0800 912 0988

What's Coming in ERDAS IMAGINE 2018


At the beginning of June, I attended the HxGN LIVE 2017 conference in Las Vegas. It was an exciting event with interesting insights into the current and future capabilities of Hexagon Products. Whilst there, I got a sneak peek at the new capability to be revealed in 2018 which I will outline for you in this article. If you wish to see an overview of the whole event, have a look at Kumar’s article here.

As we expected, the majority of development will be in the Spatial Modeler in the form of new operators. The most exciting, in my opinion, are the machine learning capabilities and the vector size and shape operators! Let me explain these in more detail:

Machine Learning

This is a new development which promises to be a very exciting tool indeed. Essentially, the machine learning algorithms allow you to automatically identify images and features. Initially, the user supplies a classified dataset to act as a training dataset. ERDAS IMAGINE uses this to build an algorithm known as a decision tree; this can then be applied to future unclassified inputs to identify features and objects recognised on the original training dataset. Figure 1 displays the feature learning operator classifying zones of a CIR image into land types such as grassland, forestry and bare earth. One of the algorithms confirmed so far is the Random Forest Classifier.

Figure 1: Workspace of Machine Learning Random Forest Classifier in IMAGINE 2018. Showing the output (top left), the spatial model, segmentation (bottom left) and input training data (bottom right)

Figure 2: Example decision tree, built using the new Machine Learning Operators in IMAGINE 2018

Vector Size and Shape Operators

A suite of GeometryCue Operators, similar to the IMAGINE Objective functionality, should be released in 2018. These will allow users to analyse the shape characteristics of polygons thus enabling effective filtering of incorrectly identified features. For example, if you wish to extract only roads, use the Compute Compactness operator and specify to keep only features with a low compactness value.

With these operators, you can also reduce the complexity of output polygons. The Orthogonalize operator computes an orthogonal polygon of best fit. Or you could use the bounding box creator to fit a rectangular surround to each polygon, both vastly reducing the vertices required.

Vector capability has never been a focus in ERDAS IMAGINE, but extending the Spatial Modeler in this way should allow better workflow automation, reducing the need for any vector post-processing.

There are at least 57 new operators for 2018 including:

ERDAS IMAGINE 2018 will be released on the 15 January 2018. As usual, we will be in contact with all of our ERDAS IMAGINE customers around that time to provide full details. Once you have tested the new operators, we would be very interested to hear what you think and share your stories in this newsletter.

Emily Winter
Geospatial Software Engineer

Add a Comment

Latest Posts
No Comments