30.5 Computer-Aided Detection and Classification Algorithm for Concealed Targets

The band ratio has been commonly used for multispectral data to reduce the effects caused by topological slope and aspects or to eliminate differential illumination effects caused by shadows (Jensen, 1996).

Let B_j and B_k be the jth and kth band image vectors. A band-ratioed image vector obtained by taking the ratio of B_j to B_k, denoted by BR_jk, is defined as follows:

(30.7)

Assume that the gray level range for all band image vectors is given by in ascending order. In case there is a pixel in B_k taking gray level value 0, the gray level of the corresponding pixel in the band ratio image vector BR_jk of (30.7) will be simply set to B_j to prevent the denominator of (30.7) from taking 0. As a result, the gray level range of the BR_jk is between 1/g_T and g_T.

Now, combining ATGP in conjunction with the BBOPC-based band selection algorithm discussed in Section 30.4 along with the band-ratioed image obtained by (30.7), the desired CADCA for concealed targets can be implemented as follows.

Computer-aided detection and classification algorithm

1. Apply BBOPC band selection algorithm to select an appropriate band set for the band ratio.

2. Apply the band ratio specified by (30.7) to generate band-ratioed images.

3. Apply ATGP to detect concealed targets.

4. Apply a target classifier to perform classification of each detected concealed targets.