Image Processing and Transmission by Means of Satellite Network

Transcription

1 Image Processing and Transmission by Means of Satellite Network Marco Molina, Roberto Bianchi, Franco Bernelli Zazzera Dipartimento di Ingegneria Aerospaziale Politecnico di Milano Via La Masa Milano, Italy Abstract In order to assess the feasibility of the transmission of images from a University minisatellite (Palamede) by means of a commercial satellite network (Orbcomm), an experimental work has been performed in order to check service availability and reliability. Orbcomm communication system is based on short messages, called GlobalGram, of about 200 bytes each. It was necessary to study the most effective way to fragment each image (already compressed by means of jpeg algorithms), in order to send it through a GlobalGram. An automatic header generation allows to put the images in the correct order when on ground, and provides additional information about the image itself (exposure time, geographical coordinates, time ). Test showed that Orbcomm system is stable, providing a low but constant data rate, with an average of about 2 Kbytes/hour. This data rate, of course, prevents the system from being used for real time operations, but, due to the fact that no information was lost during the experimental campaign, a 24-hours worldwide coverage is guaranteed, with minimum pointing requirements. A typical 50 Kbytes image will be available on ground in about 24 hours, without any need of a ground station. The data are available in the format of , which is dispatched by the system through the Internet, and this will allow a worldwide access to the Palamede satellite data, when on orbit. The tests have been performed from a static test facility; still to be investigated is the system response to antennae angular velocity, in order to understand the capability to run also in the very first phases of the mission, after Palamede deployment. Main advantage of the system is that, with rather simple interface software, and using off-the-shelf components, a reliable system for low data rate transmission for small satellites is proposed. Introduction Palamede is a microsatellite, having a total mass of less than 100 kg, developed at Politecnico di Milano. This class of satellites has become popular in the last few years, since they represent a low cost alternative to larger platforms with a sufficient operational capability. In particular, Palamede will be launched on a circular, inclined orbit, with a yet undefined launcher, with the mission objective to test new and cheap technologies for future small satellites. Most of the hardware of Palamede is not space qualified, to reduce cost. In the current configuration, the two payloads are a GPS receiver and a CCD camera to capture images of Earth, with a typical coverage of a 96 x 72 km ground surface. Images will be captured with a frame grabber card, processed by the on board computer and then sent to the ground station. Tests performed have shown that the spacecraft attitude stability must be better than 2.4 deg/s in roll and 1.9 deg/s in pitch to allow a safe image acquisition. Each image, compressed in jpeg format, will have a size between 13 and 300 Kbytes, depending on the compression ratio [1]. The communication system is managed by the on board computer, a 100 MHz Pentium class system arranged in the PC-104 configuration, with 32 MB RAM and 72 MB solid state disk. The communication system has been studied intensively, and the first option studied has been related to a conventional system with the design of a dedicated ground station [2,3]. These studies have shown the criticality of the approach, so in the final design the communication with the ground station will be granted by inter-satellite link with the Orbcomm constellation, in a so-called Dial-A-Satellite configuration. The selection of Orbcomm has been the result of a trade off between four different options: Inmarsat, ICO, Globalstar and Orbcomm. The first two options have been discarded because 1

2 of the size and power of the transmission system, while Globalstar is affected by great Doppler effects and has high subscription costs. Orbcomm constellation is constituted by 36 satellites for bi-directional communications. The system makes use of a transmitter and a receiver, and transmission is via messages. In this way, no dedicated ground station will be needed to handle Palamede communication, any PC can represent the control station. This solution is in the spirit of lowering cost, typical of microsatellites. The satellite must be equipped with a transmitter/receiver, called subscriber communicator (SC), that will communicate with the network of Orbcomm ground stations, that will redirect messages as s to the subscribed user. In this way it will be possible to transmit commands to the satellite and receive telemetry data as well as images. In this framework, the present study will deal with the problem of validating the transmission concept, verifying the effective capabilities to handle the data of Palamede. Orbcomm constellation Orbcomm is a bi-directional communication system, with global coverage, for data organized in packets [4]. The bridge between the user s SC and the Orbcomm Gateway, i.e., the Orbcomm ground stations, is represented by the LEO constellation of Microstar satellites. The system is composed of a Network Control Center (NCC) and three operational segments: the space segment, with 36 satellites, the ground segment, with the Gateway Earth Stations (GES) and the control centers located worldwide, and the user segment, composed by all transmitters that communicate with the LEO satellites. All communications must pass through the Gateway Control Center (GCC), dedicated to the transmission and monitoring of data flow, and the GES, that establishes the connection between satellites and GCC. The following sequence of actions, depicted in Figure 1, can be considered a good example of the system behavior: Figure 1: Orbcomm system 2

3 the registered user, using an software, composes a message via a terminal and downloads it to a SC; the SC sends the message to a satellite, that forwards it to the GES; the GES sends the message to the GCC, from which it is routed to an Internet provider; the recipient can download the message on his terminal, that can be a common PC, via software; the reverse order is also possible. The constellation consists in 36 satellites, scheduled to become 48 in the future, located in 6 orbital planes. The main constellation is placed on 4 orbital planes (A,B,C and D in Figure 2), with 8 satellites each. The first three planes have an inclination of 45 on the equator, while the fourth is equatorial, and the orbit altitude is 825 km. Two extra planes (F and G), inclined at 70 and 108 respectively, have 2 satellites each at an altitude of 780 km. Figure 2: Orbcomm constellation Among the services provided by Orbcomm, two are of interest for the present application: messages and GlobalGrams. Both allow the transmission of data sequences, but messages can be sent only if the satellite is in direct contact with a ground station, while GlobalGrams can be stored by the satellite and forwarded when contact with a ground station is established. Messages can however be longer than GlobalGrams, which are limited to 229 bytes. Software requirements Palamede will be managed by a Linux-based on board computer, so this is the target operating system under which the software for data transmission must run. The requirements for software devoted to image coding, segmentation, transmission 3

4 and recovery are peculiar. First, it must be mentioned that Orbcomm system can send only ASCII characters, so it has been decided to encode all data in Multipurpose Internet Mail Extension (MIME) Base64 format [5], which adds time to the transmission but is stable and robust and guarantees also a safe recovery. The limits imposed by the GlobalGram size (229 bytes), represented a second severe requirement. Each image is first encoded and then split into lines of 208 characters, in such a way that adding the appropriate header the message would be exactly of 229 bytes and would fit into one packet. An example of this result is shown in Figure 3, where the blank spaces in the header are inserted only for sake of clarity and are missing in the real header. image part Header total parts time date String of portion of encoded image Sep02 DawMDAwMDAwMDAwMDAwMDP/AABEIAE4AcAMBI gaceqedeqh/3qaeaaf/xae/aaabbq7ebaqebaqaaaaaaaaadaaecbaugbwgjc gsbaaefaqebaqebaaaaaaaaaaeaagmebqyhcakkcxaaaq6qbawieaguhbggf AwwzAQACEQ3EIRIxBUFRYRMicYEyBhSRobFCI yqvusfimzrygtfd Figure 3: example of encoded packet This operation produces a series of text files related to the same image, that are sent as Orbcomm messages in sequence. The recovery software must then be capable of rebuilding images, also if not complete, even in the event of loss of some messages. Figure 4 reports the structure of the overall transmission and recovery procedure, while Figure 5 shows how images are reconstructed in case of some packet loss, showing an acceptable performance in this respect. MIME coding JPEG image FLIGHT SEGMENT text file partition FILE0001.TXT FILE0002.TXT FILE0003.TXT FILExxxx.TXT transmission ORBCOMM transmitter text file union FILE0001.TXT FILE0002.TXT FILE0003.TXT FILExxxx.TXT software ORBCOMM receiver MIME decoding JPEG image GROUND SEGMENT Figure 4: software structure 4

5 a) original image b) image with 5 missing packets Figure 5: image reconstruction in case of packet loss Experimental results The transmission concept outlined in the previous sections has been validated in a series of experimental tests, in which both flight and ground segment were located on ground. The entire flow of data has been checked, and the time to process images, from acquisition to recovery on the user ground station, has been verified. To be sure that all tests, carried out with the transmitter located in Milano, would be significant also for the real orbit of Palamede satellite, a parallel simulated analysis has been completed creating a suitable scenario for all Orbcomm satellites, using updated Two Line Elements (TLE). Figure 6 shows that the foreseen orbit for Palamede is well covered by Orbcomm constellation, with an almost constant visibility of satellites and with only rare and short periods of lack of coverage, in the order of a few minutes. Coverage of Milano is always guaranteed. Figure 6: software scenario for Palamede and Orbcomm constellation 5

6 Ten different test sessions on ten different days have been completed, to verify in different dates the available connections between DTE and satellite constellation, and to have some statistics on the transmission process. The first result is that all messages have been received, with no loss at all. This is important because it guarantees that the chance of loosing the first packet of the image, that contains the header of the entire image and without which it is impossible to reconstruct the original file, is very low. Figure 7 reports the connections of one of the test sessions, carried out during one night. It can be noticed that not all connections are of equal duration, and that distribution over time is not uniform. However the average connection time is more than satisfactory, and the data can be transmitted with a good time distribution. Connection time in a typical 12 hour session time (h.mm.ss) connection num ber Figure 7: connections of one of the test sessions Table 1 collects the most relevant data of the transmission process, for all test sessions. The first three lines report the total number of connections established and the number of packets transmitted for each test, and the average connection time. The last four lines report the rate of transmission, in terms of packets and Bytes per hour. In particular, the first value is the ratio of the number of packets or Bytes and the time elapsed from receipt of the first and last packet. The last value instead is the same ratio but the time is net, not considering the pauses between each connection. The most important parameter, i.e. the number of Bytes transmitted per hour, is in the range between 1407 and 3791, with an average of 2127 Bytes/hour. Test session number n. of connections n. of packets avg. conn. time (mm.ss) packets/hour net packets/hour Bytes/hour net Bytes/hour Table 1: most relevant data of the transmission process 6

7 The statistics on connection time show that on average 70% of time is available for connection, which means 42 minutes per hour. This value is of course on average, and there are relevant variations from day to day, as reported in Figure 8. time (hours) Average hourly connection 0,85 0,8 0,75 0,7 0,65 0,6 0,55 0, session number Figure 8: time available for connection As far as transmission delay is concerned, i.e. time from packet transmission to receipt from the ground terminal, Figure 9 shows that also for this parameter there are consistent differences from packet to packet, with an average delay of about 7 minutes. It is interesting to remark that 90% of messages, whose length is at most 229 bytes, reach the user within 10 minutes, and 30% within 1 minute. Delay (h.mm.ss) Number of packets oct-01 oct 8_1 oct 8_2 oct 9_1 oct 9_2 oct-15 oct-16 oct-17 oct-18 oct-21 oct-22 oct-10 Figure 9: transmission delay 7

8 It must be pointed out that the system is slowed down by the polling phases required by the transmission software, phases in which the terminal must check the transmitter status to verify if new messages can be sent. Without this phase the transmission is much faster, and it has been verified that, loading the transmitter with a number of messages that do not saturate its memory, reception on the ground terminal is achieved after a few seconds, with occasional peaks of a few minutes. It must also be considered that occasionally Orbcomm satellites get saturated, due to concurrent transmission by many subscribers, and this of course introduces further delay. Figure 10 reports the effective time of reception of the single packets sent. It is interesting to notice that, regardless of the differences in the number of packets sent in each session, the frequency of their reception is almost constant. In fact, approximating each curve with a straight line, the angular coefficient would be more or less the same for all curves, showing a constant reception ratio. packet time of reception ott 8_1 ott 8_2 ott 9_1 ott 9_2 ott-10 ott-15 ott-16 ott-17 ott-18 ott-21 ott-22 Figure 10: effective time of reception of the single packets sent In summary, considering the average transmission ratio, it can be stated that a 50 kb image will be transmitted in 330 packets, and received on ground in 23 hours of connection, that correspond to about 33 hours of operation, since the net transmission time is on average 70% of total time. In the best situation occurred the same transmission would require 18 hours of operations in total, but in the worst situation the operations time required would be 50 hours. Conclusions The analysis and test completed up to now have shown that the use of Orbcomm is viable for cheap and small missions, as a substitute of a dedicated transmission system. The main advantage of the system is that, with a rather simple interface software, and using off-the-shelf components, a reliable system for low data rate transmission is proposed. The low data rate, of course, prevents the system from being used for real time operations, but this is usually not required in microsatellites. Yet to be verified is the in-orbit performance, where Doppler effects, antennae angular velocity, and lack of coverage could degrade performances. In fact, all tests have been performed from a static test facility; still to be investigated is the system response in order to understand the capability to run also in the very first phases of the mission. To this end, an experiment 8

9 based on the concepts shown in the present paper is under way on a technology demonstration mission, Rubin-2 [6], launched in December The hope is obviously to have soon a consistent set of data to verify the in-orbit performances of the system. Acknowledgments The authors would like to acknowledge the help provided by Carlo Gavazzi Space, that made available the hardware required for the experimental tests. References 1) F.Bernelli-Zazzera, C.Daeder, M.Molina, Image acquisition system on microsatellite Palamede, XVI Congresso Nazionale AIDAA, Palermo, ) M. Salvetti, Progetto del sistema di telecomunicazioni di bordo del microsatellite Palamede, M.S. thesis (in italian), Politecnico di Milano, ) R. Lombardo, Analisi e progetto di un sistema di radiocomunicazione per il satellite di classe LEO, Palamede, M.S. thesis (in italian), Politecnico di Milano, ) Unknown authors, Orbcomm system overview, ) A. Conti, A. Corsini, M. Vaglini, 6) F. Bernelli Zazzera, A. Ercoli Finzi, M. Molina, M. Cattaneo, M. Dioli, I. Bertolini, R. Bianchi, P. Sabatini, L. Crocco, F. Schiavi, A. Zucconi, In-Orbit Technology Validation for a University Microsatellite, 4 th IAA International Symposium on Small Satellites for Earth Observation, Berlin, April 2003, pp