S. Raj Chaudhury* & Dean Zollman
Physics Education Research Group
Department of Physics, Kansas State University,
Manhattan, KS 66506-2601
A few years ago the Physics Education Group at Kansas State University began exploring the use of digital video for teaching physics. In investigating the instructional use of this new technology [1,2] we have concentrated on its value for teaching topics in modern physics at the introductory level. We have examined digital video as a means to make learning physics more interesting to a broader audience than has traditionally studied physics. A larger fraction of our efforts has concentrated on directly processing the visual display which digital video allows and a smaller fraction on the quantitative analysis of the displayed information. This paper reports on our investigation of the inherent potential of digital video and our initial testing with students in an introductory physics course.
For physics teaching digital video is a technologically exciting new medium, and applications need to be designed specifically to exploit its unique capabilities. Currently many competing digital video formats -- including QuickTime [3], Video for Windows [4], Digital Video - Interactive [1], and MPEG [5] -- are available to users. It is beyond the scope of this paper to compare the strengths and weaknesses of the available digital video format. However, for the purposes of science education one needs high resolution (at least 320x240 pixels), full motion capture and access to tools that allow quantitative manipulation of the digital information contained in the video files. As the systems have developed, all of the ones listed above have reached this standard when they have sufficiently powerful computers or have additional hardware which is dedicated to the video playback. Digital Video - Interactive (DVI) which requires hardware to playback was the early leader in high quality playback and capture of digital video. Thus, it was most suited to our needs of exploring new ways to teach concepts in modern physics.
Digital Video-Interactive is currently marketed under the tradename ActionMedia II [6]. The display adapter board uses the Intel i750 chip to achieve significant differential compression and decompression of the video information. This board is necessary to playback DVI files. An optional daughter board provides full motion capture capabilities. We started our initial investigations under an earlier version of the system, the ActionMedia 750, and significant changes to both the hardware and software have occurred since that time. Until recently, this video system was limited to IBM-PC compatible computers. However, a DVI board for Macintosh computers is now available [7].
The widespread availability of digital video on the desktop has been a mixed blessing for our purposes. People are becoming more aware of the compelling nature of the medium, and video on CD-ROMs, which are an ideal way to distribute large video files, are becoming commonplace. However developers seem to have chosen to promote the software-only versions of digital video and to emphasize short sequences which are little different from playing short videotapes. They have de-emphasized the ability to complete digital image processing that were present even in the early digital video systems. Thus, we have been left with a unique problem: The multimedia industry has provided us with the means to use video in ways which has been previously unavailable, but then it did not exploit this means by providing the necessary software tools. Using the C programming language we have developed some of these tools in a Windows or MS-DOS environment.
Because of the complex nature of programming which seeks to apply digitization and image processing to video sequences, we have limited ourselves to one platform and one type of digital video. However, we believe that the most important component of our work is the unique way in which we have used digital video for teaching. As the developers of other formats of digital video bring their quality up to the standard of Digital Video - Interactive, these techniques can be applied to develop software for a particular generation or brand of hardware.
One of the problems with using a videodisc for this type of student data collection is that the video is selected by someone different from the student or the teacher. Because videodiscs are relatively expensive to create, the video is usually selected by someone far removed from the classroom in which the data analysis is taking place.
The capture capabilities of digital video systems significantly change this situation. Using computer programs that implement video capture students may store on to their hard disk a file containing video of an experiment which they themselves have performed. They may then analyze the video in a variety of ways. The analysis can use the standard techniques which were developed by our group [See, for , example, R. G. Fuller and D. Zollman, "Interactive Video for Teaching Classical Mechanics" in The many faces of teaching and learning mechanics, P. L. Lijnse, ed. (WCC, Utrecht, The Netherlands, 1985)] and refined by Jack Wilson [10] and Priscilla Laws [11]. These methods involve treating the video as a set of digitized individual frames. Data are collected from each constituent unit of the video rather than the whole. We shall refer to these techniques as traditional video analysis to distinguish them from the image processing described below. Since the video is stored as digital information on the hard disk, various techniques of image analysis and image processing can repeatedly be used on the video files with no loss of quality.
Thus real time digital video capture combines the advantages of both laserdisc and videotape. Random access is possible just as in videodiscs, and repeated viewing and processing of a video file does not cause loss of quality as in videotape. Image processing techniques applied to video are an important part of our work.
The video camera (Sony camcorder capable of 1/4000 shutter speed) and the pole from which the ball is released were both mounted on PASCO dynamics carts resting on parallel tracks. The two carts are capable of coupled or independent motion. Through a computer program students use the capture capability of the digital video system to collect video of the dropping ball in four different reference frames (three specified by us and one of their own choosing). Once the video is saved to the hard disk of the computer, another program gives students access to a VCR-like interface to the digital video hardware. Students step through the video of the event (ball drop) frame by frame and mark the location of the ball on each successive frame with the mouse pointer. A graph of height above ground versus frame number (time) is automatically plotted in an adjoining portion of the computer screen. Though the material relevant to this experiment had been covered in lecture several weeks earlier, we felt the concept learning for these students would be better reinforced by asking them to mark the location of the dropping ball themselves rather than having automatic data collection techniques do it for them. A pre-test and post-test were administered on paper to assess students' concepts of relative motion in different reference frames.
With written instructions and minimal help from proctors, students working in groups of two or three were able to complete the exercise well within the allotted time. This is particularly encouraging since many students were unfamiliar with computers and struggled with the use of the mouse. A complete analysis of learning in this environment is being completed and will be reported later [13]. While this introductory experiment does not take full advantage of the digital capabilities of DVI, it does accomplish several goals. Students become familiar with real-time video capture, with using video as an analysis tool and with analyzing events from different reference frames. The results of this initial exposure make the experiments described below, which use some automated image processes, more believable.
Once the data have been collected, a variety of graphs such as velocity and acceleration versus time can be plotted by the computer. Students can then study the dynamic variation of these quantities or how they depend on each other by looking at graphs on a computer screen. This approach to data collection and analysis is very similar to the one for one-dimensional motion used in microcomputer-based laboratories (MBL).
With Digital Video-Interactive, we can follow the same general scheme by analyzing an event which is recorded onto a fixed disk. The first step in our process is to use the real-time capture capabilities to record the event and store it. Since most of the computer's CPU time is required to collect, compress and store the video information, we do not attempt to find the coordinates of a bright point while the event is taking place. However, because the event itself has been recorded on the hard disk we can return to it and collect position versus time data very easily. These data are obtained from the individual video frames "off- line", that is after the video has been recorded. Once the space and time information is available we playback the real video along with a selection of graphs. In our present system we create four windows on the screen. In three of those windows students can select the variables they wish to see plotted. In the fourth window we play the video of the experiment. Thus, the students can see how the graph unfolds while they are also watching a recording of the event. This approach of showing the video of the event on the screen at the same time as the graph unfolds dynamically was not available with previous systems. Here, however, we can provide the students the opportunity to see how the graph changes as they watch the motion change. Because the motion can be slowed down or stopped on any frame, students have full control over the rate at which they watch the event and analyze it in terms of relevant physical phenomena.
Any event which can be recorded two dimensionally can be analyzed in this way. In addition to the standard graphs of dynamical variables versus time we can also include phase space diagrams which are particularly valuable for analyzing chaotic motion. One of the more intriguing applications of this approach is the analysis of a spherical pendulum which has magnetic interactions between itself and several magnets placed near by.
Recently there have been reports in the literature of the traditional video analysis approach used with digitized frames of motion [19]. Students playback the video one frame at a time, move a cursor over the location of the point of interest and record those coordinates. This method, as stated above, has already been thoroughly explored using analog video recorded on a videodisc. Digital video allows us to go a step further and create Video-Based Laboratories (VBL) [20] with a data collection method analogous to that in MBL. Presumably as the DVI hardware and software improve we shall be able to overcome the slight disadvantage of not being able to concurrently create graphs in real-time as the video data is being collected.
This digital video system has two advantages over MBL. It can analyze two-dimensional motion. More importantly, students can watch the graphs and a video of the event simultaneously. This approach should help students better visualize the relation between the motion of an object and the time dependence of various physical quantities related to that motion.
The demise of 8-mm film as an amateur recording format has greatly decreased the possibility that students might record events using these time- lapse techniques. While time-lapse can be accomplished using video, the recorders which have this capability are very expensive and seldom available in schools or universities. Thus, time-lapse for students is only available to those who can find an old 8-mm camera, the film needed for recording, and someone to process the film. The digital video format allows us to reintroduce this time- lapse capability rather easily. Our programmed procedure follows the basic scheme used by the older film cameras. The video camera is aimed at the event of interest. It continuously sends video information to the computer. However, the computer only captures the information at specified intervals. We can program the computer to capture one frame every 15 seconds or one frame every hour, or any other choice which we might wish. Once the individual frames are captured they can be assembled into a “video tape” on the hard disk and played back at any speed up to 30 frames per second.
Our initial programming with this technique used the ActionMedia 750 system and required a gap of 15 seconds or more between video stills captured to disk. ActionMedia II seems to allow greater flexibility in the capture rate of these frames. The combination of our technique and the ActionMedia II system provides a broad range of time-lapse capabilities. A number of commercially available software packages called video editors allow one to 'assemble' a movie in this way. [21]
We have tested this technique using a video camera connected to a microscope to view the growth of yeast cells. It seems quite effective and quite adequate to the tasks at hand. Thus, we have created a system which enables students to record the details of events which happen very slowly. Playback can enable them to watch and analyze these scenes in detail.
Our method for combining images derives from an attempt to teach the use of space-time diagrams. To create a concrete space-time diagram we wanted to display the one-dimensional motion of a real object with its space coordinate horizontally and its time coordinate vertically. Using DVI we were able to do this by selecting a horizontal strip from each of several images, "stacking" them on top of each other and displaying the resulting single image which contains spatial and temporal information about the entire event.
As an example consider a collision between two low-friction carts as illustrated by the schematic representations in Figure 1. The boxes on the left of Figure 1 show four possible video frames from this event before the collision takes place. The student would select the region of interest which in this case is the portion of the screen in which the carts appear. Once this region is selected the computer would copy that region from successive frames and place them on a single image which describes the entire event. This video-to-still image copying process is shown in Figure 1. The arrows indicate which sections of the four video frames have been copied to the single still image -- the visual space-time diagram.
The software to create this type of figure was designed to give the student a maximum control over the final image. Once the program is activated, the students may determine the width of each of the horizontal strips which will be included in the final image. They may also choose the number of video frames which elapse between each image that will be recorded on the screen. Of course, the particular video frame out of the motion sequence on which to begin building the image is also chosen by the user.
When the program is activated a video image freezes on the screen. A false color strip appears horizontally across this frozen image (indicated by the horizontal dotted bars in Figure 1). To select the area of interest the students have control over the location of the horizontal strip and its width. Using the keyboard or mouse they may move the strip vertically and also adjust the width of the strip. Once they have selected the appropriate area and indicated the elapsed time between frames which are to be recorded, the computer builds the image quickly and displays it on the screen. Asking students to draw a line connecting the front edge of each of the images of the carts shown in Figure 2 provides a concrete way to begin a discussion of space-time diagrams.
The number of consecutive pictures which can be displayed on the screen is determined by the vertical width of the interesting strip. Figure 2 shows a situation in which a very wide image was selected so only a few images can appear on the video monitor. This type of image is quite useful in helping students understand how space time diagrams can be created. They can see the entire image of the object as it is located in each of the frames. However, one does not need to see the entire image for successful analysis of the motion. Choosing a much smaller "stripe" across the video screen one can put up to 240 images on a single screen. Figure 3 shows the same collision as in Figure 2 but with 120 images. Here the motion during the entire collision between the two objects, and one of the objects with the end of the track, is shown clearly.
Once this space-time diagram is created, we can show students how it would look from other reference frames by manipulating each line of the video image so that it represents a different reference frame. The screen photographs shown in Figure 4, Figure 5 and Figure 6 indicate how the same collision with the same video data would look from other reference frames. Figure 4 is the center of mass reference frame and Figures 5 and 6 are space-time diagrams in the respective reference frame of each cart involved in the collision. This manipulation of reference frames provides students with a way to see how space-time diagrams change as they change reference frames.
Once we are able to collect data on the different reference frames, we can also manipulate the video so that we can play back the scenes as if they were recorded from different reference frames. With this synthetic video processing we can allow the student to select any of the reference frames listed above. Then, the computer can play back full-motion video as if the scene were actually recorded from that reference frame. Students can watch in real-time as the video images which they recorded from a fixed camera are played back as if the camera were in the reference frame of the initially propelled cart, the cart that was initially at rest, or in the center of mass reference frame of the two carts. They may even change, virtually, the relative masses of the two carts and see how the collision would look in a center of mass reference frame if the carts did not have identical masses. This ability to visually shift the reference frame provides students with a way to see how reference frames are important in a variety of observations and how the view of a collision changes as one moves from one reference frame to another.
The process of recording an event which has been seen in the laboratory by the students, then using image processing techniques to create space-time diagrams and to change reference frames, enables students to visualize the importance of reference frames. This visualization has not been possible with student experiments recorded by any other means. While one can move an ordinary video camera to match certain reference frames, one would be hard pressed to put a camera in, for example, the center of mass reference frame. The center of mass reference frame is a very important one for analyzing a variety of collisions ranging from those of automobiles to those of elementary particles. Thus, we believe that this technique can have an important place in aiding students, particularly those who prefer to think visually rather than numerically, with the comprehension of the important concept of reference frames in physics.
Since we began this project, we have seen the options for digital video explode. Both new hardware as well as software-only approaches have been introduced by companies such as IBM, Apple, Microsoft, Silicon Graphics and Sun Microsystems. While these companies have not agreed upon common file formats or compression algorithms, they share a common goal of using video stored digitally on a hard disk. Thus, all of the techniques described here should lend themselves to implementation on any available digital video system.
While implementation is possible in principle, severe limitations will exist with the software-only versions. First, resolution - both space and time - is poor; second, compression of video frames is limited; video will occupy a large amount of hard disk space. Finally, image processing will be slow. Thus, digital video systems with dedicated hardware are probably best suited for implementation of the techniques we have described.
Fortunately, even software-only video, such as QuickTime and Video for Windows, support optional hardware add-ons. This makes it viable to use computers running these software for student experiments with digital video.
The work was supported in part by the National Science Foundation under Grant No. MDR-9150222. Early explorations were supported by the ACIS program of IBM.
Figure 2. Creating an actual space-time diagram from stacked strips of video frames.
Figure 3. Space-time diagram for cart collision in laboratory reference frame (assimilated from 120 video frames).
Figure 4. Cart collision in center-of-mass reference frame.
Figure 5. Space-time diagram in reference frame of initially moving cart.
Figure 6. Space-time diagram in reference frame of initially stationary cart.
Dean Zollman Kansas State University Voice: 785-532-1619 FAX: 785-532-7167
