We have developed a framework, Cognitive Object Recognition System (CORS), inspired by
current neurocomputational models and psychophysical research in which multiple recognition
algorithms (shape based geometric primitives, 'geons,' and non-geometric feature-based
algorithms) are integrated to provide a comprehensive solution to object recognition and
landmarking. Objects are defined as a combination of geons, corresponding to their simple parts,
and the relations among the parts. However, those objects that are not easily decomposable into
geons, such as bushes and trees, are recognized by CORS using "feature-based" algorithms. The
unique interaction between these algorithms is a novel approach that combines the effectiveness of
both algorithms and takes us closer to a generalized approach to object recognition. CORS allows
recognition of objects through a larger range of poses using geometric primitives and performs
well under heavy occlusion - about 35% of object surface is sufficient. Furthermore, geon
composition of an object allows image understanding and reasoning even with novel objects. With
reliable landmarking capability, the system improves vision-based robot navigation in GPS-denied
environments. Feasibility of the CORS system was demonstrated with real stereo images captured
from a Pioneer robot. The system can currently identify doors, door handles, staircases, trashcans
and other relevant landmarks in the indoor environment.
Realistic building damage simulation has a significant impact in modern modeling and simulation systems especially
in diverse panoply of military and civil applications where these simulation systems are widely used for
personnel training, critical mission planning, disaster management, etc. Realistic building damage simulation
should incorporate accurate physics-based explosion models, rubble generation, rubble flyout, and interactions
between flying rubble and their surrounding entities. However, none of the existing building damage simulation
systems sufficiently faithfully realize the criteria of realism required for effective military applications. In
this paper, we present a novel physics-based high-fidelity and runtime efficient explosion simulation system to
realistically simulate destruction to buildings. In the proposed system, a family of novel blast models is applied
to accurately and realistically simulate explosions based on static and/or dynamic detonation conditions. The
system also takes account of rubble pile formation and applies a generic and scalable multi-component based
object representation to describe scene entities and highly scalable agent-subsumption architecture and scheduler
to schedule clusters of sequential and parallel events. The proposed system utilizes a highly efficient and
scalable tetrahedral decomposition approach to realistically simulate rubble formation. Experimental results
demonstrate that the proposed system has the capability to realistically simulate rubble generation, rubble flyout
and their primary and secondary impacts on surrounding objects including buildings, constructions, vehicles
and pedestrians in clusters of sequential and parallel damage events.
In this paper we present an adaptive incremental learning system for underwater mine detection and classification that
utilizes statistical models of seabed texture and an adaptive nearest-neighbor classifier to identify varied underwater
targets in many different environments. The first stage of processing uses our Background Adaptive ANomaly detector
(BAAN), which identifies statistically likely target regions using Gabor filter responses over the image. Using this
information, BAAN classifies the background type and updates its detection using background-specific parameters. To
perform classification, a Fully Adaptive Nearest Neighbor (FAAN) determines the best label for each detection. FAAN
uses an extremely fast version of Nearest Neighbor to find the most likely label for the target. The classifier perpetually
assimilates new and relevant information into its existing knowledge database in an incremental fashion, allowing
improved classification accuracy and capturing concept drift in the target classes. Experiments show that the system
achieves >90% classification accuracy on underwater mine detection tasks performed on synthesized datasets provided
by the Office of Naval Research. We have also demonstrated that the system can incrementally improve its detection
accuracy by constantly learning from new samples.
To protect naval and commercial ships from attack by terrorists and pirates, it is important to have automatic surveillance
systems able to detect, identify, track and alert the crew on small watercrafts that might pursue malicious intentions,
while ruling out non-threat entities. Radar systems have limitations on the minimum detectable range and lack high-level
classification power. In this paper, we present an innovative Automated Intelligent Video Surveillance System for Ships
(AIVS3) as a vision-based solution for ship security. Capitalizing on advanced computer vision algorithms and practical
machine learning methodologies, the developed AIVS3 is not only capable of efficiently and robustly detecting,
classifying, and tracking various maritime targets, but also able to fuse heterogeneous target information to interpret
scene activities, associate targets with levels of threat, and issue the corresponding alerts/recommendations to the man-in-
the-loop (MITL). AIVS3 has been tested in various maritime scenarios and shown accurate and effective threat
detection performance. By reducing the reliance on human eyes to monitor cluttered scenes, AIVS3 will save the
manpower while increasing the accuracy in detection and identification of asymmetric attacks for ship protection.
ATR in two dimensional images is valuable for precision guidance, battlefield awareness and surveillance
applications. Current ATR methods are largely data-driven and as a result, their recognition accuracy relies
on the quality of training dataset. These methods fail to reliably recognize new target types and targets in
new backgrounds and/or atmospheric conditions. Thus, there is a need for an ATR solution that can
constantly update itself with information from new data samples (samples may belong to existing classes,
background clutter or new target classes). In the paper, this problem is addressed in two steps: 1)
Incremental learning with Fully Adaptive Approximate Nearest Neighbor Classifier (FAAN) - A novel data
structure is designed to allow incremental learning in approximate nearest neighbor classifier. New data
samples are assimilated at reduced complexity and memory without retraining on existing data samples, 2)
Data Categorization using Data Effectiveness Measure (DEM) - DEM of a data sample is a degree to which
each sample belongs to a local cluster of samples. During incremental learning, DEM is used to filter out
redundant samples and outliers, thereby reducing computational complexity and avoiding data imbalance
issues. The performance of FAAN is compared with proprietary Bagging-based Incremental Decision Tree
(ABAFOR) implementation. Tests performed on Army ATR database with over 37,000 samples shows that
while classification accuracy of FAAN is comparable to ABAFOR (both close to 95%), the process of
incremental learning is significantly quicker.
The ever increasing volumes and resolutions of remote sensing imagery have not only boosted the value of image-based analysis and visualization in scientific research and commercial sectors, but also introduced new challenges. Specifically, processing large volumes of newly acquired high-resolution imagery as well as fusing them
against existing imagery (for correction, update, and visualization) still remain highly subjective and labor-intensive
tasks, which has not been fully automated by the existing GIS software tools. This calls for the development of novel
computational algorithms to automate the routine image processing tasks involved in various remote sensing based
applications. In this paper, a suite of efficient and automated computational algorithms has been proposed and
developed to address the aforementioned challenge. It includes a segmentation algorithm to achieve the automatic
"cleaning" (i.e. segmenting out the valid pixels) of any newly acquired ortho-photo image, automatic feature point
extraction, image alignment by maximization of mutual information and finally smoothing/feathering the edges of the
imagery at the join zone. The proposed algorithms have been implemented and tested using practical large-scale GIS
imagery/data. The experimental results demonstrate the efficiency and effectiveness of the proposed algorithms and the
corresponding capability of fully automated segmentation, registration and fusion, which allows the end-user to bring
together image of heterogeneous resolution, projection, datum, and sources for analysis and visualization. The potential
benefits of the proposed algorithms include great reduction of the production time, more accurate and reliable results,
and user consistency within and across organizations.
This paper presents a novel approach for defect detection using a wavelet-domain Hidden Markov Tree (HMT)1 model and a level set segmentation technique. The background, which is assumed to contain homogeneous texture, is modeled off-line with HMT. Using this model, a region map of the defect image is produced on-line through likelihood calculations, accumulated in a coarse-to-fine manner in the wavelet domain. As expected, the region map is basically separated into two regions: 1) the defects, and 2) the background. A level-set segmentation technique is then applied to this region map to locate the defects. This approach is tested with images of defective fabric, as well as x-ray images of cotton with trash. The proposed method shows promising preliminary results, suggesting that it may be extended to a more general approach of defect detection.
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.