Markov模型的熵与参数估计

Markov模型在标准文本和手写文字的自动识别领域中有特别重要的应用.在讨论如何由观测数据来估计模型参数的问题中,代替Baum辅助函数采用模型的熵特性引入判断函数,并利用此判断函数获得模型参数的重估算法.     

Remaining fatigue life assessment of corroded mooring chains using crack growth modelling

Failures caused by the combined actions of fatigue, corrosion and wear are important safety concerns for mooring chains used on floating structures in the oil and gas industry. Prediction of remaining corrosion fatigue life based on surface condition could therefore be a useful tool for the continued safe operation of corroded chains. This paper investigates the use of crack growth modelling for estimating the remaining corrosion fatigue life of mooring chains that exhibit significant pitting corrosion damage. A crack growth modelling approach is used to produce remaining fatigue life estimates for a selection of severely pitted mooring chains. Using fatigue crack growth rate test results for grade R4 high strength mooring chain steel, empirical crack growth laws are presented for free corrosion and cathodic protection conditions at load ratio R = 0.1. Two different methods for establishing equivalent cracks from surface scans of corrosion damage are presented. The mooring chains are proof loaded as part of their manufacturing process. Residual stresses introduced during this process have therefore been determined by finite element analysis and accounted for in the fatigue crack growth predictions. One of the equivalent crack models, accounting for the single dominant corrosion pit, provided quite accurate fatigue life predictions when compared with full scale test results.     

基于2阶HMM的智能车视觉地图定位方法

本文中针对视觉地图匹配问题，将视觉地图匹配问题转化为基于图像序列的最优视觉地图节点匹配问题，并提出基于2阶隐马尔科夫模型（hidden Markov model,HMM）的视觉地图匹配方法。在该模型中，状态变量被定义为高精度视觉地图节点，查询图像被定义为观测数据。在状态转移模型中，引入2阶模型对短时间车辆运动进行匀速运动建模，与传统的1阶HMM相比，可以提高模型的适用性与准确性。提出利用全局图像特征建立查询图像与地图节点之间的匹配关系，并从匹配的汉明距离建立发射概率模型，可有效提高地图匹配的效率。最后，通过前向算法来求解最优匹配的地图节点。为了验证算法的性能，分别在封闭工业园区、开放道路和KITTI公开数据集对算法进行验证。实验结果表明：2阶HMM模型能够有效融合车辆运动信息和图像信息，提高匹配的稳定性和精确度，算法性能明显优于传统的基于单帧匹配和序列匹配算法。