I got this data curve.
Because it's real data it's kind of shaky.
I want to differentiate the curve... this looks pretty ugly because of the shakyness.
So I went on and used ployfit and polyval to smooth the curve. As this doesn't resemble a ploynomal curve, I needed to split this curve in three and smooth them all separately and later fit them together again.
But polyval tend to overdo the edges... (the smooth one in red, the original in blue)
so when I add them together I get non smooth junctions like this one here: (I know this is extreme, but it accurs always)
when I later differentiate the curve, I get huge errors at the junctions:
any ideas to solve my problem?
I do need a clean curve for calculations and so on...
Edit
I implemented the comments:
here the results
spline isn't smooth enough for differentiation,
Savitzky-Golay Filter also not perfect
Related
I have some data that fits closely to an exponential curve on the left hand side and then flattens out on the right hand side. For the upper portion of the the curve, the data should fit closely to a straight line. I am trying to find a way to quantitatively work out exactly which part of my curve is 'most straight'.
I have tried qualitatively choosing the 'straightest' part of the data, which gives a good approximation, however I would like to do it quantitatively. I have also tried playing around in CFTOOL with to no avail.
See screenshot of data
I would like to follow a curve (with matlab or opencv) and to find the other end of it when it is cut by an empty space like this example, which is simplified to illustrate the problem:
Link to image of cut curve
Real images are more like this one: Link to real image to analyse
To follow the curve, I can use a skeleton and look at the neighbourhood. The problem is that I don't know how to find the other end efficiently.
I don't think that closing or opening operations could help because as shown on the previous image, there are other curves and the two parts of the curve are quite far from each other so it could lead to boundaries between the different curves instead of the two parts.
I was thinking about polynomial evaluation which could be a solution for simple curves but I am not sure about the precision I could get. If I use a skeleton, I have to find exactly the right pixel or to search in a reasonable neighbourhood which would take some time and once again, as there are other curves in the images, I have to be sure that I will find the good one.
That's why I am searching for an existing function which could estimate precisely the trajectory of the curve and give an usefull output to go further and find the second part of the curve.
If that kind of function doesn't exist, I'm open to any other way of analysing the problem if it can help.
I will start to explain with the first image you provided, you can implement common OpenCV function useful for detecting contour(black region in your case as you have binary image) known as cv2.findContours(), which returns the coordinates of the edges of the surface detected then you can plot each detected contour separately in a blank image to get the edge of your desired line.
Now coming to your 2nd image you have to be slightly careful while performing above analysis as there are many tiny lines. get back to me for further help
I am trying to detect corners (x/y coordinates) in 2D scatter vectors of data.
The data is from a laser rangefinder and our current platform uses Matlab (though standalone programs/libs are an option, but the Nav/Control code is on Matlab so it must have an interface).
Corner detection is part of a SLAM algorithm and the corners will serve as the landmarks.
I am also looking to achieve something close to 100Hz in terms of speed if possible (I know its Matlab, but my data set is pretty small.)
Sample Data:
[Blue is the raw data, red is what I need to detect. (This view is effectively top down.)]
[Actual vector data from above shots]
Thus far I've tried many different approaches, some more successful than others.
I've never formally studied machine vision of any kind.
My first approach was a homebrew least squares line fitter, that would split lines in half resurivly until they met some r^2 value and then try to merge ones with similar slope/intercepts. It would then calculate the intersections of these lines. It wasn't very good, but did work around 70% of the time with decent accuracy, though it had some bad issues with missing certain features completely.
My current approach uses the clusterdata function to segment my data based on mahalanobis distance, and then does basically the same thing (least squares line fitting / merging). It works ok, but I'm assuming there are better methods.
[Source Code to Current Method] [cnrs, dat, ~, ~] = CornerDetect(data, 4, 1) using the above data will produce the locations I am getting.
I do not need to write this from scratch, it just seemed like most of the higher-class methods are meant for 2D images or 3D point clouds, not 2D scatter data. I've read a lot about Hough transforms and all sorts of data clustering methods (k-Means etc). I also tried a few canned line detectors without much success. I tried to play around with Line Segment Detector but it needs a greyscale image as an input and I figured it would be prohibitivly slow to convert my vector into a full 2D image to feed it into something like LSD.
Any help is greatly appreciated!
I'd approach it as a problem of finding extrema of curvature that are stable at multiple scales - and the split-and-merge method you have tried with lines hints at that.
You could use harris corner detector for detecting corners.
I have a list of 200 points I garnered from a graph digitization software I would like to transform into a smooth curve and then into Solidworks.
My points form an ellipse (airfoil shape to be more precise), so the commands I've tried in Matlab didn't have a circular curve.
My issues are:
* Obtaining a smooth curve that doesn't necessarily pass through all points, smooth being motus operandi.
* Being able to have a elliptical curve
* Somehow being able to export this curve into Solidwords
If anyone knows the right software, command line or anything that could get me started, I would be extremely thankful.
imacube
I've used Solid Works before. It's a very powerful tool. There should be some way to draw a curved spline through these points, such as a cubic spline.
If you are using a standard(ish) airfoil, then you can use a variety of tools to plot the points without having to use a graph digitization software.
Javafoil, for instance, is one of those. Even if you know the characteristics of your airfoil, you can use this to give you a smooth set of points.
Again, if your airfoil is a naca 4-series, then these are governed by a set of equations.
But I take it that the airfoil you want a more complicated one. Let me know if I can help anymore.
like what i want is if i move my finger fast on the iphone screen , then i want like something that it make a proper curve using quartz 2d or opengl es whatever.
i want to draw a path in curve style......
i had seen that GLPaint(OpenglES) example ,but it will not help me alot , considering if your finger movement is fast.....
something like making a smooth curve.....
any one have some kind of example please tell me
thanks
Edit: Moved from answer below:
thanks to all.......
but i had tried the bezier curve algo with two control points but problem is first how to calculate the control points whether there is no predefined points....
as i mentioned my movement of finger is fast...... so most of the time i got straight line instead of curve, due to getting less number of touch points.......
now as mark said piecewise fashion, ihad tried it like considering first four touch points and render them on screen , then remove the first point then again go for next four points ex. step 1: 1,2,3,4 step 2: 2,3,4,5 like that where as in that approach i got an overlap , which is not the issue actually , but didn't get smooth curve........
but for fast movement of finger i have to find something else?????
Depending on the number of sample points you are looking at, there are two approaches that I would recommend:
Simple Interpolation
You can simply sample the finger location at set intervals and then interpolate the sample points using something like a Catmull-Rom spline. This is easier than it sounds since you can easily convert a Catmull-Rom spline into a series of cubic Bezier curves.
Here's how. Say you have four consecutive sample points P0, P1, P2 and P3, the cubic Bezier curve that connects P1 to P2 is defined by the following control points:
B0 = P1
B1 = P1 + (P2 - P0)/6
B3 = P2 + (P1 - P3)/6
B4 = P2
This should work well as long as your sample points aren't too dense and it's super easy. The only problem might be at the beginning and end of your samples since the first and last sample point aren't interpolated in an open curve. One common work-around is to double-up your first and last sample point so that you have enough points for the curve to pass through each of the original samples.
To get an idea of how Catmull-Rom curves look, you can try out this Java applet demonstrating Catmull-Rom splines.
Fit a curve to your samples
A more advance (and more difficult) approach would be to do a Least Squares approximation to your sample points. If you want try this, the procedure looks something like the following:
Collect sample points
Define a NURBS curve (including its knot vector)
Set up a system of linear equations for the samples & curve
Solve the system in the Least Squares sense
Assuming you can pick a reasonable NURBS knot vector, this will give you a NURBS curve that closely approximates your sample points, minimizing the squared distance between the samples and your curve. The NURBS curve can even be decomposed into a series of Bezier curves if needed.
If you decide to explore this approach, then the book "Curves and Surfaces for CAGD" by Gerald Farin, or a similar reference, would be very helpful. In the 5th edition of Farin's book, section 9.2 deals specifically with this problem. Section 7.8 shows how to do this with a Bezier curve, but you'd probably need a high-degree curve to get a good fit.
Naaff gives a great overview of the NURBS technique. Unfortunately, I think generating a smooth bezier on-the-fly might be too much for the iPhone. I write drawing apps, and getting a large number of touchesMoved events per second is quite a challenge to begin with. You really need to optimize your drawing code just to get good performance while recording individual points - much less constructing a bezier path.
If you end up going with a bezier or NURBS curve representation - you'll probably have to wait until the user has finished touching the screen to compute the smoothed path. Doing the math continuously as the user moves their finger (and then redrawing the entire recomputed path using Quartz) is not going to give you a high enough data collection rate to do anything useful...
Good luck!
Do something like Shadow suggested. Get the position of the touch with some frequency and then make a Bézier curve out of it. This is how paths are drawn with a mouse (or tablet) in programs like Illustrator.