`
`3,069,654
`METHOD AND MEANS FOR RECOGNIZING COMPLEX PATTERNS
`
`P. v. c. HouGH
`
`Filed March 25. 1960
`
`2 Sheets-Sheet l
`
`INVENTOR.
`,Paal M C.' Ho zíyff:
`BY
`
`ß’??afßçy
`
`Magna 2044
`TRW v. Magna
`IPR2015-00436
`
`0001
`
`
`
`Dec. 18, 1962
`
`3,069,654
`P. v. c. HoUGH
`METHOD AND MEANS FOR RECOGNIZING COMPLEX PATTERNS
`
`Filed March 25. 1960
`
`2 Sheets-Sheet 2
`
`0002
`
`
`
`3,069,654
`NETHOD AND MEANS FOR RECOGNIZÍNG
`COMPLEX PATTERNS ’
`Paul V. C. Hough, Ann Arbor, Mich., assigner to the
`United States of America as represented by the United
`States Atomic Energy Commission
`Filed Mar. 25, 1960, Ser. No. 17,715
`6 Claims. (Cl. S40-146.3)
`
`jam.
`
`3,009,654
`Patented Dec. 18, 1962
`
`2
`of the point on the line segment from the horizontal mid
`line 109 of the framelet 108.
`(3) Each line in the transformed plane is made to have ï
`an intercept with the horizontal midline 101 of the pic
`ture 100 equal to the horizontal coordinate of its respec
`tive point on the Vline segment in framelet 108.
`(cid:173) Thus, for a given reference point 110 on line segment
`102 a line 110A is drawn in the plane transform 102A.
`The reference point 110 is approximately midway between
`the top and the horizontal midline 109 of framelet 108
`and hence the line 110A is inclined to the right at an
`angle to the vertical whose tangent is approximately 1/2.
`The intersection of the line 110A with the horizontal mid
`line 101 of picture 100 is at a distance from the left
`edge of the picture 100 equal to the horizontal coordinate
`of the point 110 on line segment 102.
`It is an exact theorem that, if a series of points in a
`framelet lie on a straight line, the corresponding lines in
`the plane transform intersect in a point which we shall
`designate as a knot 112. It is therefore readily apparent
`that the rectangular coordinates of the knots 112 in
`picture '
`100 have the following properties:
`(l) The horizontal coordinates of the knots 112 equall
`the horizontal coordinates in the framelet 108 at whichv
`the straight line segments 102, 104 and 106 intercept the
`horizontal midline 109 of the framelet 108. `
`.
`(2) The vertical coordinate of the knots 112, relativel
`to the horizontal midline 101 of picture 100, is propor
`tional to the tangent of the angle of the straight line seg
`ments 102, 104 and 106 relative to the vertical.
`Thus,
`the coordinates of the knots 112 in the plane transformsy
`102A, 104A and 106A give the slopes and intercepts of
`the straight line segments
`102, 104 and 106 in framelet
`
`100.
`
`'
`
`10
`
`15
`
`20
`
`25
`
`30
`
`This invention relates to the recognition of complex
`patterns and more specifically to a method and means for
`machine recognition of complex lines in photographs or
`other pictorial representations.
`This invention is particularly adaptable to the study of
`,subatomic particle track-s passing through a viewing ñeld.
`As the objects to be studied in modern physics become
`smaller„the problem of observing these objects becomes
`increasingly more complex. One of the more useful de
`vices in observing charged particles is the bubble chamber
`wherein the charged particles create tracks along their path
`of travel composed of small bubbles approximately 0.01
`inch apart,(cid:173)’depending upon the specific ionization of the
`initiatingparticle. These tracks form complex patterns
`and ’are readily photographed with the use of a dark back
`ground. With this device, multitudinous photographs
`are produced with each photograph requiring several hours
`study by a trained observer to recognize the complex pat
`terns of the tracks. It is therefore readily apparent, that
`as the photographs increase in number, the time consumed
`by a trained observer to study them becomes excessive
`and, unless large numbers of trained observers are used,
`the reduction of data falls far behind the production rate.
`It is one object of this invention to provide a method and
`means for the recognition of complex patterns in a pic
`ture.
`It is another object of this invention to provide an irn
`proved method and means for recognizing particle tracks
`in pictures obtained from a bubble chamber.
`In general, the objects of this invention are accom
`plished by dividing the viewed representation into sufli
`ciently small sectors or framelets that the complex pattern
`is divided into substantially straight line segments. Each
`of the segments is detected and transformed into slope and
`intercept data which may be stored and later analyzed for
`the presence of desired patterns.
`. A more complete understanding of the invention will
`best be obtained from consideration of the accompanying
`drawings in which:
`FIG. l is an illustration of a plane transform repre
`sentation of straight line segments;
`PIG. 2 is a block diagram of an apparatus according to
`teachings of the present invention; and
`(cid:173)
`FIG. 3. is a detailed block diagram illustrating the elec-V
`tronic plane transform circuits of the apparatus in the
`embodiment of the present invention, shown in FIG. 2.
`55
`A geometric construction by hand is shown in FIGURE
`l(cid:173) which depicts three straight line segments 102, 104 and
`106 in a framelet 10S and their corresponding sketched
`plane transforms 102A, 104A, and 106A in picture 100.
`The geometry of construction for the plane transforms
`v60
`is accomplished according’to the following rules.
`_ (l) For a given point on a line segment in framelet
`4108, a line is drawn in the transformed plane in picture
`100.
`(2) For a point on the line at the top of the framelet
`108, the line in the transformed plane is inclined 45°
`to the right; a point on the line segment at the horizontal
`midline of the framelet 108 gives a vertical line in the
`plane transform; a pointon the line segment at the bottom'
`of-the framelet 108 gives a line in the transformed plane
`inclined at 45?V to the left. In general, the line in the
`transformed plane has' an angle relative to the vertical
`whose tangent is proportional to the vertical displacement
`
`50
`
`35
`
`40
`
`Although the foregoing description pertained to a hand`
`construction of a plane transform, it is to be understood:
`that it may be performed by adequate electronic apparatus
`or the like.
`>
`In FIG. 2, the picture containing the complex pattern",~
`such as from a photograph of a bubble chamber, is sub-A
`divided into several hundred rectangular areas or frame'
`lets. The height of each framelet is chosen small enough
`so that the portions of the pattern within each framelet
`is essentially a straight line and large enough so that the'
`line segments can be reliably distinguished from the ran-
`dom bubble background. The Width of the frameletl is(cid:173)_
`dependent upon the accuracy needed in the measurementV
`of the lateral position of the segments in the framelet. ,l
`- A television camera 210, such as of the image orthicon
`type, scans the framelet 212 containing one or more
`straight line segments composed of bubbles.
`As the scarta'.
`ning beam of the television camera 210 passes over (cid:173)av
`bubble in the line segment, the televísion‘camera 210 pro-
`drces an output pulse. For each output pulse from the
`television camera 210, electronic plane transform circuits
`214 cause a line to be drawn in a plane transform on a
`display of an oscilloscope 216 according to the geometric
`rules described for FIG. l. Thus a plane transform of
`the line segment of framelet 212 is created. The coordi
`nates of the knot in the plane transform on the display of
`oscilloscope 216 gives the slope and intercept of thefline
`segment in framelet 212 as previously sho-wn in FIG. l. ,
`A second television camera 21S, such as of the image
`'orthicon type, scans the plane transform display of os
`65
`cilloscope 216 and detects the knot with its relative co->
`ordinate data. The output of the second television cam
`era 21S containing the coordinate data of the knot is
`fed to magnetic tape recorder '220 and stored thereon.
`The magnetic tape is then fed into a computer 221, such
`as of the IBM704 type, where the coordinate data of each
`line segment is evaluated to recognize the original complex
`pattern in the picture.
`
`0003
`
`
`
`3,069,654
`
`10
`
`15
`
`20
`
`25
`
`35
`
`40
`
`3
`When a standard image orthicon television camera scans
`a. bubble chamber scene, the bubbles appear in the scan
`line as narrow regions where the video output voltage is
`much less than the background voltage on each side. The
`backgroundrvideo signal also shows considerable variation,
`and so a means must be provided for recognizing bubbles
`in a varying background, and for discriminating against
`various unwantedmarkings in the scene. A video pulse
`must satisfy two basic ’criteria to be admitted as corre
`sponding to a bubble. These are: (a) A narrowness
`criterion. The bubbles making up a track have a narrow
`and relatively constant width. Therefore, only video
`pulses of this width (within a certain tolerance) are ad
`mitted. Wider opaque regions in the scene are ignored.
`(b) A contrast threshold. The difference in light in
`tensity between the dark track and the lighter background
`on each side must be greater than a certain minimum
`value. This threshold is a parameter of the system which
`is easily adjusted. It is set to give the most reliable track
`detection and highest background rejection for any particu
`lar groups of pictures.
`Reference is now made to FIG. 3 for a detailed explana
`tion of the circuits 214 wherein the pulses from the tele
`vision camera 210 representing bubbles in the line seg
`vment inthe viewed scene are converted into the more
`useable plane transform pattern. For the purposes of
`clarity, only one detected bubble on the line segment of
`the framelet 212 will be treated although the treatment of
`allzother detected bubbles is the same.
`The video signal from the first television camera» 210
`is presented undelayed to a first input of a difference am
`plifier 222 and also delayed 0.4 microsecond to a second.
`input ofthe difference amplifier 222. The difference in
`amplitude between the two outputs of the difference am-(cid:173)
`plifier 222 represent the difference in light level at two
`points along the scan line of the first television camera
`210 separated by half the width of a bubble in the line
`segment of framelet 212. The output from the difference
`amplifier 222 corresponding to the 0.4 microsecond input
`is yfed through a 0.1 microsecond delay line to a first in
`put of a Garwin coincidence circuit 224. The other out
`put of the difference amplifier is delayed approximately
`.5 microsecond to the other input of the Garwin circuit
`soy that the two signals arrive at the coincidence circuitv
`simultaneously. Any opacity greater than twice the width
`ofthe bubble in the line segment of framelet 212 fails to
`trigger the Garwin circuit 224 and is therefore ignored.
`The output pulse amplitude of the Garwin coincidence
`circuit 224 will depend upon the difference in light inten
`sity between the bubble in the line segment and the gen
`eral background. Smaller output pulses from the Gar
`win coincidence circuit 224 will be present due to varia
`tions in intensity of the general background. These are
`eliminated by feeding the output of the Garwin coinci
`dence circuit 224 to a 0.5 microsecond monostable multi
`vibrator 226 where the bias of the trigger is set so that
`only pulses from the bubbles in the line segment of-frame
`let 212 have sufiicient amplitude to trigger the multivi
`brator 226. Thus, a single pulse output is obtained from
`the multivibrator 226 when the scanning beam of the first
`television camera 210 passes over the bubble in the line
`segment of framelet 212.
`The output pulseof the multivibrator 226 triggers a 2 '
`microsecond monostable multivibrator 228. One output
`from the monostable multivibrator 22S drives an un
`blanking amplifier 230 in which the pulse is delayed 0.3
`microsecond, clipped to a length of 1.4 microsecond, am
`plified, and applied to the cathode of the cathode ray tube
`in the oscilloscope 216. This pulse thus turns on'the
`beam of the cathode ray tube of the oscilloscope 216 for
`a period of time, approximately 1.5 microsecond, sufii
`cient to draw the line transform corresponding to the
`bubble scanned. The second output from the monostable
`multivibrator 228 drives aclipper 232 which provides a
`
`4
`0.3 microsecond pulse output at the leading edge of the
`output pulse of the monostable multivibrator 228.
`The output from the clipper 232 is fed to a set pulse
`ampliñer 234 where it is amplified and provides a 0.3
`microsecond pulse of fixed voltage, 15 volts, which is ap
`plied to the fixed line generator 236. A 2 microsecond
`output pulse is also derivedr from the clipper 232 which is
`identical to the 2 microsecond output pulse of the mono- Y
`stable multivibrator 228. This 2 microsecond output pulse
`from the clipper 232 is fed to a reset amplifier 238 Where
`it is amplified and inverted. VBoth the inverted 2 micro
`second pulse from the reset amplifier and the l5 volt out
`put pulse from the set pulse amplifier are fed simul
`taneously to the fixed line generator 236. The 15 volt
`output pulse applied to the fixed line generator 236 is
`caused to decay therein at a predetermined linear rate of '
`decay to ~(cid:173)(cid:173)15
`volts. The 2 microsecond inverted pulseY
`from the reset amplifier 238 gates the decay of the 15
`volt pulse-from the set pulse amplifier 234 and causesr it
`to be clamped at -l5 volts. The resulting 2 microsec
`ond linear decay waveform output from the fixed line
`generator 236 is amplified by the amplitierf239. and then
`applied to thevertical deñection plates-of the oscilloscope '
`216.
``
`The-0.3 microsecond pulse from clipper 232 is also fed
`to a set pulse modulator-amplifier 240 where it- is modu
`lated. The modulation is provided' by a verticalfsawtooth-V
`generator 242‘which is` synchronized with the vertical-ï
`defiection of television camera 210. The modulation‘is
`such that when the Vertical >defiection of television cam->
`30
`l era 210 is at the top of the television field,v the amplitude
`of the 0.3 microsecond pulse is 50 volts and the amplitude (cid:173)
`of the pulse drops linearly to 10 volts when the vertical
`deflection of the television camera 210 is'at the bottom ofy
`the television field. The 0.3 microsecond set'pulse from
`the set pulse modulator-amplifier 240 is fed to a variable`
`line generator 244. There, the variable amplitude of the
`set- pulse is set to 25 volts for the time whenthe vertical:
`deflection of the television camera210 is at the top of theî
`television field and 5 volts when the vertical 'deflection is
`at the bottom of the television field, intermediate points'
`decaying linearly thereto. The variable line generator
`244 causes the set pulse from the set pulse modulator
`amplifier 240 to decay therein at a predetermined rate of '
`decay and linear waveform to _25 volts for the vertical
`defiection being at the top of the television field to --5v
`volts for the vertical deflection being-at the bottom of the
`television field. The 2'microsecond inverted pulse from
`the reset amplifier 238 is applied to the variable line gen
`erator 244 simultaneously with the'0.3 microsecond set
`pulse from the set pulse modulator-amplifier 240 and
`gates the set pulse causing it to be clamped 'at the afore(cid:173)`
`mentioned negative voltages. The resulting 2 microsec
`ond variable-amplitude linear-decay output pulse from
`the variable line generator 244 is fedto an input of add
`ing circuit 246.
`The 2 microsecond linear decay pulse from the fixed
`line generator 236 is inverted by an inverting circuit'248
`and fed to an input of adding circuit 246. An output is
`taken from the horizontal deflection circuit of television
`camera 2li), amplified by the horizontal deflection ampli
`fier 250 and applied to an input of adding circuit 246.
`The adding circuit 246 acts on the three'inputs in the
`following manner. ‘If triggered when the vertical deflec-V
`tion of the television camera 210 is at the top of the tele
`vision field, the 2 microsecond output pulse of the variable
`line generator 244 starts at 25 Volts. The 2 microsecond
`inverted pulse of the line generator 236 always starts at
`_l5 volts. The adding circuit 246 sums these two pulses
`into a linear decaying sweep that starts at l0 volts and de
`cays to (cid:173)(cid:173)10 volts. If the 2 microsecond pulse of the vari
`able line generator 244 is triggered at the bottom of the
`television field of television camera’ 210, the result is a
`rising‘linear sweep starting at -10 'volts' and .rising to l0.
`volts. If the 2 microsecond pulse of the variable line
`
`50
`
`55
`
`60
`
`65
`
`70
`
`75
`
`0004
`
`
`
`5 .
`generator 244 is triggered in the center of the television
`field of television camera 210, the 2 microsecond pulse
`of the Variable line generator(cid:173) 244 starts at l5 volts, can
`celling the _l5 volt 2 microsecond inverted pulse from
`the fixed line generator 236, and results in a zero output.
`The output from the horizontal deflection amplifier 250 is
`added to the combined variable amplitude linear sweep
`of the variable line generator 244 and the fixed line gen
`erator 236, amplified by an amplifier 252, and then ap
`plied to the horizontal deflection plates of oscilloscope
`216.
`Thus, a line is drawn in the plane transform for a bub
`ble in the line segment of framelet 212. The linear sweep
`output of the fixed linear generator 236 applied to the
`vertical deflection plates of oscilloscope 216 acts in com
`bination with the linear sweep of variable amplitude pro
`duced by adding the 2 microsecond inverted linear decay
`pulse from the fixed line generator 236 and the 2 micro
`second variable amplitudes linear decay output pulse
`from the variable line generator 244 to produce a line in
`the plane transform having an angle to the vertical whose
`tangent is proportional to the vertical displacement of the
`detected bubble track in the line segment of framelet 212.
`If the detected bubble is at the top of framelet 212, the
`horizontal deñection applied to the horizontal deflection
`plates of oscilloscope 216 is initially large, positive, and
`decays linearly therefrom. lf the detected bubble occurs
`at the center of framelet 212, the horizontal detiection
`is zero and if below the center of the framelet 212, the
`horizontal deflection is initially large and negative in
`polarity from which it decays linearly. The output from
`the horizontal defiection amplifier 250 causes the spot on
`the display of oscilloscope 216 to follow the horizontal
`scanning beam of the television camera 210. When the
`horizontal scanning beam crosses the detected bubble, the
`oscilloscope spot is at the horizontal position of the de
`tected bubble and the video pulse at this instant causes
`the line transform to be drawn as heretofore described.
`The time required for the drawing of the one line in the
`transform is 1.5 microsecond. The delayed unblanking
`pulse of the unblanking pulse delay amplifier 230 gates
`the oscilloscope for this period of time. The set and re~
`set of the line generators 236 and 244 is not seen in the
`transform.
`The entire process described above is repeated each
`time the scanning beam of television camera 210 crosses
`a bubble in the line segment of framelet 212 and results
`in a plane transform being created on the oscilloscope
`display 216 as depicted in FIG. l.
`Though the above description illustrates the presenta»
`tion of only one framelet at a time to the television
`camera, as many as four framelets can be presented at
`one time. Each framelet is caused to cover the full Width
`and one-fourth the height of the television field; the re
`maining treatment of the framelets remaining the same
`as for a single framelet. It is also necessary to scan each
`picture twice at right angles to correctly recognize the
`complex patterns contained therein.
`The present invention should be readily adaptable for
`application in such areas as handwriting analysis, radar
`displays and map reading.
`Persons skilled in the art will, of course, readily adapt
`the general teachings of the invention to embodiments
`other than the specific embodiments illustrated. Accord
`ingly the scope of the protection afforded the invention
`should not be limited to the particular embodiment shown
`in the drawings and described above, lbut shall be de
`termined only in accordance with the appended claims.
`What is claimed is:
`l. A method of analyzing a complex pattern in a pic
`ture comprising dividing said picture into framelets, said
`framelets sized so that that any segment of said complex
`pattern therewithin is essentially a straight line, trans
`forming each of said segments into a plane transform,
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`and'reading the coordinate position data of each plane
`transform.
`„lí
`. 2. 'lne method of analyzing a complex pattern in a
`. picture comprising dividing said picture into framelets,
`` said framelets sized so that any segment of said complex
`pattern therewithin is essentially a straight line, ytranscrib
`ing points along each of said segments into separate
`lines, pictorially displaying said transcribed lines to form
`. a plane transform for each of said segments, the coordi
`nate position of said plane transform in said display be
`ing representative of the position of said segment in said
`framelet, and summing’the coordinate position data.
`3. A method of analyzing va complex pattern in a
`picture comprising dividing said picture into framelets,
`said framelets sized so that any segment of said complex
`pattern therewithin is essentially a straight line, transcrib
`ing points along each of said segments into separate lines,
`pictorially displaying said transcribed lines to form a
`plane transform for each of said segments, each line in
`said plane transform being positioned laterally so that
`a point on said line midway between the top and the
`bottom of said pictorial display occurs at a distance from
`the left edge of said pictorial display equal to a distance
`of said point in said segment from the left edge of said
`framelet, said line in said plane transform being inclined
`in said pictorial display at an angle to the vertical whose
`tangent is proportional to the vertical displacement of
`said point in said segment from the center of said frame
`let, and determining the coordinate position of the point
`of intersection of said lines in said pictorial display for
`each segment.
`4. A method of analyzing a complex pattern in a pic
`ture comprising dividing said picture into framelets, said
`framelets sized so that any segment of said complex pat~
`tern therewithin is essentially a straight line; transcribing
`points along keach of said segments into separate lines,
`pictorially displaying said transcribed lines to form a
`plane transform for each of said segments, each line in
`said plane transform being positioned laterally so that
`a point on said line midway between the top and the
`bottom of said pictorial display occurs at a distance from
`the left edge of said pictorial display equal to the distance
`of said point in said segment from the left edge of said
`framelet, each said line in said plane transform being
`inclined in said pictorial display at an angle to the Vertical
`whose tangent is proportional to the vertical displacement
`of said point in said segment from the center of said
`framelet; scanning said pictorial display of said plane
`transform of each of said segments and determining
`the coordinate position of the intersection point of said
`lines in said pictorial display of said plane transform,
`the lateral position of said intersection point in said pic
`torial display of said plane transform being equal to the
`lateral position at which a point in said segment on said
`framelet is equidistant from the top and bottom of said
`framelet, the vertical position of said intersection point
`in said pictorial display of said plane transform denoting
`the tangent of the angle of said segment in said framelet;
`recording the Ácoordinate data of said intersection point in
`said plane transform of each of said segments and sum
`ming said recorded data.
`5. A device for electronically transforming a straight
`line in a pictorial representation into »coordinate data corn
`prising means for scanning said representation and pro
`ducing an electrical pulse for each point scanned on said
`line, means for transforming each of said pulses into a
`separate line and for displaying each of said transformed
`lines, each of said transformed lines being geometrically
`positioned in said display with relation to the geometric
`position of its respective point in said representation, said
`transformed lines intersecting at a point in said display
`whose coordinate position is descriptive of the geometric
`position of said straight line in said representation.
`6. A device for electrically transforming a straight line
`
`55
`
`60
`
`65
`
`70
`
`75
`
`0005
`
`
`
`.3,069,554v
`8,
`in a pictorial representation into coordinate data_' com
`gering the cathode of said cathode raytube to cause said
`prising means for scanning said representation and pro
`first and second signals of each of said electrical pulsesf
`ducing an electrical pulse for each point scanned on said
`to draw a line on said cathode ray tube having a slopef
`line,'a"cathode ray tube having vertical and horizontal
`and an intercept with the horizontal midline of said
`deflection plates, means for deriving a ñrst linear decaßl
`cathode ray tube proportional to the coordinate position
`signal havingv initial «constant amplitude from each of
`of said scanned point in said pictorial representation, said
`saidV electrical pulses and'applying said ñrstrsignal to said
`lines having an intercept point whose coordinatesron said
`vertical deilection plates of said cathode rray tube, means
`cathode ray tube are proportional to the slope and the
`for deriving a second linear decay pulse having initialY
`intercept with the horizontal midline of said pictorial
`Y
`variable amplitude from eachA of> said electrical 'pulses
`,.10 representation of said straight line,
`and applying' said second signal to said horizontal de
`ñe'ction' plates of said cathode'ray tube, means for trig
`
`No references cited.
`
`0006