The categorization and aggregation of satellite imagery into discrete classes, known as land use classification, has long been a part of remote sensing science. Our image analysts are well-trained in the established sciences of image classification, and are advancing with the emerging field of object-based classification. Our unique techniques help you extract information you need from appropriate sources. Specifically, we have honed our expertise in the development of land use models for the wireless industry (a.k.a. clutter).

Traditional Classification

Traditional classification techniques divide an image into clusters that have similar spectral signatures. Our analysts use training areas to tell the software where pure pixels of a particular class are located. If ground truth (information representing actual conditions on the ground at points within a survey area) is available, it is the preferred source of training areas. If not, we use knowledge of local land cover and experience with similar projects to choose appropriate training areas. Once selected, the software runs through an algorithm to assign each pixel to a class, and finally the results are adjusted by comparing the classification to the source satellite imagery. It may take several iterations to get a classification right. This is a task for experienced image analysts, and i-cubed has over a decade of experience performing this type of work.



 Click to view larger image Clutter for Wireless

Some of the most popular and cost effective geographic data products i-cubed offers for wireless network design or optimization are clutter maps, also referred to as morphology or land-use maps.These digital maps are used in all of today’s state-of-the-art radio frequency (RF) propagation tools to model path loss, signal attenuation and frequency re-use.

i-cubed offers a range of clutter map options to match your budget and technical requirements. These maps are most commonly derived from satellite imagery using traditional classification techniques. We have perfected a technique for deriving a set of clutter classes that is highly suitable for RF propagation models.

Typically used in more populated markets, our Clutter 30 product offers 12 classes at a pixel size of 30m (1 arc-second). This uniquely-tailored land use classification scheme has been immensely popular with RF propagation engineers since we started producing it in the late ‘90s. The 12 classes include two levels of residential areas, three levels of urban areas, and an industrial/commercial class. The slicing of built-up areas into these various classes helps RF engineers accurately model radio frequency propagation over the urban landscape. Clutter 30 can be produced for just about any place in the world.


Object Based Classification

i-cubed has been closely following the new and exciting field of Geographic Object-Based Image Analysis (GEOBIA). Pixel-based methods tend to break down for imagery below 5m resolution, GEOBIA offers an alternative that aggregates pixels into homogeneous areas for subsequent classification.  Setting up the criteria for classification requires an experienced image analyst and should not be done by amateurs. For a truly accurate classification, ground truth should be collected in the field. If it’s not available, reliable information about local land use (from someone familiar with the area) is the next best alternative. Regardless of whether you have ground data or not, we are happy to talk to you about generating a classification using GEOBIA.