The mechanism of open complex objects’ thermal radiation characteristics has been elucidated. But in retrieval of LST

the high correlation of multi-angle data is the dominating obstacle for better retrieval. Taking advance of predefined knowledge can improve precision. In this paper

we divided the received radiation brightness to two parts

one is the average brightness of a land area and the other is the relative departure of each pixel to the average. We considered that the former can retrieve the average component temperature by the Matrix Expression directly and the latter can retrieve the component temperature’s relative departure by a statistical matrix. This statistical matrix is a regressive one

which still keep the linear relationship between the measured departure and retrieval departure. We utilized the least-square rule to get the regressive matrix

which includes the statistic characteristics of retrieved brightness of sensor and the land component radiation brightness. Our simulated experiment also validated this method. We point out that at present the lack of great deal of measured data of matching data from sensor and land limit the generalization of this method.