Hybrid classification methods based on consensus from several data sources are considered. Each data source is at first treated separately and modeled using statistical methods. Then weighting mechanisms are used to control the influence of each data source in the combined classification. The weights are optimized in order to improve the combined classification accuracies. Both linear and nonlinear optimization methods are considered and used in classification of two multisource remote sensing and geographic data sets. A nonlinear method which utilizes a neural network gives excellent experimental results. The hybrid statistical/neural method outperforms all other methods in terms of test accuracies in the experiments.

A new type of a neural-network architecture, the parallel consensual neural network (PCNN), is introduced and applied in classification/data fusion of multisource remote sensing and geographic data. The PCNN architecture is based on statistical consensus theory and involves using stage neural networks with transformed input data. The input data are transformed several times and the different transformed data are used as if they were independent inputs. The independent inputs are first classified using the stage neural networks. The output responses from the stage networks are then weighted and combined to make a consensual decision. In this paper, optimization methods are used in order to weight the outputs from the stage networks. Two approaches are proposed to compute the data transforms for the PCNN, one for binary data and another for analog data. The analog approach uses wavelet packets. The experimental results obtained with the proposed approach show that the PCNN outperforms both a conjugate-gradient backpropagation neural network and conventional statistical methods in terms of overall classification accuracy of test data.

Statistical classification methods based on consensus from several data sources are considered. The methods need weighting mechanisms to control the influence of each data source in the combined classification. The weights are optimized in order to improve the combined classification accuracies. Both linear and non-linear methods are considered for the optimization. A non-linear method which utilizes a neural network is proposed and gives excellent results in experiments. Consensus theory optimized with neural networks outperforms all other methods both in terms of training and test accuracies in the experiments.

A Bayesian contextual classification scheme is presented in connection with modified M-estimates and a discrete Markov random field model. The spatial dependence of adjacent class labels is characterized based on local transition probabilities in order to use contextual information. Due to the computational load required to estimate class labels in the final stage of optimization and the need to acquire robust spectral attributes derived from the training samples, modified M-estimates are implemented to characterize the joint class-conditional distribution. The experimental results show that the suggested scheme outperforms conventional noncontextual classifiers as well as contextual classifiers which are based on least squares estimates or other spatial interaction models.

In the evidential reasoning approach to the classification of remotely sensed multisource data, each data source is considered as providing a body of evidence with a certain degree of belief. The degrees of belief are represented by "interval-valued probabilities" rather than by conventional point-valued probabilities so that uncertainty can be embedded in the measures. The proposed method is applied to the ground-cover classification of simulated 201-band high resolution imaging spectrometer (HIRIS) data, from which a set of multiple sources is obtained by dividing the dimensionally huge data into smaller pieces based on the global statistical correlation information. By a divide-and-combine process, the method is able to utilize more features than conventional maximum likelihood methods.>

Consensus theory is adopted as a means of classifying geographic data from multiple sources. The foundations and usefulness of different consensus theoretic methods are discussed in conjunction with pattern recognition. Weight selections for different data sources are considered and modeling of non-Gaussian data is investigated. The application of consensus theory in pattern recognition is tested on two data sets: (1) multisource remote sensing and geographic data, and (2) very-high-dimensional remote sensing data. The results obtained using consensus theoretic methods are found to compare favorably with those obtained using well-known pattern recognition methods. The consensus theoretic methods can be applied in cases where the Gaussian maximum likelihood method cannot. Also, the consensus theoretic methods are computationally less demanding than the Gaussian maximum likelihood method and provide a means for weighting data sources differently.>

Advanced interpretation techniques used in remote sensing are surveyed in the context of earth data information systems. Included are methods for extracting information from the spectral, temporal, and spatial domains. The many types of data contained in modern earth data information systems constitute a generalized scene context which current research seeks to exploit more fully. Initial efforts in this direction are described. Significant research problems in several related areas are discussed.

This paper examines measures for evaluating the performance of algorithms for single instruction stream–multiple data stream (SIMD) machines. The SIMD mode of parallelism involves using a large number of processors synchronized together. All processors execute the same instruction at the same time; however, each processor operates on a different data item. The complexity of parallel algorithms is, in general, a function of the machine size (number of processors), problem size, and type of interconnection network used to provide communications among the processors. Measures which quantify the effect of changing the machine-size/problem-size/network-type relationships are therefore needed. A number of such measures are presented and are applied to an example SIMD algorithm from the image processing problem domain. The measures discussed and compared include execution time, speed, parallel efficiency, overhead ratio, processor utilization, redundancy, cost effectiveness, speed-up of the parallel algorithm over the corresponding serial algorithm, and an additive measure called "sprice" which assigns a weighted value to computations and processors.

A simple modification to existing probabilistic relaxation procedures is suggested which allows the information contained in initial labels to exert an influence on the direction of relaxation throughout the process. In this manner, the initial labels assume more importance than with conventional algorithms and are used in combination with the outcome of relaxation at each iteration to produce a cooperative estimate of the correct label for a particular object. Pixel labeling examples are presented which show the performance that can be obtained with the modified algorithm. The procedure is readily generalized to allow other data to influence the process.

Keywords: Remote sensing ; traitement de donnees Reference Record created on 2005-06-20, modified on 2016-08-08