Many difficult (often NP-complete) optimization problems can be solved efficiently on graphs of small tree-width with a given tree-decomposition.In this paper,it is discussed how to solve the minimum feedback vertex s...Many difficult (often NP-complete) optimization problems can be solved efficiently on graphs of small tree-width with a given tree-decomposition.In this paper,it is discussed how to solve the minimum feedback vertex set problem and the minimum vertex feedback edge set problem efficiently by using dynamic programming on a tree-decomposition.展开更多
在气候学、生态学、考古学等诸多领域,树木年轮蕴含的环境变迁与历史演替信息具有不可替代的研究价值。为解决传统年轮检测方法存在效率低、易受人为因素干扰等问题,研究提出DCW-YOLOv8年轮实例分割模型,运用可扩张残差(Dilation-wise R...在气候学、生态学、考古学等诸多领域,树木年轮蕴含的环境变迁与历史演替信息具有不可替代的研究价值。为解决传统年轮检测方法存在效率低、易受人为因素干扰等问题,研究提出DCW-YOLOv8年轮实例分割模型,运用可扩张残差(Dilation-wise Residual,DWR)注意力模块、轻量级通用上采样算子(Content-Aware ReAssembly of Features,CARAFE)、动态非单调聚焦机制边界框损失函数(Weighted Interpolation of Sequential Evidence for Intersection over Union,Wise-IoU)组合改进,并设计一种依据模型掩码获取年轮数量与宽度信息的检测方法。消融试验和前沿模型对比试验结果表明,DCWYOLOv8模型掩码平均精度mAP@0.50、mAP@0.50∶0.95提升,达到86.4%、53.6%,优于同类算法;注意力可视化比较结果表明,DCW-YOLOv8对年轮特征注意力更强;年轮信息检测结果表明,年轮数量检测准确率达到86.2%,年轮宽度检测误差在±0.5 mm范围内的占70%,总体平均误差为0.295 mm。研究提出的DCW-YOLOv8模型为年轮自动化检测提供新的思路和方法。展开更多
基金Partially supported by the National Natural Science Foundation of China( 1 0 2 71 0 65
文摘Many difficult (often NP-complete) optimization problems can be solved efficiently on graphs of small tree-width with a given tree-decomposition.In this paper,it is discussed how to solve the minimum feedback vertex set problem and the minimum vertex feedback edge set problem efficiently by using dynamic programming on a tree-decomposition.
文摘在气候学、生态学、考古学等诸多领域,树木年轮蕴含的环境变迁与历史演替信息具有不可替代的研究价值。为解决传统年轮检测方法存在效率低、易受人为因素干扰等问题,研究提出DCW-YOLOv8年轮实例分割模型,运用可扩张残差(Dilation-wise Residual,DWR)注意力模块、轻量级通用上采样算子(Content-Aware ReAssembly of Features,CARAFE)、动态非单调聚焦机制边界框损失函数(Weighted Interpolation of Sequential Evidence for Intersection over Union,Wise-IoU)组合改进,并设计一种依据模型掩码获取年轮数量与宽度信息的检测方法。消融试验和前沿模型对比试验结果表明,DCWYOLOv8模型掩码平均精度mAP@0.50、mAP@0.50∶0.95提升,达到86.4%、53.6%,优于同类算法;注意力可视化比较结果表明,DCW-YOLOv8对年轮特征注意力更强;年轮信息检测结果表明,年轮数量检测准确率达到86.2%,年轮宽度检测误差在±0.5 mm范围内的占70%,总体平均误差为0.295 mm。研究提出的DCW-YOLOv8模型为年轮自动化检测提供新的思路和方法。
