Finally, this improvement is also seen as an improvement of kappa before/after merging of 24.8/30.5 for the histogram splitting technique compared to 15.8/28.5 for ISODATA. The unsupervised histogram splitting technique presented in this paper is shown to be better than the standard unsupervised ISODATA clustering technique with an overall accuracy of 34.3/19.0% before merging and 40.9/39.2% after merging. This data set provides reference data allowing for comparisons of the efficacy of different unsupervised data analysis. The data set used in this work utilizes the publicly available data collected at Indian Pines, Indiana. Unlike current unsupervised classification techniques that rely primarily on Euclidian distance measures to determine similarity, the unsupervised classification technique uses the natural splitting of the fit parameters associated with the basis functions creating clusters that are similar in terms of physical parameters.
Histogram splitting more » of the fit parameters is then used as a means of producing an unsupervised classification. These fit parameters used to generate the basis functions allow clustering based on spectral characteristics rather than spectral channels and provide both noise and data reduction. An unsupervised classification technique is presented in this paper that utilizes physically relevant basis functions to model the reflectance spectra.
Lack of a priori knowledge regarding land cover characteristics can make unsupervised classification methods preferable under certain circumstances. Hyperspectral image analysis has benefited from an array of methods that take advantage of the increased spectral depth compared to multispectral sensors however, the focus of these developments has been on supervised classification methods. Compared to more » a hyperspectral sensor, the ISIS approach can achieve similar classification accuracy using a significantly lower number of spectral samples, thus minimizing overall sample classification time and cost. The paper concludes that an ISIS style spectral imager can acquire these optimal spectral images directly, allowing improved classification accuracy over an RGB sensor. In both cases, optimal measurements match the performance achieved by the entire hyperspectral data set. In the colon cancer example, use of optimal measurements boost the classification accuracy of critical cell constituents by 28% relative to the RGB sensor. In the prostate cancer example, the optimal measurements allow 8% relative improvement in classification accuracy of critical cell constituents over a red, green, blue (RGB) sensor. It is shown that in these applications, two to three optimal measurements are sufficient to capture the majority of classification information for critical sample constituents. The following paper investigates the application of the ISIS sensing approach in two sample biomedical applications: prostate and colon cancer screening. These optimal measurements significantly improve signal-to-noise ratio (SNR) and speed, minimize data volume and data rate, while preserving classification accuracy. By allowing the definition of completely general spectral filter functions, truly optimal measurements can be made for a given task. The Information-efficient Spectral Imaging Sensor (ISIS) approach to spectral imaging seeks to bridge the gap between tuned multispectral and fixed hyperspectral imaging sensors.
0 kommentar(er)
