Skip to main content

Full text of "High resolution sonar concept formulation"

See other formats


NPS-62-79-012PR 



FEDDOCS 
D 208.14/2: 

N PS-62-79-01 2PR 



LltSKAK ( 

RESEARCH REPORTS DIVISION 

naval postgraduate school 

MONTEREY, CALIFORNIA 93940 



NAVAL 



POSTGRADUATE SCHOOL 

Monterey, California 




H-EGH RESOLUTION SONAR CONCEPT FORMULATION 



George L. Sackman 
October 1979 



Approved for public release; distribution unlimited 
Prepared for: Coastal Technology Department 

Naval Coastal Systems Center 
Panama City, FL 32407 



DUDLEY KNOX LIBRARY 
NAVAL POSTGRADUATE SCHOOL 

MONTEREY, CA 93943-5101 ^^"T^SCHOOL 



Rear Admiral Tyler Dedman Jack R. Bo rs ting 

Superintendent Provost 



The work reported herein was supported by the Coastal Technology 
Department, Naval Coastal Systems Center, Panama City, FL 32407. 

Reproduction of all or part of this report is authorized. 

This report was prepared by: 



SECURITY CLASSIFICATION OF THIS PAGE ( When Data Entered) 



REPORT DOCUMENTATION PAGE 


READ INSTRUCTIONS 
BEFORE COMPLETING FORM 


1. REPORT NUMBER 

NPS-62-79-012PR 


2. GOVT ACCESSION NO. 


3. RECIPIENT'S CATALOG NUMBER 


4. TITLE ( and Subtitle) 

High Resolution Sonar Concept Formulation 


S. TYPE OF REPORT 4 PERIOD COVERED 

Project Report 


6. PERFORMING ORG. REPORT NUMBER 


7. authorc*; 

George L. Sackman 


8. CONTRACT OR GRANT NUMBER^ 


9 PERFORMING ORGANIZATION NAME AND AOORESS 

1 Naval Postgraduate School 
Monterey, California 939^*0 


10. PROGRAM ELEMENT. PROJECT, TASK 
. AREA 4 WORK UNIT NUMBERS 

6271 1 N; SF1 1 121 491 ;20297~09 
M61331-79-WR-90113 


11. CONTROLLING OFFICE NAME AND AOORESS 

Coastal Technology Dept. 

Naval Coastal Systems Center 

Panama P i 1-y PI 7 . .. 


12. REPORT DATE 

October 1979 


13. NUMBER OF PAGES 

17 


14. MONITORING- Agency iTAME i ADDRESS^// different from Controlling Office) 


IS. SECURITY CLASS, (ot thia report) 

Unclassified 


15 a. DECLASSI FI CATION/ DOWN GRADING 
SCH EOULE 


16. DISTRIBUTION STATEMENT (of this Report) 

Approved for public release; distribution unlimited 


17. DISTRIBUTION STATEMENT (of the abatract entered in Block 20, if different from Report) 


18. SUPPLEMENTARY NOTES 


19- KEY WORDS (Continue on reverae aide If neceaaary and Identify by block number) 




High Resolution Sonar 


20. ABSTRACT (Continue on reverse aide it neceeaary and Identify by block number) 

An assessment is made of the impact of current technological developments on 
future research in high resolution sonar. The philosophical approach is 
from the point of view of examining the rate of information flow at each stage 
through the system. It is concluded that large computer memories under 
microprocessor control and fiber optic data 1 i nlcs can be fruitfully applied 
in uture system arch i tecture. In addition, the necessity for further 
research in precision navigation systems and pattern recognition algorithms 



DD t J AN ^73 1473 EDITION OF 1 NOV 65 IS OBSOLETE 

S/N 0102-014- 6601 | UNCLASS 1 F I £ fl 

SECURITY CLASSIFICATION OF THIS PAGE (When Data Entered) 



Unci ass i f i ed 

wLLIJHITY CLASSIFICATION OF THIS PAGEfH'hgn Data Entered) 

#20 became apparent, in order to achieve reliable classification of 
underwater objects along with high area search rates. 



Unci ass i f i ed 



SECURITY CLASSIFICATION OF THIS PAGEfWien Data Eru 



I INTRODUCTION 



TABLE OF CONTENTS 



~ us&KAHr 

NAVAL POSTGRADUATE SCHnni 
MONTEREY CA 9394^510? 



II SYSTEM EXAMPLE: SIDE-LOOKING SONAR 

A. Resolution and search rate requirements 

B. Input information rate 

C. Output information rate 

III INFORMATION RATE COMPRESSION 

IV INFORMATION TRANSMISSION 

V INFORMATION STORAGE 

VI PATTERN CLASSIFICATION 

VII CONCLUSIONS 



2 

2 

3 

5 

6 
7 
9 

11 

13 



REFERENCES 



15 



HIGH RESOLUTION SONAR CONCEPT FORMULATION 



I. INTRODUCTION 



This report is a summary of work done at Naval Coastal Systems 
Center during July and August 1979 in attempting to assess the impact of 
current technological developments on future research in high resolution 
sonar. Rather than an exhaustive survey of relevant technologies, a few 
were selected as being of the highest probability for significant 
improvements in target classification capability along with greater area 
search sweep rate. The philosophical approach is from the point of view 
of examining the rate of information flow through the system at each 
stage. Technologies were evaluated on this basis and it was concluded 
that three are dominant — large computer memories, microprocessor control, 
and fiber optic data links. 

A side-looking sonar example is used to illustrate how these 
technologies might be incorporated into a future system. Many of the 
concepts considered in this report have already been incorporated in 
developmental or proposed systems at NCSL and elsewhere. This is to be 
expected and should simply be a confirmation that the limitations of 
physics will often lead research to similar trends while advances in 
technology determine the rate of progress. 



1 



II. SYSTEM EXAMPLE: SIDE -LOOKING SONAR 



A. RESOLUTION AND SEARCH RATE REQUIREMENTS 

The geometry of a side-looking sonar is constrained by the linear 
aperture oriented along the track of the support platform. Along- track- 
resolution is related to effective aperture size, while range resolution 
is related to signal bandwidth. Therefore both aperture and bandwidth 
must be maximized within the physical constraints to achieve the highest 
possible resolution. Practical limitations on physical aperture size 
provide motivation for synthetic aperture development, and the search for 
bandwidth leads to development of new signal waveforms and wideband 
transducers. Meanwhile, tactical considerations motivate efforts to 
increase area search rate, which implies increasing the sonar range capa- 
bility in spite of the limits imposed by the velocity of propagation of 
sound. A variety of clever schemes (Ref. 1,2) have been proposed to meet 
these goals generally by application of parallel processing in time, 
space, and frequency. 

If we assume that somehow larger apertures will be synthesized, 
wider bandwidth signals will be accomodated, and parallel processing will 
be employed, the rate of information flow into the system will be phenom- 
enal. Even with the resolution and search rates already obtainable, the 
rate of information flow is so high that the system capacity is being 
overloaded at every stage, from the preamplifiers to the operator. 



2 



In order to quantify the problem as to relative magnitude, consider 
a side-looking sonar with constant resolution cell size at all ranges of 
10 x 10 cm. This cell size is representative of the linear resolution 
required for reliable classification (on the order of 1/10 of the object 
size), as reported in several studies (Ref. 3). Synthetic aperture tech- 
niques inherently provide constant resolution cell size because the 
effective aperture length is proportional to range. Time variable focus 
of a physical aperture can also approach constant resolution. 

If an area coverage rate of one square nautical mile per hour is 
taken as a benchmark which should be equaled or exceeded, the total num- 
ber of resolution cells covered per second is approximately 

2 

2 km x 2 km x 100 cells/m ^ , . 

3 ^ 00 sec/hr * '0 cell s/second 

B. INPUT INFORMATION RATE 

If the reflected power from each cell can be distinguished to eight 
"shades of grey", (or equivalently the signal/reverberation ratio is 
9 dB) , then each cell represents approximately three "bits" of informa- 
tion. 1 The information rate R^ at this point in the system is there- 
fore on the order of 3 x 10^ bits/sec. 

1 Information can be measured quantitatively in terms of log 2 of the 
number of distinguishable states of the data, represented by a binary 
number. The size of the binary number required measured in binary digits 
("bits") gives the information content of the data. See Goldman (Ref. 4) 



3 



Another way of looking at the input information rate is to consider 
a linear aperture of length L. If the system bandwidth is W(Hz), inde- 



pendent samples of the signal can be measured approximately every 1/W 
seconds, or equivalently every c/W meters in space (where c = sound 
velocity). Therefore the aperture can gather LW/ c independent samples of 
the wavefront at any instant of time. The number of distinguishable 
states is approximately S + N, hence the information rate is 



Note that the information rate varies linearly with the aperture length, 
but only as the logarithm of the S/N ratio, while it is proportional to 
the square of the bandwidth. 

Furthermore, the input information rate to the aperture R a will be 
greater than the rate computed above for the beamformer output, R^. 

Even though an ideal beamformer neither creates nor destroys information 
(in the simplest case it can be considered as performing a linear 
Fourier Transform on the aperture data) in practice there is certain to 
be some loss of information in the beamforming process. Therefore, the 
rate at the aperture will be even greater than at the beamformer output 
where resolution cells contain the information. 



N 




R a = log 2 (1 + S/N) 




S + N 



4 



C. OUTPUT INFORMATION RATE 



At the other extreme, an operator can make only a very modest number 
of reliable classification decisions per unit time. As a guess, the max- 
imum number might be on the order of one decision per minute, each of 
which represents a classification between a small number of object cate- 
gories. Allowing for 16 categories, an information rate for classifica- 
tion R c is 

1og ? 16 

R c ■ 60 sec/m in ' ’ /15 bit/seC 

Therefore, an information rate compression of 15x3x10)5 = 4.5x10)6 
must take place in the system. 



5 



III. INFORMATION RATE COMPRESSION 



Information rate compression by definition implies some sort of 
memory or storage, since the input rate greatly exceeds the output rate 
and information must be "digested*' somewhere. Many bits must be integra- 
ted together to come up with patterns which can be classified. 

In a typical present day system, the compression occurs primarily in 
the eye and brain of the operator, since the information rate on the dis- 
play is virtually of the same order of magnitude as the rate at the aper- 
ture. Essentially raw data is delivered in great quantity to the opera- 
tor, who must bear the entire burden of sorting, rejecting, selecting and 
classifying it. There is a fundamental theorem of information theory by 
Shannon (Ref. 5) that each system has a certain "channel capacity" or 
maximum information rate measured in equivalent bits/second. The theorem 
states that if the actual data rate is significantly less than the 
channel capacity, the output decision error rate can be reduced 
arbitrarily by proper processing. Conversely, if the data rate exceeds 
the channel capacity, the probability of decision error increases toward 
unity very rapidly. Even though this theorem was derived for very 
precisely defined conditions, the basic idea most certainly applies in 
general, including to human operators. Therefore it should be clear that 
some sort of information rate compression is absolutely necessary in a 
high/resolution sonar system, which in turn implies the need for a large 
memory. 



6 



IV. INFORMATION TRANSMISSION 



Since the information rate at the aperture is on the order of 10^ 
bits/sec., there is a question of architecture as to how this processing 
of data will be distributed throughout the system. Locating the memory 
and its attendant support in an underwater body where space is at a pre- 
mium does not seem desirable. On the other hand, if the memory is loca- 
ted remotely from the aperture, there is a corresponding need for high 
data rate transmission between aperture and memory. It is conceivable, 
for example, to connect the preamplifiers directly to A/D converters and 
transmit the transducer data from the aperture (which must be at some 
depth in the water) to the beamformers which could be aboard a surface 
platform. In fact such a scheme has been used in low frequency passive 
towed arrays where the data rate is very low compared to high resolution 
sonar. However, at the present stage of technology it appears to be more 
feasible to do the beamforming in the vicinity of the aperture. The 
higher frequencies associated with high resolution sonar discourage di- 
rect A/D conversion, even though the system bandwidth may be an order of 
magnitude less than the carrier frequency. The short wavelength of the 
high carrier frequency typically used is required to achieve the desired 
angular resolution, essentially by coherently processing the phase infor- 
mation in the wavefronts. Another approach has been proposed (ref. 2, 
op.cit.) using wideband signals to obtain bearing information from 
doppler time compression, using a lower frequency. 

Even after beamforming, however, the information rate is quite high 
as discussed in the example of the side-looking sonar. However, at this 



7 



point it becomes worthwhile to consider transmitting the data from the 
underwater body to a topside processor* The primary motivation for this 
is to reduce the size and expense of the underwater portion of the system 
since reduced size facilitates handling, and reduced expense makes loss 
or damage to the most vulnerable portion of the system less costly. It 
might eventually be possible to make the underwater portion of the system 
semi-expendable, with recovery being an option in its employment* 

A fiber optic wideband link between the underwater body and the top- 
side processor would be a promising application of new technology to high 
resolution sonar. Fiber optics offers the potential of low cost, low 
loss, and redundant and/or parallel data transmission in a very small 
diameter cable (Ref. 6). It is even possible to consider spooling the 
cable at both ends as is done with XBT wire and expending it along with 
the underwater portion of the system, rather than dragging it through the 
water. 



8 



V. INFORMATION STORAGE 



At a rate on the order of 10^ bits/sec., it is obvious that some 
buffering will be required at both ends of the data link. Some pre- 
processing could also be done before transmitting the data up the link. 
However, it looks more attractive to the author to concentrate most of 
the processing topside. Development of the system and later modifica- 
tions to processing based on operation would be easier to incorporate, 
and size and weight and cost of the underwater body could be minimized. 

Assuming that 10^ bits/sec. is flowing into the memory, some pre- 
liminary arithmetic and logic operations could be performed by microproc- 
essor devices to begin thresholding and batching the data for the first 
steps in rate compression. At this stage, most of the reverberation 
could be eliminated, and only portions of the field of view containing 
clusters of target-like features would be retained. It seems necessary 
that multiple looks at candidate clusters in both time and aspect angle 
will be required by any conceivable system to provide enough clues for 
classification (Ref. 7). For this reason, regions of memory should prob- 
ably be organized on the basis of addresses being assigned to specific 
regions of physical space in the field of view, so that as more data 
about each region is accumulated it would be associated with previous 
data on the same region. This memory organization has been found to be 
essential in medical ultrasonics in order to provide image quality neces- 
sary for diagnosis, etc. In medical systems, the aperture location data 
is obtained mechanically from encoders connected to the linkage support- 
ing the scanning transducer (Ref. 8). This requirement is closely 



9 



related to the navigational data necessary for forming a synthetic aper- 
ture, hence any methods applicable to one problem apply more or less 
directly to the other. That is, the essential requirement is to know 
exactly where the aperture is located with respect to inertial space. 

This requirement appears to the author as the most crucial problem to be 
solved in order to achieve significant improvement in high resolution 
sonar. The requirements on technology are analyzed in a recent technical 
report by Griffith, et al. from The Analytical Sciences Corp. (Ref. 9). 

In order to reconstruct the image field, the aperture field must be known 
precisely over long distances. 

It seems that some fundamental research concentrated on improving 
navigational resolution for this application is called for as soon as 
possible. Inertial systems, acoustic doppler correction techniques 
and/or other alternatives such as laser doppler velocimeters (Ref. 10) 
might be pursued as candidate systems. 



10 



VI. PATTERN CLASSIFICATION 



It should be recognized at the outset that only a limited number of 
categories of object can be classified in the available time in typical 
search scenarios. Therefore, the objective of the data analysis is to 
extract sufficient clues for classification as "rapidly as possible. 

These clues will be in the form of measurable parameters, such as rela- 
tive level of energy reflection from each resolvable element in a candi- 
date cluster. The reflected energy will be dependent upon aspect angle 
and coherent interference phenomena (speckle, scintillation, glint) which 
must be processed by suitable techniques. With sufficient computer capa- 
bility, statistical measures of the image features can be formed, calcu- 
lating the relative probability of occurrence of echoes of each strength. 
Use of this data to generate a suitable amplitude transfer function 
before thresholding provides a maximum contrast image of the field of 
view, enhancing image features (Ref. 11). For example, maximum and mini- 
mum values occuring in patterns such as highlights and shadows can be 
compared to time averages and spatial averages. Only clusters meeting 
certain criteria would be validated for display, in order to avoid over- 
loading the operator. 

Interactive graphics with the operator in the loop is desirable to 
keep the data flow to the operator at a rate which he can process. A 
family of patterns, any one of which could be classified as the same 
object, could be stored in ROM in microprocessors. Clusters falling into 
any of these patterns could be presented to the operator in an alphanu- 
meric or symbolic code. Only after the operator selects a particular 



11 



symbol with a cursor would the entire cluster be presented to him for 
examination. It is not proposed that automatic classification would do 
the whole job. There is too much potential variety of patterns in high 
resolution sonar for an automatic classifier to deal with them all. How- 
ever, it is also impossible for an operator to examine the entire field 
of view and do all the pattern classification by himself. Therefore, a 
large computer memory is required to store the data, and a large number 
of microprocessor operations are required to accomplish the sorting and 
thresholding. 

Fortunately, technology is advancing rapidly in this direction, as 
evidenced by electronic TV games and home computers which incorporate 
ROM's and interactive graphics. However, measures of "goodness” for pat- 
tern recognition cannot be obtained very consistently using human opera- 
tors, because of difficulty in modeling the psychophysical phenomenon of 
perception (Ref. 12). Statistics gathered by testing humans is typically 
very unsatisfactory, with laboratory results not transferring well to 
operating conditions. On the other hand, the application of automatic 
pattern recognition theory to typical high resolution sonar images will 
require considerable effort to bring it to practical application. Multi- 
ple aspect data must be integrated in some fashion because the data at a 
single look and single aspect is insufficient to support the number of 
object classification categories that will be required. The need here 
does not seem to be for hardware technology as much as for software 
(algorithms) validated by application to a realistic data base. 



12 



VII 



CONCLUSIONS 



This analysis has shown that technological developments in computer 
memory, microprocessor control, and fiber optic transmission of large 
amounts of information can be fruitfully applied to high resolution sonar 
system architecture as sketched in Figure 1. However, in the course of 
this study, the key role of technology in high precision navigation 
systems became apparent. In order to accumulate sufficient information 
about underwater objects to achieve reliable classification, aperture 
data must be processed in multiple beams with variable focus, as sketched 
in Figure 2. It is essential that the spatial relationship of the aper- 
ture to an inertial reference be maintained precisely. Presently avail- 
able inertial systems probably fall short of the requirement since they 
are designed for other applications. Update corrections by acoustic 
doppler or perhaps other means such as optical doppler using a rangegated 
or dual-beam laser might be considered for this problem (if the under- 
water platform is operated close enough to the bottom to allow the laser 
to overcome the optical attenuation of the water). In any case, the 
status of technology is unevenly developed in its readiness for applica- 
tion to high resolution sonar. Memories and microprocessors are here 
today, and fiber optics tomorrow, but high resolution navigation lags 
behind and seems to be a necessity to capitalize on the other develop- 
ments. 



Pattern recognition algorithms also must be developed in order to be 
applied to high resolution sonar images. Data and techniques from radar 
and optics may not be directly applicable because of differences in back- 



13 



ground and object characteristics. Very few resolvable elements or clas- 
sification clues are available for typical objects of interest, so that 
the scene must be analyzed for several time intervals and several aspect 
angles to increase the number of clues. This returns the issue to the 
precision navigation system, which appears to be the most severe problem 
to be overcome in order to achieve the stated goals. 



14 



REFERENCES 



1. Loggins, C.D., F.J. Higgins, & J.T. Christoff, "Synthetic Aperture 
Side-looking Sonar (U)", Journal of Underwater Acoustics, Vol. 24, No. 4, 
October 1974, CONFIDENTIAL. 

2. Skinner, D.P. , "Doppler Azimuth Discrimination", NCSC Technical Note 
TN 464, August 1978, revised November 1978, Naval Coastal Systems Center, 
Panama City, FL 32407. 

3. Duda, R.O. , & P.E. Hart, Pattern Classification and Scene Analysis , 
Wiley Interscience, 1973. 

4. Goldman, S. Information Theory , Prentice-Hall 1953. 

5. Shannon, C.E. , "Communication in the Presence of Noise", Proc. 

I.R.E., Vol. 37, pp 10-21, January 1949. 

6. Eastley, R.A. , & W.H. Putnam, "Fiber-optic Components for an Optical 
Data Link to Interconnect the Submerged Hydrophone Package of a 5 Kilo- 
meter Depth Sonobuoy",N0SC TR432, 27 Mar 79, Naval Ocean Systems Center, 
San Diego, CA 92152. 

7. Freedman, A., "The High Frequency Echo Structure of Some Simple Body 
Shapes", Acustica Vol. 12, pp 10-21, 1962, also published in Underwater 
Sound , V.M. Albers, Editor, Dowden, Hutchinson & Ross 1972. 

8. McGinness, M.G. , "Methods and Terminology for Diagnostic Ultrasound 
Imaging Systems", Proc. IEEE, Vol. 67, No. 4, pp 641-653, April 1979. 

9. Griffith, E.W., E.M. Geyer, & M.A. Chory , "Multifrequency Vernier 
SA/SLS Image Quality Evaluation and Motion Sensing System Error Budget 
(U)", TR 1269-2, The Analytic Sciences Corporation, 18 Jul 79, 
CONFIDENTIAL. 

10. a) Kroeger, R.D. , "Motion Sensing by Optical Heterodyne Doppler 
Detection from Diffuse Surfaces", Proc. IEEE, (Correspondence) Vol. 53, 
pp 211-212 1965. 

b) Durrani, T.S. & C.A. Greated, "Theory of Laser Doppler Velocity 
Tracking", IEEE Trans. AES Vol. 10, No. 4 1974. 

c) Durst, F. Principles and Practice of Laser Doppler Anemometry 
Academic Press 1976. 

11. Andrews, H.C. & B.R. Hunt, Digital Image Restoration, Prentice-Hall 
1977. 

12. Aschenbrenner , C.M. , "Problems in Getting Information Into and Out 
of Air Photographs", Photogramm. Engr., Vol. 20, No. 3, pp 398-401, 1954. 



15 



EXAMPLE : SYSTEM ARCHITECTURE 



> 

APERTURE 






SIGNAL 

CONDITIONING 

adc/mux 




FIBER OPTIC 
LINK 






THRESHOLD 




OPERATOR 



FIGURE 1--A computer memory is shown comb i ned- wi th a fiber optic data 
link to the aperture and interactive graphics output to the operator. 
The memory is organized on the basis of coordinates x,y in inertial 
space. 



16 



EXAMPLE: SHADOWGRAPH OR SYNTHETIC APERTURE 




UNIFORM RESOLUTION 
CELL SIZE 



VARIABLE FOCUS/ VARIABLE SIZE APERTURE 




SUB-APERTURES/ MULTIPLE ASPECT 




MULTIPLE PRE-FORMED BEAMS 



FIGURE 2--The aperture Is shown maintaining constant resolution 
by size and focus with range (echo return time). Additional 
target classification information is obtained by multiple 
aspect, multiple echo beamforming. 



17 



Distribution List 



Naval Coastal Systems Center 10 

Panama City, FL 32407 

ATTN: Dr. M.J. Wynn, Code 790 

Commander, Mine Warfare Command 1 

Naval Base 

Charleston, SC 29408 

ATTN: Mr. D.L. Folds, Code 006 

Dudley Knox Library 2 

Naval Postgraduate School 
Monterey, CA 93940 

Office of Research Administration 1 

Naval Postgraduate School 
Monterey, CA 93940 

Chairman, Department of Electrical Engineering 1 

Naval Postgraduate School 
Monterey, CA 93940 

Professor George L. Sackman 2 

Department of Electrical Engineering 
Naval Posgraduate School 
Monterey, CA 93940 



1)189649 



• 8 * 4.9 I* 

/U