Channel selection of atmospheric remote sensing.

ABSTRACT

We introduce a channel selection method for atmospheric remote-sensing problems described by a Fredholm integral equation of the first kind. Whether one set of channels (CH) is more suitable than another (CH') can be judged by whether (1) the degree of predominance (DP) value of CH is larger than that of CH', i.e., if the number of channels is the same and (2) the number of channels of CH is more than that of CH', if the DP values of both are acceptable. One can calculate the DP of the unknown function f (y) for a set of remote-sensing channels by DP = [1 + (Rf̃(a)(2) - 1) R(d)(2)](-1/2), Rf̃(a)(2) = R(c)(2)[ R(b)(2) + R(a)(2)(1 + R(b)(2))], where R(a), R(b), R(c), and (1 - R(d)(2))(1/2) of this channel set represent the influences on the ability to recover the unknown function caused by various measurement errors, the noise parameter, the relativity of the kernel functions, and the blindness of remote sensing means, respectively. Our channel selection method can be simplified to a conventional method when there are no differences in the relative measurement errors, no blind components of the unknown function and no noise parameters in the kernel function.