The idea of using a video camera for observing the night sky and searching for meteors was born on IMC' 92
in Smolenice. There two Canadian astronomers presented a poster about their video equipment and how they
manage to find and analyse meteors on their video tapes. One of the most significant facts was, that they
reach dreamlike limiting magnitudes with their system. On the other hand they could observe only a small area
of the sky.
So four of as, a team of young meteor observers at Archenhold Observatory Berlin, decided to reproduce a
similar system for wide angle sky observations with the already existing equipment at our observatory.
The very first use of MOVIE (which is the name of our system) was on August 12, 1992. Although the sky was full of clouds and the full moon illuminated the whole scenery, we were able to catch our first dozen of meteors. We learnt, that the power of the video equipment is comparable with a normal visual observer. Because of such successful results we improved the hardware, started to work on a computer based video analysis and used the equipment on some other occasions. In November '92 we recorded another 20 meteors, in January MOVIE survived (like all the visual observers) 15 degrees Celsius below zero and catched more than 60 meteors during the Quadrantids, and in August '93 the video equipment was during the maximum of the Perseids 15 hours in operation. Therefore we estimate the number of recorded meteors to be bigger than 500.
Now, one year after the first light for MOVIE, I want to give a report about video observation of meteors and our experiences with the used technology. It starts with a description of the hardware. Later the problem of searching and analysing meteors is discussed and at the end some unsolved problems and alternatives are mentioned.
Video observation has a lot of advantages over visual or photographic observations:
As the name already implies, you need a video camera for recording the sky. The only requirements are, that
the camera should be as sensitive as possible and that it has to record the time automatically. A simple Camcorder
already fulfil the task excellent.
The general part of the system should be an image intensifier. So far as we know there doesn't exist a video camera, that has enough power to record the sky with a satisfying limiting magnitude alone. It should be a three level multiplier with an output screen diameter of more than 2 centimetres and as little electron noise as possible. To get such a multiplier is a real problem and much more difficult then getting the whole rest of the equipment. The very best solution would be an image intensifier of the newer generation, a so called channel plate. Unfortunately these multipliers are much to expensive for amateur astronomers and therefore we have to look for other opportunities. In our case a three level image intensifier was already available at Archenhold Observatory Berlin. An alternative should be old equipment from the army, which is sometimes available at a good price and quality.
To complete the system a good lens for the multiplier is necessary. Here every photo lens with a good focal ratio and a short focal length is useful.
Figure 1 shows the general structure of our MOVIE. The central component is the mentioned image intensifier without any own optics. To record a possible big part of the sky, the lens at the front side is generally a 2,8/20mm wide angle photo lens, that projects the sky on the intensifiers input screen. On the other end a Camcorder films the output screen in macro mode. All three parts are mounted on a special holding device with a base for an equatorial mounting.
With MOVIE we are able to record stars, that are fainter than 6th magnitude. The field of view has an diameter of about 60 degrees and the time is recorded with an accuracy of about 0,1s.
At the end of an observation night you have a videotape with some hours night sky and a few meteors on it.
Now you are confronted with the problem of searching for the meteors on the tape. If you made parallel visual
observations, you already know the times of most of the meteors. If not, there are only two ways out:
You watch the tapes by yourself or you let a computer do the work for you.
The first method is very simple but you have a lot of work with it, because it takes you at least as much time as the length of the video. Nevertheless you will have some advantage in comparison with an visual observation. You can sit in your warm flat at a good time, you have the replay function of your video recorder if you are not quite sure of an event, and at the end you get very accurate data from every visible meteor.
The real advantage is of course the automatic search with the help of a computer. We already work several month in solving this problem, but up to now the result is not completely satisfying.
The idea is to digitize a greater number of video images per second and test, whether they contain a meteor or not. This task needs a high performance computer with a frame grabber card for digitizing the video signal.
There is one big problem, that make the automatic search nearly impossible: It is the strong noise of the image intensifier especially in the middle of the camera field. In our case this electron noise can be as bright as stars of the 2nd magnitude. Besides most of the meteors appear as a point on a single image due to the very short integration time. So you cannot simply test, whether there is a bright new spot on the picture or not, but you need a more skilful method for the search.
The algorithm we use at the moment is shown in Figure 2. At first we calculate the difference between the actual and the previous image. By doing this all stars and other constant objects disappear. The next step is a resolution reduction from a 512x512 to a 128x128 Pixel picture by averaging over 16 Pixel. As the result of this most of the statistic noise disappears. Next we subtract from the image a mask with the average noise, because there is generally much more electron noise in the centre of the image than on the edges. At the very end we check, if there is a bright and slightly longish object in the resulting picture. That might be a meteor!
To catch also faint and short meteors, your algorithm has to be very clever and your computer therefore very
fast. Our experiences show, that only the fastest 486 PCs with clock rates of 50 MHz or higher have the
needed power to digitize several images per second and analyse them in real time. You might of course use a
video recorder in slow motion, but then the recorder would be much more expensive and the evaluation of a
single videotape would last a whole day.
With our assembler program we are able to digitize and analyse 8 half images per second on a 66 MHz 486 PC. This speed seems to be enough to find meteors, but we need two passes for checking one video tape. Up to now the rate of recognized meteors is not very high. Due to the strong noise of our image intensifier only meteors up to the 1st or 2nd magnitude will be found. So here is still a lot of work for the future.
This last part of video observation and computer based analysis is not so difficult like the meteor search. You
can estimate the brightness of the meteor, fixing its time and duration and plot it into a gnomonic map for the
calculation of trail coordinates and the velocity. Even more accurate is the analysis with a computer. You
need the same equipment as for the computer based meteor search and of course a special computer program.
Figure 3 shows the way, how we get a meteor picture from the video tape and what happens after that. At first we add some images from the background to get an image without noise. Next we compose the single pictures from the meteor in a suitable way to get one image from the whole meteor trail. After that we use different methods to reduce the noise in the resulting pictures. The next step is the composition of both images and the subtraction of the background brightness. As the result we have a more or less beautiful picture from the meteor as you can see it for example in Figure 4. This picture can be stored in a compressed format (JPeg) and it is a basis for the calculation of the meteor trail coordinates.
This is done by a special part of the program, that searches for bright stars and analyses the absolute position of the meteor considering the strong distortion of the field of view.
At the end we get the position of an meteor with an accuracy of only a few arc minutes. All the data are stored in IMO's PosDat format to make them available for further reductions.
At the moment we are in the last stage of software development and hardware tests for using the full power
of MOVIE. Especially the problems of automatic meteor search and a fast analysis of meteors are essential in
the future. Only if we can make the analysis easier, this observation method will have a great future.
One possible way might be the use of a better equipment to avoid some of the problems. Here I see 3 alternatives at the moment: