Capturing Sensors
One of the most used sensors for capturing a 3D scene is Kinect Azure. This device allows your camera system to capture detailed scenes in front of you using ToF technology. The device has 2 cameras with which it evaluates the whole scene and sends the obtained 3D images to the computer via USB interface. The output of this device can be imagined as a 3D video that carries information not only in the X and Y axis but also in the Z axis.
Hardware specifications
The Azure Kinect camera consists of an RGB camera and an infrared camera. The RGB camera offers various resolution modes where the highest possible resolution is 3840 × 2160px at 30 Hz.
The infrared camera has the highest resolution of 1024 × 1024px and uses the ToF principle. In addition, both cameras support different field of view modes.
Azure Kinect also has an IMU sensor, consisting of a triaxial accelerometer and a gyroscope, with which the device can estimate its own position in space. Microsoft also offers a Body Tracking SDK with the device that can detect and track the movements of multiple users, each with 32 joints.
This development package is available for Windows and Linux operating systems and the C ++ and C # programming languages.
Contrary to the skeleton definition of the former generation of Kinect, the current definition includes multiple joints in the face, such as the ears and eyes.
Body tracking
Azure Kinect supports capturing multiple people at once in real time. Each person is identified by their own temporary identifier, which allows the classification to re-recognize the same person even after leaving the scene and returning. The skeleton is formed by points in space that are connected. Their position and orientation are estimated in millimetres and expressed as a normalized quaternion.
The scanned skeleton consists of 32 individual connected joints that spring from the centre of the body to the limbs. Each connection between these points can be called a bone and represents the connection between the parent point and the offspring point.
The following image shows all available joints and bones of the complete skeleton represented by Azure Kinect.
