IMU Overview

Accelerometer #

The accelerometer detects the instantaneous acceleration of the camera. The data provided by the accelerometer determines whether the camera is getting faster or slower, in any direction, with a precise value in meters per second squared (m/s²).

When an accelerometer is static, it is still measuring an acceleration of 9.8m/s² which corresponds to the force applied by the Earth’s gravity. This force has always the same direction, from the camera to the center of the earth. The gravity allows us to compute the camera’s absolute inclination and detect events like free falls.

Gyroscope #

A gyroscope measures the angular velocity of the camera in degrees per second (deg/s). When combined with the accelerometer, both sensors can estimate the orientation of the camera at a high frequency.

Output Data #

The following information is accessible from the camera sensor stack:

Output Data	Description	Units
Accelerometer
linear_acceleration	Acceleration force applied on all three physical axes (x, y, and z), including the force of gravity. Values are corrected from bias, scale and misalignment.	m/s²
linear_acceleration_uncalibrated	Acceleration force applied on all three physical axes (x, y, and z), including the force of gravity. Values are uncalibrated.	m/s²
linear_acceleration_covariance	Measurement noise of the uncalibrated linear acceleration of the accelerometer. Provided as a 3x3 covariance matrix.
Gyroscope
angular_velocity	Rate of rotation around each of the three physical axes (x, y, and z). Values are corrected from bias, scale and misalignment.	deg/s
angular_velocity_uncalibrated	Rate of rotation around each of the three physical axes (x, y, and z). Values are uncalibrated.	deg/s
angular_velocity_covariance	Measurement noise of the uncalibrated angular velocity of the gyroscope. Provided as a 3x3 covariance matrix.
Orientation
pose_covariance	Measurement noise of the pose orientation. Provided as a 3x3 covariance matrix.
camera_imu_transform	Transform between IMU and Left Camera frames.

Using the API #

Accessing the IMU data can be done through the SensorsData class. Data is stored in the class SensorsData::IMUData which amongst others contains the accelerometer and gyroscope values.

You can retrieve IMU data with the following code:

C++
Python
C#

SensorsData sensors_data;
SensorsData::IMUData imu_data;

// Grab new frames and retrieve sensors data
while (zed.grab() == ERROR_CODE::SUCCESS) {
  zed.getSensorsData(sensors_data, TIME_REFERENCE::IMAGE); // Retrieve only frame synchronized data

  // Extract IMU data
  imu_data = sensors_data.imu;

  // Retrieve linear acceleration and angular velocity
  float3 linear_acceleration = imu_data.linear_acceleration;
  float3 angular_velocity = imu_data.angular_velocity;
}

sensors_data = sl.SensorsData()

# Grab new frames and retrieve sensors data
while zed.grab() == sl.ERROR_CODE.SUCCESS :
  zed.get_sensors_data(sensors_data, sl.TIME_REFERENCE.IMAGE) # Retrieve only frame synchronized data

  # Extract IMU data
  imu_data = sensors_data.get_imu_data()

  # Retrieve linear acceleration and angular velocity
  linear_acceleration = imu_data.get_linear_acceleration()
  angular_velocity = imu_data.get_angular_velocity()

SensorsData sensors_data = new SensorsData();
IMUData imu_data = new IMUData();

RuntimeParameters runtimeParameters = new RuntimeParameters();
// Grab new frames and retrieve sensors data
while (zed.Grab(ref runtimeParameters) == ERROR_CODE.SUCCESS) {
  zed.GetSensorsData(ref sensors_data, TIME_REFERENCE.IMAGE); // Retrieve only frame synchronized data

  // Extract IMU data
  imu_data = sensors_data.imu;

  // Retrieve linear acceleration and angular velocity
  float3 linear_acceleration = imu_data.linearAcceleration;
  float3 angular_velocity = imu_data.angularVelocity;
}

Pose #

One of the key reasons for having an accelerometer and gyroscope in the camera is that their data can be fused to estimate camera orientation. The accelerometer provides gravity orientation, while the gyroscope estimates the rotation applied to the camera. Fused at high frequency, the combination of both sensors provides a robust orientation estimation.

IMU pose can be retrieved in imu_data.pose.

Code Example #

For a code example, check out the Getting Sensor Data tutorial.