BluePink XHost

Servere virtuale de la 20 eur / luna. Servere dedicate de la 100 eur / luna - servicii de administrare si monitorizare incluse. Colocare servere si echipamente de la 75 eur / luna. Pentru detalii accesati site-ul BluePink.

General Remote Sensing Principles - Part 2

(So, dear students, learn Engleesh, please!)

You know, love story, you don't know, I'm sorry!

Constantin Nitu

Generic Remote Sensing Approach

Although NOAA/AVHRR satellite data can be used at a low resolution to monitor vegetation the remotely sensed data of choice is LANDSAT TM or SPOT having a spatial resolution of 30m x 30m. This imagery can be corrected radiometrically and geometrically. Vegetation classification takes place with auxillary data such as aerial photography, ground check plots and a good set of control points. Additional vector graphics such as roads and streams, if available, could assist the vegetation delineation process. The image classification process can either be supervised using training sites or unsupervised. The following flowchart illustrates the process:

A Geographic Information System (GIS) is needed to perform spatial analyses such as overlays, buffers, surface modelling and creation of digital elevation models. Remotely sensed data, together with aerial photography, ground plots and other vector data (roads, streams, etc), can be used simultaneously to produce, for example, rangeland vegetation classes, determine rangeland productivity and hence grazing capacity. These various data have different levels of resolution associated with them as follows:

Ultimately the goal is to produce digital maps containing data NOT previously available. Remotely sensed data can yield useful maps at a predetermined scale if the smallest unit of land, that is to be recognized as a distinct entity, can be defined and is larger than the resolution associated with the original imagery.

IMAGE PROCESSING

Image processing involves the manipulation of images to:

• extract information
• emphasize/de-emphasize certain aspects of the information containing in an image
• perform statistical or other analyses to extract non-image information.

Some common digital image processing operations are:

1. histogram generation/equalization
2. contrast enhancement
3. density slicing
4. band ratios
5. multispectral classification (supervised/unsupervised)
6. neighborhood averaging/filtering
7. edge enhancement
8. principal components
9. Fourier transforms
10. high-pass and low-pass filtering

Image processing operations are of two types: enhancement and classification.

Enhancement: There are three types of enhancements: spectral, spatial and multi-channel. Spectral enhancements involve the construction and display of band histograms as well as stretching the bands. Spatial enhancements consist of image filtering and warping, texture enhancement and image convolutions/compression. Examples of filters are: edge detection, high pass, low pass and Laplacian. Filter kernels consist of a matrix of weights that are assigned to the reflectance values of a 3 x 3 or 5 x 5 group of pixels with the reference pixel in the middle. These filters can be edited. For instance, an edge detection filter would look as follows:

-1 -1 -1
-1 +8 -1
-1 -1 -1

Multi-channel enhancements include algebraic operations such as band ratios, principal component analysis and RGB cluster analysis (compression).

Band Ratios

Description:

Rationing is the division of the digital radiance value of one band by another on a pixel by pixel basis.

Advantages:

1. The major advantage of ratio images is that they convey the spectral or color characteristics of image features, regardless of variations in scene illumination conditions. A ratioed image of the scene effectively compensates for the brightness variation caused by varying topography and/or changing atmospheric conditions and emphasizes the color content of the data.
2. Ratio images can also be used to generate false color composites by combining three monochromatic ratio data sets. This has the advantage of combining data from more than two bands and presenting the data in color, which further facilitates the interpretation of subtle reflectance differences.

***CAUTION

1. Ratioing images can sometimes be "intensity blind". That is, dissimilar materials with completely different absolute radiances but similar slopes in their reflectance curves may appear identical. This can be very troublesome when these materials are contiguous and of similar image texture.
2. The images must have noise removed prior to ratioing, because ratioing enhances noise patterns.
3. Ratioing cancels only those factors that operate equally in the bands and not those that are additive. For example, atmospheric haze is an additive factor that might have to be removed prior to ratioing to give acceptable results.

Common Ratios and their Properties

TM1/TM2

Especially for deciduous forest are highly correlated, so the ratio image has low contrast.

TM3/TM4

Features such as water and roads are shown in higher tones. This is because they reflect high in the red band (TM3) and little in the near infrared band (TM4). Vegetation appears in darker tones because of the relatively low reflectance in the red band and high reflectance in the near-infrared.

TM5/TM2

Vegetation appears in light tones because of its relatively high reflectance in the mid-infrared band (TM5) and its comparatively lower reflectance in the green band (TM2). However, certain vegetation types do not follow this trend due to their particular reflectance characteristics and they appear in very dark tones. So this ratio is useful for vegetation discrimination.

TM3/TM7

Roads and other cultural features appear in lighter tone due to their relatively high reflectance in the red band (TM3) and low reflectance in the mid-infrared band (TM7). Also differences in water turbidity are readily observable.

TM5/TM4

Coniferous forest damage shows clearly, with high ratios characterizing high damage sites and low ratios showing low damage sites. This means the dryer a leaf becomes, the higher the ratio. This is due to the fact that when a leaf becomes dryer its reflectance increases in the mid-infrared band (TM5) in contrast to the near-infrared band (TM4) that is relatively unaffected by changes in moisture content.

TM7/TM4

It is very sensitive to total leaf-water content of the vegetative canopy. In general mid-infrared band (TM7) is low when moisture is high, due to absorption and near-infrared band (TM4) is unaffected by changes in moisture content. Therefore it is effectively used for forest damage estimation (works exactly as TM5/TM4) and vegetative biomass estimation. High TM7 values during a period of active growth implies a low productivity site because of moisture stress or less vegetation biomass at that site.

TM4*TM5*TM7/constant = 0-255

This is a very good discriminant of coniferous vegetation, deciduous vegetation, rock outcroppings and water.

Classification: There exist either supervised or unsupervised classification schemes. Supervised classifications include the identification, selection and manipulation of training fields or sites. Once the spectral signature of a training area has been saved a quick alarm can be displayed. This permits the identification of other training fields having the same signature. Examples of classification operations are: Euclidean, statistical, and parallelipiped classifiers, contingency tables, classification thresholding and overlaying. For example, a Euclidean classifier would compute group centroids based on distances between pixels having similar reflectance values. A statistical classifier would rely on correlation coefficients.

Image Gallery

DAIS-1 false color 1-m resolution imagery (12-bit no stretch)

DAIS-1 false color 1-m resolution imagery (12-bit stretched)

0.5-m resolution imagery of residential area, Bellevue, WA

1/3-m high resolution natural color imagery of commercial buildings, Honolulu, Hawaii

1/3-m high resolution natural color imagery of racetrack, Honolulu, Hawaii

1/3-m resolution 4-3-2 imagery of forested area in Idaho

1/3-m resolution 3-2-1 imagery of commercial buildings, King Co., WA

1/3-m resolution 4-3-2 imagery of commercial buildings, King Co., WA

Commonly Used Acronyms:

• MSS: Multispectral Scanner System
• TM: Thematic Mapper
• IR: Infra-Red
• NIR: Near Infra-Red
• DN: Digital Numbers
• SAR: Synthetic Aperture Radar
• AVHRR: Advanced Very High Resolution Radiometer
• IFOV: Instantaneous Field Of View
• NOAA: National Oceanic and Atmospheric Admin.
• WRS: Worldwide Reference System
• EOS: Earth Observing System

LITERATURE CITED

Cracknell, A.P. and L.W.B. Hayes. 1991. Introduction to Remote Sensing. Taylor & Francis Inc., Bristol, PA. 293 p.

ERDAS Inc. 1988. Image Processing System. ERDAS User's Guide.

EPS Ltd. 1994. PAMAP GIS version 4.1. Victoria, B.C. Canada

Holben, B.N. and C.O. Justice. 1981. An Examination of Spectral Band Ratioing to Reduce the Topographic Effect on Remotely Sensed Data. International Journal of Remote Sensing. 2(3): 115-133.

Jensen, J.R. 1986. Introductory Digital Image Processing. A Remote Sensing Perspective. Prentice-Hall, New Jersey. 379 p.

Justice, C.O., S.W. Wharton, and B.N. Holben. 1981. Application of digital terrain data to quantify and reduce the topographic effect on Landsat data. International Journal of Remote Sensing. 2(3): 213-230.

Lillesand, T.M. and R.W. Kiefer. 1994. Remote Sensing and Image Interpretation. Third Edn. John Wiley and Sons. New York, NY. 750 p.

Nguyen, P.T. and D. Ho. 1988. Multiple source data processing in remote sensing. pp 153-176. In Digital Image Processing in Remote Sensing. J.P. Muller, editor. Taylor and Francis Ltd, London. 275 p.

Nitu, C., Nitu, C. D., Tudose, C., Visan, M. 2002. Sisteme informationale geografice si cartografie computerizata (Roumanian), University of Bucharest Publishing House, 278 p.

Sabins Jr, F.F. 1986. Remote Sensing. Principles and Interpretation. W. H. Freeman and Company, New York. 449 p.

Schowengerdt, R.A. 1983. Techniques for Image Processing and Classification in Remote Sensing. Academic Press Inc. 249 p.

Spanner, M.A., L.L. Pierce, S.W. Running and D.L. Peterson. 1990. The seasonality of AVHRR data of temperate coniferous forests: Relationship with leaf area index. Remote Sens. Environ. 33: 97-112.

Walsh, S.J. 1987. Variability of Landsat MSS spectral responses of forests in relation to stand and site characteristics. International Journal of Remote Sensing. 8(9): 1289-1299.

Woodcock, C.E. 1982. Reducing the Influence of Topography on the Classification of Remotely Sensed Data. Masters thesis, University of California, Santa Barbara. 69 p.

Woodcock, C.E., A.H. Strahler, and T.L. Logan. 1980. Stratification of forest vegetation for timber inventory using Landsat and collateral data. p. 1769-1707. In Proceedings of the 14th International Symposium on Remote Sensing of the Environment. San Jose, Costa Rica. Vol. 3. pp. 1273-1928.

Acest document a fost realizat de Nitu Constantin.

Penultima actualizare: 12 septembrie 2001 (la o zi dupa WTC)

Ultima actualizare: 11 septembrie 2002 (la un an dupa WTC)

Trimite comentarii si sugestii la: cnitu@rol.ro