Jay Hackett

Non-Negative Maximum Likelihood ICA for Blind Source Separation of Images and Signals with Application to Hyperspectral Image Subpixel Demixing

The use of independent component analysis (ICA) methods for blind source separation of signals and images has been demonstrated in many applications and publications. While many ICA based algorithms for source separation exist, few impose physical constraints on the recovered independent components and the mixing matrix. Of particular interest is the non-negativity of the recovered independent components and the recovered mixing matrix. Such constraints are important for example when trying to do subpixel demixing on hyperspectral images. In this article, we propose a constrained non-negative maximum-likelihood ICA (CNML-ICA) algorithm that tackles the limitations of some existing non-negative ICA methods.

A qualitative and quantitative method for predicting sentiment toward deployed U.S. forces

The ability to automatically predict likelihood of reaction to specific events and situational awareness is important to many military and commercial applications. Gauging population sentiment for targeted response areas and having the ability to predict or control sentiment within these areas is invaluable. Review of reception towards deployed forces must be analyzed, especially in areas vital for U.S. national interests. Predicting population behavior is critical for success and must include a qualitative as well as a quantitative solution. Additionally, a feedback mechanism is needed for periodically updating reception towards presence of U.S. Forces over time. We propose a method for predicting sentiment towards deployed U.S. Forces in near real time, to efficiently propitiate manpower resources, allocate equipment assets, and reduce cost of analyses. Sentiment prediction is becoming an increasingly important and feasible task based on social media, open source data, physical imagery and abundance of video data feeds. Predicting reaction to events can be time consuming. Locating the most likely affected areas is very tedious, requiring much human labor effort, and it is often difficult to obtain the best information on a timely basis. An efficient tool would be helpful to rapidly parse text that has been extracted from an intelligent algorithm in order to evaluate the population sentiment for the targeted area. Multiple data inputs and artificial intelligence (AI) algorithms are required in order to support sound decision making theory. The goal of our system, called GlobalSite, is to deliver trustworthy threat analysis systems and services that understand situations, while being a vital tool for continuing mission operations information.

An efficient visualization method for analyzing biometric data

We introduce a novel application for biometric data analysis. This technology can be used as part of a unique and systematic approach designed to augment existing processing chains. Our system provides image quality control and analysis capabilities. We show how analysis and efficient visualization are used as part of an automated process. The goal of this system is to provide a unified platform for the analysis of biometric images that reduce manual effort and increase the likelihood of a match being brought to an examiner’s attention from either a manual or lights-out application. We discuss the functionality of FeatureSCOPE™ which provides an efficient tool for feature analysis and quality control of biometric extracted features. Biometric databases must be checked for accuracy for a large volume of data attributes. Our solution accelerates review of features by a factor of up to 100 times. Review of qualitative results and cost reduction is shown by using efficient parallel visual review for quality control. Our process automatically sorts and filters features for examination, and packs these into a condensed view. An analyst can then rapidly page through screens of features and flag and annotate outliers as necessary.

Automated model-based image registration for tactical image exploitation

Automatic and timely image registration and alignment for producing highly accurate geodetic coordinates is of interest to tactical systems involved in battlespace awareness. We present an approach to registration that applies rigorous photogrammetric techniques to sensor geometry models to achieve registration accuracy of only a few pixels. Image collection is fully modeled in terms of its static geometry including aircraft and sensor parameters. The registration process not only aligns imagery, but also significantly reduces geoposition errors when multiple images are used. A normalized cross- correlation is applied to align image pixels through adjustments to the initial collection geometry. Our process is fully automatic and requires no operator intervention. This technique has a side benefit that the amount of time to register images is somewhat independent of the image size. Registration can be applied to imagery from disparate sensors, such as Synthetic Aperture Radar (SAR), Electro- Optical (EO), Multi-Spectral, and Infrared, in a multi- sensor fusion approach to reduce geodetic errors. This approach is implemented on standard Commercial-Off-The-Shelf hardware and has been tested on SAR and EO imagery at near real-time processing rates.

Quality metric for automated image registration performance prediction

Automated image geo-registration of military and defense related imagery can sometimes produce an unsuccessful result due to poor image quality, cloud cover, supporting data errors, and sensor phenomenology. In addition, there are many possible image processing algorithms that further compound the problem of prediction. An accurate mathematical model that is able to incorporate all these parameters and can predict the outcome of a registration event is not feasible. What is proposed here is a probabilistic approach to the problem. A robust quality metric that is able to determine the success of an autonomous registration will be discussed.

Cross-sensor registration optimization study

Multi-sensor and multi-spectral data fusion is becoming a very useful technology to solve a host of defense and commercial imaging and computer vision problems. Many of the techniques that can be used to fuse multi-sensor image data require co-registration or alignment of pixels between image bands. We have performed a non-parametric study to determine which multi-spectral bands should be chosen for optimum pixel level alignment. The data used during this study is composed of two aerial multi-spectral sensors (one with 3 visible bands and one with 5 bands in the visible and short wave infrared and one synthetic aperture radar sensor in the X-band. The study is presented in a scientific manner to allow for objective analysis of the results. A similarity measure and normalization approach was developed to allow for direct comparison between all combinations of visible, short wave infrared, and SAR phenomenology. All combinations of data alignment are performed and analytical results are extracted, analyzed, and statistically plotted. Variations in time of day of collection, atmospheric transmission, and collection path length are investigated. This approach has applicability for band selection in both manual and automatic registration techniques that are used to co-register multi-sensor data.