Solar Transit of Venus Video 2012
Solar Transit of Venus, 2012
I'm finally getting my shots of the transit of Venus together as a video. I had a bit of trouble with the set of photos. I wasn't as prepared as I would have liked to have been. The day of the transit started off cloudy, so I didn't bother getting ready, and then the clouds parted. I ran out with the ED80 telescope and a simple alt-az mount, and started snapping photos. It didn't take me long to realize that I was going to have problems tracking the entire transit with the alt-az mount, so I ran back inside and grabbed the EQ5 polar mount. After a quick (and very rough) polar alignment, I swapped mounts. That's what caused the field rotation you see about half way through the video. Luckily, the clouds mostly stayed away during the transit. The video cuts off before the end of the transit; I could have shot a bit more, but I setup too close to the house, and didn't think about the chimney getting in the way (that's the chimney that cuts off the final shot).
Once I had the sequence of photos, I didn't at first think they would make much of a video. Since I started with a non-motorized alt-az mount (and tracked by hand), and then switched to a poorly-polar-aligned polar mount, the solar disk moved quite a bit from shot to shot. My first approach was to try to align the frames in Photoshop, but I gave up after I realized that was going to take forever. My second stab at making a video started with an attempt to align the shots by cross-correlation, which I quickly coded up in Matlab. But, since the entire disk is not visible in each shot, the cross-correlation did a very poor job of aligning the shots. I tried edge-finding with cross-correlation, but that wasn't much better. I then thought about using a feature-locator algorithm (like SIFT) to find points to align the images, but didn't pursue that approach since the few features to track weren't visible in all frames. Finally, I hit on the idea of using a Hough circle finding algorithm. That worked very well, but was horribly slow (my Matlab implementation was a little rough). After letting the Hough alignment run over night, I had the stack of aligned frames that I assembled into this video.
Solar Transit of Venus Video 2012
Solar Transit of Venus, 2012
I'm finally getting my shots of the transit of Venus together as a video. I had a bit of trouble with the set of photos. I wasn't as prepared as I would have liked to have been. The day of the transit started off cloudy, so I didn't bother getting ready, and then the clouds parted. I ran out with the ED80 telescope and a simple alt-az mount, and started snapping photos. It didn't take me long to realize that I was going to have problems tracking the entire transit with the alt-az mount, so I ran back inside and grabbed the EQ5 polar mount. After a quick (and very rough) polar alignment, I swapped mounts. That's what caused the field rotation you see about half way through the video. Luckily, the clouds mostly stayed away during the transit. The video cuts off before the end of the transit; I could have shot a bit more, but I setup too close to the house, and didn't think about the chimney getting in the way (that's the chimney that cuts off the final shot).
Once I had the sequence of photos, I didn't at first think they would make much of a video. Since I started with a non-motorized alt-az mount (and tracked by hand), and then switched to a poorly-polar-aligned polar mount, the solar disk moved quite a bit from shot to shot. My first approach was to try to align the frames in Photoshop, but I gave up after I realized that was going to take forever. My second stab at making a video started with an attempt to align the shots by cross-correlation, which I quickly coded up in Matlab. But, since the entire disk is not visible in each shot, the cross-correlation did a very poor job of aligning the shots. I tried edge-finding with cross-correlation, but that wasn't much better. I then thought about using a feature-locator algorithm (like SIFT) to find points to align the images, but didn't pursue that approach since the few features to track weren't visible in all frames. Finally, I hit on the idea of using a Hough circle finding algorithm. That worked very well, but was horribly slow (my Matlab implementation was a little rough). After letting the Hough alignment run over night, I had the stack of aligned frames that I assembled into this video.