KEYWORDS: Signal to noise ratio, Image analysis, Detection and tracking algorithms, Bridges, Image processing, Video, Error analysis, Sensors, Electroluminescence, Control systems
Many trackers make use of correlation techniques to provide an estimate of the shift between incoming imagery and a stored reference image. One efficient method for estimating this shift is based on a least squares approach that makes use of gradient and difference imagery to avoid the computationally expensive construction of a correlation surface. A problem with this method is that it tends to underestimate image shifts when there is significant noise in the reference image-which is often the case. An alternative method makes use of a generalized least squares approach that takes the noise in the reference image into account when estimating the image shift. This paper describes these two correlation algorithms and presents the results of an empirical comparison of their performance under varying noise conditions for a variety of test imagery.
The function of the tracker image processing in our airborne FLIR tracking system is to continually measure the position of the tracked target relative to the image center. These positions are used by the line of sight control system to maintain the target in the center of the imaging sensor FOV. In order to perform an end-to-end test of the accuracy of the target positions measured by the tracker, we developed a verification method that makes use of instrumentation data and digital video recorded during flight test. The instrumentation data contains the track errors, which we wish to verify. A post-flight image analysis tool was developed to accurately measure the track positions in the flight recorded digital video for comparison to the recorded flight track errors. This tool made use of a correlation technique combined with a non-linear optimization procedure. The off-line track position estimates generated by this tool were then compared to those produced in flight by the tracker. This paper describes the algorithms used in the post-flight analysis to estimate the track positions from the digital video. The results of applying this verification technique to both a correlation tracker and a centroid tracker are then described.
The central problem of image based centroid tracking is that of target segmentation, that is, determining which pixels belong to the target and which belong to the background. Once the target pixels are identified, the centroid (the center of gravity) of these pixels can be used as the estimated target position. The underlying assumption made in centroid tracking is that the target image contains intensity values that are unlikely to occur in the background. Based on this assumption, the centroid tracker uses three concentric gates to determine which pixels are target pixels. The areas bounded by these gates form three disjoint regions; the inner region, the track region, and the outer region. An inner histogram is collected over inner region that should contain mostly target pixels. An outer histogram is collected over the outer region that should contain only background pixels. These histograms are then used to generate a probability map that indicates the probability that a pixel with a given intensity is part of the target. This probability map is then used to segment the target and find its centroid. This paper describes the methods used to generate the probability map and its use in the centroid tracking algorithm. The performance of this algorithm is compared to that of the previously used dual-threshold segmentation algorithm.
The Fitts correlation algorithm is known to be optimal in measuring small image translations. As the position measurement portion of an image-based tracker, it yields excellent track fidelity over a sequence of images in which the tracked target exhibits small shifts from the predicted position. Real-world experience, however, informs us that the tracked target rarely confines itself to small-scale translational motion in the image plane, and sometimes changes size during the track period. When does the steady-state Fitts correlation need a boost from auxiliary algorithms to maintain target lock? We describe the envelope within which the Fitts correlation tracker can be expected to report target position accurately. Tolerances for target translation, in-plane rotation, and size change are parameterized by the correlation length of the tracked target. A successful image-based tracker using Fitts correlation for position measurement will require additional logic or modifications to the image-processing algorithm to enable it to operate outside the bounds described.
This paper describes two applications of a model based target recognition approach that employs an efficient interpretation tree search algorithm for matching 3D model features to sensed features extracted from 2D imagery. The algorithm is tolerant to missing and incomplete features and makes optimal use of geometric constraints to greatly reduce search time while guaranteeing an optimal match. When a target is recognized, a minimum error estimate of its location is made. The algorithm requires that a rough estimate of the sensor view point be available prior to matching. Two applications of this algorithm are described, one for FLIR and one for SAR processing. For the FLIR application 3D models are constructed using linear space curves, quadratic space curves, and quadric surfaces. The features extracted from the imagery are straight line segments, representing the projection of linear space curves, and conic sections, representing the projection of quadratic space curves, and quadric surface limbs. For the SAR application the target models specify the 3D location of model scatterers, and the sensed features are the 2D locations of the returns detected in the SAR image. In the SAR application, the target was assumed to be mobile so special processing was necessary to handle the initial view point uncertainty. Experimental test results of these applications are described.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.