摘要
慢性创面愈合时间较长,需进行长期监测。针对传统创面分割方法主要依赖于医生的经验和肉眼观察,存在主观性强、准确度不足等缺点,本文开发了一种基于改进UperNet的创面组织分割算法,从而实现准确自动分割,提高监测效率。同时,引入KNet卷积核更新模块,用于动态调整卷积核权重,并集成卷积块注意力模块CBAM提高整体模型性能。结果表明:该模型的分割精度好高,具有可行性。
Chronic wound healing time is longer,need long-term monitoring.In view of the shortcomings of traditional wound segmentation methods,which mainly rely on doctors' experience and gross observation,such as strong subjectivity and low accuracy,this paper developed a wound tissue segmentation algorithm based on improved UperNet to achieve accurate automatic segmentation and improve monitoring efficiency.At the same time,KNet convolutional kernel update module is introduced to dynamically adjust the weight of convolutional kernel,and CBAM is integrated to improve the overall model performance.The results show that the segmentation accuracy of this model is high and feasible.
作者
盛沈君
张在房
SHENG Shenjun;ZHANG Zaifang
出处
《计量与测试技术》
2025年第2期135-138,143,共5页
Metrology & Measurement Technique
