Article Details

Abstract

Animal recognition is of great significance and role in the fields of ecological protection, agricultural management and intelligent monitoring. However, the manual recognition of animals often depends on the individual's experience and knowledge, which is subjective, inefficient and difficult to apply on a large scale. In order to solve this challenge, this paper focuses on an RGB image recognition method based on Deep Learning. Based on YOLOv7, K-means clustering algorithm is introduced to optimize the target candidate box.The experimental results show that compared with YOLOv3, YOLOv5, YOLOv7 and YOLOv10, the proposed model has the best performance in all evaluation indicators, with an accuracy of 73.80%, mAP@0.5 of 74.60%, and mAP@0.5:0.95 of 66.60%. Compared with the baseline method YOLOv7, the Precision（P）, Recall（R）, mAP@0.5 and mAP@0.5:0.95 are increased by 0.50%, 3.40%, 1.30% and 2.00%, respectively. In summary, this study achieves efficient and accurate identification of a variety of wild animals and domestic animals, and provides strong technical support for ecological protection, agricultural management, and intelligent monitoring.

Keywords

wild animal recognition; Target Detection; Attention Mechanism

References

1. [1] Liang Xiuhao, Yang Liping, Liao Wangjiao, etc Ecological recognition of oil tea pests based on convolutional neural networks [J]. Guangxi Forestry Science, 2023, 52 (3): 361-366
2. [2] Zhang Zuyu, Teng Yonghe, Qin Yuanli, etc Research on Automatic Extraction of Land Cover Information from UAV Image Data Based on U-Net Model [J]. Guangxi Forestry Science, 2022, 51 (4): 516-519
3. [3] Chen Qi, Lin Xin, Bai Aokun Tree species recognition method based on improved YOLOv8 drone remote sensing image [J]. Guangxi Forestry Science, 2024, 53 (4): 523-529
4. [4] Yang Minglun, Zhang Xu, Guo Ying, etc Wild animal image recognition based on YOLOv5 infrared camera [J]. Advances in Laser and Optoelectronics, 2022, 59 (12): 382-390
5. [5] Wu Xuntao, Bu Shuhai, Li Xianmin, etc Wildlife monitoring in Taibai Mountain Nature Reserve based on infrared camera technology [J]. Journal of Northwest Forestry University, 2017, 32 (3): 170-175
6. [6] Xu Qiujing, Qi Dawei Texture feature parameters of Northeast tiger based on gray level co-occurrence matrix [J]. Journal of Northeast Forestry University, 2009, 37 (7): 125-127+129
7. [7] Zhang Gongbo, Gu Yuliang, Zhu Hegui Design of a simulation system for animal image classification and recognition based on color features [J]. Ship Electronic Engineering, 2017, 37 (5): 81-85
8. [8] Li Jianwei, Zheng Yongguo Research on Animal Recognition Based on CNN [J]. Software Guide, 2019, 18 (3): 30-33
9. [9] Qi Jiandong, Zheng Shangzi, Chen Ziyi, etc Construction of an improved model for infrared camera wildlife image recognition in Beijing based on ConvNeXt [J]. Forestry Science, 2024, 60 (8): 33-45
10. [10] Gong Yinan, Tan Mengyu, Wang Zhen, etc Artificial Intelligence Recognition of Infrared Camera Animal Images Based on Deep Learning: A Case Study of Northeast Tiger and Leopard National Park [J]. Acta Zoologica Sinica, 2019, 39 (4): 458-465
11. [11] Cheng Zhean Automatic recognition of terrestrial wildlife in Inner Mongolia based on deep convolutional neural networks [D]. Beijing: Beijing Forestry University, 2019
12. [12] Xiang Qiumin Significance detection algorithm and application of wildlife monitoring images [D]. Beijing: Beijing Forestry University, 2017
13. [13] WANG C Y，BOCHKOVSKIY A，LIAO H-Y M. YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors [C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Vancouver：IEEE，2023：7464-7475.
14. [14] REDMON J，FARHADI A. YOLOv3: An Incremental Improvement [J/OL].arXiv:1804.02767 [cs.CV].[2024-10-26].https://doi.org/10.48550/arXiv.1804.02767.
15. [15] WANG A，CHEN H，LIU L H，et al. YOLOv10: Real-Time End-to-End Object Detection [J/OL].arXiv:2405.14458 [cs.CV].[2024-10-15].https://doi.org/10.48550/arXiv.2405.14458.

Funding

Guangxi Normal University for Nationalities Project (2022FW073)

About the authors

Meng Susu (1994-), female, Zhuang nationality, from Laibin, Guangxi, full-time teacher, master's degree, research direction: deep learning, image recognition; Kang Jiarong (1989-), male, Han ethnicity, born in Ji'an, Jiangxi, is a full-time teacher with a master's degree. His research interests include image processing and embedded systems; Yang Xiuzeng (1974-), male, Dong ethnic group, from Huaihua, Hunan, professor, master's degree, research direction: artificial intelligence optimization algorithms.