K.J. Palmer
K.J. Palmer

Reputation: 145

Android IMU accuracy

I'm working on a project with the goal of tracks animals from a hill. In the first field season we used using OpenTags and a calibrated camera, which worked but required extensive processing/calibration and-crucially-did not provide any visual feedback when the system went awry.

Since one of the goals of the project is to make it easy for other biologists to repeat,I'm investigating using the gyro, magnetometer, and accelerometer already in smartphones. Then, using an app or some simple post processing transfer this to roll, pitch, and yaw.

My question is regarding the accuracy of the smartphone vs an OpenTag. In order to track animals at distances greater than 200m, the system must have sub-degree accuracy and, more importantly, low noise. I've been looking at the Sensor Kinetics app but I find that the pitch/yaw/ heading values tend to drift more than is acceptable. I'm not sure if this is a filtering issue or an issue with the sensors themselves.

Does anybody know where I might find this type of data and/or recommend phones and apps that easily provide these data with little additional processing?

Thanks!

Upvotes: 3

Views: 2987

Answers (1)

Sevron
Sevron

Reputation: 64

All systems will drift, and sub-degree is considered very accurate today. If you are trying to track the locations of these animals only with inertial sensors (i.e. without GPS or some other type of geo tag), the accel and gyro sensors in modern cellphones will be dominated by noise after several minutes or 10s of minutes if you're lucky. You likely will need to somehow recalibrate every now and again, with some kind of beacons for triangulation. Most inertial tracking applications today use a combination of step-counting, along with some understanding of the moving body's gait length, in order to track distance traveled, along with WiFi or Bluetooth beacons. Earth's magnetic field can be used for immediate direction orientation. Companies like http://www.tendegrees.net are trying to do this type of thing for indoor navigation and pedestrian dead reckoning.

Upvotes: 4

Related Questions