This paper presents an algorithm, based on principal component analysis for the detection of biological threats using
General Dynamics Canada's 4WARN Sentry 3000 biodetection system. The proposed method employs a statistical
method for estimating background biological activity so as to make the algorithm adaptive to varying background
situations. The method attempts to characterize the pattern of change that occurs in the fluorescent particle counts
distribution and uses the information to suppress false-alarms. The performance of the method was evaluated using a
total of 68 tests including 51 releases of Bacillus Globigii (BG), six releases of BG in the presence of obscurants, six
releases of obscurants only, and five releases of ovalbumin at the Ambient Breeze Tunnel Test facility, Battelle, OH.
The peak one-minute average concentration of BG used in the tests ranged from 10 - 65 Agent Containing Particles per
Liter of Air (ACPLA). The obscurants used in the tests included diesel smoke, white grenade smoke, and salt solution.
The method successfully detected BG at a sensitivity of 10 ACPLA and resulted in an overall probability of detection of
94% for BG without generating any false-alarms for obscurants at a detection threshold of 0.6 on a scale of 0 to 1. Also,
the method successfully detected BG in the presence of diesel smoke and salt water fumes. The system successfully
responded to all the five ovalbumin releases with noticeable trends in algorithm output and alarmed for two releases at
the selected detection threshold.
KEYWORDS: Land mines, System on a chip, Sensors, Neural networks, Holography, General packet radio service, Image classification, Mining, Ground penetrating radar, Detection and tracking algorithms
This paper evaluates the performance of a holographic neural network in comparison with a conventional feedforward backpropagation neural network for the classification of landmine targets in ground penetrating radar images. The data used in the study was acquired from four different test sites using the landmine detection system developed by General Dynamics Canada Ltd., in collaboration with the Defense Research and Development Canada, Suffield. A set of seven features extracted for each detected alarm is used as stimulus inputs for the networks. The recall responses of the networks are then evaluated against the ground truth to declare true or false detections. The area computed under the receiver operating characteristic curve is used for comparative purposes. With a large dataset comprising of data from multiple sites, both the holographic and conventional networks showed comparable trends in recall accuracies with area values of 0.88 and 0.87, respectively. By using independent validation datasets, the holographic network’s generalization performance was observed to be better (mean area = 0.86) as compared to the conventional network (mean area = 0.82). Despite the widely publicized theoretical advantages of the holographic technology, use of more than the required number of cortical memory elements resulted in an over-fitting phenomenon of the holographic network.
KEYWORDS: General packet radio service, Sensors, Land mines, Forward looking infrared, Algorithm development, Palladium, Data fusion, Detection and tracking algorithms, Mining, Standoff detection
This paper explores the concepts of multisensor data fusion based on the Dempster-Shafer (DS) evidential theory in order to achieve a mine versus false-alarm (FA) classification of landmine targets. Initially, a decision-level DS algorithm is proposed to combine the evidence from multiple sensors of the landmine detection system developed by the General Dynamics Canada Limited (GD Canada). Subsequently, a feature-level DS fusion algorithm is employed to operate on a set of features reported by the ground penetrating radar (GPR) sensor of the system. The data used in the present study was acquired from the Aberdeen Proving Grounds (APG) test site in USA as part of the Ground Standoff Mine Detection System (GSTAMIDS) trials.
The proposed decision-level DS algorithm yielded a probability of detection (pd) of 92.53% at a false-alarm rate (FAR) value of 0.0697 FAs/m2. The Pd and the FAR performance results achieved by using the decision-level DS algorithm are comparable with the results obtained using three other decision-level fusion algorithms that were previously developed by GD Canada based on heuristic, Bayesian inference, and Voting fusion concepts. On the other hand, the feature-level DS fusion, when tested with the information presented by the GPR sensor only, resulted in a higher Pd value of 78.54% as compared to the corresponding result of 61.43% obtained by using the heuristic algorithm. The GPR sensor is one of the three scanning sensors present on the system.
The present paper proposes image analysis methods for the detection and classification of landmine targets in images acquired using a ground penetrating radar sensor. The detection methodology initially employs a preprocessing step based on principal component analysis principles. The preprocessed image is further subjected to a multilevel density slicing operation to generate a map of iso-intensity contours in the image. Salient regions, that correspond to true targets as well as false-alarms in the image, are then segmented by establishing hierarchical intensity links within the framework of iso-intensity contours based on parent-to-child nodal relations. Features are proposed to classify mines and FAs based on size, shape, contrast, and texture of the segmented regions.
We propose a method for detection of masses in mammographic images that employs pyramidal or hierarchical decomposition and Gaussian filtering operations as pre-processing steps. A procedure is then developed to segment the mass portions by establishing gradual intensity links from the central portions of masses into the surrounding areas in the image. The proposed mass detection algorithm was tested with 39 cases (28 benign and 11 malignant) selected from the Mammographic Image Analysis Society database. The technique achieved a success rate of 91% in detecting the malignant tumors and 68% in detecting the benign masses in the test set. The segmented mass portions were evaluated in terms of their benign versus malignant discriminant capabilities by computing two gradient- based features and texture features based on gray-level co- occurrence matrices (GCMs). The features were computed using a ribbon of pixels across the mass boundaries. The GCM-based texture features in combination with the gradient-based features resulted in the best benign versus malignant classification of the mass regions segmented by the proposed algorithm, with an area of 0.84 under the receiver operating characteristics curve.
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.