Supervised Classification

Supervised classification involves creating a "seed" of spectral data that will be used as the core pixels for creating land cover classification. Unlike unsupervised classification, only minimal user interaction is needed.  Familiarity with the area is still an important aspect as the analyst still needs to know what land features are in the ares such as lakes, roads, urbanized and residential, forest (deciduous and/or conifer) and grass.

Unsupervised Classification

Unsupervised classification is a process of classifying that uses a computational algorithm to "cluster" pixels of similar density.  Clusters are then classified by name by an analyst.  It is important the the analyst be aware of and familiar with the subject image.  The image below was originally clustered into 50 classes.  The analyst (me) then named each class and later merged all similar classes into only five remaining categories.

Thermal and Multispectral Analysis

Thermal imaging takes advantage of the thermal layer (band 6).  This thermal layer is intensified visually the higher the temperature of an object.  The image below is displayed using RGB layers 5,7,3.  The bright yellow indicates the fire and the bluish hue indicates the smoke.  The black object just northeast of the fire is a lake.

Multispectral Analysis

I hope I got these right.  This week we worked with multispectral analysis.  By adjusting color bands and contrast, we can better see variations of color which can further assist in our analysis of the image and the features that are present.

Gray scale image with Layer_4 enhanced shows the brightest spots on the image

Multispectral image with enhanced layers

High contrast between two bodies of water

Spatial Enhancement

Remote sensing covered the topic of spatial enhancement and the different methods/filters by which spatial enhancement can be achieved.   The deliverable below was created using a number of tools ERDAS Imagine including the Fourier transform editor to remove striping lines created by a faulty LANDSAT 7 satellite.  The Convolution filter was used to sharpen the image further.  That sharpened image was then opened in ArcMap where the Focal Analyst tool was used to create a more blended image without compromising the image itself.

Introduction to ERDAS Imagine

This week we took our first look at ERDAS Imagine.  Definately a lot of setup involved with this program.  Sometimes that's a good thing, sometimes not.  It's good if you know what you are doing.  As with any program, time and experience will make this program as easy to use as any other.  

Ground Truthing of LULC Classification

Ground truthing is the process by which LULC classification are verified by taking a sample of the classifications and performing an in-situ or as my industry calls it "boots on the ground" verification.  Classifying LULC via aerial photography can only be so accurate.  High resolution photos may not be available and lower resolution may not be clear enough to determine what an object is.  Experience of the classifier can compensate, but there are still other instances where the ground truthing is the better classification route.  For example, there could be areas that were rezoned from residential to commercial.  In an aerial photo, the location would look like it should be zoned residential, but ground truthing reveals that what once was a home is now a office for a lawyer or a dentist.  In the map below, I chose to do was is called systematic sampling.  By ground truthing a sample of the classification, you can determine how accurate your overall assessment should be.

Remote Sensing - Land Use Land Cover Classification

Land use is defined as the way that the land is used by humans whereas land cover defines the way the land is vegetated.

Remote Sensing - Visual Interpretation

Module 2 of Remote Sensing begins our foray into identifying features on an aerial photograph.  Having learned of the various films that are used and viewing the resulting photographs it is easy to see that all have their purposes.  This week we identified the different levels of tone and texture as shown in map 1, and we took a stab at identifying features based on size and shape, pattern, shadow, and association as shown in map 2.  We also compared two photos of the same area, one true color and one false color.

Map 1: Tone and Texture

Map 2: Size and shape, shadow, pattern, and association