Tutorial - pdil/photo-density GitHub Wiki

0. Create photo geodata file

The Java Extractor folder contains the .java file which handles the creation of the data set of photo locations that are plotted on a map. A sample of the data set is as follows:

id	year	month	day	file	lat	lon
1	2013	Sep	29	IMG_1407.jpg	35.0679083	-106.63624
2	2013	Oct	11	IMG_1700.jpg	35.067775	-106.63557
...	...	...	...	...	...	...
1498	2015	Apr	7	IMG_5410.jpg	35.08214167	-106.6247944

The Java code uses the metadata-extractor library file available here: https://github.com/drewnoakes/metadata-extractor

1. Install and load `ggmap`

install.packages("ggmap")
library(ggmap)

ggmap contains the required functions for easily creating map plots from Google Maps images.

2. Import and prepare data

Import the file containing photo locations:

# import data from csv
# headers are id, date, file, lat, lon
geo_data <- read.csv("data/geoDataTable.csv", header = TRUE)

Missing values in the data set were automatically entered as -999 by the Java code, so we replace them with the R-friendly NA here.

# missing values were entered as -999
geo_data[] <- lapply(geo_data, function(x) {replace(x, x == -999, NA)})

The images we are interested in are located in the Southwestern United States, so we create our bounding box using the longitude and latitude of this region. The longtidue and latitude will essentially map to the x-y coordinates of our plot.

# lat/lon coords of "southwest region" bounding box
sw_top <- 50
sw_bottom <- 25
sw_left <- -130
sw_right <- -100

Now that we have defined the region we can extract only the images that are in this region to be plotted.

# select only images with locations in selected region
sw_data <- geo_data[geo_data$lon <= sw_right & geo_data$lon >= sw_left 
                    & geo_data$lat <= sw_top & geo_data$lat >= sw_bottom, ]

3. Plot map with 2D density

The get_map() function in ggmap allows us to obtain a Google Maps image of the region we define:

# create map image
swmap <- get_map(location = c(lon = -115, lat = 37.5), zoom = 5, maptype = "roadmap")

Finally, we can plot the image locations on this map using ggmap. This package follows the grammar of grpahics used in ggplot2 so the same logic and functions apply here.

# use ggmap to create map with points, 2D density, and legend
sw <- ggmap(swmap, darken = 0.1) 
sw <- sw + geom_point(data = sw_data, alpha=0.4, aes(x = lon, y = lat), colour = "blue", size = 3)
sw <- sw + stat_density2d(data = sw_data, aes(fill = ..level..), alpha = 0.25, geom = "polygon", bins = 14)
sw <- sw + scale_fill_gradientn(colours = colorFunc(20), name = "Density")
sw <- sw + xlab("Longitude") + ylab("Latitude")
sw <- sw + theme(legend.position = c(0.1, 0.2), legend.background = element_rect(colour = "black"),
                 panel.border = element_rect(colour = "black", fill = NA))

Save and view the completed map!