摘要
本文阐述了概率松驰法的基本原理;着重讲座了把它用作子空间法后处理过程的情况下初始概率和相关系数的估计问题。实验表明:概率松驰法可以进一步提高识别结果的准确度,同时可以消除一些噪声的影响。
This paper develops in detail the principle of probabilistic relaxation and puts emphasis on the problem of estimating initial probabilities and correlative coefficients when it is used as the postprocessing of subspace pattern recognition. Experiment shows that probabilistic relaxation can improve accuracy of image classification and remove some noises.
出处
《测绘科学技术学报》
1993年第3期45-48,共4页
Journal of Geomatics Science and Technology
关键词
概率松驰法
相关单数
初始概率
Probabilistic relaxation, Correlative coefficient, Initial probability
