Tongue color feature recognition algorithm based on YOLOv8(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2026年第1期
- Page:
- 132-140
- Research Field:
- 医学人工智能
- Publishing date:
Info
- Title:
- Tongue color feature recognition algorithm based on YOLOv8
- Author(s):
- LIU Xiaokuo; TANG Lijun; SUN Zhaowen
- School of Physics and Electronic Science, Changsha University of Science and Technology, Changsha 410000, China
- Keywords:
- Keywords: tongue color detection adaptive feature extraction hierarchical feature fusion dynamic adaptive convolution mechanism
- PACS:
- R318；TN911.7
- DOI:
- DOI:10.3969/j.issn.1005-202X.2026.01.017
- Abstract:
- Abstract: Given the subtle differences between different tongue colors and the suboptimal performance of YOLOv8 algorithm when directly applied to tongue color detection and classification, a high-precision, real-time tongue color recognition and classification algorithm named YOLOv8-AFD is proposed. This algorithm integrates the adaptive multi-layer convolution block with the C2f module of YOLOv8, which not only expands the receptive field for feature extraction while capturing global context information, but also enables adaptive channel feature recalibration to enhance the networks feature extraction capability. A multi-scale feature extraction module (FF-Block) is incorporated into the YOLOv8 network architecture for further enhancing the feature extraction ability across different scales and boosting the classification accuracy of tongue color features. Additionally, the convolution operation of YOLOv8 is replaced with a dynamic adaptive convolution module (DyCBAM-Conv) to automatically and dynamically adjust the weights of different channels according to the feature distribution of the input image, thereby adaptively highlighting the important tongue color features and suppressing irrelevant background information. Experimental results show that the improved model achieves a mean average precision of 92.6% for tongue color recognition on a dataset collected from a traditional Chinese medicine hospital. Compared with other commonly used algorithms, the proposed YOLOv8-AFD exhibits significant improvements to varying degrees and has a superior performance in tongue color feature recognition.
Last Update: 2026-01-27