A. M. Starostenko, V. N. Lagutkin

A method of optimal selection bands and determining their parameters for the observation of extended objects with low signal on the multispectral images obtained by remote sensing is developed. Optimization criterion is the maximum probability of correct detection for fixed false alarm probability. The technique takes into account not only the relations between the parameters of the signal, background and noise, but instrumental limitations that arise when designing observation equipment. One of the ways to increase the probability of detection of extended objects with low signal in different viewing conditions is the use of multispectral sensors, but simply increasing the number of spectral bands does not always lead to increase the probability of detection. To increase the effectiveness of observation, and as a result of computing and hardware limitations do not allow to increase the number of spectral channels over a certain number, it is necessary to solve the problem of choosing the optimal set of ranges. If in addition to spectral information using spatio-temporal filtering, the number of ranges can be reduced. Method considered in the paper, allows you to select ranges for the detection extended objects with low signal by their contrast with the background of the underlying surface. Selecting the spectral channels used criteria based on the likelihood function. The calculation for the aircraft at different altitudes (at effective threshold of 4.6-10-2), has identified the following sets of spectral bands: {sj} = {[1.525, 1.775], [3.375, 3,625] [4,425, 4,675]}, the parameters of data sets: s({sj}) = 0.2808, C({sj}) = 1, m = 3, the number of executed steps of the recursive enumeration was 26,518,585.