Showing papers by "David A. Landgrebe published in 1982"

The relationship of sensor parameters to applications data analysis

Abstract: A stochastic model for the data acquisition system in a multispectral scanner system, like the one utilized by the LANDSAT satellites, is presented. A list of noise sources which are known or presumed to have a significant effect in the information extraction process was constructed. Since the shot noise introduced by the photodetectors in the sensor system is signal level dependent, an atmospheric model was adopted which could adequately describe the amount of radiation that gets into the sensors based on the atmospheric transmittance. An analysis was carried out to find the output spectral statistics in terms of the input signal statistics and the system parameters. This was integrated into a set of FORTRAN programs that when supplied with, the class statistics, the noise levels introduced by the sensor system, the atmospheric transmittance, and the atmospheric path radiance, can be used to estimate the classification performance. In order to show the beneficts of this model a series of runs were performed in which the Thematic Mapper multispectral scanner was the system under consideration.

Probabilistic Relaxation on Multitype Data

Abstract: Classification of multispectral image data based on spectral information has been a common practice in the analysis of remote sensing data. However, the results produced by current classification algorithms necessarily contain residual inaccuracies and class ambiguity. By the use of other available sources of information, such as spatial, temporal and ancillary information, it is possible to reduce this class ambiguity and in the process improve the accuracy. In this paper, the probabilistic and supervised relaxation techniques are adapted to the problem. The common probabilistic relaxation labeling algorithm (PRL) , which in remote sensing pixel labeling usually converges toward accuracy deterioration, is modified. Experimental results show that the modified relaxation algorithm reduces the labeling error in the first few iterations, then converges to the achieved minimum error. Also a noniterative labeling algorithm which has a performance similar to that of the modified PRL is developed. Experimental results from Landsat and Skylab data are included.