0

Which classification is best for Crop mapping? Random Forest or Maximum Likelihood?

closed as primarily opinion-based by wetland, Erica, aldo_tapia, Dan C, MaryBeth Dec 20 '17 at 14:19

Many good questions generate some degree of opinion based on expert experience, but answers to this question will tend to be almost entirely based on opinions, rather than facts, references, or specific expertise. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    I don't think that it is opinion based because the algorithms have been well studied. – radouxju Dec 20 '17 at 14:31
4

This is a hard question to answer, as it is dependent on so many variables:

  • How many different species of vegetation are you planning on mapping i.e. the level of detail.
  • The parameter selection process for RF.
  • Pixel based and object based classifications.

Most of the literature states that very similar results can be obtained from both classifier approaches when a non-complex scene is being used, however if the scene is complex then RFs are superior. Maximum Likelihood has been around for a long time and has been research extensively. It can offer satisfactory results and is fairly easy to implement.

Random Forests are newer in comparison and offer a powerful technique for remote sensing classification. RF classification uses a large number of decision trees to get to the final result. Each tree is created using a random sample selection. A random subset of input predictors is used at every tree to split it making a new node. The final result is gathered from the majority vote created by all the trees in the process.

One advantage of using RF is that it can give you an indication of the relative importance of the input variables used. This blog explains this very well https://www.r-bloggers.com/random-forest-variable-importance/.

Many studies involving crop mapping have noted the inferiority of ML to other non-parametric classifier such as RF (Huang et al., 2002; Dixon and Candade, 2008; Yang et al., 2011; Nitze et al., 2012).

References

  1. Huang, C., Davis, L. C. & Townshend, J. R. G, 2002. An assessment of support vector machines for land cover classification, International Journal of Remote Sensing 23, pp. 725-749.
  2. Dixon, B. & Candade, N., 2008. Multispectral landuse classification using neural networks and support vector machines: one or the other, or both?, International Journal of Remote Sensing 29(4), pp. 1185-1206.
  3. Yang, C., Everitt, J. H. & Murden, D., 2011. Evaluating high resolution SPOT 5 satellite imagery for crop identification, Computers and Electronics in Agriculture 75, pp. 347-354.
  4. Nitze, I., Schulthess, U. and Asche, H., 2012. Comparison of machine learning algorithms random forest, artificial neural network and support vector machine to maximum likelihood for supervised crop type classification. Proc. of the 4th GEOBIA, pp.7-9.

Not the answer you're looking for? Browse other questions tagged or ask your own question.