Search

This paper describes the design of a control and management network (orderwire) for a mobile wireless Asynchronous Transfer Mode (ATM) network. This mobile wireless ATM network is part of the Rapidly Deployable Radio Network (RDRN). The orderwire system consists of a packet radio network which overlays the mobile wireless ATM network, each network element in this network uses Global Positioning System (GPS) information to control a beamforming antenna subsystem which provides for spatial reuse. This paper also proposes a novel Virtual Network Configuration (VNC) algorithm for predictive network configuration. A mobile ATM Private Network-Network Interface (PNNI) based on VNC is also discussed. Finally, as a prelude to the system implementation, results of a Maisie simulation of the orderwire system are discussed., Comment: author's web page at http://www.crd.ge.com/people/bush

An optimal frequency domain textural edge detection filter is developed and its performance evaluated. For the given model and filter bandwidth, the filter maximizes the amount of output image energy placed within a specified resolution interval centered on the textural edge. Filter derivation is based on relating textural edge detection to tonal edge detection via the complex low-pass equivalent representation of narrowband bandpass signals and systems. The filter is specified in terms of the prolate spheroidal wave functions translated in frequency. Performance is evaluated using the asymptotic approximation version of the filter. This evaluation demonstrates satisfactory filter performance for ideal and nonideal textures. In addition, the filter can be adjusted to detect textural edges in noisy images at the expense of edge resolution.

Texture is known to be important in the analysis of radar images for geologic applications. It has previously been shown that texture features derived from the grey level co-occurrence matrix (GLCM) can be used to separate large scale texture in radar images. Here the influence of sensor parameters, specifically the spatial and radiometric resolution and flight parameters, i.e., the orientation of the surface structure relative to the sensor, on the ability to classify texture based on the GLCM features is investigated. It was found that changing these sensor and flight parameters greatly affects the usefulness of the GLCM for classifying texture on radar images.

Information extraction and transmission techniques for synthetic aperture radar (SAR) imagery were investigated. Four interrelated problems were addressed. An optimal tonal SAR image classification algorithm was developed and evaluated. A data compression technique was developed for SAR imagery which is simple and provides a 5:1 compression with acceptable image quality. An optimal textural edge detector was developed. Several SAR image enhancement algorithms have been proposed. The effectiveness of each algorithm was compared quantitatively.

Texture is an important image characteristic. A variety of spatial domain techniques were proposed for extracting and utilizing textural features for segmenting and classifying images. for the most part, these spatial domain techniques are ad hos in nature. A markov random field model for image texture is discussed. A frequency domain description of image texture is derived in terms of the power spectral density. This model is used for designing optimum frequency domain filters for enhancing, restoring and segmenting images based on their textural properties.

A statistical model is developed that portrays an imaging radar as a noisy communication channel with multiplicative noise, and the model is used to evaluate the average amount of information that can be extracted about a target from its radar image. The average information content is also used to define a measure of radiometric resolution for radar images. It is shown that the information content and the resolution capabilities of an imaging radar reach a limit beyond which an increase in scene dynamic range does not improve the information content or the resolution. This limitation results from the multiplicative nature of the noise introduced in the imaging process.

Airborne C- and L-band radar data acquired over a test site in western kansas were analyzed to determine corn-field identification accuracies obtainable using single-channel, multichannel, and multidate radar data. An automated pattern-recognition procedure was used to classify 144 fields into three categories: corn, pasture land, and bare soil (including wheat stubble and fallow). Corn fields were identified with accuracies ranging from 85 percent for single channel, single-date data to 100 percent for single-channel, multidate data. The effects of radar parameters such as frequency, polarization, and look angle as well as the effects of soil moisture on the classification accuracy are also presented.

Texture is known to be important in the analysis of radar images for geologic applications. It has previously been shown that texture features derived from the grey level co-occurrence matrix (GLCM) can be used to separate large scale texture in radar images. Here the influence of sensor parameters, specifically the spatial and radiometric resolution and flight parameters, i.e., the orientation of the surface structure relative to the sensor, on the ability to classify texture based on the GLCM features is investigated. It was found that changing these sensor and flight parameters greatly affects the usefulness of the GLCM for classifying texture on radar images.

An optimum (global) textural edge detection operator based on statistical models for texture was developed. Also the effects of the imaging process on the textural patterns of a scene as it appears in the image were analyzed.

Standard image processing techniques which are used to enhance noncoherent optically produced images are not applicable to radar images due to the coherent nature of the radar imaging process. A model for the radar imaging process is derived in this paper and a method for smoothing noisy radar images is also presented. The imaging model shows that the radar image is corrupted by multiplicative noise. The model leads to the functional form of an optimum (minimum MSE) filter for smoothing radar images. By using locally estimated parameter values the filter is made adaptive so that it provides minimum MSE estimates inside homogeneous areas of an image while preserving the edge structure. It is shown that the filter can be easily implemented in the spatial domain and is computationally efficient. The performance of the adaptive filter is compared (qualitatively and quantitatively) with several standard filters using real and simulated radar images.

The development is examined of a new edge detector which is shown to perform adequately in the non-Gaussian multiplicative noise environment which characterizes radar images. This edge detector operates over larger local neighborhood and is less susceptible to noise than previous edge detectors and is therefore more suitable for radar. In addition, a radar image noise model is employed for the design of this new operator. This edge detector is unique in that it is assumed that every local area belongs to either the class of local areas not containing edges or to the class of local areas containing edges. Each pixel's local neighborhood is then assigned to one of these two classes using a statistical hypothesis (a likelihood ratio) test. It is demonstrated that this algorithm is useful for detecting edges in radar images.

The presently derived and validated statistical model for radar images containing many different homogeneous fields predicts the probability density functions of radar images of entire agricultural scenes, thereby allowing histograms of large scenes composed of a variety of crops to be described. Seasat-A SAR images of agricultural scenes are accurately predicted by the model on the basis of three assumptions: each field has the same SNR, all target classes cover approximately the same area, and the true reflectivity characterizing each individual target class is a uniformly distributed random variable. The model is expected to be useful in the design of data processing algorithms and for scene analysis using radar images.

The use of automated algorithms for removing aliases in NOSS-SCAT data is reported. The algorithms used for alias removal consist of histogram analysis, local averaging and curve fitting. The histogram analysis is used to determine the degree of homogeneity of the wind field defined by the largest probability alias vector at each grid point. The alias directions are compared with the preferred direction at each grid location and one of the multiple aliases is chosen as the true direction.

The noise properties of the radar image formation process are used in the present modeling of a communication channel in which the desired target properties are the information transmitted, and the final image represents the received signal. The average information rate over this communication channel is calculated together with appropriate bounds and approximations, and is found to be small on a per-sample basis. As a result, many samples must be averaged to allow for the discrimination, or classification, of several levels of target reflectivity. These information rate properties are consistent with known results concerning target detection and image quality in speckle, and the rate is applicable to the definition of radar image radiometric resolution. Radiometric resolution is functionally related to the degree of noncoherent averaging performed by the sensor.