Introduction

Handheld and body-worn devices, like the PERCEPT Body Camera, are designed to be lightweight and easy to carry around and operate. It is technically challenging to obtain sharp video without motion blur when a camera is moving around instead of being fixed on a tripod or other mount. 

Motion-based image stabilization

To overcome these difficulties and provide a better video experience, we’ve equipped the PERCEPT Body Camera with an advanced motion-based image stabilization solution. At the heart of the system is the 9-axis motion sensor, embedded within the electronics of the body camera. This sensor combines an accelerometer, a gyroscope and a magnetometer, each of them monitoring a specific type of movement on all 3 axes (hence the term “9-axis”). The accelerometer detects the change of speed (acceleration) within a linear motion and gravitational forces, the gyroscope measures the rate of rotation in space (roll/pitch/yaw), while the magnetometer measures the terrestrial earth’s magnetic fields. The body camera’s intelligent firmware fuses raw data from these sensors and creates an absolute orientation vector. This is added to each frame of the captured video via a patented process where QR code markings are embedded into each frame.

Advantages compared to other stabilization solutions

During playback, the operator can select zones and regions of interest (ROI) in the video feed coming from the body camera to create “virtual cameras”. Each “virtual camera” is then individually stabilized. The way this works is very simple: when body camera movement is detected from the information in the QR codes, the portion of the image sensor used to record the image in that ROI shifts in the opposite direction to compensate for the motion. This is feasible in a PERCEPT Body Camera since the Panomorph lens captures the full hemispheric field-of-view in front of it, so all the visual information required to shift the image from the “virtual camera” is already there. Basically, the field-of-view of each “virtual camera” is not affected by the stabilization process, as opposed to electronic stabilization employed by a regular camera, where some cropping of the captured image is required to get the desired effect (typically you lose about 10% of the captured image in this case).

Another major advantage of using this type of digital image stabilization is the fact that calibration can be conveniently performed and is not as prone to de-calibration as traditional optical image stabilization (OIS) mechanisms. The client software performing the image stabilization can be easily maintained and improved, with features added via updates, without replacing any hardware on the body camera itself. On the other side, an embedded OIS system needs to be tailored for each camera module with a specific lens. This tailoring requires very sensitive calibration, which can be easily skewed if the device is dropped or otherwise receives some sort of impact. The PERCEPT Body Camera was developed with robustness and durability in mind, so the choice of going with a digital stabilization instead of an OIS system was the optimal one from the very beginning. And, since the stabilization is done on the client side, it will not have any impact on the autonomy of the body camera, while an always-on OIS mechanism could potentially drain the battery very fast if used extensively.

Conclusion

The benefits of stabilization are obvious when the person wearing the PERCEPT Body Camera is, for instance, riding a bike/motorcycle, chasing on foot another person or simply walking up or down some stairs. It is much easier to focus on areas of interest to understand scene details (behavior of a suspect, movement of objects in the scene, a vehicle passing by, etc.) on one or more stabilized zones of the video and this can sometimes mean the difference between taking a good or a bad decision during a real-time event or critical