I have two different date of Landsat image.I want to apply image regression on Landsat images for land use change detection. Any body have any idea regarding this please share with me. Is this possible to do with erdas model maker??
Thanking you
Image regression is a vague term. Can you be more specific? Are you looking to measure change across space or time or both?
You only have toe images so using regression on two data points (at the pixel level) would be highly questionable.
I have to apply regression technique in mean and variance between pixel values for different dates. Before applying regression technique on the image I have done NDVI. The earlier NDVI image will be considered as independent variable and recent NDVI image will be dependent variable. After that I want to apply linear relationship between these two images. Then the predicted image will be substracted from the base image. I don't know how to do this regression analysis in ERDAS, ArcGIS or ENVI ? I was trying to do in ERDAS model but it did not work.
Have you used any of the three functions under 'Spatial Analyst' tool in ArcMap ToolBox? With ArcMap ver. 10.3.1 I can see the following tools, Exploratory Regression, Geographically Weighted Regression (GWR) and the Ordinary Least Squares (OLS) linear regression tools.
In ERDAS Imagine, I could not find a ready made model for performing regression analysis. If you have a regression equation which shows the relationship between the dependent and independent variable, which in your case will be two NDVI Images then this equation can be used to create your own model in ERDAS Imagine Model Maker. Check attached links.
However, I tend to agree with Prof. Hamilton because the results generated from such a regression analysis would be more of a statistical operation rather than having any real significance. I say this because the spatio-temporal dynamics of lets say changes in vegetation could be dependent on so many factors like climate, land use, anthropogenic impact, etc. etc. and to predict such an outcome based on just two images would be indeed questionable!
Do you have any research articles which you can share that support your hypothesis?
http://www.helsinki.fi/maantiede/geoinformatiikka/opetus/2005_2006/GISKeijo_Final.pdf
https://www.academia.edu/9217875/Air_Pollution_Modeling_from_Remotely_Sensed_Data_Using_Regression_Techniques
- Badges
- Science topic
- Similar topics
- Computer Science
- Images