues such as high fusion between individuals and their surroundings, as well as high inter-class similarity. To address these challenges, an improved EfficientNet model is proposed, incorporating self-calibrating convolution and bilevel routing attention. The self-calibrating convolution adaptively builds remote space and inter-channel dependencies around each spatial location. The ability to recognize detailed features is enhanced by explicitly combining richer contextual information. This effectively mitigates the recognition challenges posed by high inter-class similarity. Meanwhile, the bilevel routing attention combines the top-down global attention strategy and the bottom-up local attention strategy to solve the problem of high integration between individuals and their environment. Experiment results show that the accuracy of the proposed model reaches 95.56% in the task of leopard individual recognition, which is significantly higher than the original EfficientNet. These findings validate the effectiveness and superiority of the proposed model in dealing with leopard individual recognition task.

Key words: individual recognition, self-calibrating convolution, bi-level routing attention, deep learning, self-built data set

中图分类号:

TP391.4

杨婉, 陈爱斌, 赵莹, 武阅, 甑鑫, 肖治术. 基于双层路由注意力和自校准卷积的豹个体识别[J]. 应用科学学报, 2025, 43(2): 348-360.

YANG Wan, CHEN Aibin, ZHAO Ying, WU Yue, ZHEN Xin, XIAO Zhishu. Leopards Individual Recognition Based on Bi-level Routing Attention and Self-Calibrated Convolution[J]. Journal of Applied Sciences, 2025, 43(2): 348-360.

参考文献

[1] 肖治术, 李欣海, 王学志, 等. 探讨我国森林野生动物红外相机监测规范[J]. 生物多样性, 2014, 22(6): 704-711. Xiao Z S, Li X H, Wang X Z, et al. Developing camera-trapping protocols for wildlife monitoring in Chinese forests [J]. Biodiversity Science, 2014, 22(6): 704-711. (in Chinese)
[2] 钟俊杰, 钮冰, 陈沁, 等. 深度学习在野生动物保护中的应用[J]. 兽类学报, 2023, 43(6): 734-744. Zhong J J, Niu B, Chen Q, et al. Application of deep learning in wildlife conservation [J]. Acta Theriologica Sinica, 2023, 43(6): 734-744. (in Chinese)
[3] Dwivedi Y K, Hughes L, Ismagilova E, et al. Artificial intelligence (AI): multidisciplinary perspectives on emerging challenges, opportunities, and agenda for research, practice and policy [J]. International Journal of Information Management, 2021, 57: 101994.
[4] Jordan M I, Mitchell T M. Machine learning: trends, perspectives, and prospects [J]. Science, 2015, 349(6245): 255-260.
[5] 赵婷婷, 周哲峰, 李东喜, 等. 基于改进的Cifar-10深度学习模型的金钱豹个体识别研究[J]. 太原理工大学学报, 2018, 49(4): 585-591, 598. Zhao T T, Zhou Z F, Li D X, et al. Individual identification of leopard based on improved Cifar-10 deep learning model [J]. Journal of Taiyuan University of Technology, 2018, 49(4): 585-591, 598. (in Chinese)
[6] Zeng D, Veldhuis R, Spreeuwers L. A survey of face recognition techniques under occlusion [J]. IET Biometrics, 2021, 10(6): 581-606.
[7] Wang P, Fan E, Wang P. Comparative analysis of image classification algorithms based on traditional machine learning and deep learning [J]. Pattern Recognition Letters, 2021, 141: 61-67.
[8] Bernal J, Kushibar K, Asfaw D S, et al. Deep convolutional neural networks for brain image analysis on magnetic resonance imaging: a review [J]. Artificial Intelligence in Medicine, 2019, 95: 64-81.
[9] Si C Y, Chen W T, Wang W, et al. An attention enhanced graph convolutional LSTM network for skeleton-based action recognition [C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 1227-1236.
[10] Esteva A, Chou K, Yeung S, et al. Deep learning-enabled medical computer vision [J]. NPJ Digital Medicine, 2021, 4(1): 5.
[11] Cha S, Lim J, Kim K, et al. Deepening the accuracy of tree species classification: a deep learning-based methodology [J]. Forests, 2023, 14(8): 1602.
[12] Atila Ü, Uçar M, Akyol K, et al. Plant leaf disease classification using EfficientNet deep learning model [J]. Ecological Informatics, 2021, 61: 101182.
[13] Liu J J, Hou Q, Cheng M M, et al. Improving convolutional networks with self-calibrated convolutions [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 10093-10102.
[14] Zhu L, Wang X J, Ke Z H, et al. BiFormer: vision transformer with bi-level routing attention [C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 10323-10333.
[15] Lee T, Na Y, Kim B G, et al. Identification of individual Hanwoo cattle by muzzle pattern images through deep learning [J]. Animals, 2023, 13(18): 2856.
[16] Takaya K, Taguchi Y, Ise T. Individual identification of endangered amphibians using deep learning and smartphone images: case study of the Japanese giant salamander (Andrias japonicus) [J]. Scientific Reports, 2023, 13(1): 16212.
[17] Luo C Y, Cheng S Y, Xu H, et al. Human behavior recognition model based on improved EfficientNet [J]. Procedia Computer Science, 2022, 199: 369-376.
[18] Zhu X Z, Cheng D Z, Zhang Z, et al. An empirical study of spatial attention mechanisms in deep networks [C]//2019 IEEE/CVF International Conference on Computer Vision, 2019: 6687-6696.
[19] Scherer D, Müller A, Behnke S. Evaluation of pooling operations in convolutional architectures for object recognition [C]//International Conference on Artificial Neural Networks, 2010: 92-101.
[20] 宋大昭, 王卜平, 蒋进原, 等. 山西晋中庆城林场华北豹及其主要猎物种群的红外相机监测[J]. 生物多样性, 2014, 22(6): 733-736. Song D Z, Wang B P, Jiang J Y, et al. Using camera trap to monitor a North Chinese leopard (Panthera pardus japonesis) population and their main ungulate prey [J]. Biodiversity Science, 2014, 22(6): 733-736. (in Chinese)
[21] Mathis A, Biasi T, Schneider S, et al. Pretraining boosts out-of-domain robustness for pose estimation [C]//IEEE/CVF Winter Conference on Applications of Computer Vision, 2021: 1858-1867.
[22] Song H, Kim M, Lee J G. SELFIE: refurbishing unclean samples for robust deep learning [C]//International Conference on Machine Learning, 2019: 5907-5915.

基于双层路由注意力和自校准卷积的豹个体识别

Leopards Individual Recognition Based on Bi-level Routing Attention and Self-Calibrated Convolution

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	张蕾, 李萄苑, 李明飞, 邓海敏, 谢诚. 基于“面-廓-色”多通道特征融合的多肉叶片繁育评分算法[J]. 应用科学学报, 2026, 44(2): 316-329.
[2]	吴文强, 陈爱斌, 李潇瑶. 基于特征融合注意力和对比学习的森林图像去雾[J]. 应用科学学报, 2026, 44(1): 97-109.
[3]	鲁桢, 吴建斌. 基于字符识别的无载体隐写[J]. 应用科学学报, 2025, 43(4): 684-693.
[4]	邵子龙, 漆林, 陈昆, 许玉斌, 秦昆, 余长慧. 密集多实例建筑物场景变化检测[J]. 应用科学学报, 2025, 43(4): 656-671.
[5]	汪婉灵, 熊邦书, 欧巧凤, 余磊, 饶智博. 基于暗区域引导的低照度图像增强[J]. 应用科学学报, 2025, 43(2): 245-256.
[6]	吴祥, 肖剑, 吉根林. 基于多尺度特征融合和注意力机制的视频异常检测方法[J]. 应用科学学报, 2025, 43(2): 234-244.
[7]	胡汪鑫, 成英超, 何玉林, 黄哲学, 蔡占川. 脉冲神经网络基准测试及类脑训练框架性能评估[J]. 应用科学学报, 2025, 43(1): 169-182.
[8]	邓抒憧, 陈爱斌, 戴子健. 基于多路激励和金字塔切分注意力的鸟类行为识别[J]. 应用科学学报, 2025, 43(1): 154-168.
[9]	王莹笑, 杨彦红, 谭云峰. 面向视频的人脸特征计算方法[J]. 应用科学学报, 2025, 43(1): 137-153.
[10]	刘珈宁, 张思佳, 张正龙, 王祎涵, 安宗诗. 基于多跳机制的扩散图谱推荐模型[J]. 应用科学学报, 2025, 43(1): 123-136.
[11]	贾志洋, 许兆, 冷艳梅, 闻新, 龚浩宇. 基于并行优化CBAM的轻量级故障诊断模型[J]. 应用科学学报, 2025, 43(1): 94-109.
[12]	段新涛, 白鹿伟, 徐凯欧, 张萌, 保梦茹, 武银行, 秦川. 基于卷积神经网络的轻量高效图像隐写[J]. 应用科学学报, 2025, 43(1): 80-93.
[13]	王冰轮, 方巍. 基于U-Net改进的日平均2 m气温订正方法[J]. 应用科学学报, 2025, 43(1): 51-65.
[14]	唐意恒, 王永雄, 王哲, 张晓理. 基于子领域适应和时空学习的脑电情感识别[J]. 应用科学学报, 2024, 42(6): 1016-1026.
[15]	贺加贝, 周菊香, 甘健侯, 吴迪, 温晓宇. 基于多任务学习的课堂表情分类模型[J]. 应用科学学报, 2024, 42(6): 947-961.