Process Noise Matrix Calculation - Kalman Filtering - matrix

I am learning Kalman Filtering.
I am facing difficulties on Process Noise Matrix Q. I understand what it is, but I do not know how to model it.
Can anyone help me with a reference or tutorial to understand?
I modeled my Q taking into consideration this tutorial.

Related

Vector order statistics - color image edge detection

I'm trying to implement specific algorithm from vector order statistics, does anyone implemented that before? Im using matlab, but any source of example would be very helpful.
Im trying MVR - minimal vector range, but im stuck in understanding the alghorith and how to make a code from this paper.
Original paper: https://tspace.library.utoronto.ca/bitstream/1807/10051/1/Venetsanopoulos_11349_535.pdf

Inverse gait analysis using processing

I am developing a code to do inverse gait analysis,that is, given the knee and hip joint angles, I am trying to recreate a walking animation on processing. But I am stuck at the maths of it.
So basically I am trying to make a stickman walk based on given lower limb joint angles.
And I for the life of it cant figure out the math.Here is a sample of the data I am working with

Image segmentation with watershed thresholding

I have implemented the marker-less (so not like OpenCV) watershed algorithm proposed in a 1991 paper by Vincent and Soille.
I have also implemented a distance transform algorithm to apply it before watersheding.
It works well in a good number of cases, but sometimes it produces a little oversegmentation. I also corrected some of this with gauss filtering the distance-transform image.
I am planning to correct this by applying thresholding to the watersheds. Therefore considering watershed of only some height greater than a threshold
Considering that this paper is quite old (1991) I am wondering if anyone knows of papers or resources that explain something similar to what I intend to do.
Notes:
1) I am not using OpenCV. I am implementing all by myself from papers
2) I am going for a marker-less watersheding.

How to apply kalman filter on rotation matrix?

I would like to apply a Kalman filter on data I receive from an attitude determination algorithm. I keep getting a rotation matrix as an output.
I am really new to kalman filters. I have read all the formulas and more or less understand them, but still not enough to apply them.
I have a non-linear system.
Could anyone explain how kallman filters can be applied on rotation matrices?
(without too complex and much maths)

Kalman filter for car's tracking path

I am having a set of Points like Point(x,y). After the car gone through so many ways in the same road it is almost messing the resulting map. I heard that Kalman filter can make a sigle path from the available paths.
Can any body say how to make it? I am not from computer science. So please explain me about that concept and those matrices. Then I will code them. Please anybody enlighten me about the concept.
As far as I know is the Kalman filter capable to combine several sources of the same information to get a more precise measurement of the observed variable. It could be possible to combine also with the same measruement device measured multiple times.
Here is a good introduction:INTRO, AnotherOne
I don't know if this question is still active, but if your intressted in learning more about the Kalman filter I can strongly recomend this short matlab script. Even if you dont have matlab installed it should be about the simplest example on Kalman your likely to find!
I don't see how exactly a Kalman filter would be applied here.
I would approach this problem either by image processing, so a thick path would be reduced to a thin line or by successive linear regression on the path segments.
You're probably trying to use the detected coordinates of a car, to determine where a road is, when there is no roadmap information available. Trying to create a road when there is no road, right?
The Kalman Filter is meant to smoothen values obtained from a sensor. When a sensor detects a car, the sensor may not give the car's actual position. It will contain some errors in x and y coordinates.
You have to feed these x,y values to the Kalman filter while the data is being obtained from the sensor. Or at least in the correct order that it was obtained from the sensor.
The Kalman filter will give you the estimated values (smoothened values) of x and y positions, which will tell you approximately the correct position of the car.
Assuming that the car is travelling in the middle of the road, these estimated (filtered) x,y values are what you can take as the midpoints of the road.
I saw your question only now. I know it's late, but I hope that helped?

Resources