Land Use-Cover Change Detection Using Knowledge based approaches: Remote Sensing and GIS
Knowledge-based Image Analysis
Feature space plot in two dimensions depict the distribution of all the pixels in the scene using bands 3 and 4 (Figure 1). Feature space plot provide great insight into the actual information content of the image and the degree of between band correlation. Therefore, this feature space partitions can be used as the actual decision logic during the classification (Jensen, 1996). Classification using spectral reflectance classes wee done for all images 1988 and 1995.
Figure 1: Feature space of Bands 3 and 4
Modified Spectral Classification
Modified spectral classification was introduced to compensate the conventional radiometric correction. Prior knowledge about land cover spectral and relations with land use are important, however, in some cases land cover change in land use. Base on spectral reflectance of bands 3 and 4 new value for bands 3 and 4 was calculated using this formula: (Figure 2).
Figure 2: Modification of feature space bands 3 and 4
NBk = ((((band k-ak) / (bk -ak)/10)+1)>10,10 (band k-ak)/(bk-ak)/10) +1)
Where a
k= lower value of band k, b
k = upper value of band k, to make the value of bands 3 and 4 are equal for every image, so land cover classes based on spectral reflectance could be determined. These transformations results in comparable with feature space plots and images as well.
Land Cover classification
| Band 4 |
Band 3 |
| 3 |
3 |
6 |
6 |
6 |
9 |
9 |
12 |
12 |
12 |
| 3 |
3 |
6 |
6 |
6 |
9 |
9 |
12 |
12 |
12 |
| 3 |
3 |
6 |
6 |
8 |
8 |
11 |
11 |
11 |
11 |
| 2 |
4 |
5 |
5 |
8 |
8 |
11 |
11 |
11 |
11 |
| 2 |
4 |
5 |
5 |
7 |
7 |
11 |
11 |
11 |
11 |
| 1 |
4 |
7 |
7 |
7 |
7 |
11 |
11 |
11 |
11 |
| 1 |
4 |
7 |
7 |
7 |
7 |
11 |
11 |
11 |
11 |
| 1 |
10 |
10 |
10 |
10 |
10 |
11 |
11 |
11 |
11 |
| 1 |
13 |
10 |
10 |
10 |
10 |
11 |
11 |
11 |
11 |
| 14 |
13 |
13 |
13 |
10 |
10 |
11 |
11 |
11 |
11 |
Figure 3 Feature space of TM bands 3 and 4 (number indicate cover code : see table 1 for class description)
After the image had been compared, all possible combination of values in bands 3 and 4 were classified according a box classifier (figure 3). With this classifier, the reflectance was 'translate' to cover classes. Using the result of Bronsveled (1994), and personal knowledge of study area gathered during field work, vegetation and soil characteristic can be estimated on the basis of reflectance data in red and infrared. A 'belief' factor was assigned to each label of land cover for each class as land use (table 1).
Table1 cover classes with assigned beliefs(blu)for each label of land use- cover
| code |
cover |
for |
rubfor |
op |
rub |
mh |
ra |
gl |
urb |
cons |
cl |
wb |
| 1 |
Dark veg |
0.8 |
0.8 |
0.6 |
0.6 |
0.5 |
0.3 |
0.1 |
0.4 |
0.1 |
0.1 |
0.8 |
| 2 |
Mid veg |
0.8 |
0.8 |
0.8 |
0.6 |
0.6 |
0.8 |
0.1 |
0.5 |
0.1 |
0.1 |
0.5 |
| 3 |
Light veg |
0.3 |
0.3 |
0.5 |
0.8 |
0.6 |
0.8 |
0.6 |
0.3 |
0.3 |
0.3 |
0.1 |
| 4 |
D veg+soil |
0.8 |
0.8 |
0.8 |
0.8 |
0.5 |
0.5 |
0.1 |
0.5 |
0.1 |
0.1 |
0.5 |
| 5 |
M veg +soil |
0.3 |
0.3 |
0.5 |
0.5 |
0.8 |
0.5 |
0.3 |
0.8 |
0.3 |
0.3 |
0.3 |
| 6 |
l veg+soil |
0.1 |
0.1 |
0.3 |
0.3 |
0.3 |
0.8 |
0.8 |
0.5 |
0.1 |
0.1 |
0.1 |
| 7 |
M soil +veg |
0.5 |
0.5 |
0.5 |
0.5 |
0.8 |
0.5 |
0.3 |
0.8 |
0.3 |
0.3 |
0.1 |
| 8 |
l veg+veg |
0.3 |
0.3 |
0.5 |
0.5 |
0.7 |
0.7 |
0.8 |
0.5 |
0.5 |
0.3 |
0.1 |
| 9 |
m soil+veg |
0.1 |
0.1 |
0.3 |
0.3 |
0.3 |
0.8 |
0.5 |
0.3 |
0.3 |
0.3 |
0.1 |
| 10 |
Dark soil |
0.3 |
0.3 |
0.3 |
0.3 |
0.3 |
0.1 |
0.1 |
0.3 |
0.3 |
0.5 |
0.7 |
| 11 |
Light soil |
0.1 |
0.1 |
0.1 |
0.1 |
0.1 |
0.5 |
0.5 |
0.6 |
0.8 |
0.6 |
0.1 |
| 12 |
V light soil |
0.1 |
0.1 |
0.1 |
0.1 |
0.1 |
0.1 |
0.1 |
0.3 |
0.5 |
0.8 |
0.1 |
| 13 |
Water |
0.5 |
0.3 |
0.3 |
0.3 |
0.3 |
0.5 |
0.3 |
0.5 |
0.1 |
0.1 |
0.8 |
| 14 |
W/shade |
0.3 |
0.3 |
0.3 |
0.3 |
0.1 |
0.5 |
0.1 |
0.5 |
0.1 |
0.1 |
0.8 |
Cover-use lable:
For= forest, rubber and secondary forest,
rub= rubber, mh = mix horticulture
Ra= recreational area; gl = grass land ; urb = urban;
cons = construction; cl =cleared land wb= water body
The cover class- use maps 1988 and 1995 could be derived form new bands 3and 4 |
Land Use Change
To determine the land use change from land cover change, the average value of land cover chances in years 1 and 2 could be calculated, and the maximum value was assigned for selected land use change class. The land cover and land use change could be calculated as follows:
| BLUC1(date 1-date 2)= |
BLU(date1)+BLU(date2)
------------------------------
2 |
| BLU2 (date1-date2) = |
BLUC1(date1-date2) +BSLUC(date1-date2)
-------------------------------------------------------------
2 |
BLUCI=blief land cover change based on image data.
BLUC2=blief land use change based on BLUCI and belief factor on "succession" of land use change.
Visual Interpret ion
On screen digitizing for visual interpretion was done for image 1995. Image characteristic such as tone, texture colour and pattern are translated into land use attributes. The translation process (transfer function) is guided by local knowledge (e.g land use map, land cover map or agriculture statistic) which were collected during field work or background studies. Polygons are drawn around features (forest, oil palm, rubber, urban, cleared land, grass land, homogenous area, ect) and a label was assigned to each polygon, characterizing it by attributes (the legend).
