AUTOMATIC GEOLOCATION OF AERIAL PHOTOGRAPHY
Road Image Processing
The goal of this step is to prepare the road images for correlation analysis with processed satellite images. We will attempt to create a boolean matrix or roads to simplify and reduce the data storage needs
Road Dataset
All our road map images come from Mapbox and are based off the OpenStreetMap project. This website allows us to export images at precise GPS coordinates that show only the features we'd like (no labels). Additionally, consistent colors unique to each feature lends these images well to color filtering.
Figure 1: Road Map of San Francisco, CA [1]
Color Filter
To filter the road images we use a basic color filter. The function examines the color at each pixel of the image. If the color is within a certain range of oranges or white, the pixel in the output image is set to value 1. Otherwise, the pixel is set to value 0. The color selection was given a moderate range in order account for slight mixing of colors at the edges, an artifact of image storage methods. The result of this transformation are shown in Figure 2.
Figure 2.1: Road Map Color Filter Input
Figure 2.2: Road Map Color Filter Output
The color filter concludes the processing of the roads in our final implementation. The results provide a sufficient base with which we can correlate the output of our satellite image processing
[1] Mapbox. [Online]. Available: http://mapbox.com.