Charlene E. Caefer

Improved covariance matrices for point target detection in hyperspectral data

Algorithms for point target detection in hyperspectral images use the inverse covariance matrix in order to separate a detected pixel from it surrounding noise. The inverse covariance matrix can be implemented from all the pixels or from the close surroundings of the examined pixel. We compare the different methods and conclude which method brings the best results.

Optimization of point target tracking filters

We review a powerful temporal-based algorithm, a triple temporal filter (TTF) with six input parameters, for detecting and tracking point targets in consecutive frame data acquired with staring infrared (IR) cameras. Using an extensive data set of locally acquired real-world data, we used an iterative optimization technique, the Simplex algorithm, to find an optimum set of input parameters for a given data set. Analysis of correlations among the optimum filter parameters based on a representative subset of our database led to two improved versions of the filter: one dedicated to noise-dominated scenes, the other to cloud clutter-dominated scenes. Additional correlations of filter parameters with measures of clutter severity and target velocity as well as simulations of filter responses to idealized targets reveal which features of the data determine the best choice of filter parameters. The performance characteristics of the filter is detailed by a few example scenes and metric plots of signal to clutter gains and signal to noise gains over the total database.

Analysis of false alarm distributions in the development and evaluation of hyperspectral point target detection algorithms

We analyze the efficacy of various point target detection algorithms for hyperspectral data. We present a novel way to measure the discrimination capability of a target detection algorithm; we avoid being critically dependent on the particular placement of a target in the image by examining the overall ability to detect a target throughout the various backgrounds of the cube. We first demonstrate this approach by analyzing previously published algorithms from the literature; we then present two new dissimilar algorithms that are designed to eliminate false alarms on edges. Trade-offs between the probability of detection and false alarms rates are considered. We use our metrics to quantify the improved capability of the proposed algorithms over the standard algorithms.

Point target detection in consecutive frame staring infrared imagery with evolving cloud clutter

The problem of detection of aircraft at long range in a background of evolving cloud clutter is treated. A staring infrared camera is favored for this application due to its passive nature, day/night operation, and rapid frame rate. The rapid frame rate increases the frame-to-frame correlation of the evolving cloud clutter; cloud-clutter leakage is a prime source of false alarms. Targets of opportunity in daytime imagery were used to develop and compare two algorithm approaches: banks of spatio-temporal velocity filters followed by dynamic-programming-based stage-to-stage association, and a simple recursive temporal filter arrived at from a singular-value decomposition analysis of the data. To quantify the relative performance of the two approaches, we modify conventional metrics for signal-to-clutter gains in order to make them more germane to consecutive frame real data processing. The temporal filter, in responding preferentially to pixels influenced by moving point targets over those influenced by drifting clouds, achieves impressive cloud-clutter suppression without requiring subpixel frame registration. The velocity filter technique is roughly half as effective in clutter suppression but is twice as sensitive to weak targets in white noise (close to blue sky conditions). The real-time hardware implementation of the temporal filter is far more practical.

Automated clustering/segmentation of hyperspectral images based on histogram thresholding

A very simple and fast technique for clustering/segmenting hyperspectral images is described. The technique is based on the histogram of divergence images; namely, single image reductions of the hyperspectral data cube whose values reflect spectral differences. Multi-value thresholds are set from the local extrema of such a histogram. Two methods are identified for combining the information of a pair of divergence images: a dual method of combining thresholds generated from 1D histograms; and a true 2D histogram method. These histogram-based segmentations have a built-in fine to coarse clustering depending on the extent of smoothing of the histogram before determining the extrema. The technique is useful at the fine scale as a powerful single image display summary of a data cube or at the coarser scales as a quick unsupervised classification or a good starting point for an operator-controlled supervised classification. Results will be shown for visible, SWIR, and MWIR hyperspectral imagery.

Temporal filters for tracking weak slow point targets in evolving cloud clutter

Abstract A class of temporal filters is presented for use with a staring infrared camera in detecting and tracking weak point targets moving slowly in evolving cloud clutter. The generic temporal filter, originally suggested by the singular value decomposition of consecutive frame data, is a zero mean damped sinusoid which can be recursively implemented in the complex plane. From this filter type, a composite triple temporal filter (TTF) is developed, consisting of two sinusoids of different periods in sequence followed by a third (averaging) filter. The TTF achieves impressive cloud clutter suppression by responding strongly to pixel temporal responses caused by moving point targets and weakly to responses caused by cloud edges moving into or out of pixels. An extensive database of local airfield scenes with targets of opportunity taken with two laboratory staring IR cameras was used in the design and testing of the filters. Issues and trade-offs in choosing the parameters of the TTF are explored by comparing two specific forms of the filter: the first based on a damped sinusoid with a period of 16 frames followed by one with a 10 frame period; the second filter has corresponding periods of 40 followed by 30 frames. The first TTF is very effective with targets having velocities from 0.1–0.5 pixels/frame in daytime drifting cloud scenes. However, target signal-to-noise values of ⩾6 are required for detection in white noise (close to blue-sky conditions). The second TTF is more sensitive to slower, weaker targets in blue-sky or cloudless night scenes; however, in order to operate in daytime cloud scenes, spatial enhancements are required. Results are detailed for some representative scenes and given as well for the total database as signal-to-clutter gain plots based on a newly formulated antimedian metric.

Point target detection in consecutive frame staring IR imagery with evolving cloud clutter

We treat the problem of long range aircraft detection in the presence of evolving cloud clutter. The advantages of a staring infrared camera for this application include passive performance, day and night operation, and rapid frame rate. The latter increases frame correlation of evolving clouds and favors temporal processing. We used targets of opportunity in daytime imagery, which had sub-pixel velocities from 0.1 - 0.5 pixels per frame, to develop and assess two algorithmic approaches. The approaches are: (1) banks of spatio-temporal velocity filters followed by dynamic programming based stage-to-stage association, and (2) a simple recursive temporal filter suggested by a singular value decomposition of the consecutive frame data. In this paper, we outline the algorithms, present representative results in a pictorial fashion, and draw general conclusions on the relative performance. In a second paper, we quantify the relative performance of the two algorithms by applying newly developed metrics to extensive real world data. The temporal filter responds preferentially to pixels influenced by moving point targets over those influenced by drifting clouds and thus achieves impressive cloud clutter suppression without requiring sub-pixel frame registration. It is roughly twice as effective in clutter suppression when results are limited by cloud evolution. However when results are limited by temporal noise (close to blue sky conditions), the velocity filter approach is roughly twice as sensitive to weak targets in our velocity range. Real-time hardware implementation of the temporal filter is far more practical and is underway.

Segmentation of hyperspectral images from the histograms of principle components

Further refinements are presented on a simple and fast way to cluster/segment hyperspectral imagery. In earlier work, it was shown that, starting with the first 2 principal component images, one could form a 2-dimensional histogram and cluster all pixels on the basis of the proximity to the peaks. Issues that we analyzed this year are the proper weighting of the different principal components as a function of the peak shape and automatic methods based on an entropy measure to control the number of clusters and the segmentation of the data to produce the most meaningful results. Examples from both visible and infrared hyperspectral data will be shown.

Temporal filtering for point target detection in staring IR imagery: I. Damped sinusoid filters

In an earlier conference, we introduced a powerful class of temporal filters, which have outstanding signal to clutter gains in evolving cloud scenes. The basic temporal filter is a zero-mean damped sinusoid, implemented recursively. Our final algorithm, a triple temporal filter, consists of a sequence of tow zero-mean damped sinusoids followed by an exponential averaging filter along with an edge suppression factor. The algorithm was designed, optimized and tested using a real world database. We applied the Simplex algorithm to a representative subset of our database to find an improved set of filter parameters. Analysis led to two improved filters: one dedicated to benign clutter conditions and the other to cloud clutter-dominated scenes. In this paper, we demonstrate how a fused version of the two optimized filters further improves performance in severe cloud clutter scenes. The performance characteristics of the filters will be detailed by specific examples and plots. Real time operation has been demonstrated on laboratory IR cameras.

Point target detection in segmented images

To perform point target acquisition in multispectral and hyperspectral images, it is often advantageous to compare the signature of the investigated pixel to a known target signature. To do this properly, it is necessary to estimate the expected mean and covariance matrix of an investigated pixel in a particular location, based on its local surroundings. The degree to which this pixel signature differs from the estimated background then becomes the data, which is matched to the desired target signature. The standard method for such an analysis is the RX algorithm of Reed and Yu. The mean is normally estimated from the local environment of the pixel; the covariance matrix can either be estimated globally or in some local window. In recent research, we have considered how to improve the algorithm by eliminating edge points as potential false alarms. In the present work, a prior segmentation of the image before processing is utilized. While our estimate for the mean continues to be based on the immediate neighbors of the investigated pixel, our estimate of the covariance matrix is now based on the covariance matrix of the segment to which the adjacent pixels belong. In this way, we get a more accurate estimate of the covariance matrix. Results on real multispectral and hyperspectral images with embedded targets in several spectral regions are presented and improvement is demonstrated.