TEXTURAL ANALYSIS OF CEPSTRUM IMAGES OF SUBSURFACE STRUCTURES

Consideration is being given to the problem of discriminating objects hidden under upper layers of the ground at depths comparable to the probing pulse duration. Based upon the cepstrum analysis a subsurface radar signal processing technique has been suggested. It is shown that, however the shape of the probing signal spectrum might be, the r esponses from point targets in the cepstrum images of subsurface ground layers make up the texture whose distinctive features enable objects to be detected and identified. As far as real systems are concerned, the false alarm probability is estimated at fixed SN ratios. Additionally, a study has been made of the issue as to the minimal length of a training sample, which is needed to obtain specified operating characteristics of a target detection algorithm. It is also revealed that for a preset SN ra tio there is little point in extending the training sample length to an optimal value.