33

I would have to say that the most complete software environment for Machine Learning and nonparametric modeling is R. This is a big field in statistics, spanning K-NN, Kernel smoothing, General Additive Models, weak learners, support vectors, neural nets, semi-parametric spline regression, imputation, etc... I would highly recommend reading: Hastie, T., R. ...


23

I'd strongly recommend scikits-learn for Python. It supports supervised and unsupervised classification and the documentation is excellent (particularly check out the Machine Learning for Astronomical Data Analysis tutorial and the accompanying YouTube video (note: this is 3 hours long)). The project is under active development, with the last version being ...


11

To simplify the raster it might be worth looking at gdal_sieve, it's available under the "Raster" menu. See: http://www.gdal.org/gdal_sieve.html N.


11

A good overview of machine learning techniques in R is the machine learning taskview. It offers a host of different algorithms, recommended by the experts.


10

There is a considerable body of literature on individual crown detection in spectral and lidar data. Methods wise, perhaps start with: Falkowski, M.J., A.M.S. Smith, P.E. Gessler, A.T. Hudak, L.A. Vierling and J.S. Evans. (2008). The influence of conifer forest canopy cover on the accuracy of two individual tree measurement algorithms using lidar data. ...


8

If I understand you correctly, you are looking for a supervised classification procedure. Some theoretical background: http://rst.gsfc.nasa.gov/Sect1/Sect1_17.html This is certainly possible through grass: http://grass.osgeo.org/wiki/Image_classification#Supervised_classification_2 As an alternative you could also look at saga (I'm not saying it is better, ...


7

I poked around on the State of Hawaii GIS site but most of the layers were too general or too old. The NOAA Coastal Services Center has some Landsat ETM derived land-cover data for Hawaii. There's also some 2.4m Quickbird-derived land-cover data for all islands except the Big Island. Hope that helps!


6

Your question assumes that machine learning algorithms for land classification are somehow distinct from software used for other machine learning applications. There are some applications that require special treatment because of unusual characteristics, but there is no reason I know of to think that land use needs special treatment. If land use data can ...


6

Creating landuse database, you need sensor technologies to detect and classify objects. i think you already know that: There are two main types of remote sensing: passive remote sensing and active remote sensing. Passive sensors detect natural radiation that is emitted or reflected by the object or surrounding areas. Reflected sunlight is the most ...


6

The only high resolution global land cover that I know is the one done by the PR of China. In Europe they seemed to use CORINE as an ancillary data, so it is difficult to judge the consistency accross the world, but it has a spatial resolution of 30 m. http://www.globallandcover.com/GLC30Download/index.aspx http://unstats.un.org/unsd/default.htm You ...


6

In optical remote sensing in the visible spectrum you cannot see through clouds. So there is nothing you can do, except to wait for images without clouds. Cloud masks are (as far as i know) used to exclude clouded areas from (for example) landcover classification, because results there would be incorrect anyways. edit As Aaron mentioned, you can sometimes ...


6

The size of the area is not the primary issue for the selection of a satellite sensor resolution. You should rather think about the size of the objects that you are mapping and their contrast with the background (see this paper about sub-pixel detection). My "rule of thumb" is that you do not detect accurately an object that is less than four pixels if you ...


6

There is a lot of ways to do this. If you want to preserve class ID, you need to mask the desired class. For example: var crop_mask = cover.eq(13); // create a mask for crops var crop = cover.mask(crop_mask); // mask it Map.addLayer(crop,{},'Crops'); As Rodrigo E. Principe and Nicholas Clinton suggest: var crop = cover.updateMask(cover.eq(13)); // mask it ...


6

You can use the aggregate function in the raster package to do this. For example, to produce a raster at half the resolution of the original, showing the proportion of cells covered with land class 1: aggregate(r, fact=2, fun=function(vals, na.rm) { sum(vals==1, na.rm=na.rm)/length(vals) }) To do this for all land classes: cov_pct <- lapply(unique(r)...


5

It does vary by Country - Turkey is poor because the data they used is minimal. For France, Germany, UK, Ireland the data accuracy is vastly better. If you want the accepted paper on the project "The Corine Land Cover (CLC2000) database received a thumbs up for accuracy from an assessment of the project, details of which were released by the EEA today. ...


5

Split tools has an error in ArcGIS 10 SP2. The tool makes the split, but leaves all features empty. Esri was registered this error and recommends for now, if you want use the split tool, downgrade ArcGIS to SP1. I suggest you visit this link http://resources.arcgis.com/es/gallery/file/geoprocessing/details?entryID=6C5D9A77-1422-2418-7F6C-01564409B1AF , where ...


5

Did you have a look at eCognition? With their new Version (8.9) they provide Random Forests algorithm within a GUI environment. You can create nice process trees and include object features.


5

You can find Global Land Cover 2000 dataset produced by JRC (an EU organization) http://bioval.jrc.ec.europa.eu/products/glc2000/products.php


5

Yes, there are free object-oriented (segmentation) software available. A few that come to mind are Spring, ITK, Orfeo toolbox and GRASS GIS. I would however point out that image segmentation is a poor direction to peruse when trying to model fractional cover. A segmentation algorithm is designed to minimize within unit variance and maximize between unit ...


5

It is a difficult thing that you are attempting. Small subtle changes in reflectances caused by different acquisition dates will cause major errors to arise when using your approach. You will have to do more preprocessing of your data, in order to have your approach be reliable. Normalizing the other years to your reference will most likely help, but it may ...


5

I would look at the support of you individual classes. If support for a given class is marginal in your fit model, the error may propagate in very undesirable ways. I would also consider fitting a series of binary models and predicting probabilities of each class separately. You could then perform a sensitivity test on the probabilities and evaluate if ...


5

You can make a paletted raster by assigning a colortable in the legend. If you have a raster called r and a data frame like yours above called ctab, with value and red/green/blue colour values, you can do something like this: > ctable = rep(NA,max(ctab$value)+1) > ctable[ctab$value+1] = rgb(ctab$red,ctab$green,ctab$blue,maxColorValue=255) > ...


5

Suppose two LULC rasters with 6 classes each one: library(raster) library(rasterVis) r <- raster() set.seed(123) lc1 <- setValues(r, sample(1:6, 64800, replace = T)) lc2 <- setValues(r, sample(1:6, 64800, replace = T)) To detect landcover changes, the basic approach is to using logical tests: changeDet1 <- lc1 != lc2 The result is 1 when ...


4

there are lots of method for calculating land use change in gis systems. if you want to use Arcgis, you should check out Confusion Matrix Analysis for your work. with confusion matrix you can also measure urban sprawl, too.. it is of course necessary to do some research. Wikipedia defination: In the field of artificial intelligence, a confusion matrix is ...


4

You can also do land classification with DTclassifier (Decision Tree classifier) plugin for QGIS. It provides simple interface for classification of raster data using decision trees, to perform within QGIS.


4

There is a group out of Duke University that have developed some interesting script tools for ArcGIS, including random forest models. Marine Geospatial Ecology Tools


4

To generalize, try running a majority filter. This is available in saga (and grass as well, check markusN his answer). An explanation for how it works from arcgis: http://edndoc.esri.com/arcobjects/9.2/net/shared/geoprocessing/spatial_analyst_tools/majority_filter.htm


4

The FAO released the GLC-Share which provides a set of major thematic land cover layers resulting by a combination of “best available” high resolution national, regional and/or sub-national land cover databases. Metadata and download link here: http://www.fao.org/geonetwork/srv/en/main.home?uuid=ba4526fd-cdbf-4028-a1bd-5a559c4bff38 Remind that both this, ...


4

eCognition is the best software for that, I did that using other software but eCognition its better. Here is the reference to literature on the subject: Karlson, M., Reese, H., & Ostwald, M. (2014). Tree Crown Mapping in Managed Woodlands (Parklands) of Semi-Arid West Africa Using WorldView-2 Imagery and Geographic Object Based Image Analysis. ...


4

I am assuming that by mode you mean the most frequent class? You can use the R function "table" to calculate the frequencies of a vector. x <- c(1,1,2,3,4,4,4,4) table(x) Then use which.max to return the class associated with the most frequent class. To return the actual class name you need to wrap the statement in names. which.max( table(x) ) names( ...


Only top voted, non community-wiki answers of a minimum length are eligible