The Space-Based Visible Program: Signal Processor Software Enhancements
The SBV sensor has an onboard signal processor that distills several frames of full-bandwidth CCD data into a few stars and a few satellite streaks, as illustrated earlier in Figure 4. The CCD camera collects a series of raw frames, or exposures, and transmits them to the signal processor. The signal processor then extracts a set of stars and the streak end points from the camera data, and those extracted data are downloaded to the SPOCC. In general this process has worked well, although some difficulties have occurred. Since the MSX satellite is situated near the inner Van Allen radiation belt, protons emanating from the sun are often detected by the CCDs on the SBV sensor, as described in the sidebar entitled "Radiation in Space." These proton events are manifested as short, bright detections that under certain conditions can either corrupt the formation of an RSO streak or, in some cases, appear to the signal processor like a valid streak.
FIGURE 4. (a) SBV raw full-frame CCD exposure and (b) an associated signal processor image. The onboard signal processor reduces the volume of raw data by as much as a factor of a thousand, thus allowing for more effective downloading of image data across narrow-bandwidth telemetry links.
Figure 17 shows an example of two such proton detections. The frameset in the figure was taken with the cover of the SBV telescope closed. Thus all star-like and streak-like apparitions in the frameset are caused by proton events on the focal plane. The number of such events varies from as few as four events per frame in benign regions of the orbit to as many as nine hundred events per frame in the South Atlantic Anomaly. Approximately 32% of the framesets collected by the SBV sensor show false streaks due to such proton events.

FIGURE 17. A frameset with the cover of the SBV sensor closed, showing false streak detections caused by proton events. Protons, which are always present in SBV framesets, usually have event signatures that are significantly different from those of RSOs. Under certain conditions, these proton detections are registered as valid streaks by the signal processor, and they must be filtered by both on-orbit and ground-based software prior to the data-reduction process.
Filtering out false streaks due to proton events is currently done by processing software on the ground. Figure 18 compares the characteristic signatures of a false streak caused by a proton event and a valid streak caused by an RSO. The proton event is a temporally short bright event, clearly unlike the signature left by a typical RSO. Software has been added to the ground-based data-reduction process to detect and remove false streaks on the basis of these differences in characteristic signatures [16, 22].

FIGURE 18. Characteristic intensity signatures, given in digital numbers (dn), of a false streak caused by a proton event (top) and a valid streak caused by an RSO (bottom). By analyzing variations in pixel intensities across the entire streak, ground-based processing software can distinguish false streaks from valid streaks.
Tests on the ground have shown that two additional problems occur because of the proton events. Proton events can overlie valid streaks, in which case the detected streak can be corrupted and the suppression algorithm described above fails. In addition, data collection in a region of intense proton events can overwhelm the signal processor, rendering it unable to detect any streaks. These problems can be corrected only by modifications to the onboard processing software in the signal processor. This modified software was developed and tested on the ground, and Table 1 shows the results of this testing.
Table 1. Differences in Streak Processing for Old and New Signal Processing Software* | ||||
Number of Looks | Number of Streaks | False Streaks | Number of Valid Streaks | |
Old signal processing software | 195 | 298 | 41% | 177 |
New signal processing software | 195 | 241 | 15% | 204 |
* Following launch and an initial analysis of the streak data, we determined that false streaks could be reduced on board, prior to downloading, if the signal processor was reprogrammed. New signal processing software was developed and uploaded to the MSX spacecraft, resulting in a reduction in the number of false streaks and an increase in the number of valid streaks. |
We examined 195 framesets, or looks, and we processed streaks through both the old signal processing software and the revised signal processing software. As the table shows, the number of detected streaks actually decreased with the implementation of the revised signal processor code, a result that occurred entirely because of the substantial reduction in the percentage of false streaks detected. It is also evident from the table that the actual number of valid streaks increased with the revised software. Both of these results achieve the desire outcome-a higher percentage of object detections. Because of the positive results of ground-based testing, we uploaded the revised signal processing software to the MSX spacecraft and significantly improved the performance characteristics of the SBV sensor [22].
Next Topics: Software System and Automation,
Data Reduction, & Health and Status
Return to SBV Table of Contents
| Top | Home | Quick Links | Search |
Direct comments and questions to: spocc@ll.mit.edu
Last modified: August 15, 1999
© MIT Lincoln Laboratory. All rights reserved.