ASPRS

PE&RS April 1998

VOLUME 64, NUMBER 4
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
JOURNAL OF THE AMERICAN SOCIETY FOR PHOTOGRAMMETRY AND REMOTE SENSING

Peer Reviewed Articles

273 A Modified Contextual Classification Technique for Remote Sensing Data
K.M.S. Sharma and A. Sarkar

Abstract
Download Full Article
Conventional techniques of classification make use of the spectral information at each pixel to predict the class of that independently of the observations at other pixels. Contextual techniques, on the other hand, utilize the information from other neighboring pixels also. Two methods of contextual classification exist - one for low-resolution data and one for high-resolution data. A new method is proposed by combining these two methods. The new method is compared with the Gaussian maximum-likelihood classification and the two methods of contextual classification for low-resolution and high-resolution data. Classification algorithms are compared using normalized classification accuracies and the Kappa statistics. 

285 Exact Binomial Confidence Interval for Proportions
Jeffrey T. Morisette and Siamak Khorram

Abstract
Download Full Article
In remote sensing accuracy assessment applications, the confidence interval is commonly used as a way to establish an appropriate sample size. However, confidence intervals are also informative when included in the accuracy assessment report. Many reports and papers give accuracy figures and leave out confidence intervals. In those cases where a confidence interval is constructed, the standard approach is to derive the interval through the use of a normal approximation of the binomial distribution or by referring to exact tables. This note briefly discusses the benefit of using confidence intervals. The main objective is to describe the calculation of an exact equal-tail confidence interval for the proportions of correctly classified pixels. The exact confidence interval is not based on a normal approximation but, instead, uses the relationship between the binomial and F distributions. While the derivation of the exact confidence interval relies on a somewhat involved mathematical relationship between the probability density functions, in practice the exact confidence interval is a relatively straightforward formula. With the percentiles of "F" Distributions now commonly available in hand calculators and spreadsheet (and other) programs, the exact confidence interval is easily calculated. After presenting the formula for the exact confidence interval, we will present an example using both the normal approximation and the exact confidence interval.

285 Forest Area Estimation Using Sample Surveys and Landsat MSS and TM Data
F. Deppe

Abstract
Download Full Article
Sample surveys data, and Landsat Multispectral Scanner System (MSS) and Landsat Thematic Mapper (TM] image data were used to establish forest area estimates within a test site area in the state of Rio Grande do Sul, southern Brazil. Two methods were applied: direct expansion and regression estimator methods. For the former, an area frame sampling scheme was used and, for the latter, estimates were corrected by means of classification results. Reference data were compiled for two periods through aerial photography interpretation (API) and direct field observations. The direct expansion method produced rapid and independent area estimates while results from the regression estimator method showed narrower confidence intervals, indicating higher accuracy. Mean relative efficiency figures for forest classes were 3.63 using the MSS data set and 7.73 using the TM data set. Findings of this study demonstrated the potential and advantages of using Landsat image data for forest area estimation.  

293 An Efficient and Accurate Method for Mapping Forest Clearcuts in the Pacific Northwest Using Landsat Imagery
Warren B. Cohen, Maria Fiorella, John Gray, Eileen Helmer, and Karen Anderson

Abstract
Download Full Article
Two variations of image differencing were compared. The first was based on unsupervised classification, repeated five times, using five sequential date-pairs of difference images between 1972 and 1993. Referred to as merged image differencing, this method required merging the results from five separate time intervals into a single map of forest harvest activity. The other method involved a single unsupervised classification of the full sequential difference image data set, and was referred to as simultaneous image differencing. A thorough harvest map error assessment using an independent reference database was compared to two methods of assessment based on visual interpretation of the Landsat data used to develop the difference images. Results indicate that harvest activity was mapped using merged image differencing with greater than 90 percent accuracy, and that visual methods of error assessment using the Landsat images gave nearly identical results with those of the independent reference data. Simultaneous image differencing resulted in a map that was consistent with merged image differencing, and was considerably more cost-effective to implement. 

301 Effectiveness of the MCC Method in Detecting Oceanic Circulation Patterns at a Local Scale from Sequential AVHRR Images
Jay Gao and Matthew B. Lythe

Abstract
Download Full Article
The maximum cross-correlation coefficient (MCC) method is a recently devised automatic approach for detecting translational motions from remotely sensed data, and has been commonly used to estimate motion velocities. This paper aims to evaluate the effectiveness of this method in detecting oceanic circulation patterns on a local scale from sequential AVHAR images. It is found that the MCC-derived results from one image pair are indicative of the general flows only. Incoherent flows caused by non-translational motions result in the detected circulation pattern being in loose agreement with the cruising-observed circulation pattern (COCP). Averaging of the directions detected from multiple image pairs slightly improves this agreement. The main factors constraining the performance of the MCC method are identified as rotational and strained motions, isothermal fields, and non-advective processes. Their impact is drastically minimized if the detected results are tested at a significance level of 90 percent or higher. Their elimination from the flow fields causes the retained directions to be more uniform. The larger the number of images used in a detection, the closer the correlation of the detected results with the COCP,especially at a higher significance level. The flow field averaged from three pairs and tested at the 99 percent significance level is most closely correlated with the COCP at a coefficient of 0.728 that is underestimated by 8 percent due to the quality of the COCP. 

309 Mapping Altered Rocks Using Landsat TM and Lithogeochemical Data: Sulphurets-Brucejack Lake District, British Columbia, Canada
J.R. Harris, A.N. Rencz, B. Ballantyne, and C. Sheridon

Abstract
Download Full Article
Landsat Thematic Mapper (TM) data have been used to detect the presence of altered rocks associated with ore mined deposition. In arid environments, the spectral signatures of diagnostic minerals are often not masked by water, vegetation, or surficial materials. There are many studies in which TM data have been used to locate hydrothermally altered rocks; in near desert regions, but few successful examples can be found that relate to glaciated, vegetated terrain at high latitudes. In this paper, TM data, in concert with lithogeochemical data and field observations, are used to detect and map altered rocks in the Sulphurets-Brucejack Lake district of north western British Columbia.

Areas of iron-pyrite and phyllic (sericitic) alteration as well as areas of intense silicification have been located using band ratioing techniques. The chemical nature of these alteration zones has been established through comparison with lithogeochemical data. These alteration styles reflect the underlying rock type, as would be expected, but also appear to reflect two different styles of alteration, perhaps reflecting different mineralizing systems. The western half of the study area, characterized by iron-pyritic alteration, reflects a base metal (Cu porphyry and minor Au) system while the eastern half, characterized by areas of silicification and clay alteration, may represent a precious metal [Au, As, Sb] system.

323 Scale-Space Methods for Image Feature Modeling in Vision Metrology
C.S. Fraser and Juliang Shao

Abstract
Download Full Article
A vision-based measurement system applicable for use in two-dimensional (2D) industrial inspection is described. The system embodies a five-stage approach to feature modeling comprising the use of multi-scale functions for filtering and feature location, image structure representation through a fusing of filter responses. extraction of edge information, edge building, and a determination of geometric feature attributes. The purpose of feature modeling is to describe and represent feature geometry and dimensions. In an inspection context, information on conformance to design can be provided by using this approach. The application of the system to the tasks of on-line steel plate and wheel rim inspection are described and, following an explanation of the use of scale-space methods for feature modeling, results of experimental tests are summarized.

GPS-Controlled Strip Triangulation Using Geometric Constraints of Man-Made Structures
Hamid Ebadi and Michael A. Chapman

Abstract
Download Full Article
Conventional block adjustments have been widely used to determine both photogrammetric ground point coordinates and the exterior orientation parameters of photography for mapping purposes. Ground control points are necessary to relate the image coordinate system to the object space coordinate system and to ensure the geometric stability of the conventional photogrammetric block as well as to control error propagation. The major impact of cost and time consumption for ground control establishment on any mapping project is the primary reason that photogrammetrists have been looking for a replacement for ground control by auxiliary date (e.g., Global Positioning System). This paper describes a new technique for GPS-controlled single strip triangulation using geometric constraints of man-made structures (e.g., high voltage towers. high rise buildings) located approximately along the flight line. The effects of the different GPS measurement accuracies were also investigated. Both the precision and the reliability analyses of the GPS bundle strip adjustment with these constraints were carried out on simulated and real data.
Top Home