Ultraviolet-visible (UV-Vis) spectroscopy is a well-established technique for real-time analyzing contaminants in finished drinking water and wastewater. However, it has struggled in surface water because surface water such as river water has more complex chemical compositions than drinking water and lower concentrations of nutrient contaminants such as nitrate. Previous spectrophotometric analysis using absorbance peak at UV region to estimate nitrate in drinking water performs poorly in surface water because of interference from suspended particles and dissolved organic carbon which absorb light along similar wavelengths. To overcome these challenges, the paper develops a machine learning approach to utilize the entire spectral wavelengths for accurate estimation of low concentration of dissolved nutrients from surface water background. The spectral training data used in this research are obtained by analyzing water samples collected from the US-Canada bi-nationally regulated Detroit River during agricultural seasons using A.U.G. Signals' dual channel spectrophotometer system. Confirmatory concentrations of dissolved nitrate in these samples are validated by laboratory analysis. Several commonly used supervised learning techniques including linear regression, support vector machine (SVM), and deep learning using convolutional neural network (CNN) and long short-term memory (LSTM) network are studied and compared in this work. The results conclude that the SVM with linear kernel, CNN with linear activation function, and LSTM network are the best regression models, which are able to achieve a cross validation root-mean-squared-error (RMSE) less than 0.17 ppm. The results demonstrate effectiveness of the machine learning approach and feasibility of real-time UV-Vis spectral analysis to monitor dissolved nutrient levels in the surface watersheds.
The presented work is an extension of previous work carried out at A.U.G. Signals Ltd. The problem is approached herein for vessel identification/verification using Deep Learning Neural Networks in a persistent surveillance scenario. Using images with vessels in the scene, Deep Learning Neural Networks were set up to detect vessels from still imagery (visible wavelength). Different neural network designs were implemented for vessel detection and compared based on learning performance (speed and demanded training sets) and estimation accuracy. Unique features from these designs were taken to create an optimized solution. This paper presents a comparison of the deep learning approaches implemented and their relative capabilities in vessel verification.
Knowing the exact growth stage of agricultural crops can be valuable information for crop management and monitoring. In Canada, canola fields are particularly vulnerable for crop disease development during their flowering stage, especially when the fields are under persistent wet conditions. Clubroot and sclerotinia are diseases that can occur in canola when these two factors come together. Remote sensing can provide an interesting tool for the monitoring of crop phenological stages over large agriculture landscapes. Reliable and frequent access to data is needed to determine field-specific growth stages. Given their all-weather capability, radar sensors are optimal for monitoring such a dynamic crop parameter. In 2014, Agriculture and Agri-Food Canada collected crop phenology information over multiple canola fields in the area of Carman, Manitoba. Coincidental to ground data collection, fully polarimetric RADARSAT-2 and dual-polarimetric TerraSAR-X satellite data were acquired over the study site. In collaboration with A. U. G. Signals Ltd., a methodology will be developed and validated for the identification of inflorescence emergence and flowering in canola fields. Analysis of the polarimetric datasets from this study determined that several polarimetric parameters were sensitive to the emergence of flower buds and the flowering stage in canola. The alpha angle and entropy in both the C- and X-band were able to identify these growth stages, in addition to any of the reflectivity ratios and differential reflectivity responses that incorporated an HV response. The RADARSAT-2 scatter diversity, degree of purity and depolarization index also demonstrated great potential at identifying canola flower emergence and flowering. These latter polarimetric parameters along with the reflectivity ratios may be advantageous given their ease in implementation within a larger risk assessment satellite-derived methodology for canola crop disease.
Training classifiers individually, and then fusing their results, has the potential to improve classification accuracy; often,
dramatic improvements are realized. In this paper we examine how training classifiers using multiple polarimetric
features such as the Cloude-Pottier decomposition, even and odd bounce and the Polarimetric Whitening filter and then
fusing their results affects performance of ship classification. We explore and compare two currently competing
technologies of classifier bagging and classifier boosting for classifier fusion and introduce a new approach which
conducts a search through solution space to configure an optimal classifier given a library of classifiers and features. A
related and important facet of this work is feature selection and feature reduction methods. We explore how the selection
of different features affects classification performance. We also explore estimates of the classifier error and provide
estimates for noise bounds on the data and compare performance of the different methods compared to the noise present
in data.
The objective of this paper is to develop novel classification structures for military targets detection and recognition by
employing different fusion techniques. In real applications, the great diversity of materials in the background areas and
the similarity between the background and target signatures result in high false alarm rates and large miss classification
errors. In this paper, three new systems are proposed using different fusion techniques: pixel level fusion, decision
fusion, and classification fusion employing confidence vectors. These new developed systems are tested using an
experimental data to show its effectiveness.
The probability distribution function (pdf) used to model Synthetic Aperture Radar (SAR) clutter is an important design element in Constant False Alarm Rate (CFAR) detection; the mean of the local CFAR window is taken as the first moment of the pdf. This study presents research examining the relationship between clutter statistics and radar resolution cell size in the Convair-580 (CV-580) C-SAR and RADARSAT-2 systems. The experiment consisted of decreasing the resolution of a HV polarized, high-resolution, CV-580 sea SAR image and determining the best fit pdf for the corresponding clutter. The same methodology was used on standard- and fine-beam-mode RADARSAT-2 HV images. It was found that the GΓ pdf could be fitted very well to the experimental data for all CV-580 and RADARSAT-2 resolutions. Furthermore, the highest resolution SAR data was Weibull distributed, and decidedly non-Gaussian, in all cases. The medium resolution CV-580 image was very closely modelled by the Lognormal distribution while the Rayleigh distribution (Gaussian statistics) proved highly suitable for modelling the lowest resolution SAR data. The test results presented in this paper may be useful to SAR researchers.
KEYWORDS: Signal to noise ratio, Time-frequency analysis, Radar, Image fusion, Signal processing, Transform theory, Data fusion, Interference (communication), Wavelets, Signal detection
We describe a new fusion method for time-frequency distribution (TFD) that increases the ability to detect and classify time-varying signals while suppressing signal-dependent artifacts and noise. This is achieved by applying a least-squares algorithm to estimate the second-order approximation Volterra series coefficients of the outputs of selected TFDs. These coefficients are used for the fusion of the selected TFDs and generate a new TFD. The proposed fusion method is compared with four other fusion methods in terms of resolution and signal-to-noise ratio (SNR) in the time-frequency (TF) plane. Five representative TFDs are fused to generate a new TFD and their performances are analyzed. The results show that the new fusion method considerably increases sharpness (resolution) and strength (SNR) of the signal in the TF plane and, furthermore, achieves better signal description over other fusion methods and the traditional TFDs.
When target motion is confined to a two-dimensional plane during coherent processing intervals, the adaptive joint time-frequency algorithm is shown to be an effective method for achieving rotational motion compensation in ISAR imaging. We illustrate the algorithm using both simulated and measured experimental radar data sets. The results show that the adaptive joint time-frequency algorithm performed very well in achieving a focused image of the target. Results also demonstrate that adaptive joint time-frequency techniques can significantly improve the distorted ISAR image over what can be achieved by conventional Fourier transform methods when the rotational motion of the target is confined to a two-dimensional plane. This study also adds insight into the distortion mechanisms that affect the ISAR images of a target in motion.
This paper is devoted to change detection for precision agriculture. It presents a new change detection methodology and an overview of an online image processing tool (www.signalfusion.com) with direct applicability for crop classification. The change detection algorithms discussed in the paper will be implemented on the web. They preserve the fine details of the pixel changes and furthermore may be used for crop classification. Image registration, calibration/normalization, pixel level transformations, clutter suppression and pixel level change detection are discussed and new technologies are presented. The effectiveness of the proposed technology is being tested on real SAR data with partial truthing.
In this paper we present a distributed processing system that will be used for manual and/or automated change detection using data from databases and from online sensors. The automated change detection algorithm described herein is based on a method developed by Armenakis et al. This technique is applied on two level image classification. Its extension to multiple level classification and change detection is also being discussed. The paper presents two other change detection methodologies that are based on the Principal Component Analysis and wavelet techniques. Finally, it discusses the effect of matched filters for improving the change detection performance. Experimental results are provided using RADARSAT images which have been registered with the automated registration algorithm of AUG Signals that is currently available under the distributed procesing system www.signalfusion.com.
In this paper a new approach for clutter and target characterization is proposed. The method is based on the use of Markov chains for representing the samples of both the clutter and the target. The mathematical representation of the clutter and the target is based on the transition matrix of an irreducible Markov chain. This kind of representation incorporates a full description of the underlying pdf as well as any order of statistical correlation. Among the useful and meaningful parameters of the transition matrix are its eigenvalues. In natural signals, transition matrices have only a small number of their elements with significant value. This fact can be used to device relatively simple Markov chain models for clutter representation. The target statistics can also be modeled by means of a Markov chain model. However, in this case, the model may be simpler since the target samples or pixels are highly correlated and their values are restricted to a smaller range compared to those of the clutter.
This paper presents a new concept for Time-Frequency estimation, which is based on algorithmic fusion. It is shown that algorithmic fusion increases considerably the detectability of signals while suppresses artifacts and noise. The paper reviews a sample of representative Time-Frequency algorithms. Their performance is studied from a qualitative and quantitative point of view. For simplicity, we have considered the Mean-Squared Error (MSE) as a measure of performance in quantitative performance evaluation studies. The algorithmic fusion is presented using a self adaptive signal and noise dependent or independent approach, while the fusion is performed using the first two terms of the Volterra Series expansion. Simplistic algorithmic fusion methods on time-frequency results (e.g. weighted averaging or weighted multiplication), are special cases of the proposed fusion technique. Experimental results are presented from simulated and real High Resolution (HR)-SAR data. Real HR-SAR data were used from the experiments performed by the Defence Research Establishment (DRDC)-Ottawa.
In this paper, tradeoff studies on several pixel level fusion algorithms and on their performance evaluation criteria are presented. Electro-optical (EO) and SAR sensors are dissimilar and produce images with very low degrees of correlation. These images are initially registered at subpixel level accuracy. The fusion is performed using the following pixel level fusion algorithms: Principal Component Analysis (PCA), Averaging (Ave), Laplacian Pyramid, Filter Subtract Decimate (FSD), Ratio Pyramid, Contrast Pyramid, Gradient Pyramid, Discrete Wavelet Transform (QWT), Shift Invariant DWT (SIDWT) with Haar, Morphological Pyramid, and the recent image fusion method developed by AUG Signals Ltd. A MATLAB based dedicated image fusion toolbox, that includes several pixel level fusion, restoration and registration algorithms, has been recently developed by AUG Signals. This toolbox is used for the tradeoff studies.
In this paper we present a new Web-based application for registering multi-sensor satellite images for image fusion operations. It is a distributed processing system which offers automatic or semi-automatic image registration and it is intended to provide a service to the Canadian Geospatial Data Infrastructure (CGDI) users through the GeoConnections Discovery Portal, formerly CEONet. It will be also provided on the web page of A.U.G. Signals Ltd.(www.augsignals.com) which will be connected to CEONet and CGDI. This innovative technology of A.U.G. Signals has all the advantages of current registration techniques, plus is can estimate reference (control) points automatically at high degree of accuracy and with zero false alarms. Users who apply existing remote sensing software tools, such as PCI or IDL/ENVI, with geo-referenced points for registration, may employ the A.U.G. Signals software to further improve the registration accuracy of their images. Geo-referenced control points may also be used with the proposed software. The proposed service is expected to evolve and expand other distributed processing initiatives of current interest, such as the emerging GRID technologies under development in United States and Europe and the Canadian high-speed network CA*Net3 and be part of the US OGC Web based Initiative.
This paper's objective is to present a new, computationally efficient method for automatic exploration, detection and recognition. The automatic mineral homogeneous region separation algorithm developed by A.U.G. Signals in cooperation with the Canadian Space Agency (CSA) using AVIRIS data and mineral signatures from the Nevada's (U.S.) Cuprite site is described. The hyperspectral data and spectral signatures were provided by the Canada Center for Remote Sensing (CCRS). The algorithm is able to successfully divide the image in regions where the mineral composition remains constant. Hence, it can be used for reducing the noise is estimating the abundance parameters of the minerals on a pixel-by-pixel basis, for image region selection and hyperspectral image labeling for data storage and/or selective transmission. This may be another form of lossless hyperspectral image compression. Through the presented approach we are able to: a) divide a hyperspectral image into regions of adaptivity where pixel unmixing algorithms are able to extract the abundance parameters with higher degree of confidence, b) increase the signal to noise ration (SNR) of the present spectral signatures in a region and c) apply the proposed hyperspectral homogeneous region separation for data reduction (hyperspectral image compression). Experimental and theoretical results and comparisons/tradeoff studies are presented.
This paper presents a systematic new methodology for restoration of infrared images. The approaches described herein are applicable to single frame and multi-frame or hyper-spectra infrared images. The restoration problem is performed in two stages: (1) noise reduction and (2) linear blur and image estimation and restoration. The additive and multiplicative noise reduction is statistically optimal and improves the estimation of blurring function and restored image. For the restoration process we discuss alternate methods and provide the framework for error free restoration by eliminating the well known singularity problems that are often present in inverse solutions with singularities. Some initial results are presented.
MeteorWatch is a concept for the observation of small meteor events from a microsatellite in low earth orbit. To achieve high spatial resolution (about 1 km), fast update rate (up to 50 Hz), and large instantaneous coverage (107 km2), a distributed sensor is appropriate. The MeteorWatch sensor design has about 300 independent detection modules linked by a data bus to a central controller and image processor. Each detection module has a camera, digitizer, controller, image preprocessor, and bus interface. In operation, each detection module decides on the probability that a particular image has a meteor. Meteor event rates are expected to be low compared to the data rate, so that preprocessing at the detector modules reduces traffic on the data bus to the central controller. Image sequences with probable meteors are sent to the central controller for further processing and extraction of the meteor parameters. This paper gives an overview of MeteorWatch and describes the image processing approach, including partitioning of the tasks between the detection modules and the central image processor, the selection of clutter-rejection algorithms and the limits of detection for small meteors.
This paper presents a comparison of chaotic and statistical CFAR detectors for detection of manmade point targets from SAR. Detection of small manmade targets in SAR or IR clutter is an important area of interest in many applications such as ocean surveillance, search and rescue, remote sensing, mine detection, etc. It has been shown that IR and radar clutter exhibit chaotic rather than purely random behavior. From the chaotic point of view, a neural network predictor has been developed using Radial Basis Functions (RBF) to detect small targets embedded in natural clutter. In this paper, we present tradeoff studies between the above chaotic CFAR detector and purely statistical detectors such as the Cell Averaging, Order Statistics, and Optimal Weibull. The tradeoff studies are performed on real data with real or simulated targets. It is shown that adaptive chaotic RBF detectors many outperform statistical detectors in real clutter environments.
In this paper, we present the formulation of the problem for recognition of targets from hyperspectra images. It is shown that conventional recognition techniques may be extended to hyperspectra images for the distribution process. It is also shown that the recognition process is directly proportional to the number of multispectra frames that represent each target. For the discrimination process we propose a parametric and a nonparametric process in which both are extensions to Fisher and Fukunaga-Mantock methods respectively. Examples that show the composition of hyperspectra features are presented.
KEYWORDS: Sensors, Target detection, Wavelets, Signal detection, Signal to noise ratio, Radar, Signal processing, Sensor performance, Radar signal processing, Electronic filtering
The objective of this paper is to present a new coherent adaptive Constant False Alarm Rate (CFAR) wavelet detector which can be used as an additional independent detector for effective CFAR detection of point targets. It is shown through examples that this detector may provide a reliable estimate of the clutter mean which in turn is used, when multiplied by a constant, to determine the CFAR detector cutoff point for the target detection process. The detector is coherent and furthermore, the real and imaginary parts of the clutter and target are processed independently and their results are combined. As shown through an experimental example, coherent detectors offer better performance over amplitude detectors at high Signal to Noise Ratios (SNRs). At low SNRs their performance approximates that of amplitude detectors due to the fact that phase information is very sensitive to noise at low SNRs and its contribution becomes insignificant.
In this paper we present a method to obtain a maximum likelihood estimation of the parameters of the Generalized Gamma and K probability density functions. Explicit closed form expressions are derived between the model parameters and the experimental data. Due to their nonlinear nature global optimization techniques are used for solving the derive expressions with respect to clutter model parameters. Experimental results show in all attempted cases that the resulting expressions are convex functions of the parameters. In addition to the maximum likelihood solution we present two other solutions. One is based on moment and the other on histogram matching.
This paper reviews general methodologies for hyper-spectra image processing and provides a systematic way of hyperspectra data exploitation. Although the paper reviews the most recent hyperspectra processing techniques, which are available in the open literature, it focuses on those that have been recently developed by the authors These approaches often complement work presented by others. Since the field of hyperspectra processing is relatively new, and is growing rapidly, it is a field rich of research areas with many unsolved problems. Its significance in military, and more generally in remote sensing applications, is tremendous. Furthermore, this paper has the objective to offer a quick look to the many approaches, to put in light the authors' recent developments in this field, and to serve as a background for new advances.
In this paper, a general philosophy about feature windows based applications and a windows-based application are presented for analysis, algorithm development, testing and validation studies for signal, image and data processing (SIDP) for Space-based Surveillance applications. This dedicated facility is called AUG-SIDP. It performs several specialized tasks such as blur estimation, restoration, CFAR detection, clutter modeling, registration, pixel and data level fusion, target tracking and classification. It is still in the development and testing phase.
KEYWORDS: Target detection, Sensors, Image processing, Image segmentation, 3D acquisition, Point spread functions, Motion estimation, 3D image processing, Digital filtering, Detection and tracking algorithms
In this paper we present a new technique for the estimation of the velocity of moving targets using sequential frames. This estimation process may be used to estimate a potential set of velocities of moving targets which in turn may be used by three-dimensional (3-D) directional matched filters. It may also be used as a target trajectory estimation technique. The method is based on a local probability density matching segmentation technique with spatiotemporal associations. Experimental results are presented.
In this paper we present a method to obtain a maximum likelihood estimation of the parameters of the generalized gamma and K probability density functions. Explicit closed form expressions are derived between the model parameters and the experimental data. Due to their nonlinear nature global optimization techniques are proposed for solving the derived expressings with respect to clutter model parameters. Experimental results show in all attempted cases that the resulting expressions are convex functions of the parameters. In addition to the maximum likelihood solution we present two other solutions. One is based on moment and the other on histogram matching. The Cramer-Rao lower bound is also derived and used for performance comparisons.
This new paper presents a nonlinear filtering technique for speckle noise reduction when the statistical clutter model is known. It is assumed that the noise or image statistical contributions are not known. The method has been successfully used for speckle noise reduction on electro-optical and SAR images for pre-filtering in multi-sensor image registration applications. A detailed software description of this speckle and noise reduction algorithm is also presented.
This paper addresses the composition of multispectra images for target detection and recognition. It will provide the reader with tools with which multispectra images may be decomposed and new superimages may be composed to emphasize desired features of targets of interest. The proposed methodology consists of blur and noise estimation and reduction processes in association with registration, decomposition and resolution improvements. Examples from real data are being presented.
In this paper we present an approach for association and registration of multi-sensor images. With no loss of generality, we consider herein the association and registration of SAR and visible satellite images. Although the current approach as presented is a semi-automatic registration technique, there is sufficient promise for the development of a fully automatic multi-sensor multi-spectra registration approach. Some experimental results are presented.
A method to compose multi-sensor multi-spectra images is presented in this paper. In addition, a method that can estimate the motion of small targets between two frames of the same spectra band also is shown. Objects which were not present in one frame are automatically detected by utilizing this technique. For the composite multi-sensor and/or multi- spectra images, the Karhunen-Loeve (K-L) and Gram-Schmidt (G-S) orthogonalization techniques are used in combination with our blur estimation and 3-D restoration methods. For the motion estimation or target detection in a set of two frames of the same spectral band, the G-S orthogonalization is used. Results procured from real data of satellite images will be offered fittingly in this presentation.
A new linear discriminant technique that results in better classification performance over existing techniques is presented in this paper. This new approach is formulated in a similar manner to that of the Fisher linear discriminant. However, the matrix which corresponds to within classes has been replaced by a new matrix. This matrix takes into consideration the cross-correlation properties of the classes of interest. It has been shown through simulations that this matrix replacement results in a better classification performance over other linear discrimination methods, including the Fisher discriminant. Finally, the proposed new discriminant is presented in parametric and non-parametric forms, and is found to exhibit better classification in both cases over other parametric and non- parametric methods, respectively. With this new approach, the non-parametric method will prove to be more successful than its parametric counterpart.
In this paper, we study the effects of orthogonalization on sequential, multisensor, and multispectral satellite images for CFAR point target detection incorporting fusion techniques. Although the K-L orthogonalization offers the best CFAR detection performance, it requires central fusion. A version of the G-S orthogonalization method, which preprocesses data in a pipeline form, offers a comparable CFAR detection to that of the K-L method. Point target CFAR detection is carried out by employing various fusion approaches on the orthogonal data. Sensor level fusion with quality information is shown to be preferable when the proposed sequential G-S orhtogonaltization is applied. The proposed CFAR approach is applied to dissimilar sensors and avoids overloading the communication channel transmitting only in the case of target detection. Trade-off studies and experimental results on real and simulated data are presented.
In this paper, we present a new statistical model to describe infrared images. The proposed pdf is a compound model derived from the Gaussian and the GT-pdf. Closed form expressions for the statistical moments of the model are derived. For specific values of its parameters the model ends up to be simpler pdfs. The proposed compound pdf models the thermal radiation from the background, the sunlight scattering as well as scintillating effects. We propose a parameter estimation technique which is based on the equivalence of experimental and theoretical moments. Experimental results are provided for model validation in real data as well as for demonstration of the segmentation procedure.
In this paper the three-dimensional (3-D) restoration problem is addressed for multi- frame/multi-spectra space-based infrared image processing. The conventional restoration approaches are reviewed and new 3-D restoration methods are proposed. Issues related to multispectra restoration are also discussed. Experimental results are presented using sequential frames obtained from high altitude by an array of infrared sensors.
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.