Convolutional neural network approach for buried target recognition in FL-LWIR imagery

Abstract

A convolutional neural network (CNN) approach to recognition of buried explosive hazards in forward-looking long-wave infrared (FL-LWIR) imagery is presented. The convolutional filters in the first layer of the network are learned in the frequency domain, making enforcement of zero-phase and zero-dc response characteristics much easier. The spatial domain representations of the filters are forced to have unit l2 norm, and penalty terms are added to the online gradient descent update to encourage orthonormality among the convolutional filters, as well smooth first and second order derivatives in the spatial domain. The impact of these modifications on the generalization performance of the CNN model is investigated. The CNN approach is compared to a second recognition algorithm utilizing shearlet and log-gabor decomposition of the image coupled with cell-structured feature extraction and support vector machine classification. Results are presented for multiple FL-LWIR data sets recently collected from US Army test sites. These data sets include vehicle position information allowing accurate transformation between image and world coordinates and realistic evaluation of detection and false alarm rates.

Shallow buried improvised explosive device detection via convolutional neural networks

“…In [54], the authors proposed a cellular neural network (which must not be confused with a CNN) for the 3D thermal modelling of soil. [55] makes use of a CNN for spotting patterns in thermal images of buried items. Other studies using neural networks to analyse GPR data can be found in [56], which proposes an ANN approach and a fuzzy approach, and [57], where a Fast R-CNN is employed.…”

A New ECT Image Reconstruction Algorithm Based on Convolutional Neural Network

“…The traditional convolution layer contains a large number of convolution kernels, which makes the convolution layer structure very complex. Since the convolution process takes up most of the time of this algorithm, for sub-sampling layer the entire CNN computing time accounted for less than 1% [17][18]. So the speed convolution layer can effectively improve the efficiency of the algorithm.…”

Fusion of iECO image descriptors for buried explosive hazard detection in forward-looking infrared imagery

“…To address this shortcoming, we have explored, extended and created a number of image space features and descriptors, including convolutional neural networks (CNNs) 10 , improved Evolution COnstructed (iECO) features 11 , "soft" (importance map weighted) features 12 , histogram of cell-structured Gabor energy filter and Shearlet filter bank responses 8,11 , histogram of gradients (HOG) 13 and local binary pattern (LBP) [14][15][16] and "soft" edge histogram descriptor features 8,9 . Anderson et al 9 proposed additional anomaly evidence map features in FLIR, which include features from maximally stable extremal regions (MSERs) 17 and Gaussian mixture models (GMMs) 18 for change detection.…”