A comparison of AVIRIS and Landsat for land use classification at the urban fringe
Summary (2 min read)
Introduction
- In rapidly urbanizing areas, such as the Front Range of Colorado, maps fast lose their validity.
- Less traditional sensors may provide additional information that can improve mapping accuracy.
- The Airborne Visible Infrared Imaging Spectrometer , for example, produces images with 224 spectral bands between 0.4 and 2.45 m, compared to six bands for Landsat (not including the thermal band) and three for SPOT’s Multispectral Imager (XS).
Sensor Specifications and Classification Accuracy: The Case of the Urban Fringe
- Among the factors that may influence classification accuracy are the Ground Sampling Distance (GSD), number of spectral bands, and Signal-to-Noise Ratio (SNR) of a sensor.
- Generally, it is thought that GSD is the most important factor for classification accuracy of built environments (Forster, 1985).
- SNR, which varies sensor-by-sensor and band-by-band and pixel-by-pixel, may also influence classification accuracy.
- Finally, the number of spectral bands may influence accuracy of land use classification.
Image Processing
- An AVIRIS flight line was acquired for 30 September 1999 along the northern Front Range of Colorado.
- Page 814 takes advantage of recent advancements in atmospheric radiative transfer, resulting in highly accurate atmospheric corrections (Qu et al., 2000; Qu et al., 2002).
- An AVIRIS image was compared to synthetic Landsat ETM+ image derived from AVIRIS.
- To estimate the noise levels of ETM+, the following model was used (John Barker, NASA/Goddard Space Flight Center, personal communication): SNR DN (a b * DN )^.5, (1) where DN is the digital number of a pixel, and a and b are coefficients for each band calibrated on ETM+ data from 06 September 2002.
- A MNF transform, similar to a principal components transform, derives a series of uncorrelated bands and segregates noise in the data.
Classification Methodology
- There are myriad classification methods, each with different properties.
- Unsupervised classification automatically separates land use into a number of computer-defined categories.
- Linear spectral mixing derives pixel-by-pixel measures of abundance for spectrally pure materials.
- Using ENVI, images were classified into eight classes with the ML classifier.
- The ground truth image was geometrically registered to the AVIRIS image using a 1-degree polynomial and bilinear resampling with 20 ground control points.
Results
- Accuracy of a supervised classification of land use typically ranges between 60 percent–90 percent depending on the classification scheme, the classifier, and the image itself.
- The accuracy assessment verified that the AVIRIS classification was superior to that of the synthetic ETM+ image.
- Importantly, the results of the synthetic ETM+ classification were virtually identical to a similar classification of ETM+ without added noise.
- On the off-diagonal, numbers show the change in misclassification; a negative number indicates that the classification does not confuse these classes as often with AVIRIS data.
- Page 817 versus vegetated land uses; commercial/industrial versus fallow, dry rangeland, and residential; and, urban irrigation versus irrigated crops and grassland.
Discussion
- The classified images contained similar types of misclassifications, often due to the well-identified problem of heterogeneity in urban land covers (Forster, 1985).
- Residential areas were sometimes confused with vegetated land uses because both have mixtures of soil and vegetation.
- The false positives decreased, in some cases dramatically, again perhaps because subtle signatures in the spectrum distinguished easily confused classes.
- The producer’s accuracy for fallow, water, and dry rangeland decreased with AVIRIS.
- In these fairly homogenous land uses, perhaps the additional bands in AVIRIS simply added noise and provided no additional useful information over synthetic Landsat.
Conclusion
- A supervised classification of AVIRIS was more accurate than one of synthetic Landsat ETM+ for land use classification at the urban fringe.
- Which imagery a researcher should use, provided both are available, largely depends on the purpose of the study.
- On the other hand, using AVIRIS produced a greater number of false positives for commercial/industrial land and performed poorly in classifications of relatively homogenous, less-vegetated land uses, such as fallow and dry rangeland.
- Since classification accuracy is dependent on a number of factors besides sensor specifications, caution should be used in extending the conclusions of this study to other areas.
Did you find this useful? Give us your feedback
Citations
2 citations
Cites background from "A comparison of AVIRIS and Landsat ..."
...For classes with similar but separable spectra, this is a reasonable assumption [17]....
[...]
2 citations
2 citations
Cites background from "A comparison of AVIRIS and Landsat ..."
...t i i l l , , , i , i l grouping band range 6~12, 13~21, 24~33, 40~54, 123~143, and 177~220 of the Indian Pine dataset [21]....
[...]
...The generated six bands are to simulate blue, green, red, near infrared, short wave infrared channels by grouping band range 6~12, 13~21, 24~33, 40~54, 123~143, and 177~220 of the Indian Pines dataset [21]....
[...]
References
4,586 citations
"A comparison of AVIRIS and Landsat ..." refers background in this paper
...…well documented in domains such as mineralogical mapping (Chabrillat et al., 2002; P H OTO G R A M M E T R I C E N G I N E E R I N G & R E M OT E S E N S I N G J u l y 2004 8 1 3 R.V. Platt was formerly with the Department of Geography, UCB 260, University of Colorado at Boulder, Boulder, CO 80309....
[...]
...While the advantages of high SNR are well documented in domains such as mineralogical mapping (Chabrillat et al., 2002; P H OTO G R A M M E T R I C E N G I N E E R I N G & R E M OT E S E N S I N G J u l y 2004 8 1 3 R.V. Platt was formerly with the Department of Geography, UCB 260, University of Colorado at Boulder, Boulder, CO 80309....
[...]
...A pair-wise comparison of the kappa statistics (Rogan et al., 2002; Congalton and Green, 1998) for the two classifications shows that these results are significantly different from each other with a z-value of 78.32....
[...]
4,154 citations
"A comparison of AVIRIS and Landsat ..." refers background or methods in this paper
...Less traditional sensors may provide additional information that can improve mapping accuracy....
[...]
...The classification system was a modification of Anderson Level II (Anderson et al., 1976) and used training samples from the following land use categories: residential, commercial/industrial, water, irrigated cropland, fallow, dry rangeland, grassland, and irrigated urban....
[...]
3,039 citations
"A comparison of AVIRIS and Landsat ..." refers background in this paper
...The classifier determines the probability that a pixel belongs to each class and then assigns the pixel to the class with the highest probability (Richards, 1999)....
[...]
2,576 citations
"A comparison of AVIRIS and Landsat ..." refers background or methods in this paper
...Furthermore, this advantage comes almost entirely from the large number of sensor spectral bands rather than the high Signal-to-Noise Ratio (SNR)....
[...]
...After creating the synthetic Landsat image, a Maximum Noise Fraction (MNF) transform was performed on the AVIRIS cube and synthetic Landsat images to reduce processing time and noise, (Green et al., 1988....
[...]
1,729 citations
"A comparison of AVIRIS and Landsat ..." refers background in this paper
...Keywords AVIRIS, Landstat ETM+, Colorado, Urban Fringe, Signal to Noise Ratio Disciplines Environmental Indicators and Impact Assessment | Environmental Sciences This article is available at The Cupola: Scholarship at Gettysburg College: https://cupola.gettysburg.edu/esfac/8...
[...]
...Sensors with a large number of continuous spectral bands, such as AVIRIS, are called hyperspectral imagers (Green et al., 1998)....
[...]