An image processing technique is described for the detection of mines in IR imagery. The proposed technique is based on a third-order neural network, which processes the output of a wavelet packet transform. The technique is inherently invariant to changes in signature position, rotation and scaling. The sell-known memory limitations that arise with higher-order neural networks are addressed by (1) the data compression capabilities of wavelet packets, (2) projections of the image data into a space of similar triangles, and 93) quantization of the "triangle space". Using these techniques, image chips of size 28 x 28, which would require O (10~9) neural net weights, are pressed by a network having O(10~2) weights. ROC curves are presented for mine detection in real and simulated imagery.
展开▼
