Statistical Models of Targets and Clutter for Use in Bayesian Object Recognition

Srivastava, Anuj

doi:10.1016/b978-012119792-6/50137-6

Cited by 3 publications

(5 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Each template is denoted by I α , to be consistent with the notations used in, 11 where α represents the class of the object α ∈ {car, jeep, tank, plane, . .…”

Section: Mathematics Of Template Transformationsmentioning

confidence: 99%

“…In the context of ATR, the observation z k can be thought of as an array or a lattice of observed pixel intensities obtained by a camera at time scan k, while x k can be interpreted as the (corresponding) array or lattice of pixel intensities obtained from the hypothesized scene generated using the transformed templates. Physically, the likelihood function serves as the sensor or the measurement model, 9,11 while the prior probability density can be used to set prior distributions for each of the parameters to be estimated. Typically, prior density modeling is relevant for problems where ATR is performed on videos or dynamic scenarios.…”

Section: Problem Formulationmentioning

confidence: 99%

“…In, 9,11 Bayesian inference is applied to estimate the position, pose, class and number of targets from simulated FLIR and EO camera videos that are generated using CAD models of targets and by the application of sensor models. Hypothetical scenes that are parameterized by the position, pose, class and number of targets are generated, where the templates are arranged and transformed based on the parameters, and the candidate scene that maximizes the a posteriori probability density/ likelihood function is obtained by carrying out a numerical search along the parameter vector space.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Pose and class estimation of stationary ground vehicles using a scale augmented parameter set

et al. 2011

View full text Add to dashboard Cite

Automatic target recognition (ATR) has historically entailed the problems of detection, classification and tracking of ground or air targets from high resolution (imaging) sensor data as well as low resolution (radar) sensor data. A popular approach to solving the ATR problem is Bayesian inference, where detection (position and pose), classification and tracking are solved via a parameter estimation framework. The present paper offers a treatment of a subset of the aforementioned problem, which can be stated as "given imaging data of a stationary ground target and assuming that the target centroid's position is known in pixel coordinates, how can one estimate its pose (orientation) and class?" Furthermore, we also address the problem of scale invariance, i.e. how to ensure that, for instance, a target that appears smaller in an image is not misclassified to a class that has a similar sized template in the database? This problem is a very significant one since it is realistic to expect target templates in the database to be only of a certain size and of targets in the observed image to be smaller or larger depending on its relative distance to the camera. Hence, by proposing the use of scale as an additional parameter to be estimated, it is shown, via simulations, that this inclusion enhances the accuracy of class estimation.

show abstract

“…Each template is denoted by I α , to be consistent with the notations used in, 11 where α represents the class of the object α ∈ {car, jeep, tank, plane, . .…”

Section: Mathematics Of Template Transformationsmentioning

confidence: 99%

Section: Problem Formulationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Pose and class estimation of stationary ground vehicles using a scale augmented parameter set

et al. 2011

View full text Add to dashboard Cite

show abstract

“…Does visual search mean an algorithm for finding and identifying a specific object or class of objects in an image? The extensive literature on automated target recognition (ATR) exemplifies this philosophy [1]. Or, does visual search imply a general framework for finding information from visual data, but without object-specific guidance?…”

Section: Introductionmentioning

confidence: 99%

Foveated Visual Search for Corners

Arnow

Bovik

2007

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

We cast the problem of corner detection as a corner search process. We develop principles of foveated visual search and automated fixation selection to accomplish the corner search, supplying a case study of both foveated search and foveated feature detection. The result is a new algorithm for finding corners, which is also a corner-based algorithm for aiming computed foveated visual fixations. In the algorithm, long saccades move the fovea to previously unexplored areas of the image, while short saccades improve the accuracy of putative corner locations. The system is tested on two natural scenes. As an interesting comparison study, we compare fixations generated by the algorithm with those of subjects viewing the same images, whose eye movements are being recorded by an eye tracker. The comparison of fixation patterns is made using an information-theoretic measure. Results show that the algorithm is a good locater of corners, but does not correlate particularly well with human visual fixations.

show abstract

“…Does visual search mean an algorithm for finding and identifying a specific object or class of objects in an image? The extensive literature on automated target recognition (ATR) exemplifies this philosophy 1 . Or, does visual search imply a general framework for finding information from visual data without object-specific guidance?…”

Section: Introductionmentioning

confidence: 99%

Finding corners in images by foveated search

Arnow

Bovik

2006

SPIE Proceedings

View full text Add to dashboard Cite

We develop a new approach to finding corners in images that combines foveated edge detection and curvature calculation with saccadic placement of foveal fixations. Each saccade moves the fovea to a location of high curvature combined with high edge gradient. Edges are located using a foveated Canny edge detector with spatial constant that increases with eccentricity. Next, we calculate a measure of local corner strength, based on a product of curvature and gradient. An inhibition factor based on previous visits to a region of the image prevents the system from repeatedly returning to the same locale. A long saccade i s move thes fovea to previously unexplored areas of the image. Subsequent short saccades improve the accuracy of the location of the corner approximated by the long saccade. The system is tested on two natural scenes and the results compared against subjects observing the same test images through an eyetracker. Results show that the algorithm is a good locator of corners.

show abstract

Statistical Models of Targets and Clutter for Use in Bayesian Object Recognition

Cited by 3 publications

References 15 publications

Pose and class estimation of stationary ground vehicles using a scale augmented parameter set

Pose and class estimation of stationary ground vehicles using a scale augmented parameter set

Foveated Visual Search for Corners

Finding corners in images by foveated search

Contact Info

Product

Resources

About