摘要
利用多时相NOAA/AVHRR热红外数据构成像元级的时间序列,根据不同像元上时间序列曲线的距离和相似度进行聚类分析;对传统的模糊C-均值聚类算法进行改进,在算法中引入指标权重,对不同质量的数据赋予不同的指标权重。试验表明,改进后的算法扩大了应用范围,克服了单幅图像常存在的云干扰,实际效果明显。
This paper deals with the improvement of the fuzzy C-means cluster method with weighted data by applying it to multidate NOAA/AVHRR thermal data. With the values at one point on many time images, we can construct a time series vector. In the new algorithm using the vector similarity as a new fuzzy membership expression, we can classify the thermal data. The authors also give each value a power according to its identification as a real value of surface or a value of cloud. The result shows that the new algorithm has achieved great improvement in precision and flexibility in comparison with the ISODATA method.
出处
《国土资源遥感》
CSCD
2004年第4期7-10,共4页
Remote Sensing for Land & Resources
基金
浙江省自然科学基金项目(编号:400032)。
