What is Automatic classification?

What is Automatic classification?

Automatic classification is a set of procedures used to automatically place points in known classes. Automatic classification requires that the user has already pre-classified their data and uses this as a template. Automatic Classification can be done using templates from hand-classified images or from previous automatic classifications, but it is important that these have been checked for accuracy to ensure the correct data is being assigned new labels. Automatic classification software will find new objects not included in the original training examples and classify them with an unknown label.

There are two methods of Automatic classification: supervised and non-supervised learning. In supervised learning Automatic Classification, examples have both a point cloud and a label, whereas non-supervised Automatic Classification only has a cloud example available to train on where

Why use Automatic classification with point cloud software

Automatic classification is very useful when processing massive amounts of data. Automatic Classification provides a good method for assigning known labels to points in the point cloud.

How does Automatic classification work?

Non-supervised Automatic Classification works by finding objects in the scene that are similar in appearance or closely located to other known objects with their label and assigns this label to them, thus Automatic classifying these objects. They can also be Automatic classified by finding minimum spanning trees between groups of their neighboring pixels, which are similar in appearance or intensity, thus Automatic classifying these pixel groupings into regions/objects with similar properties. The supervised Automatic Classification algorithm creates templates from labeled examples then searches through the unlabelled examples for matches with high enough similarity. Automatic Classification is then able to assign labels to these matches. To classify the unclassified examples, Automatic Classification finds the closest matching template and assigns that label.

Practical uses of Automatic classification with point cloud software Automatic classification can be used in many situations such as: –

  • Mining – Automatic classifying minerals on a rock sample or oil/gas on an image of a region
  • Forestry – Automatic classifying trees and counting how many of each kind there are on a plot of land
  • Agriculture – Automatic classifying crop types and growth stages so they can be more efficiently harvested and maintained
  • Astronomy – Automatic Classifying celestial objects based off their projected spacial coords which allows scientists to track known celestial objects accurately