We present a robust geometric active contour model to track targets in video sequences captured from mobile cameras.
The target's contour is tracked on each frame of the sequence by both region and boundary information. The region
information is formulated by minimizing the Bhattacharyya coefficient between the color histogram of reference target
and that of the background. For the boundary information, we use gradient vector flow field to attract the contour from
either side of the target's boundary. The contour's evolution is implemented using the level set method. For each frame
coming from the sequence, template matching is performed before the curve evolution process to locate the region of
interest. The robustness and effectiveness of the proposed algorithm is demonstrated on real sequences.
KEYWORDS: Cameras, Reconstruction algorithms, Image restoration, 3D modeling, 3D image processing, Calibration, Stereolithography, 3D image reconstruction, Visualization, 3D acquisition
A new approach to recover 3D structure from the uncalibrated image sequence is presented. Unlike previous methods,
the method recovers 3D models' initial values by enforcing the geometric constraints of the straight line in the scene and
then gets accurate results by bundle adjustment. The focal length and the principal point of each camera are supposed to
be unknown intrinsic parameters. The similarity invariance of the straight line that vertical straight line pairs in the scene
keep vertical in similarity transformation are taken account of as a geometric constraint. A projective reconstruction
performed by an iterative factorization algorithm is upgraded to a Euclidean one with the constraint. Bundle adjustment
refines a visual reconstruction to produce jointly optimal structure and viewing parameter estimates when we get the
ideal initial values from the method above. Experiments results on synthetic and real image sequences verified the new
approach's precise and efficiency.
The mean shift algorithm is an efficient technique for tracking 2D blobs through an image. The scale of the mean shift
kernel is a crucial parameter. Classic Mean shift based tracking algorithm uses fixed kernel-bandwidth, which limits the
performance when the object scale exceeds the size of the tracking window. Although some modified algorithms can
settle the problem of object zooming in a way, these algorithms are helpless to the object rotation. Based on the analysis
of the scale-space theory and the current Mean shift algorithms, a scale and rotation adaptive mean shift tracking
algorithm is proposed. Experimental results show that the new method can effectively and accurately obtain the best
description of the target areas for the first frame, and the new mean shift tracking algorithm can adapt to any kind of
object's movements.
KEYWORDS: Detection and tracking algorithms, Image processing, Distortion, Target recognition, Image quality, Algorithm development, 3D acquisition, 3D image processing, 3D vision, Signal to noise ratio
The target tracking method based on correlation in image sequence is always invalid because of the magnitude or shape
distortion and occlusion. In this paper a robust and highly accuracy matching algorithm called Matching Based on Valid
Invariant Feature Part (MBVF) is proposed, which combines the target image invariant feature description, matching of
feature, and recognition of valid feature. The feature descriptor is formed from a vector containing the values of all the
grad magnitude and orientation entries that belong to the divided parts of target area. The features is robust to image
rotation, distortion, addition of noise, change in 3D viewpoint, and change in illumination. The first step of the algorithm
is to build the invariant feature descriptor of the target area in the referenced image. At the second step, a coarse position
of the target is calculated using the traditional forecast and correlation method. And the invariant feature descriptors of
all the possible points of the tracked target in image to be tracked are built also. Next, by comparing the invariant feature
of the referenced target and the tracked target the valid feature parts of the feature are recognized. At last, similitude
function is calculated according the valid feature parts in both images, which give the final fine position of the target in
the tracked image. Experiment results show that the MBVF can deal with the target tracking and positioning problems in
image sequence process and stereo image analysis automatically and accurately.
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.