Estimating Geometric Location (time/distance)

[Davidezrajay]

Problem Each tree photo is tagged with a location. However, the Lat and Long based geo locations have anomalies on a number of different phones. In a tracking session, it is common for a phone to have a number of accurate points and one or two points that are tagged with inaccurate locations.

Example Problem

Tree id 3491297 was given a random location by the phone's GPS, and has subsequently been manually flagged for being located in the ocean.

The user tracks 4 trees. Tree #4, Time Stamp: 14:45:03, POINT (-13.275777666666666 8.484414333333332) Tree #3, Time Stamp: 14:41:32, POINT (-13.275915666666666 8.484348333333333) Tree #2, Time Stamp: 14:40:03, POINT (-13.250369000000005 8.014978000000003) Tree #1, Time Stamp: 16:39:00, POINT (-13.277494999999998 8.482369)

Possible Solutions 1

1. Review location data and time stamped data.
2. Use time and distance calculations between points to identify anomalies.
3. Flag the points that are in the incorrect location.

For example: Distance between #2, #3 and #4 is a distance of about 27km covered in 63 seconds and 89 seconds. The distance that someone can walk in Meters is: 1.788 meters per second, hence it would take about 5.9 hours to walk between the two.

The estimated geometric location is probably about half way between point #1 and #3.

Possible Solution 2 Compare locations with track files

Supporting data:

• Time stamp
• Geo Location
• Track File

Deliverable: Scrubbed geo location data

1. Flag any data point that has traveled faster than a fast walk in a "tracking session"
2. Flag any specific devices that this occurs on.
3. Create a airflow script that runs periodically to population the Estimated Geometric Location to correct the flagged data point based on proximity to sequential data points.

Required Resources:

1. Export of tree capture geo location data
2. Database schema