生物多样性 ›› 2026, Vol. 34 ›› Issue (2): 25256.  DOI: 10.17520/biods.2025256  cstr: 32101.14.biods.2025256

• 技术与方法 • 上一篇    下一篇

基于Diff-SCC模型的偏态分布野生动物识别方法

纪林1,2,3, 邓宸迅1,2,3, 王丽凤1,2,3, 王德港1,2,3, 王建涛4, 于永永4, 张军国1,2,3*   

  1. 1. 北京林业大学工学院, 北京 100083; 2. 林木资源高效生产全国重点实验室, 北京 100083; 3. 北京林业大学生物多样性智慧监测研究中心, 北京 100083; 4. 内蒙古乌兰坝国家级自然保护区管理局, 内蒙古赤峰 025450
  • 收稿日期:2025-07-02 修回日期:2026-01-12 接受日期:2026-02-28 出版日期:2026-02-20 发布日期:2026-03-23
  • 通讯作者: 张军国
  • 基金资助:
    国家自然科学基金(32371874); 科技部雄安新区科技创新专项(2023XAGG0065); 北京市自然科学基金项目(6192019)

A wildlife recognition method for skewed distributions based on the Diff-SCC model

Lin Ji1,2,3, Chenxun Deng1,2,3, Lifeng Wang1,2,3, Degang Wang1,2,3, Jiantao Wang4, Yongyong Yu4, Junguo Zhang1,2,3*   

  1. 1 School of Technology, Beijing Forestry University, Beijing 100083, China 

    2 State Key Laboratory of Efficient Production of Forest Resources, Beijing 100083, China 

    3 Research Center for Biodiversity Intelligent Monitoring, Beijing Forestry University, Beijing 100083, China 

    4 Administration of Ulanba National Nature Reserve, Chifeng, Inner Mongolia 025450, China

  • Received:2025-07-02 Revised:2026-01-12 Accepted:2026-02-28 Online:2026-02-20 Published:2026-03-23
  • Contact: Junguo Zhang

PDF (PC)

14

摘要： 随着人工智能技术的快速发展, 利用深度学习方法对野生动物图像进行自动识别, 已成为野生动物调查保护的关键手段。实际采集的野生动物图像数据通常呈现一种偏态分布特征, 即少数高频类别物种样本充足, 而大多数低频类别物种样本稀缺, 影响模型的整体识别性能。针对这一问题, 本文提出一种基于Diff-SCC模型的偏态分布野生动物识别方法。首先, 该方法利用大语言模型生成类别的丰富语义描述, 引导扩散模型生成额外样本, 同时引入多尺度负样本筛选策略, 从像素空间、特征空间及语义空间3个维度进行图像质量评估和筛选, 提升低频类别的特征多样性并平衡数据分布。其次, 本文在主干网络ResNet50中引入SCConv模块以减少空间与通道建模过程中的冗余特征, 并增强模型对前景区域的感知能力, 从而提高模型对低频类别的识别性能。最后, 本文在自建数据集ULB-12和公开野生动物数据集NACTI上开展对比实验以验证模型的性能。实验结果显示, Diff-SCC模型在上述两个数据集上的整体识别准确率分别达到78.71%和80.84%, 低频类别的识别准确率相较基线模型分别提升9.96%和9.99%。上述结果验证了Diff-SCC在处理偏态分布数据集上的有效性, 能够为野生动物智能监测与保护提供可靠的技术支撑。

关键词: 野生动物, 图像识别, 偏态分布, 扩散模型, 特征重建

Abstract

Aims: With the rapid development of artificial intelligence, deep learning has become a key tool for automating wildlife image recognition and advancing intelligent ecological monitoring. However, real-world wildlife image datasets typically exhibit a skewed distribution, in which a few common species have abundant samples, while most species are underrepresented, thereby limiting the overall recognition performance of the model. 

Methods: To address this issue, this study proposed a novel method for wildlife recognition named Diff-SCC, which integrated data generation using a diffusion model and feature reconstruction. Specifically, rich semantic descriptions of low-frequency categories were first generated using a large language model to guide the diffusion model in synthesizing additional samples. A multi-scale negative sample filtering strategy was then introduced to assess image quality from pixel, feature, and semantic levels, enhancing the diversity and balance of low-frequency categories’ features. Furthermore, an SCConv module was incorporated into the backbone network to improve spatial and channel modeling, focusing more effectively on foreground regions while reducing redundant computation. 

Results: This paper conducted comparative experiments on a self-built wildlife dataset from Ulanba National Nature Reserve, which comprised 12 wildlife categories, and on the public wildlife NACTI dataset. Results showed that the proposed Diff-SCC model achieves overall recognition accuracies of 78.71% and 80.84% on the two datasets, respectively. Notably, the recognition accuracy of low-frequency classes improved by 9.96% and 9.99% over the baseline model, demonstrating the effectiveness of the proposed method in handling skewed data and recognizing rare species. 

Conclusion: The Diff-SCC model proposed in this study demonstrates strong capability in mitigating the challenges of skewed distributions in wildlife image classification. It offers a reliable and practical solution for intelligent wildlife monitoring and contributes to the advancement of biodiversity conservation.

Key words: wildlife, image classification, skewed distributions, diffusion model, feature reconstruction

备案号：京ICP备16067583号-7
Copyright © 2026 版权所有 《生物多样性》编辑部
地址: 北京香山南辛村20号, 邮编：100093
电话: 010-62836137, 62836665 E-mail: biodiversity@ibcas.ac.cn