In light of the recent collisions between the fragments of comet Shoemaker-Levy and the planet Jupiter, there has developed an increased awareness among the public of the possibility the earth may encounter a large meteoroid in the near future. In order to prevent a major catastrophe from occurring, it becomes necessary to systematically search for and identify earth crossing meteoroids and comets before they collide with the earth [1]. Very often cometary bodies that have earth crossing orbits have been associated with meteor streams that are seen on an annual basis as meteor showers. Several of the major meteor showers that have been studied and whose orbital elements are known, have been associated with periodic comets [2] such as comet Halley (Orionids), comet P/Swift-Tuttle (Perseids), and asteroid 3200 Phaethon (Geminids). This meteoric phenomenon has its origins in the material released from the outgassing of a comet whose ejected dust particles get distributed along the comet's orbit. If the cometary orbit and thus its debris stream cross the orbit of the earth, then we witness a meteor shower. In addition, large non-cometary meteoroid bodies have been suspected to also have smaller particles strewn along their orbits due to impacts with other meteoroids thereby also possessing a debris stream. It is proposed that detection of the larger parent bodies would be possible through the study of the orbits of the small particle meteor streams associated with comets and asteroids. If the meteor streams are visible from earth as meteor showers, then the parent body must also be earth crossing and potentially dangerous.
Although many major meteor streams have been well studied and their parent bodies identified, there still exist a large number of minor streams, unverified streams, and sporadic meteors for which no association is available. This is due to a lack of accurate measurements in sufficient quantity and quality to identify and/or verify meteor streams and their orbits. The primary cause is that weak meteor streams compete with sporadic meteors in numbers of observed events, making radiant association difficult due to poor statistics. In addition, there is the unavailability of low cost, low-light sensitive, automated meteor detection equipment for the field that would be more sensitive and accurate than visual observers. Although a tremendous amount of the meteoric data till now has been provided by the amateur visual meteor observer network, in order to study the minor meteor streams to sufficient detail, an automated meteor detection system that exceeds human visual magnitude limits is required. The use of such a system in the systematic study and evaluation of meteors, meteor streams and their orbits, would help in addressing the issues associated with detecting potentially dangerous earth crossing comets and asteroids as well as furthering basic knowledge of meteoric phenomena.
Meteor astronomy to the present time has relied on a number of observing techniques and technologies to monitor the near earth small particle environment. These have included a world wide network of visual observers, small frame camera and Schmidt camera photography, and radio/radar measurements. Visual observation by amateur meteor astronomers has provided the bulk of the information currently available about meteor stream structure, density, mass distribution, and yearly variability. Double station photographic techniques have provided orbital information yielding cometary and asteroid associations of meteor streams with earth crossing objects. Radar observations have extended the observing coverage to 24 hours per day and has provided information on the sub-micron sized particle distribution of meteor streams.
With the advent of modern video and computer technology, another class of instrumentation could be added to those techniques currently utilized in the field of meteor astronomy. The system to be described herein involves the use of low cost, off the shelf imaging hardware, a personal computer, the application of image processing algorithms, and the development of software for a real-time automated meteor detection capability. The instrumentation is referred to as the Automated Meteor Detection System (AMDES). The basic system consists of a fast, wide field lens, image intensifier, and CCD camera all coupled to a real-time detection capability in a PC-sized computer system. It will provide the following capabilities and enhancements over current observational approaches.
Note that the magnitude limits and resolution capabilities were based on a system with a 10° field of view. Adjusting the focal length of the leading objective lens can provide either higher resolution and fainter meteor detection or wider fields of view with a loss in limiting detectable magnitude.
AMDES can provide automated measurements, enhanced detectability, and improved accuracy in a small and inexpensive package that is easily portable and simple to setup at any site worldwide.
Compared to current systems available, AMDES does have many advantages. Visual observations suffer from human eye limitations and individual subjective biases. The limit of the human eye under optimal seeing conditions is at best magnitude +6.5 and then only over less than a 5° field of view [3]. Outside that angular region the sensitivity drops off dramatically yielding a brighter limiting magnitude with respect to look angle. The detection angular capability thus widens with increasing meteor brightness, but offsetting that is the reduction in the number of meteors visible above a certain limiting magnitude. The result is that most meteors seen by visual observers lies in the +3 to +4 magnitude range. Secondly, the human eye can detect only in the visual frequency band and misses some of the meteor spectral energy in the near infra-red. Third, even experienced meteor observers can have variable detection thresholds during the night, require observational breaks, and make errors in estimation or recording vital information. Fourth, the accuracy of hand plotted meteor tracks are on the order of one-half degree and highly dependent on the observer's knowledge of the sky and the ability to remember positions of an event that occurs in less than one second.
Photographic observations correct for the difficulties of accurately plotting meteor tracks and are capable of measuring meteor paths to within tens of arc seconds. The limit of photographic methods lies in the sensitivity of film to record down to only magnitude +2 for small frame 35mm cameras [4]. Such a reduced brightness threshold results in fewer meteors recorded and thus fewer measurements available for statistical studies. In addition, proper timing of an event in a several minute exposure can be difficult to accomplish. The development of film and the scanning of negatives is also a time consuming process. Currently there are no batteries of cameras set up in the U.S. for the purposes of meteor research but there does exist a network of cameras operating in Europe.
Radar and radio observations can work around the low detection statistics by observing the smaller sized meteoroid particles that enter the earth's upper atmosphere and leave an ionization trail that scatter and reflect electromagnetic waves. Since these smaller sized particles are far more numerous than those that produce visual meteor trails, many events can be recorded. In addition, radar can operate day and night giving twenty-four hour coverage with good statistics. The latest generation of radar meteor detectors such as operating in Austrailia, can determine a reasonable orbit. The drawbacks are the extreme expense involved of $1 million per system, selection effects limiting cross sectional and evolutionary studies of streams, and no systems operating in the northern hemisphere that produce accurate orbital data.
Video observations have only recently reached a point where off-the-shelf hardware is available at extremely low cost making the wide area distribution of imaging systems possible. Current efforts along these lines have resulted in detection limiting magnitudes of 8th using second generation image intensifiers coupled with CCD cameras [5,6]. The major drawback is that the night's observations are videotaped and then played back over several hours requiring a human observer to spend a large fraction of time searching for meteor events on a TV monitor. Many of the same disadvantages of direct visual observations apply again to this method of meteor observation. Further improvements beyond this current capability are necessary in order to make video systems competitive with visual and photographic observational techniques.
AMDES can provide the capabilities necessary to address the basic deficiencies indicated above. Improved night sky coverage, deeper magnitude limits for meteor detection, and higher accuracy in track can all be used to advance the current state of the knowledge in meteor astronomy. In addition, full automation of the detection processing would result in more consistent data by removing human biases induced by fatigue, errors in track due to large look angles, or the inherent signal losses of storing and retrieving images off video tape. With the use of a common set of equipment and algorithms, a uniform basis of sensitivity and detection would become available.
Currently, there is no work being done in this country in the field of meteor research by professional astronomers. The field has been left largely to the realm of the amateur observer. Worldwide there is a network of meteor observers that are organized by the International Meteor Organization (IMO) where interesting work has been done and could be further pursued using data supplied by AMDES. A better understanding of meteoroid origins, evolution, cometary associations, stream structure, thickness, mass distribution, existence of sub-radiant streams, and identification of stream composition would be possible. With a larger number of events captured in a dual station system, one could identify new streams and verify only marginally detected minor streams. Such detection could lead to the discovery of a meteoric stream orbit that could be associated with an unknown but potentially dangerous earth crossing parent body. Given the estimated orbit, a search could be mounted for the parent body over a much reduced portion of sky than present day asteroid hunting programs currently scan.
The system could also be configured for extremely low cost per unit (2000 $US) as a fireball/meteorite fall detection system. This would help in identifying the origins of some of the larger fragments of infalling meteoroids and where other potential sources of earth crossing asteroids may arise. In addition, the study of sporadic meteors through the study of their orbits, origins, and annual variability, would lead to a better understanding of these apparently random events. Does the sporadic background, which makes up a significant portion of observed meteors, have some common origin that presents a heretofore unknown collision threat? All these observations and studies would be feasible with fielded pairs of AMDES worldwide setup up as dual station meteor monitoring networks.
The objective for AMDES would entail four stages of development. Portions of the first stage such as the hardware configuration for the imager have been examined and have already been exercised in the field by a number of researchers. The critical test for AMDES at this point in time is to demonstrate the proof of concept of automated video meteor detection on a low cost computer platform. The four stages are outlined as follows:
The critical element in the AMDES project development is in the construction of a functional detection system. This would involve the coupling of a low light level imaging system with a real-time image analysis computer system. The proposed system block diagram is illustrated in figure 1. The configuration includes an imaging system with mount, monitor, recorder and computer for real-time/post processing. The elements of each sub-system break down as follows:
The basic components of the imager sub-system are filter, lens, image intensifier, transfer optics and a low light level frame rate CCD camera. The front end filter can be employed as a means to enhance contrast for poor lunar lighting conditions or for spectral analysis using standardized color filters. By using a R60 red filter, the detrimental effects of bright moon lit nights can be minimized by enhancing the meteor's most detectable spectral region in the near infra-red. In addition, spectral colors could be obtained simultaneously on multiple AMDES to do composition studies specific to each meteoroid stream.
The objective lens is one of the most critical elements in the imager and controls the ability to achieve a high input signal to noise ratio into the intensifier. It is best to boost the gain at this stage of the optical path to minimize noise by using fast lenses with short f-ratios. Use of a readily available and high quality fast camera lens will mitigate the need for higher gain in the intensifier stage with its associated higher noise levels. Specification of the required field of view, whether it be all-sky to telescopic, will determine the focal length required for the lens. Issues of lens vignetting are of far less importance than speed of the optics since in the design configuration proposed only the central portion of the lens is actually imaged by the intensifier. Typical system characteristics for various lenses are given later in table 1. The lens chosen should be free of spherical aberration and also be coma corrected. Spectrally the lens should be clear in the visible and near infra-red. Note that standard video camera lenses should not be used as they employ infra-red blockers for proper color balance and would reduce sensitivity of the system to the near infra-red.
As part of the lens configuration, an electronic focuser could be added for automated remote focusing in situations where lenses may be interchanged often and hands-off operation of the focus is a requirement. For cases where a single lens would be used exclusively, the focuser would not be necessary as the lens could be pinned permanently to the correct focus. Finally a coupler stage is necessary to mate the lens' mounting system to the C-mount threaded barrel of the image intensifier (1", 32tpi).
The image intensifier proposed for this work is a three stage multi-channel plate generation 2.5 intensifier tube with automatic gain control. Gains can vary from 10^4 to 10^5 with good linearity across the tube, low cost of 3500 $US for a scientific grade unit, and light in weight. Spectral response is in the 400 - 950 nanometer range which pushes further into the near infra-red than second generation tubes. The near infra-red sensitivity of these tubes has been conjectured to aid in the detectability of fainter meteors. These tubes have already been used in a number of astronomical applications. The output from the intensifier which is at visible wavelengths must be focused onto a CCD chip via a set of transfer optics. The transfer lens arrangement must be optimized for a given chip size which, in the past, has been composed of a coated optics six element f/1.1 flat field design.
The CCD detector is a low-light sensitive, high resolution, black and white frame rate camera with peak spectral response in the visible wavelengths. Its output is a standard NTSC TV signal with fully interleaved images produced at a 30 Hz rate. This rate provides sufficient temporal resolution to capture a meteor event across several frames and allow for estimation of velocity. In addition, there is no non-imaging dead time associated with downloading frames from camera to computer. The collection of each frame separately with time tagging eliminates the problems associated with meteor event time estimation, correlation between dual station measurements, and chopping wheel inaccuracies. The alternative of using an integrating type CCD camera has the advantage of containing a complete meteor track on one image and no frame grabber required with the computer. However, the disadvantages that ruled out its use are lower light sensitivity than frame rate cameras, slower download time to the computer resulting in dead observing time, and higher levels of integrated background noise.
The mounting system for the imager can consist of a simple tripod for fixed azimuth and elevation orientation or a sophisticated remote operated equatorial mount with tracking drive. The latter would have the capability to slew to a particular radiant position through a remotely operated command linkup. The monitor's purpose would be for the early development stages where verification of proper signal receipt and manual focusing on site would be necessary. This would consist of a portable 5 inch B/W TV monitor. Though the purpose of AMDES is to automatically record only meteor events via computer, in the early development stages is would be necessary to simultaneously record the images on tape. The best video recorder on the market for this purpose is a Hi-8mm VCR recorder which retains image resolution down to that available in the original signal. The tapes could be played back into the computerized detection processing sub-system to examine alternate detection and processing algorithms with only small losses in signal fidelity.
The basic computer proposed for this effort involves the use of a 90 MHz Pentium processor. Alternate processors such as the PowerPC have sufficient computational speed for the necessary image processing, but currently lack supporting hardware such as the frame grabber board needed as an interface to the CCD camera. The basic computer would be equipped with 32 Mbytes RAM for storage of multiple frames for summation on the fly, at least a 500 Mbyte hard disk, a floppy for data output, and remote communication capability for each imager.
The frame grabber board must be capable of grabbing NTSC fully interleaved 510 x 492 pixel images at a 30 Hz rate and transfer the complete images over the computer's PCI bus to the computer memory with no dead time. The individual frames will be summed by the CPU while the next set of images are collected by the frame grabber board. This requires a board with concurrent image grabbing and data transfer capability (asynchronous processing). The summation is done to develop a linear track on the image for input to the meteor detection algorithm. The computer processor and data bus should be fast enough to sum these images in real time and exercise the detection algorithm at the 30 Hz rate of the camera output. An alternative is to use a summation capable frame grabber board which off loads the computational load from the computer processor at the cost of losing the individual frames and much higher overall system cost. With this type frame grabber, the velocity information can still be backed out by electronically chopping the image (leaving out every nth frame in the image summation).
The real-time processing will require the development of software for the frame grabber control and image download running in parallel with the fast integer summation and linear track detection algorithms. The detection algorithm will be based on either a Hough transform line searching algorithm working on the summed image or a motion detection algorithm working across several individual frames. Part of this algorithm will involve the development of a noise subtraction algorithm to enhance the signal track to noise ratio and increase the detection probability. Once a detection is made, image storage of all the contributing individual frames will also be done in real time with a time/date stamp accurate to 33 msec.
The post-processing algorithms that will operate on images containing a meteor detection can be exercised in an off-line mode. These algorithms involve the development and incorporation of software for star field identification, pointing direction determination, plate constant evaluation, application of correction terms, meteor track coordinate estimation, magnitude estimation, velocity estimation, radiant association, and report/gnomic projection generation. The final set of information is sufficiently reduced in data bandwidth that results could be transferred via floppy or downloaded over a remote hookup. For dual station work, additional algorithms for meteor event correlation, radiant association, and orbital element determination would be necessary [7].
A prototype imager sub-system has been constructed and is currently operating having achieved the following levels of capability under skies with limiting visual magnitude of +5.5:
LENS FIELD OF VIEW LIMITING MAGNITUDE
------------- --------------- ------------------
28 mm f/1.8 12° x 16° +8.0 stellar
50 mm f/1.4 6.5° x 9° +9.0 stellar
These results neglect gains that could be obtained from using faster lenses and applying image processing enhancement. It also neglects any detectability gains due to the near infra-red sensitivity of the intensifier as this test only evaluated stellar images and not meteor tracks. This also neglects losses from less integration time per pixel for a moving meteor trail relative to a stationary stellar source. Further work needs to be done to determine the cost/capability trade-off for an optimally priced system.
Currently, a complete single station AMDES with the capabilities listed above can be obtained for 8000 $US in hardware costs. To develop the real-time detection software and construct a complete single integrated system, the total cost is estimated to be 30000 $US. For dual station work, the cost for initially developing software to interface multiple station data with the orbit determination algorithms is estimated to be 50000 $US. After the initial software development, the costs would be limited to analysis and hardware purchases. For improved limiting magnitudes, finer angular resolution, better detectability, far more expensive imaging components can be obtained thus raising the single unit hardware costs.
Given the current state of the art in video and computer technologies, it is proposed that a fully automated meteor detection and monitoring system could be developed and fielded at low cost. Multiple systems could be distributed worldwide to provide 24 hour night coverage of meteor activity with operation of the systems done from a remote location. The capabilities of AMDES significantly improves upon that of visual, photographic, and radar techniques by providing both greater quantity and quality of meteoric event data useful for orbit estimation. The data collected would be used to advance the state of research in meteor astronomy, aid in identifying potential orbital parameters for large earth crossing meteoroids and comets, and establish a more active role for meteor research within the United States.