Landsat Applications to Agricultural and Urban Feature Analysis

Project Menu







Classificaton Procedure Methods

Explanation of Methods for Landsat Images

• All images are Landsat 4-5 TM downloaded from EROS website using GLOVIS.

• Images were first imported into Idrisi Taiga then cropped to show area of interest using the REFORMAT>WINDOW function.

• Each band for each image was processed for haze reduction using GIS ANALYSIS>MATHEMATICAL OPERATORS>SCALAR. Only bands 1-5, 7 were used for the images processed using the hyperspectral analysis modules in Idrisi.

• ISOCLUST was used for cluster analysis and each band 1-5, & 7 was included in the analysis. The number of iterations used was 3, and the number of clusters was 7, minimum sample, used default 60.

• When ISOCLUST did not work, ISODATA was used. Number of clusters 7- 10, MAX number of clusters – 15, MAX iterations – 20. The results were not as good using ISODATA so CLUSTER was used and then reclassed the images a number of times in order to get the results needed.

• RECLASS function was used on each ISOCLUST, some multiple times to refine image.

• HYPERUSP function was used on isotmp and clustertmp files with good result when reclassed. Some images had to be reclassed a few times in order to get the best results.

Method Terminology


• MOSAIC – automates color balancing when adjacent overlapping images are joined into a single larger image.

• REFORMAT – allows for data and file type of a file, reorient an image or vector file, change the extent of the study area, change resolution, generalize the level of detail in the file, join files together, and conver files from raster to vector and vice versa.

• WINDOW- extracts a rectangle sub-area of a larger image to create a new smaller image.

• GIS Analysis – has the ability to perform analyses based on geographic location.

• MATHEMATICAL OPERATORS – a set of mathematical tools necessary for compete map algebra.

• SCALAR – does scalar arithmetic on images by adding, subtracting, multiplying, dividing or exponentiating the pixels in the input image by a constant value.

• ISOCLUST – an iterative self-organizing cluster analysis procedure using a predetermined number of clusters.

• ISODATA – provides an unsupervised classification of input images using an iterative self-organizing data analysis technique.

• RECLASS – performs image reclassification.

• HYPERUSP – provides unsupervised classification for hyperspectral image data. It was used with non-hyperspectral data in this project. NOTE – we derived interesting results using the isotmp or clustertmp files with one class so then we tried using 6 classes for all of the bands 1-6. Band number 5 “hyperusp_atl_10_5” worked well so then we did a RECLASS function and the results appeared promising. Each HYPERUSP set of files has one seemingly useful image out of six to work with.

Class Project for ES771 - Remote Sensing
Emporia State University
Brenda Zabriskie, Mary Reardon, Jay Doolittle
December 09, 2010