3 (yellow) Shrubs, Less dense forest
4 (orange) Grass
5 (cyan) Bare soil, built-up areas
6 (blue) Turbid water, bare soil, built-up areas
7 (red) Bare soil, built-up areas
8 (white) Bare soil, built-up areas
Figure 3.7 Image to do supervised (left) and define training data in classification (right)
3.6.5 Accuracy Assessment
Accuracy assessment is the comparison of a classification with ground truth data to evaluate how well the classification represents the real world. Classification error occurs when a pixel (or feature) belonging to one category is assigned to another category. Accuracy assessment is performed by comparing the map created by remote sensing analysis to a reference map based on a different information source. In order to be compared, both the map to be evaluated and the reference map must be accurately registered geometrically to each other. They must also use the same classification scheme. Accuracy of image classification is most often reported as a percentage correct.
Figure Classification accuracy assessment report
Figure Overall classification accuracy
3.7 Algorithm Model
After pre-processing, other measures need to be taken with the processing steps comprises two steps modeler model uses an algorithm to detect the particulate matter (PM10). Data is classified from the higher, middle and lower, then the other is classified as a cloud because the software cannot separate which one the land and cloud by using the modeler model in Erdas Imagine software.
The digital numbers (DN) of the four visible bands (Band 1, Band 2, Band 3 and Band 4) of Landsat 8 OLI and three visible bands (Band 1, Band 2 and Band 3) of Landsat 5 TM were extracted corresponding to the locations of in-situ PM10 measurements and converted into radiance and then to reflectance. The algorithm for determined PM10 for Landsat 5 TM and Landsat 8 OLI are different.
3.7.1 Landsat 5 TM
In Landsat 7ETM+ was given by (Liu, etc al., 1996) as
Τr = aerosol optical thickness (Molecule)
Pr(θ) = Rayleigh scattering phase function
μv = Cosine of viewing angle
μs = Cosine of solar zenith angle
Assume that the atmospheric reflectance due to particle, Ra, was also linear with the τa of a factor, K0 . This assumption was reasonable because Liu, et al., (1996) also found the linear relationship between both aerosol and molecule scattering.
Atmospheric reflectance was the sum of particle reflectance and molecule reflectance, Ratm, (Vermote, et al., 1997).
Ratm = Ra + Rr
Ratm = atmospheric reflectance
Rp = particle reflectance
Rr = molecule reflectance
The optical depth was given by Camagni and Sandroni, (1983), as equation. From the equation, rewrite the optical depth for particle and molecule as equation
t = σps
τ = optical depth
σ = absorption
s = finite path
The result was extended to a three-band algorithm as equation below
A = Particle concentration (PM10)
Ratmi = Atmopsheric reflectance, i = 0, 1 and 3 are the band number
ej = algorithm coefficients, j = 0, 1, 2, … are then empirically determined
Form the equation, founded that PM10 was linearly related to the reflectance for band 1 and band 2. This algorithm was generated based on the linear relationship between τ and reflectance.
3.7.2 Landsat 8 OLI
Algorithm number 13 selected to be our proposed algorithm due to its highest correlation coefficient of (0.834) and lowest root mean square error (RMSE) value of (11.836) between the measured and calculated PM10 values. The accuracy and validation of proposed algorithm results were performed using PM10 ground measurements and calculated PM10 by our algorithm. The relationship between extracted spectral reflectance from Landsat 8 OLI satellite image with PM10 ground measurements were examined and investigated through correlation analysis.
Table 3.2 Regression results (R) and (RMSE) using different forms of algorithms. (*) Calculated PM10 by algorithms, (**) b1, b2, b3 and b4 are the reflectance values for band1,band2, band3 and band4 (Geophys Remote Sens, 2014)
1 PM10(*) = 2.26 b(**)1 – 2.267 0.799 12.87
2 PM10= 2.04 b2 – 4.406 0.785 13.263
3 PM10= 1.81 b3 – 17.728 0.802 20.986
4 PM10= 1.39 b4 – 13.099 0.77 18.772
5 PM10= 3.56 b1 – 1.17 b2 – 0.255 0.79 13.127
6 PM10= 1.21 b1 + 0.98 b3 – 12.903 0.83 20.986
7 PM10= 1.56 b1 + 0.51 b4 – 9.458 0.789 13.156
8 PM10= 0.64 b2 + 1.27 b3 – 15.226 0.805 20.986
9 PM10= 1.24 b2 + 0.60 b4 – 11.170 0.832 11.879
10 PM10= 4.36 b1 – 3.50 b2 + 1.51 b3 – 12.615 0.81 20.986
11 PM10= 0.41 b2 + 2.27 b3 – 0.66 b4 – 16.174 0.802 20.985
12 PM10= 0.99 b1 + 1.62 b3 – 0.45b4 – 13.481 0.817 20.986
13 PM10= 4.72 b1 – 4.19 b2 + 3.07 b3 – 1.02
b4 – 13.871 0.834 11.836
3.8 Data analysis
Last product will produce analysis from processing data. In data analysis will produce a landuse changes within three years on area study, Klang, Selangor. Then, generate the final map for PM10 using algorithm in three periods (different years). Besides that, it also shows the correlation between landuse development and PM10 in three years in graph form. It will relate the PM10 since landuse development in that year.
This entire chapter explains about the research flow work from early stage of the methodology. A lot of alternatives can be used in order to manage data collection and analysis tasks. The data and information about the PM10 are used in this project to create an analysis base on the aim and objectives of the research by using ERDAS Imagine software. In this research planning is very important thing to be done. Due to the time constraint, it is needed to be complete the work orderly.
...(download the rest of the essay above)