临床医生可通过观察眼底视网膜血管及其分支对人体是否患有疾病进行早期诊断,但由于视网膜中的血管错综复杂,模型在分割时会出现对微细血管分割精确度不足的问题。为此,提出一种结合残差模块Res2-net以及高效通道注意力机制(efficient channel attention,ECA)的D-Linknet模型。首先,利用Res2-net代替基础模型中的残差模块Res-net以提升每个网络层的感受野;其次,在Res2-net中添加一种结合压缩激励(squeeze and excitation,SE)和门通道(gated channel transformation,GCT)的注意力机制模块,改善处于复杂背景下的血管分割效果和效率;在网络的解码层加入ECA确保模型计算的性能,避免因降维导致的精度下降;最后,融合改进的模型输出图与掩膜图细化分割结果。在公开数据集DRIVE、STARE上进行分割实验,模型准确度(accuracy,AC)分别为97.11%、96.32%,灵敏度(sensitivity,SE)为84.55%、83.92%,曲线下方范围的面积(area under curve,AUC)为0.9873和0.9766,分割效果优于其他模型。实验证明了算法的可行性,为后续研究提供科学依据。
Estimating the capacity factor is a crucial and challenging phase in developing any wind energy project. The capacity factor is estimated based on both wind speed characteristics and wind turbine performance. Distribu...
详细信息
Estimating the capacity factor is a crucial and challenging phase in developing any wind energy project. The capacity factor is estimated based on both wind speed characteristics and wind turbine performance. Distributions such as Weibull, Rayleigh, and Kernel density estimation are regularly employed to predict wind speed, whereas cubic, polynomial, quadratic, and linear power curves are commonly used to describe wind turbine performance. The study intends to investigate the most appropriate method to estimate wind energy capacity factors in Taiwan. The techniques are evaluated by comparing the capacity factors estimated with the actual capacity factor recorded in the current operating wind farms in Taiwan. However, it appears that the distribution of wind speed in Taiwan is bimodal rather than unimodal. Thus, a mixture of Weibull distributions with a time-series approach is developed and integrated into the comparative analysis.
暂无评论