Particle filtering how to implement the target tracking in the algorithm is how to implement the tracking step by step, seeking expert help

I'll explain in layman's terms that Particle Filtering (PF) is applied roughly like this: (In fact, the theory of target tracking is the real-time valuation of state vectors)

Featured with a bunch of samples, assuming there are N of them, and initially give them the same weights 1/N.

This system state-transfer equation, which is generally non-linear, is the only thing that we need to know what we need to do to propagate the state values from this moment to the next. state values to propagate to the next moment. To do this, the N sample values are transferred through the state to get the sample predictions for the next moment, including the process noise factor.d

The system also has a nonlinear observation equation, through which we get the true observation Z. At this point, the N sample predictions are brought in to get Z'.

Based on the degree of difference between Z' and Z, a decision is made on the level of confidence in this sample, the closer the better of course, and then these levels of confidence are weighted and normalized.

Resampling link, these samples according to the weights for random sampling (the higher the weights, of course, the more likely to be sampled. For example, the value of the next moment, there are four samples said to be equal to 1, there are two samples said to be equal to 1.5, then there is a 2/3 probability that it is equal to 1. This is really complicated to explain, generally do up to 200-300 samples to get the value is close to the same, but also to set a 2/3n threshold to prevent the particle scarcity, that is, to prevent all samples from obtaining the same posteriori estimation results), the obtained values as close as possible to what really happened.

Loop 2 to 5