Which sensors to use for Volumetric capture? Azure Kinect or Kinect v2.
The thing all computer vision and volumetric capture specialists were waiting for in 2019 was the new Azure Kinect camera from Microsoft. As it’s previous versions Kinect Xbox One and Kinect 360 it will be used by to obtain depth data including real-time volumetric capture.
Why everyone is excited about Azure Kinect?
The new Azure Kinect comes with 12MP CMOS rolling shutter sensor. This gives a lot of freedom to build versatile applications including volumetric capture rigs. The new sensor gives a variety of frame rates and resolutions from internal RGB and depth sensors. Its synthetic RGBD resolution of 4096×3072 generates an astonishing 12 million points every 66.6 milliseconds. When you compare Azure to Kinect Xbox One which gives only 2 million points per frame Azure seams like a sensors from the future. With this you can actually capture a very good quality volumetric video. One other function which comes handy is hardware synchronisation via 3.5 mm jacks. It allows multiple Azure Kinect camera frames to be synced. You can check the full specs here.
Portable volumetric capture
Kinect Xbox One was made for body tracking experiences while gaming. Therefore it was supposed to be mounted in one place. It weighted 960 grams and was 249 x 66 x 67 mm size. Also, it had to be connected to a Windows PC via two cables. The new Azure Kinect is half the size, weight and it comes with a USB-C type cable. However, the cable is only 1.5m long and we quicly noticed that it is not enough for volumetric capture set up. We recommend using NEWNEX cables for extending and having a larger capture area. Have a look at full hardware specs for portable set up here.
Volumetric capture need software… and hardware.
Kinect Azure DK hardware and Developers Kit are not compatible with Kinect v2. The new Kinect camera comes with different Kinect Azure Viewer app. However, it mainly allows you to inspect depth and RGB but that is it. Apart of This new viewer there are close to non plug and play software for Kinect cameras. You can find some open source code for Kinect v2 and try to put it all together but it does not solve the main problem of cleaning the raw point cloud in real time and calibrating all sensors together. What you will end up with is a noisy one sided capture. EF EVE Volumetric Capture software is the number one solution which allows real time data cleaning and automatic calibration of up to 8 sensors. It is one stop solution to start using multiple Kinect Azure and Kinect v2.
Azure Audio capabilities
Kinect Azure comes with astonishing 7-mic circular array and extra layer of functionality. While the old Kinect v2 only had 4 mics.
- Sensitivity: -22 dBFS (94 dB SPL, 1 kHz)
- Signal to noise ratio > 65 dB
Kinect has shifted from just a gaming peripheral to a stand alone 3D camera. Kinect now has 6 times better resolution that Kinect V2. It is twice the size and has much better audio. It is considered the best off the shelf sensor for volumetric video.