论文部分内容阅读
本文首先用KLT反演了D.A.Stellingwerf教授带来的奥地利Kobernausser森林55个训练区的多波段遥感数据的压缩结果。进而通过对因子负荷量的分析,指出MSS-4与MSS-5,MSS-6与MSS-7分属两类。当采用下列回归方程:1/y=A+B/x,y=A·e~BX,y=A·x~B,进行拟合时,MSS-4与MSS-5之间,MSS-6与MSS-7之间的相关极为显著:对本次试验而言F>2000,R>0.99。这说明MSS-4与MSS-5互不独立,两者提供的信息有相当一部分重复;同样MSS-6与MSS-7也存在这样的关系。这就提示我们在今后数据处理中宜适当注意。而目前,则可利用这种相关性,对某种情况下遗漏的象元的灰阶值进行有效的补码,实践计算证明:MSS—4与MSS-5,MSS-6与MSS-7之间的相互补码,其相对误差小于6％,效果良好。
In this paper, we first use KLT to inverse the compression result of multi-band remote sensing data of 55 training areas in the Kobernausser forest in Austria brought by Professor D.A.Stellingwerf. And then through the analysis of factor load, it is pointed out that MSS-4 and MSS-5, MSS-6 and MSS-7 belong to two categories. When the fitting is performed using the following regression equation: 1 / y = A + B / x, y = A · e ~ BX, y = A · x ~ B, MSS-4 and MSS- The correlation with MSS-7 is extremely significant: F> 2000 for this test, R> 0.99. This shows that MSS-4 and MSS-5 are not independent of each other, both of which provide a considerable part of the duplication of information; the same MSS-6 and MSS-7 also exist in this relationship. This suggests that we should pay due attention to data processing in the future. At present, we can make use of this correlation to effectively complement the grayscale values of missing pixels in some cases. Practice shows that MSS-4 and MSS-5, MSS-6 and MSS-7 Complementarity between the two, the relative error of less than 6%, the effect is good.