IS-1500 User Guide
Thales Visionix, Inc.
MNL- 0024 (D)
Page 15 of 59
2.2.2
VINS Natural Feature Tracking
To track off of natural features, the IS-1500 uses a computationally lightweight Vision-aided Inertial Navigation System (VINS)
algorithm. When an image from the InertiaCam camera is received by the sfHub software, it is processed to isolate particular pixel
patches that the system considers unique and identifiable. Frequently the patches will be found in regions of sharp contrast, such as
corners or edges. These pixel patches are the natural features. Figure 7 shows the Optical Data display with Image Transfer and VINS
Image enabled. It illustrates the features found by VINS while viewing a collection of objects. Each blue box outlines a pixel patch
feature with a unique ID in green.
Figure 7 – VINS Features
As the camera continues to provide subsequent frames, they are analyzed to relocate the natural features. If they can’t be located or
become compromised, they are replaced with new features. Reasons for this might be that the feature has moved out of the frame
and is no longer visible, or that it is simply unrecognizable. (Note that many of the features in Figure 7 are partially identified by a
glare, such as 13686 on the tree branch of the ceramic plate. If the light causing the reflection were blocked, that feature would likely
become unrecognizable.) The system also has a preference for ‘natural’ (immobile) features. If a feature previously thought to be
motionless is eventually determined to be mobile, it will soon be discarded in favor of a stationary feature.
With VINS, data on each feature is not saved to permanent memory. Data is only maintained for currently tracked features. This
means that if the camera is turned away for several seconds and brought back to the same field of view, the previous features will not
be identified again. Even if the same pixel patch is identified as a feature, it will have a new feature ID and any data describing it will
be generated from scratch.