TIDUD31B May 2017 – September 2019
The amount of time needed by ARM R4F to process input point cloud and deliver target information is a function of number of targets currently tracked, and number of measurements (points in input point Cloud) received. Figure 13 shows the results collected (blue) and a bound defined (red).
We observed that processing time increases linearly with number of tracking objects. With fixed number of tracking objects, complexity increases linearly with number of input points. Taking the worst case of number of points per frame of 250, we can derive a bound of about ≈ 200µs per tracking object. With frame time of 50ms, tracking 20 targets shall be completed in 4ms, which will consume <10% of CPU.