Tuned basis function (TBF) is a powerful technique for classification of two classes by transforming them into a new space, where both classes will have complementary eigenvectors. A target discrimination technique can be described based on these complementary eigenvector analyses under two classes: (1) target and (2) background clutter, where basis functions that best represent the desired targets form one class while the complementary basis functions form the second class. Since the TBF does not require pixel-based preprocessing, it provides significant advantages for target tracking applications. Furthermore, efficient eigenvector selection and subframe segmentation significantly reduce the computation burden of the target tracking algorithm. The performance of the proposed TBF-based target tracking algorithm has been tested using real-world forward looking infrared video sequences.
A novel hyperspectral target detection technique based on Fukunaga-Koontz transform (FKT) is presented. FKT offers significant properties for feature selection and ordering. However, it can only be used to solve multi-pattern classification problems. Target detection may be considered as a two-class classification problem, i.e., target versus background clutter. Nevertheless, background clutter typically contains different types of materials. That’s why; target detection techniques are different than classification methods by way of modeling clutter. To avoid the modeling of the background clutter, we have improved one-class FKT (OC-FKT) for target detection. The statistical properties of target training samples are used to define tunnel-like boundary of the target class. Non-target samples are then created synthetically as to be outside of the boundary. Thus, only limited target samples become adequate for training of FKT. The hyperspectral image experiments confirm that the proposed OC-FKT technique provides an effective means for target detection.
Target detection is one of the most important topics for military or civilian applications. In order to address such detection tasks, hyperspectral imaging sensors provide useful images data containing both spatial and spectral information. Target detection has various challenging scenarios for hyperspectral images. To overcome these challenges, covariance descriptor presents many advantages. Detection capability of the conventional covariance descriptor technique can be improved by fusion methods. In this paper, hyperspectral bands are clustered according to inter-bands correlation. Target detection is then realized by fusion of covariance descriptor results based on the band clusters. The proposed combination technique is denoted Covariance Descriptor Fusion (CDF). The efficiency of the CDF is evaluated by applying to hyperspectral imagery to detect man-made objects. The obtained results show that the CDF presents better performance than the conventional covariance descriptor.
Nowadays food inspection and evaluation is becoming significant public issue, therefore robust, fast, and environmentally safe methods are studied instead of human visual assessment. Optical sensing is one of the potential methods with the properties of being non-destructive and accurate. As a remote sensing technology, hyperspectral imaging (HSI) is being successfully applied by researchers because of having both spatial and detailed spectral information about studied material. HSI can be used to inspect food quality and safety estimation such as meat quality assessment, quality evaluation of fish, detection of skin tumors on chicken carcasses, and classification of wheat kernels in the food industry. In this paper, we have implied an experiment to detect fat ratio in ground meat via Support Vector Data Description which is an efficient and robust one-class classifier for HSI. The experiments have been implemented on two different ground meat HSI data sets with different fat percentage. Addition to these implementations, we have also applied bagging technique which is mostly used as an ensemble method to improve the prediction ratio. The results show that the proposed methods produce high detection performance for fat ratio in ground meat.
Powerful image editing tools are very common and easy to use these days. This situation may cause some forgeries by adding or removing some information on the digital images. In order to detect these types of forgeries such as region duplication, we present an effective algorithm based on fixed-size block computation and discrete wavelet transform (DWT). In this approach, the original image is divided into fixed-size blocks, and then wavelet transform is applied for dimension reduction. Each block is processed by Fourier Transform and represented by circle regions. Four features are extracted from each block. Finally, the feature vectors are lexicographically sorted, and duplicated image blocks are detected according to comparison metric results. The experimental results show that the proposed algorithm presents computational efficiency due to fixed-size circle block architecture.
The performance of the kernel based techniques depends on the selection of kernel parameters. That’s why; suitable parameter selection is an important problem for many kernel based techniques. This article presents a novel technique to learn the kernel parameters in kernel Fukunaga-Koontz Transform based (KFKT) classifier. The proposed approach determines the appropriate values of kernel parameters through optimizing an objective function constructed based on discrimination ability of KFKT. For this purpose we have utilized differential evolution algorithm (DEA). The new technique overcomes some disadvantages such as high time consumption existing in the traditional cross-validation method, and it can be utilized in any type of data. The experiments for target detection applications on the hyperspectral images verify the effectiveness of the proposed method.
Hyperspectral imagery (HSI) is a special imaging form that is characterized by high spectral resolution with up to hundreds of very narrow and contiguous bands which is ranging from the visible to the infrared region. Since HSI contains more distinctive features than conventional images, its computation cost of processing is very high. That’s why; dimensionality reduction is become significant for classification performance. In this study, dimension reduction has been achieved via VNS based band selection method on hyperspectral images. This method is based on systematic change of neighborhood used in the search space. In order to improve the band selection performance, we have offered clustering technique based on mutual information (MI) before applying VNS. The offered combination technique is called MI-VNS. Support Vector Machine (SVM) has been used as a classifier to evaluate the performance of the proposed band selection technique. The experimental results show that MI-VNS approach has increased the classification performance and decrease the computational time compare to without band selection and conventional VNS.
Often sensor ego-motion or fast target movement causes the target to temporarily go out of the field-of-view leading to reappearing target detection problem in target tracking applications. Since the target goes out of the current frame and re-enters at a later frame, the re-entering location and variations in rotation, scale, and other three-dimensional orientations of the target are not known, thus complicating the detection and tracking of reappearing targets. A new training-based target detection algorithm has been developed using tuned basis functions (TBFs). The detection algorithm uses target and background information, extracted from training samples, to detect possible candidate target images. The detected candidate target images are then introduced into the second algorithm, called clutter rejection module, to determine the target re-entering frame and location of the target. The second algorithm has been designed using the spatial domain correlation-based template matching (TM) technique. If the target re-enters the current frame, the target coordinates are detected and tracking algorithm is initiated. The performance of the proposed TBF-TM-based reappearing target detection algorithm has been tested using real-world forward-looking infrared video sequences.
Image segmentation is an important and difficult computer vision problem. Hyper-spectral images pose even more
difficulty due to their high-dimensionality. Spectral clustering (SC) is a recently popular clustering/segmentation
algorithm. In general, SC lifts the data to a high dimensional space, also known as the kernel trick, then derive
eigenvectors in this new space, and finally using these new dimensions partition the data into clusters. We demonstrate
that SC works efficiently when combined with covariance descriptors that can be used to assess pixelwise similarities
rather than in the high-dimensional Euclidean space. We present the formulations and some preliminary results of the
proposed hybrid image segmentation method for hyper-spectral images.
In many production systems, the products are inspected by human operators who observe faults with their naked eye
while most of the other manufacturing activities are automated. However, manual inspection is slow and yields
subjective results. To defeat this problem, image processing based visual control systems have been integrated to the
production systems. The visual system performance depends on the robustness of the image processing techniques.
Especially, the thresholding technique plays crucial role if you are inspecting scratches on the products. Since utilizing
the constant threshold fails in many cases, we have proposed an adaptive thresholding technique based visual inspection
system to detect production faults rapidly and efficiently without hampering the manufacturing process. The proposed
visual system also includes rotation invariant properties, which is important to get high speed processing.
Pattern recognition in hyperspectral imagery often suffers from a number of limitations, which includes
computation complexity, false alarms and missing targets. The major reason behind these problems is that the
spectra obtained by hyperspectral sensors do not produce a deterministic signature, because the spectra
observed from samples of the same material may vary due to variations in the material surface, atmospheric
conditions and other related reasons. In addition, the presence of noise in the input scene may complicate the
situation further. Therefore, the main objective of pattern recognition in hyperspectral imagery is to maximize
the probability of detection and at the same time minimize the probability of generating false alarms. Though
several detection algorithms have been proposed in the literature, but most of them are observed to be
inefficient in meeting the objective requirement mentioned above. This paper presents a novel detection
algorithm which is fast and simple in architecture. The algorithm involves a Gaussian filter to process the
target signature as well as the unknown signature from the input scene. A post-processing step is also included
after performing correlation to detect the target pixels. Computer simulation results show that the algorithm
can successfully detect all the targets present in the input scene without any significant false alarm.
Often sensor ego-motion or fast target movement causes the target to temporarily go out of the field-of-view leading to reappearing target detection problem in target tracking applications. Since the target goes out of the current frame and reenters at a later frame, the reentering location and variations in rotation, scale, and other 3D orientations of the target are not known thus complicating the detection algorithm has been developed using Fukunaga-Koontz Transform (FKT) and distance classifier correlation filter (DCCF). The detection algorithm uses target and background information, extracted from training samples, to detect possible candidate target images. The detected candidate target images are then introduced into the second algorithm, DCCF, called clutter rejection module, to determine the target coordinates are detected and tracking algorithm is initiated. The performance of the proposed FKT-DCCF based target detection algorithm has been tested using real-world forward looking infrared (FLIR) video sequences.
Fukunaga-Koontz Transform based technique offers some attractive properties for desired class oriented dimensionality reduction in hyperspectral imagery. In FKT, feature selection is performed by transforming into a new space where feature classes have complimentary eigenvectors. Dimensionality reduction technique based on these complimentary eigenvector analysis can be described under two classes, desired class and background clutter, such that each basis function best represent one class while carrying the least amount of information from the second class. By selecting a few eigenvectors which are most relevant to desired class, one can reduce the dimension of hyperspectral cube. Since the FKT based technique reduces data size, it provides significant advantages for near real time detection applications in hyperspectral imagery. Furthermore, the eigenvector selection approach significantly reduces computation burden via the dimensionality reduction processes. The performance of the proposed dimensionality reduction algorithm has been tested using real-world hyperspectral dataset.
Target detection and tracking algorithms deal with the recognition of a variety of target images obtained from a multitude of sensor types, such as forward-looking infrared (FLIR), synthetic aperture radar and laser radar.1,2 Temporary disappearance and then reappearance of the target(s) in the field-of-view may be encountered during the tracking processes. To accommodate this problem, training based techniques have been developed using combination of two techniques; tuned basis functions (TBF) and correlation based template matching (TM) techniques. The TBFs are used to detect possible tentative target images. The detected candidate target images are then introduced into the second algorithm, called clutter rejection module, to determine the target reentering frame and location of the target. The performance of the proposed TBF-TM based reappeared target detection and tracking algorithm has been tested using real-world forward looking infrared video sequences.
Target tracking in forward looking infrared (FLIR) video sequences is challenging problem due to various limitations such as low signal-to-noise ratio, image blurring, partial occlusion, and low texture information, which often leads to missing targets or tracking non-target objects. To alleviate these problems, we propose the application of quadratic correlation filters using subframe approach in FLIR. The proposed filtering technique avoids the disadvantages of pixel-based image preprocessing techniques. The filter coefficients are obtained for desired target class from the training images. For real time applications, the input scene is first segmented to the subframes according to target location information from the previous frame. The subframe of interest is then correlated with correlation filters associated with target class. The obtained correlation output contains higher value that indicates the target location in the region of interest. The simulation results for target tracking in real life FLIR imagery have been reported to verify the effectiveness of the proposed technique.
We have previously shown that polarization enhancement of fingerprint images during the enrolment process improves the performance of the verification and identification processes. In this paper, we present a design and analysis of a new synthetic discriminant function (SDF) for rotation/scale invariant polarization-enhanced fingerprint system. Performance comparison between the proposed SDF and an SDF obtained for traditional fingerprint systems is included.
The performance of target detection and tracking algorithms generally depends on the signature, clutter, and noise that are usually present in the input scene. To evaluate the effectiveness of a given algorithm, it is necessary to develop performance metrics based on the input plane as well as output plane information. We develop two performance metrics for assessing the effects of input plane data on the performance of detection and tracking algorithms by identifying three regions of operation—excellent, average, and risky intervals. To evaluate the performance of a given algorithm based on the output plane information, we utilize several metrics that use primarily correlation peak intensity and clutter information. Since the fringe-adjusted joint transform correlation (JTC) was found to yield better correlation output compared to alternate JTC algorithms, we investigate the performance of two fringe-adjusted JTC (FJTC)-based detection and tracking algorithms using several metrics involving the correlation peak sharpness, signal-to-noise ratio, and distortion invariance. The aforementioned input and output plane metrics are used to evaluate the results for both single/multiple target detection and tracking algorithms using real life forward-looking infrared (FLIR) video sequences.
Neural network based image processing algorithms present numerous advantages due to their supervised adjustable weight and bias coefficients. Among various neural network architectures, dynamic neural networks, Hopfield and Cellular neural networks have been found inherently suitable for filtering applications. These kind of neural networks present two important features; supervised learnable and optimization properties. Using these properties, dynamic neural filtering technique has been developed based on Hopfield neural networks. The filtering structure involves adjustable a filter mask and 2D convolution operation instead of weight matrix operations. To improve the supervised training properties, Widrow-recurrent learning algorithm has been proposed in this paper. Since the proposed learning algorithm requires less computation, consumption time in the training stage has been decreased considerably compared to previous reported supervised techniques for dynamic neural filtering.
The parallel processing capability and adaptive filtering features of dynamic neural networks offer highly efficient feature extraction and enhancement capability for fingerprint images. The most important aspect of the fingerprint enhancement is the extraction of relevant details with respect to distributed complex features. For this purpose, an efficient dynamic neural filtering technique has been proposed in this paper. After the enhancement process, fingerprint identification is/has been achieved using joint transform correlation (JTC) algorithm. Since the fringe-adjusted JTC algorithm has been found to yield significantly better correlation output compared to alternate JTCs, we used it in this study. The identification test results are presented to verify the effectiveness of the proposed enhancement and identification algorithms.
One of the most important challenges of fingerprint identification is the extraction of relevant details against distributed complex features. The parallel processing capability and learnable filtering features of cellular neural networks offer highly efficient feature extraction and enhancement capability for fingerprint images. In this paper, we propose to utilize the Widrow learning algorithm with a cellular neural network to efficiently enhance fingerprint details during the enrollment part. To evaluate the performance of the verification-identification part, enhanced fingerprint images are introduced into the fringe-adjusted joint transform correlator architecture for verification of an unknown fingerprint from a database. Comparison between the original and enhanced fingerprint identification and verification results is provided through computer simulation.
We propose a novel decision fusion algorithm for target tracking in forward-looking infrared (FLIR) image sequences recorded from an airborne platform. The algorithm allows the fusion of complementary ego-motion compensation and tracking algorithms to estimate the position of the target in the current frame among a sequence of frames of FLIR imagery. We identified three modes that contribute to the failure of the tracking system: (1) the sensor ego-motion failure mode, which causes the movement of the target beyond the operational limits of the tracking stage; (2) the tracking failure mode, which occurs when the tracking algorithm fails to determine the correct location of the target in the new frame; (3) the reference-image distortion failure mode, which happens when the reference image accumulates walkoff error, especially when the target is changing in size, shape, or orientation from frame to frame. The strategy in our design is to prevent these failure modes from producing tracking failures. The overall performance of the algorithm is guaranteed to be much better than any individual tracking algorithm used in the fusion. One important aspect of the proposed algorithm is its recoverability: the ability to recover following a failure at a certain frame. The experiments performed on Army Missile Command AMCOM FLIR data set verify the robustness of the algorithm.
An important step in the fingerprint identification system is the extraction of relevant details against distributed complex features. Identification performance is directly related to the enhancement of fingerprint images during or after the enrollment phase. Among the various enhancement algorithms, artificial intelligence based feature extraction techniques are attractive due to their adaptive learning properties. In this paper, we propose a cellular neural network (CNN) based filtering technique due to its ability of parallel processing and generating learnable filtering features. CNN offers high efficient feature extraction and enhancement possibility for fingerprint images. The enhanced fingerprint images are then introduced to joint transform correlator (JTC) architecture to identify unknown fingerprint from the database. Since the fringe-adjusted JTC algorithm has been found to yield significantly better correlation output compared to alternate JTCs, we used it for the identification process. Test results are presented to verify the effectiveness of the proposed algorithm.
In this paper, a new Hopfield neural network based supervised filtering technique is proposed. The learnable filtering architecture has been developed by modifying the Hopfield network structure using 2D convolution instead of weight-matrix multiplications. This feature offers high speed learning and testing possibility for image feature extraction process. The learning property of the filtering technique is provided by using a recurrent learning algorithm. The proposed technique has been implemented using joint transform correlator. The requirement of non-negative data for optoelectronic implementation is provided by incorporating bias technique to convert the negative data to non-negative data. Simulation results for the proposed technique are reported for feature extraction problems such as edge detection, and vertical line extraction.
The performance of a target tracking algorithm is directly related to global motion compensation performance, if the imaging sensor systems are not stable. Especially, forward looking infra-red (FLIR) video sequences are detrimentally affected by camera motion since the infrared camera mounted on an airborne platform suffer from abrupt discontinuities in motion. Since this global motion could cause the movement of the target outside the operational limits of the tracking algorithm, each frame in FLIR sequences has to be recovered by motion estimation technique. In this paper, a normalized cross correlation based template matching algorithm has been developed to accurately estimate and compensate the global motion before the application of the tracking algorithm. Then, the automatic target tracking algorithm has been applied using fringe-adjusted joint transform correlator (FJTC) based target detection and tracking technique.
Moving target tracking is a challenging task and is increasingly becoming important for various applications. In this paper, we have presented target detection and tracking algorithm based on target intensity feature relative to surrounding background, and shape information of target. Proposed automatic target tracking algorithm includes two techniques: intensity variation function (IVF) and template modeling (TM). The intensity variation function is formulated by using target intensity feature while template modeling is based on target shape information. The IVF technique produces the maximum peak value whereas the reference target intensity variation is similar to the candidate target intensity variation. When IVF technique fails, due to background clutter, non-target object or other artifacts, the second technique, template modeling, is triggered by control module. By evaluating the outputs from the IVF and TM techniques, the tracker determines the real coordinates of the target. Performance of the proposed ATT is tested using real life forward-looking infrared (FLIR) image sequences taken from an airborne, moving platform.
A technique has been formulated based on hetero-associative target detection strategy that recognizes and tracks multiple dissimilar or hetero-associative targets from gray-scale image sequences taken from a moving aircraft in real time. Fringe-adjusted joint transform correlation combined with the proposed hetero-associative filter is used to enhance the correlation performance and thus ensures strong and equal cross-correlation peak for each element of the selected class. Tracking is accomplished by combining the analysis of single image frame with the determination of the motion from consecutive image frames. For efficient performance, the desired targets are identified prior to be tracked by correlating successive frames using the proposed filter which is an enhanced version of the fringe-adjusted filter. The optimality of the tracking performance is tested by MATLAB software.
In this paper, we propose a novel decision fusion algorithm for target tracking in forward looking infrared (FLIR) image sequences recorded from an airborne platform. The algorithm allows the fusion of complementary ego-motion compensation and tracking algorithms. We identified three modes that contribute to the failure of the tracking system: (1) the sensor ego-motion failure mode, which causes the movement of the target more than the operational limits of the tracking stage; (2) the tracking failure mode, which occurs when the tracking algorithm fails to determine the correct location of the target in the new frame; (3) the distortion of the reference image failure mode, which happens when the reference image accumulates walk-off error, specially when the target is changing in size, shape or orientation from frame to frame. The proposed algorithm prevents these failure modes from developing unrecoverable tracking failures. The overall performance of the algorithm is guaranteed to be much better than any individual tracking algorithm used in the fusion. The experiments performed on the AMCOM FLIR data set verify the robustness of the algorithm.
Cellular Neural Networks (CNN) provides fast parallel computational capability for image processing applications. The behavior of the CNN is defined by two template matrices. In this paper, adjustment of these template-matrix coefficients have been realized using supervised learning algorithm based on back-propagation technique and wavelet function. Back-propagation algorithm has been modified for dynamic behavior of CNN. Wavelet function is utilized to provide the activation function derivation in this learning algorithm. The supervised learning algorithm is then executed to obtain a compact CNN architecture, called as Wave-CNN. The proposed new learning algorithm and Wave-CNN architecture performance have been tested for 2D image processing applications.
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.