Consider a scenario of tracking multiple targets using a single camera sensor, and then a fault occurs in the camera sensor. Do the existing target tracking algorithms guarantee the quality of tracking?
Any degradation in the sensor measurement due to malfunctions in the camera hardware structure is fault. Fault differs from noise or other external sources that affects the measurements. Also fault is different from failure. Failure means the camera doesn't work anymore.
Hello. The question as you have posed it is a bit too loose to answer precisely. It naturally depends on the type of fault or (subsystem) failure. As far as tracking is concerned, there is no general way to guarantee quality of tracking under a fault. By its nature a fault will degrade system performance, for instance, by causing a loss of detectability, which can lead to track loss if it continues for long enough. Your system has no inherent fault tolerance if it is based on a single sensor. You need redundancy to tolerate faults.
Note that I would not consider model mismatch to be a fault, although in some sciences this may be the case. In tracking, model mismatch is often caused by a target manoeuvre, i.e., a significant change in dynamical behaviour. In condition monitoring, this might correspond to a fault. Both cases involve tracking of a dynamical system.
Please have a look at the references below, they cover multitarget tracking algorithms, manoeuvring target tracking and fault detection in navigation systems (for high integrity).
The GPS navigation integrity problem is a bit easier than your problem since you generally have more satellites than you need and you can exclude faulty ones, at least one at a time.
Article Taxonomy of multiple target tracking methods
Technical Report A Survey of Manoeuvring Target Tracking Methods
Conference Paper Optimising the algorithm design for high-integrity relative ...
Hello. The problem of single target tracking in the presence of random sensor faults have been discussed in
"Extended Kalman filtering with stochastic nonlinearities and multiple missing measurements". (link below)
I want to know why such an important problem does not exist for multiple target tracking using camera sensor?
http://www.sciencedirect.com/science/article/pii/S0005109812002555
- Badges
- Science topic
- Similar topics
- Correction