Hi David,

The more important question should be which method to use, rather than which software. ArcGIS, ERDAS, ENVI, QGIS, GRASS, ILWIS, R, etc, etc. (there are many more) can all do what you want to do, and the quality of your output is not dependent on which of these platforms you use.

First figure out which method of classification method best suits your objectives and the data available to you. So, should you be using supervised or unsupervised classification or as suggested in another answer, segmentation based classification (works best if you have very high resolution imagery)?

Then pick any software that provides this capability and one that you are comfortable using. Standard supervised and unsupervised classification methods and their variants can be found in most software platforms. Segmentation based methods will be restricted to fewer options.

Yes, it should be a good option. I think GrassGIS, which is a QGIS application and is free, might be worth consideration.

GRASS GIS is what I use . Here you have a wide variety of algorithms to choose from.  You can make supervised, unsupervised and image segmentation .  The last one is very useful in case you have little information about  the area of study and wish fro example to segment  the image in order to find potential training sites .  QGIS is also a good option . Take a look at the semi-automatic classification plugin for QGIS.

Hi David,

The more important question should be which method to use, rather than which software. ArcGIS, ERDAS, ENVI, QGIS, GRASS, ILWIS, R, etc, etc. (there are many more) can all do what you want to do, and the quality of your output is not dependent on which of these platforms you use.

First figure out which method of classification method best suits your objectives and the data available to you. So, should you be using supervised or unsupervised classification or as suggested in another answer, segmentation based classification (works best if you have very high resolution imagery)?

Then pick any software that provides this capability and one that you are comfortable using. Standard supervised and unsupervised classification methods and their variants can be found in most software platforms. Segmentation based methods will be restricted to fewer options.

Thanks!

Is there book or article that mentions in which conditions different methods are used?

Lillesand and Kiefer is one of my favorites., but there are plenty of them.

Another one of my favorites is  Robert schowengerdt ! 

erdas imagine is the best SW for image processing

Hy David, i agrre with Sumeet, Erdas is more specialized, other GIS can do it yes, but Erdas is one of the best.

Hi! David, particularly I use ENVI CLASSIC + IDL. 

Hi David,

I am assuming you’re using ArcGIS’s Image analyst extension, if that’s the case then it’s more than capable to carry out the classification (Unsupervised or Supervised). The concern here should be the accuracy assessment carried out after the image classification process.

I have used ArcGIS, ERDAS Imagine, ENVI and ILWIS but, like Varun mentioned earlier, there are also other factors to be considered depending on the theme of your project. 

Try also eCognition from Trimble. You have the possibility to request a free trial. 

BEST SOFTWARE IS ERDAS IMAGINE. GO FOR BOTH SUPERVISED. AND UNSUPERVISED CLASSIFICATION

Hi David, I have been used Forest Canopy Density (FCD) software to classify vegetation and sucsess with accuracy more 80%. good luck 

One more option for classification is the OTB provided by CNES integrated into QGIS.

Hi David,

Remote sensing softwares (ENVI, ERDAS ..) are very powerful for most of the classifications.

ERDAS IMAGINE is the best!

ERDAS IMAGINE would be the best where you could a supervised or non-supervised classification.

In this case you need to make unsupervised or supervised classifications. It is more appropriate a image processing software rather than a GIS. You must use Erdas, ENVI as is suggested before

eCognition from Trimble, ArcGis, Spring from Inpe (Brasil).

And QGis

You can go for python(programming perspective). A simple 10-15 lines can give you a nice classification result.

Step by Step procedure(Supervised Classification):

1: Collect the training data from your satellite image. You can take help of multispec to view the data in various bands and selecting the training area.

2. Convert the DN values to reflectance and do atmospheric correction(see LANDSAT guide).

3: Create a classifier using training area. The algorithms you can use are SVM, Kneighbors, Guassian etc. in python in scikit library you can get all these algo. Just simple  1-4 lines to run it. Overall my classification experience says normally SVM gives nice result. But can always check the classification accuracy by making confusion matrix and check the producers and users accuracy.

3) plot the classified image.

Libraries to import(Python):

from scipy import misc

import scipy

import numpy as np

from sklearn.metrics import confusion_matrix

from sklearn.naive_bayes import GaussianNB

from sklearn import svm

from sklearn import neighbors

import matplotlib.pyplot as plt

import matplotlib as mpl

import os

from compiler.ast import TryExcept

from sklearn.metrics import precision_score

from sklearn.metrics import recall_score

import matplotlib.cm as cm

import math

from scipy.stats import itemfreq 

Hope this will help.

I have used ERDAS to make a supervised classification of several Landsat images, and it worked pretty well, we implemented an expert classification and a posterior map corroboration on the field.

However when choosing a suitable method, you should take into account account the extent of your study area, data source and quality, resolution a purpose of your study.

Best regards

Article Ecological classification system and ecosystems maps: Concep...

Hola David, coincido con Enrique, ERDAS es el más especializado, también con la suite de ERDAS viene ErmApper, que sirva para los mismo , análisis satelital, Yo te recomiendo ERDAS ya que es el que tiene mas experiencia, los acabados son muy profesionales.

Aunque el de Arcgis no es malo, dependerá de tu versión, para la 10.2.1 existe un módulo para imágenes Landsat 8 (si es el caso) , que te permite obtener firmas espectrales y hacer clasificación, la verdad funcionan muy bien. Te dejo liga.

https://www.arcgis.com/home/item.html?id=f331af8569fb4f19bd74954eddc44bd3

Saludos

It really depends on what kinds of remote sensing techniques you would like use: object-based or pixel-based classification.

For object-oriented classification, E-cognition software provided by Trimble is very good to generate image segmentation. However, a license is quite costly. See the link

http://www.ecognition.com/suite/ecognition-developer

For pixel-based classification, Erdas or ENVI are both good for image classification. However, they are also very expensive for obtaining a license.

For student, I recommend IDIRISI provided by Clark University, US for image processing with quite reasonable cost for a license. If you are a student, you also get a discount for obtaining a license.

https://clarklabs.org/products/

Hope answers could help some.

All the best;

Dat

Depending on your goals and your computer you can either use QGIS's semi automatic classification plugin or ERDAS's supervised classification and accuracy assesment.

Any of them can be best up to your objective and data spesification.

"No free lunch theorem"

for classification you can use ERDAS IMAGINE software. In ERDAS there is supervised classification option as well as unsupervised classification. I used supervised classification. even you may go for on screen digitization for classification, if you have good idea of your study area.

ERDAS IMAGINE 8.7

Great suggestions. Some of the above mentioned solutions are implemented into IMPACT toolbox, a free and open source potable tool developed by the Joint Research Center. It includes fully automatic classification, image segmentation, post classification and mapping tool and much more. Have a try ... 

http://forobs.jrc.ec.europa.eu/products/software/

Technical Report IMPACT: Portable GIS toolbox for image processing and land c...

Hola David, ya había escrito una respuesta, pero ahora detallando lo que menciona Dario Simontti, existen varios software libres para hacer esto, el que menciona Dario es muy bueno, otro mas es White Box Gat (te dejo liga), es de una Universidad de Canadá, trabaja en windows y mac.

http://www.uoguelph.ca/~hydrogeo/Whitebox/

Otro gratuito y multiplataforma es LeoWorks de la Agencia Espacial Europea (te dejo liga).

http://leoworks.terrasigna.com/index.php?s=download_leoworks

Al final, la mejor forma de clasificar no te la dará un software en específico, te la dará la calidad de la imagen (escala), la veracidad y cantidad de tus muestras si es que haces clasificación supervisada o bien tu conocimiento del territorio y experiencia si es que haces clasificación no supervisada.

Saludos

softwares free (QGIS e Spring)

ENVI and QGIS

Grass gis

We've used MARS for automatic clasification and edit point' clouds of LIDAR sensors. The experience is ok, but edition and control error is compulsory.

Hola David

Podés usar SOPI. Es software libre desarrollado por la CONAE (Comisión Nacional de Actividades Espaciales) de Argentina. Podés hacer clasificación supervisada y no supervisada y multinivel. Tenés álgebra de bandas y herramientas de edición vectorial. Todo en español. Tambén tenés cursos en plataforma virtual. Descargalo de: https://sopi.conae.gov.ar/

Saludos

Leo

https://sopi.conae.gov.ar/

Hola ya varios colegas han sugerido sig´s, pero estoy de acuerdo primero define tu escala espacial de trabajo, para luego selecional el tipo de imágenes y su resolución, de esto último depende la confiabilidad de tu información, de paqueteria de los mas sencillos es QGIS.

MultiSpec is free of charge

https://engineering.purdue.edu/~biehl/MultiSpec/download_win.html