Using Satellite Imagery For Good: Detecting Communities In Desert And Mapping Vaccination Activities

Government of Punjab has been revamping vaccination program to increase its geographical coverage and to better manage vaccination staff spread across whole Punjab. To analyze the coverage of the vaccination activities especially in the areas far away from the urban centers of the Punjab, we need information about locations and size of communities in those regions. Since this information is not readily available, we have employed deep learning to detect house-like-structures on the freely available low resolution satellite imagery (visible spectrum) of the Cholistan Desert.

Introduction

Our algorithm detects areas that are clustered together (to represent housing-clusters) and polygon fitting is performed so that we can create manageable representations (you can visualize our results on the interactive map below and technical details are discussed in the last section). Finally we map vaccination and public school location data over these detected regions. For vaccination data we calculate both count of vaccination activity per quarter per region, as well as coverage to understand how geographically diverse activities were performed in those region. Our analysis shows that although not all the detected regions appear to be covered, large number of them are. Some get vaccination activity in all the quarters, others are missed in one or two of the quarters. Large number of detected regions had at-least one public school as per government record. Certain detected regions were found to be getting very few or none of vaccination activity, their location and cluster indicates to us that there could be anomaly in the data, we hope to verify and correct that in our future version. We are in process of conducting such survey for whole Punjab, on much higher resolution satellite imagery.

 

View detected segments on the interactive map

In today's era, geo-located maps fuel a considerable portion of the digital economy. From ride sharing services like Uber to finding restaurants in your neighborhood, we are dependent upon comprehensive, up to date and accurate maps. Aside from these commercial examples, accurate maps play a vital role in running many functions of the government. They are used for city planning, monitoring development and also in organizing rescue operations in times of disasters. The sheer usefulness and economic importance of these maps have pushed both government and non-government bodies to invest in creating, managing and updating these maps. Ideas like using mobile-phone-activity to geolocate roads or directly crowdsourcing (Open Street Map, Wikimapia, etc..) the map information have been used to build and update maps. In the developing countries, government maintained maps are mostly not geo-located or are not readily accessible in the digitized format. This is true especially for remote regions of the developing countries, where mostly crowd-sourced information fails to be helpful. Satellite imagery provides opportunity to extract much of the information like urban and non-urban population, roads, crop yield without placing humans on ground. However, using humans to tag information by eyeballing the satellite imagery is a painstakingly difficult task.

Correlating vaccination activity with detected communities

Vaccination Activity (August 2015 - October 2015)

Vaccination Activity (November 2015 - January 2016)

Vaccination Activity (February 2016 - April 2016)

Vaccination Activity (May 2016 - July 2016)

Vaccination Activity (August 2016 - October 2016)

Vaccination Activity (November 2016 - December 2016)

Pakistan has been aggressively running vaccination programs to decrease child mortality rate and eradicate diseases like polio. Government trained vaccinators travel to different parts of cities and far away villages to inoculate children in cities and far away villages. Punjab Information Technology Board (PITB) has distributed smartphones to these officials and a phone based application records information about the inoculation activity. Using this vaccination activity data, we analyze how often and in what ratio vaccinators visit different identified regions. We have only used data that is starting from August 2015 till December 2016, and have divided them into the interval of 3 months. Number of vaccinators who accessed a detected cluster is computed and a histogram is plotted to show the distribution of assigned vaccinators in the Cholistan region. Further a buffer of 200 meters is created around each location tagged by the vaccinator and then intersection of these buffers is computed resulting in an approximate movement of vaccinators. A percentage coverage is computed using movement of vaccinators in fixed area of detected segments.

As per Government's vaccination programme each vaccinator maintains a record that contains age of children with there pending vaccines. Following this record vaccinators visit their assigned vicinities to complete a vaccine's round according to its specificity. During each activity a histogram is plotted showing number of vaccinators who accessed a detected community. These numbers justify the fact that regions with larger areas were accessed the most.

Histogram of vaccinators per segment.

The movement of vaccinators covering an area in each community is proportioned over the area of detected region. These coverage percentages are graphed giving us segments that are covered most. A pattern of coverage can be seen based on the movement of vaccinators over the span.

Histogram of area covered per detected segment during an activity(three month interval).

Mapping public schools data on detected segments

Just like vaccination activity information, government schools data was correlated with the detected regions to understand accessibility to education. It was discovered that a large number of the detected regions did have the government school in them. Red areas show polygons with no schools, where as the shades of green represent increasing number of schools with light green being the least(one or two) and dark green with most(more than 10) schools.

Strength of public schools mapped on detected segments.

Technical Details

We have used village finders satellite imagery data in order to train our network. Image tiles are divided in small chips of size 128 and further resized to 64. Data augmentation methods were used before and during training. The probability map achieved after last convolution layer is bilinearly interpolated to achieve input size. A threshold of 0.5 is applied to generate a binary image. Polygons are fitted on these binaries resulting in detected communities. The deep learning program is written in python using Chainer framework.
A shortcoming is that our algorithm cannot predict height or even type of the building. Therefore some of the building detected are themselves private schools run. Furthermore some built structures are missed that is because they seem to have merged into the landscape when viewed at certain zoom level.

Future Work

Satellite imagery offers a vast research area. To take full advantage of the diversity these images contain, we plan to study the variety of bands that hyperspectral imagery presents. Detecting built structures successfully has inaugurated problems like house counting, tracking urbanization and detecting damaged structures. We intend to deeply analyze the detected segments in order to locate hospitals, schools and parks near each community.

We are grateful to Punjab Information Technology Board (PITB) especially Ms. Maria Badar for sharing vaccination activity data as well as public schools data. We would also like to thank Muhammad Umer Ramzan, ITU's management and their website creators for helping us build this page.