MARCO Processing stages

Each image below represents the same input image data after various layers of processing.  I'd recommend clicking on an image to bring up the gallery view, since it's much easier to follow the continuity of the data that way.


Raw greyscale image input, the image is rotated because of the way the docking camera is oriented on my test rig.

Transition filtering: Each pixel is fed through an iterative algorithm which picks out sharp transitions in the image.  The high contrast of the black and white target make this a pivotal first step for culling away irrelevant image data.
Grouping: Based on the results of the previous layer, the raw data rows are grouped as regions, each of which has a series of parameters such as average color, size, endpoints, and variance which are considered in the next layers.

Scoring: Each region is fed through an algorithm specially designed to determine how likely it is that a particular region is the center of the target.   There is quite a bit of noise in this stage of the process, which is fine.  It's difficult to tell, but the pixels in the center of the target are actually the brightest white, as they should be.

Sliding-window Optimization:  A range window of definite width is superimposed over the rows of the scored region data.  The window slides from the top row down until it reaches the bottom of the image.  Particular start & end points of the window which yield the highest score are recorded and the median of the highest-scoring region centroids within the window is considered the x-coordinate of the target centroid.

An additional filter reprocesses the region data to find white regions with little variance.

The boundaries of high-scoring white regions close to the determined centroid are examined.  If 3/5 of the regions surrounding the centroid are correctly bounded to the edges of the target, all of the points passed to the next stage are valid.
Right now I'm working on the elliptical least-squares fitting function which determines the distance and attitude of the target, so I'll post further results as I work my way through that process.

Comments

Post a Comment

Popular posts from this blog

The Future of MARCO

So I took some time, got a real job at NASA and I've been out traveling a bit, and I only just recently had some time to settle down and take a good look at this project from a more seasoned perspective.  For reference, when I set out to make this happen, I didn't know python, I had little experience with linux, and "computer vision" was little more than a buzzword to me.  I knew the math worked out, however I had a hard time translating the principles of MARCO from the realm of math where everything is clean and orderly to the real world where colors sometimes don't look like they should and the sun shines more brightly on some days than on others. I was not well-prepared to take on this project, and it's a good thing I started as early as I did because there's no way in hell I would have been able to achieve the fairly-positive results I got with less time. Be that as it may, I'm now in an incrementally better position to tackle this task, and I ...
Keep reading

Preliminary Results and higher-order error sources

Image
I have some test results today, and I'm quite happy with how they turned out.  Quick refresher of the parameters: range refers to the target distance in centimeters, a_ang is the rotation of the major axis of the projected ellipse, and b_ang is a measurement of the target's rotation about the major axis, both in radians. The convention when dealing with rotated ellipses is to measure the angle of rotation of the major axis from +j (beh) clockwise (BEH!) but also, since I went through the trouble of rotating these images so that they're not sideways, a_ang is measured clockwise from -i.  The elliptical fit is programmed to run along the inside edge of the larger target circle. {'range': 42.27010789888579, 'a_ang': -0.04591864992309163, 'b_ang': 0.3644755561635681} Standard up front shot, I'm impressed that the software picked up that the target is actually tilted upwards relative to the camera though I think .36 radians (20 ish degree...
Keep reading

A Working Filter Actuator

It may have taken a week and an unwieldy hardware solution, but I now have an operational filter actuator.  hooray The transistors I tried just simply could not supply enough current in series to the solenoid when biased by the 3.3V digital output of a GPIO pin, so I threw together a servo-operated switch that flicks the filters back and forth pretty reliably.  I've been testing the switch and actuator by setting the scripts that operate them on an infinite loop (see video) and after a couple of solder joint breaks and design revisions the assembly finally seems to be straightened out. In the meantime, I have continued working on the startup scripts. Of course I have scripts operating both lights and the actuator, and I am integrating them all into a sort of pre-mission check to verify functionality.  I have managed to take images using both the IR and visible light filters, and the pictures look good.  All that remains before the real hard part of this project i...
Keep reading