Search and detection are two everyday jobs of many biological systems, performed almost innately, either consciously or subconsciously and necessary for survival. Search and target detection, in particular, are the first stages in visual observation tasks associated with military target acquisition, industrial inspection, traffic control, and many more applications. These tasks currently are often performed with the aid of an electro-optic viewing system. In these tasks, search is defined as the process by which an observer surveys his surroundings, and detection is the process of successfully declaring a desired target as such—more precise interpretations of these terms will be found in what follows and in the included papers.
For more than 50 years, the U.S. Army RDECOM CERDEC Night Vision and Electronic Sensors Directorate (NVESD) has been studying the science behind the human processes of searching and detecting, and using that knowledge to develop and refine its models for military imaging systems. Modeling how human observers perform military tasks while using imaging systems in the field and linking that model with the physics of the systems has resulted in the comprehensive sensor models we have today. These models are used by the government, military, industry, and academia for sensor development, sensor system acquisition, military tactics development, and war-gaming. From the original hypothesis put forth by John Johnson in 1958, to modeling time-limited search, to modeling the impact of motion on target detection, to modeling target acquisition performance in different spectral bands, the concept of search has a wide-ranging history. Our purpose is to present a snapshot of that history; as such, it will begin with a description of the search-modeling task, followed by a summary of highlights from the early years, and concluding with a discussion of search and detection modeling today and the changing battlefield. Some of the topics to be discussed will be classic search, clutter, computational vision models and the ACQUIRE model with its variants. We do not claim to present a complete history here, but rather a look at some of the work that has been done, and this is meant to be an introduction to an extensive amount of work on a complex topic. That said, it is hoped that this overview of the history of search and detection modeling of military imaging systems pursued by NVESD directly, or in association with other government agencies or contractors, will provide both the novice and experienced search modeler with a useful historical summary and an introduction to current issues and future challenges.
The problem solved in this paper is easily stated: given search parameters (p∞, τ)
that are known functions of time,
calculate how the probability a single observer acquires a target grows with time. This problem has been solved
analytically. In this paper we describe the analytical solution and provide derivations of the results. Comparison
with perception experiments will be reported in a future publication and hopefully will support the results presented
here. The provided solution is applicable to any scenario where the search parameters are changing with time and
are specified. In particular, the solution can be used to estimate the probability of target acquisition as a function of
time: (1) when the sensor-target range is changing, (2) for a slewed sensor where the target is alternately in and out
of the field of view, and (3) for a sensor that switches between wide and narrow fields of view.
To investigate the benefits of multiband infrared sensor design for target detection, search and detection experiments were conducted in the midwave infrared (MWIR) and longwave infrared (LWIR) wavebands in both rural and urban battlefields. In each battlefield environment, real imagery was collected in both bands by a single sensor using the same optics for both bands, resulting in perfect co-registration of the imagery. In order to study the performance impact of the spectral content, and not diffraction or other sensor-specific differences, the images were processed as needed so that differences in resolution due to diffraction were mitigated. The results of perception experiments, including detection probabilities, search times, and false alarm data, were compared between the wavebands.
A model has been developed that predicts the probability of detection as a function of time for a sensor on a
moving platform looking for a stationary object. The proposed model takes as input P (calculated from
NVThermIP), expresses it as a function of time using the known sensor-target range and outputs detection
probability as a function of time. The proposed search model has one calibration factor that is determined
from the mean time to detect the target. Simulated imagery was generated that models a vehicle moving
with constant speed along a straight road with varied vegetation on both sides and occasional debris on the
road and on the shoulder. Alongside, and occasionally on the road, triangular and square shapes are visible
with a contrast similar to that of the background but with a different texture. These serve as targets to be
detected. In perception tests, the ability of observers to detect the simulated targets was measured and
excellent agreement was observed between modeled and measured results.
This paper provides an overview of research in search and detection modeling of military imaging systems. For more than forty-five years the US Army Night Vision and Electronic Sensors Directorate (NVESD) and others have been working to model the performance of infrared imagers in an effort to link imaging system design parameters to observer-sensor performance in the field. The widely used ACQUIRE model accomplished this by linking the minimum resolvable contrast of the sensor to field performance. From the original hypothesis put forth by John Johnson in 1958, to modeling time limited search, to modeling the impact of motion on target detection, to modeling target acquisition performance in different spectral bands, search has a wide and varied history. This paper will first describe the search-modeling task and then give a description of various topics in search and detection over the years. Some of the topics to be discussed will be classic search, clutter, computational vision models and the ACQUIRE model with its variants. It is hoped that this overview will provide both the novice and experienced search modeler alike with a useful summary and a glance at current issues and future challenges.
When modeling the search and target acquisition process, probability of detection as a function of time is important to war games and physical entity simulations. Recent US Army RDECOM CERDEC Night Vision and Electronics Sensor Directorate modeling of search and detection has focused on time-limited search. Developing the relationship between detection probability and time of search as a differential equation is explored. One of the parameters in the current formula for probability of detection in time-limited search corresponds to the mean time to detect in time-unlimited search. However, the mean time to detect in time-limited search is shorter than the mean time to detect in time-unlimited search and the relationship between them is a mathematical relationship between these two mean times. This simple relationship is derived.
This research compares target detection in the longwave and midwave spectral bands in urban environments. The Night Vision and Electronic Sensors Directorate (NVESD) imaged one hundred scenes at several Army Military Operations in the Urban Terrain (MOUT) sites during day and night. Images were resized to make the field-of-view (FOV) for each scene approximately the same. These images were then presented in a time-limited search perception experiment using military observers. Probabilities of detection were compared between the two spectral bands. Results from MOUT search were compared with previous modeling efforts.
KEYWORDS: Probability theory, Received signal strength, Curium, Target detection, Visual process modeling, Infrared imaging, Optical engineering, Forward looking infrared, Data modeling, Performance modeling
Traditional FLIR performance analysis uses analytic models to predict sensor performance characteristics such as modulation transfer function (MTF) and minimum resolvable temperature (MRT). These characteristics are then used in conjunction with empirical criterion such as the Johnson cycle criteria to predict the performance of observers using the modeled sensor. In general, such an analysis suffers from inadequate descriptions of the effects of the background and incomplete descriptions of the observer detection mechanism. Accurate predictions of field performance in a particular setting require the expensive collection of imagery for metric analysis or perception tests. In this paper, an image-based approach is investigated. Using an advanced FLIR simulation, synthetic image sets are generated under controlled conditions. Using these image sets, image metrics are calculated and predictions of target detectability are made using a contrast-to-clutter model and a computational vision model. These predictions are compared to results obtained using a traditional range performance analysis from the MRT based ACQUIRE model. An assessment of the advantages and disadvantages of the image-based approach is given.
Recently the standard Night Vision search and detection thermal models have been challenged with the need to address scenarios which are quite different than those for which the models were originally intended. For example, there is a need to address the characteristics of the target and background in much more detail than was previously required. This paper will discuss and illustrate a selection of new concepts and formulations being considered by the Night Vision and Electronic Sensors Directorate for incorporation in improved search and detection thermal models. Included are new proposed formulations for the mean time for detection for both field-of-view and field-of-regard search and new concepts and formulas for thermal contrast signatures of targets and clutter characterization of backgrounds. Each new concept will be individually explained in detail with mathematical formulations and imagery examples. These formulations are then combined to illustrate how they can form new model metrics which can be used to predict both the static and dynamic probability of detection. The new candidate model formulations will be matched against available measured data to show the potential improvement in predictive capability offered.