United States Patent [19]
`Lee et a1.
`
`[54] DETERMINATION OF AMBIENT LIGHT
`LEVEL CHANGES IN VISUAL IMAGES
`[75] Inventors: George C. Lee, Williamsville, N,Y.;
`Xianyi Sun, Beijing, Taiwan
`[73] Assignee: The Research Foundation of State
`Univ. of N.Y., Albany, NY.
`[21] Appl. No.: 580,629
`[22] Filed:
`Sep. 11, 1990
`
`[51] Int. Cl.5 ............................................. .. G06K 9/00
`[52] US. Cl. ........................................ .. 382/1; 382/37;
`382/4; 358/105
`[58] Field of Search ..................... .. 382/1, 4, 2, 37, 38,
`382/39, 42; 358/105, 108, 109, 101
`References Cited
`U.S. PATENT DOCUMENTS
`
`[56]
`
`4,337,481 6/l982 Mick et al. ........................ .. 358/105
`4,408,224 l0/l983
`4,455,550 6/1984
`4,679,077 7/1987
`4,737,847 4/1988
`Primary Examiner—Michael Razavi
`{S 7]
`ABSTRACT
`A visual image comparison method is provided, includ
`
`lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll
`US005151945A
`Patent Number:
`5,151,945
`Sep. 29, 1992
`Date of Patent:
`
`[11]
`[45]
`
`ing the steps of: obtaining a ?rst digital representation of
`a ?rst visual image. where the ?rst digital representation ‘
`comprises a ?rst plurality of pixels, and each pixel has a
`gray scale indicative of light intensity; obtaining a sec
`ond digital representation of a second visual image,
`where the second digital representation comprises a
`second plurality of pixels, and each pixel has a gray
`scale indicative of light intensity; and selectively mak
`ing a ?rst predetermined number of comparisons of
`corresponding pixels from the ?rst and second digital
`representations to determine whether a difference in
`ambient light intensity exists between the ?rst and sec
`ond visual images, and, selectively making a second
`predetermined number of comparisons of correspond
`ing pixels from the ?rst and second digital representa
`tions if and only if no difference in ambient light inten
`sity exists between the ?rst and second visual images,
`and indicating an alarm condition when a percentage of
`the second predetermined number of comparisons result
`in pixels having a difference in gray scale of a predeter
`mined amount. Apparatus is also described for the pur
`pose of implementing the method.
`
`3 Claims, 7 Drawing Sheets
`Micro?che Appendix Included
`‘(1 Micro?che, 21 Pages)
`
`1 1
`/
`
`I
`
`CAMERA ——‘-—>
`I
`
`IO
`j
`
`I
`
`VISUAL
`IMAGE
`COMPARATOR
`
`13
`2
`
`'
`
`‘
`
`CONTROLLED
`VICE
`DE
`8
`
`|
`1
`
`,/12
`
`Page 1 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`US. Patent
`
`Sep. 29, 1992
`
`Sheet 1 of 7
`
`5,151,945
`
`VISUAL
`IMAGE
`COMPARATOR
`
`k
`V
`
`CONTROLLED
`DEVICES
`
`Izw):
`
`MONITOR
`
`FIG. 1 A
`
`ANALOG TO
`DIGITAL
`CONVERTER
`
`CONTROLLER
`
`EXECUTOR
`
`ISA-A
`
`CONTROLLED
`DEVICES
`
`Page 2 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`US. Patent
`
`Sep. 29, 1992
`
`Sheet 2 of 7
`
`5,151,945
`
`20 /
`
`INPUT DIGITAL IMAGE
`INTO SECOND MEMORY
`(OLD IMAGE) '
`
`INPUT DIGITAL IMAGE
`INTO FIRST MEMORY
`(NEW IMAGE)
`
`CHANGE IN
`ABIENT LIGHT
`INTENSITY ;>
`
`SUF FICI ENT DIFFERENCE
`
`SIGNAL ALARM
`
`TRANSFER DIGITAL
`IMAGE FROM FIRST
`MEMORY TO SECOND
`MEMORY
`
`26
`
`,
`
`FIG.2
`
`Page 3 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`US. Patent
`
`Sep. 29, 1992
`
`Sheet 3 0f 7
`
`5,151,945
`
`COLUMNS
`
`.-
`
`|
`
`\ \
`
`a V
`
`~ ~
`
`\
`
`-
`
`\
`
`-
`
`ROWS
`
`FIG.3A
`
`M240,“ .. .
`
`.
`
`. .. .. .. ::::'.:.': X(24O,256)
`
`COLUMNS
`
`. . . - . 1 ' . -
`
`-
`
`. .
`
`-
`
`. - . . -
`
`YW)
`
`-:::
`
`YBJ) ::::: ‘
`
`".::::::
`
`9 . . . ..
`
`ROWS
`
`Page 4 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`US. Patent
`
`Sep. 29, 1992
`
`Sheet 4 of 7
`
`5,151,945
`
`5lO.n.
`
`5V
`
`F IG.4A
`
`RESET
`OSC
`
`Page 5 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`Page 6 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`Page 7 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`US. Patent
`
`Sep. 29, 1992
`
`Sheet 7 0f 7
`
`5,151,945
`
`I
`\ REI
`
`CONTROLLED
`SWITCH
`
`8
`REZ
`
`CONTROLLED
`AC POWER SUPPLY
`
`I IOV
`AC ""
`
`FIG.4E
`
`Page 8 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`1
`
`DETERMINATION OF AMBIENT LIGHT LEVEL
`CHANGES IN VISUAL IMAGES
`
`In accordance with 37 C.F.R. 1.96. a micro?che ap
`pendix is to be considered a portion of the entire “writ
`ten description” of this invention in conformance with
`35 U.S.C. 112. The appendix includes one micro?che
`having 21 data frames.
`
`5,151,945
`2
`in ambient light. Yoshida suggests displacing the cap
`turing of the two video images in time by an amount
`which is negligible with respect to the ambient changes
`in brightness. For example. Yoshida suggests that dis
`placing the capture of the images by 15 seconds to 1
`minute is suitable to overcome the effects of gradually
`changing brightness. Unfortunately, this attempt to
`solve the problem is limited in its usefulness in that it is
`dependent upon the rate of change of the ambient light
`intensity. While one time setting may be suitable for
`slowly changing intensity levels (such as might occur at
`sunset, dawn, etc.), this same time setting may be unsuit
`able for rapid changes (such as clouds passing overhead
`during a thunderstorm, or sudden dimming of lights in
`an of?ce, etc.).
`What is needed, then, is a surveillance system which
`is not only immune to false alarms caused by changes in
`ambient light conditions, but also functions indepen
`dently of the speed with which these ambient changes
`occur.
`
`20
`
`BACKGROUND OF THE INVENTION
`The present invention relates generally to video sur
`veillance methods and apparatus and, more particularly,
`to automatic surveillance systems which detect changes
`in a ?eld of view over time and indicate an alarm condi
`tion accordingly.
`The use of video cameras at remote locations for
`surveillance by video monitors is well known. In some
`circumstances, constant human supervision or monitor
`ing is required. A typical example of this manual sur
`veillance method would be the remote placement of
`cameras in a retail store to detect shoplifting; another
`example would be a camera in a bank activated during
`working hours to monitor a robbery attempt. Many
`security systems typically employ a plurality of video
`25
`cameras situated throughout a facility, with a central
`monitoring location where a human guard keeps watch.
`These manual systems are predecessors within the ?eld
`of the present invention.
`Human interaction in surveillance is extremely expen
`sive. In some circumstances, then, it is economical and
`desirable to replace the human observer with an auto
`matic surveillance system, or at least to alleviate the
`guard from the burden of constant supervision, freeing
`him to perform other useful work. Automatic surveil
`lance systems have evolved, therefore, to handle situa
`tions which do not require constant human supervision.
`An example of this application would be the monitoring
`of an empty room at night, where an automatic system
`would sense the entry of an intruder and sound an
`alarm. It may be desired to monitor an outdoor parking
`lot, or perhaps the entrance or exit of a building. Other
`applications include monitoring products or workpieces
`on an assembly line, etc. In a multitude of applications,
`automatic surveillance methods and systems are more
`economical and even more reliable than syst’ems requir
`ing constant human interaction.
`A common problem encountered by all automatic
`surveillance systems, both indoor and outdoor, involves
`false alarms triggered by changes in ambient light inten
`sity. For example, in monitoring an outdoor scene such
`as a parking lot, a cloud passing overhead may substan
`tially affect ambient light conditions and trigger a false
`alarm. Even in indoor applications, many of?ces em
`ploy automatic light dimming circuits which dim the
`lights in the evening, causing problems for automatic
`surveillance systems.
`Attempts to solve the false triggering problem are
`well documented in the art. One well-known technique
`involves the use of automatic exposure lenses or cam
`eras to compensate for ambient light intensity varia
`tions. Unfortunately, this method is limited to only
`small variations in intensity. Another alleged solution is
`proposed by Yoshida in U.S. Pat. No. 4,408,224 (Oct. 4,
`1983). Yoshida broadly discloses a surveillance method
`which includes the comparison of two digitized video
`image signals taken of a “place scenery” at different
`points in time. To solve the problem caused by changes
`
`30
`
`SUMMARY OF THE INVENTION
`A visual image comparison method is provided, in
`cluding the steps of obtaining a ?rst digital representa
`tion of a ?rst visual image, where the ?rst digital repre
`sentation comprises a ?rst matrix having a plurality of
`pixels, and each pixel has a gray scale value indicative
`of light intensity; obtaining a second digital representa
`tion of a second visual image, where the second digital
`representation comprises a second matrix having a plu
`rality of pixels, and each pixel has a gray scale value
`indicative of light intensity; and selectively making a
`?rst predetermined number of comparisons of corre
`sponding pixels from the ?rst and second digital repre
`sentations to determine whether a difference in ambient
`light intensity exists between the ?rst and second visual
`images, and, selectively making a second predetermined
`number of comparisons of corresponding pixels from
`the ?rst and second digital representations if and only if
`no difference in ambient light intensity exists between
`the ?rst and second visual images, and indicating an
`alarm condition when a percentage of the second prede
`termined number of comparisons result in pixels having
`a difference in gray scale of a predetermined amount.
`The ?rst and second visual images can be images ob
`tained at different locations simultaneously, images
`taken of a single location at different times, or images
`obtained at different locations at different times. An
`apparatus is also provided to implement the method of
`the invention.
`A primary object of the invention is to provide a
`visual image comparison method which functions inde
`pendently of ambient changes in light intensity between
`the visual images being compared.
`A secondary object is to provide a visual image com
`parison method which functions independently of the
`time rate of change of ambient light intensity between a
`?rst and second visual image being compared.
`
`BRIEF DESCRIPTION OF THE DRAWINGS
`FIG. 1A is a general block diagram of a video surveil
`' lance apparatus which utilizes the present invention.
`FIG. 1B is a general block diagram similar to FIG.
`1A but expanded to show the major elements of the
`visual image comparator.
`FIG. 2 is a flow diagram illustrating the general
`method of the invention.
`
`65
`
`Page 9 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`5,151,945
`
`10
`
`25
`
`3
`FIG. 3A represents a ?rst digital representation of a
`?rst visual image, and FIG. 38 represents a second
`digital representation of a second visual image, which
`two images are compared by the present invention.
`FIGS. 4A-4E illustrate a schematic circuit diagram
`of an alternative embodiment of the invention which
`utilizes a dedicated “hard" circuit.
`
`4
`NEW) 15 is transferred to RAMwLD) 16, and the new
`digital image is stored in RAMA-Em 15. Controller 17
`controls the transferring of the old image from RAM(.
`NEW) 15 to RAM(01_D)16 and then compares the old and
`new images. If a suf?cient difference exists between the
`images, controller 17 sends an alarm signal to executor I
`18. Executor l8 activates various auxiliary alarm de
`vices as discussed previously.
`DETAILED DESCRIPTION OF THE
`In the preferred embodiment depicted in FIG. 1B,
`INVENTION
`both the old and new digitally represented images are
`stored in separate memories prior to comparison. It
`The present invention is a visual image comparison
`should be readily apparent, however, that it is also pos
`system which enables automatic surveillance of a loca
`sible to compare temporally displaced images by only
`tion over time, or simultaneous monitoring of two or
`storing the earlier image in memory and then compar
`more identical or nearly identical objects. The system
`ing a present or new image to the old in real time, elimi
`described herein may be used in homes, museums,
`nating the need for one of the memories. It is also possi
`stores, of?ces, and other commercial establishments as
`ble to compare two distinct images simultaneously re
`well as in hotels, airports, and other special places re
`quiring security measures. The invention may also ?nd
`ceived by two video cameras at the same time, eliminat
`applications in industry such as, for example, the moni
`ing both memories. For example, one camera could be
`toring of a workpiece on an assembly line. In addition to
`focused on an image (such as a ?ngerprint) while a
`security surveillance applications, the system may also
`second camera scans other images looking for a match.
`be used to compare two visual images (?ngerprints,
`An inverter circuit connected to controller 17 or execu
`etc.) simultaneously to determine if the images are the
`tor 18 would sound an alarm only when a match is
`found. Similarly, on an assembly line, one camera could
`same or nearly the same.
`The system is used in conjunction with a video cam
`be focused on a static image of a workpiece as it should
`era or other means of providing an analog visual image.
`appear at a certain step in assembly, while a second
`A visual image comparator method and apparatus ana
`camera is timed to monitor workpieces on the actual
`lyzes the analog visual images and sounds an alarm if a
`assembly line. If the two images don’t match, a suitable
`warning would be given to indicate a possible product
`suf?cient difference exists between two different im
`defect.
`ages.
`7
`Adverting once again to the preferred embodiment
`The apparatus of the invention is outlined in block
`depicted in FIG. IE, it should be noted that VIC 10
`form in FIG. 1A to show how the invention interacts
`with auxiliary equipment. Visual image comparator
`may be implemented in apparatus form in one of at least
`(VIC) 10 represents the present invention, which oper'
`two ways. In a ?rst embodiment, VIC 10 may comprise
`ates on visual images provided by video camera 11 or,
`software run by a conventional computer such as an
`alternatively, by optional video monitor 12. When a
`IBM PC® or compatible computer. In a second em
`change in non-ambient light intensity or when motion
`bodiment, VIC 10 may comprise a dedicated circuit
`specially designed to implement the method of the in
`occurs within the ?eld of view of camera 11 (or within
`monitor 12), VIC 10 signals an alarm, represented by
`vention. In either case, the method of comparison is the
`controlled devices 13in FIG. 1A. Controlled devices 13
`same, and this method is described herebelow:
`40
`may be any device capable of indicating an alarm (bell,
`The Visual Image Comparison Method
`whistle, buzzer, light, etc.) or it may even comprise a
`video monitor which automatically displays the chang
`The present invention broadly comprises a visual
`.image comparison method, comprising the steps of:
`ing video image when a change occurs. For example, in
`one application contemplated by the inventors, a video
`obtaining a ?rst digital representation of a ?rst visual
`45
`image comprising a ?rst plurality of pixels, where each
`camera is focused on the entrance to a residence. As
`pixel has a gray scale indicative of light intensity; ob
`someone approaches the entrance, the motion is de
`tected by the system. The system may be programmed
`taining a second digital representation of a second visual
`image comprising a second plurality of pixels, where
`to display the video camera image on a television set (or
`each pixel has a gray scale indicative of light intensity;
`to display the image as a “picture-within-a-picture” on
`50
`selectively making a ?rst predetermined number of
`the television) to indicate the arrival of a visitor. When
`comparisons of corresponding pixels from the ?rst and
`the television set is turned off, the system may be pro
`second digital representations to determine whether a
`grammed to turn on lights or sound audible alarms, etc.
`difference in ambient light intensity exists between the
`(or to turn the television on to display the changing
`?rst and second visual images, and, selectively making a
`video image). In yet another application, the camera
`second predetermined number of comparisons of corre
`might be focused on an infant’s crib to monitor the baby
`sponding pixels from the ?rst and second digital repre
`at sleep. If the infant awakens, moves or becomes dis
`tressed, the parents can be alerted accordingly.
`sentations if and only if no difference in ambient light '
`A preferred embodiment of the invention which
`intensity exists between the ?rst and second visual im
`shows VIC 10 in more detail is shown conceptually in
`ages, and indicating an alarm condition when a percent
`age of the second predetermined number of compari
`FIG. 1B. VIC 10 includes A/D converter 14, RAM(.
`sons result in pixels having a difference in gray scale of
`NEW) 15, RAM(0LD) 16, controller 17 and executor 18.
`A/D converter 14 converts the analog video signal
`a predetermined amount.
`provided by camera 14 into digitized signals. The digital
`FIG. 2 illustrates by flow diagram the general
`method of the invention. To begin the process, it is
`representation of a ?rst visual image so obtained is then
`assumed that a frame of digitized data representative of
`stored in RAMWEW) 15, which is a random access mem
`a second visual image is already stored in a second
`ory. At a subsequent instant in time, a new image signal
`is obtained. The digital representation stored in RAM(.
`memory (box 21). A new image is then digitized and a
`
`30
`
`35
`
`65
`
`Page 10 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`5,151,945
`
`15
`
`25
`
`30
`
`5
`new frame of digitized data is stored in a ?rst memory
`(box 22). The new and old digital representations of ?rst
`and second visual images, respectively. are then com
`pared (box 23). A decision is made as to whether a
`difference in ambient light intensity exists between the
`two images. Such a difference would occur, for exam
`ple, if the sun suddenly disappeared behind a cloud; if a
`tree branch was moved by the wind in front of the
`camera lens; at sunset or sunrise, or if the lights in a
`room were turned on or off, etc. In other words, such a
`change would likely be distributed somewhat through
`out the entire visual ?eld. The essence of the invention
`is to distinguish between such ambient intensity
`changes, and other changes, such as might be caused by
`someone or something entering or leaving the visual
`?eld. The invention will indicate an alarm condition for
`the latter condition, but will not sound a false alarm for
`the former condition.
`If an ambient intensity difference is found to exist in
`box 23, control is passed to box 26, where the digital
`image previously stored in the ?rst memory is trans
`ferred to the second memory (and the image previously
`stored in the second memory is erased). The method
`then proceeds back to box 22 where a new image is
`obtained and stored and the process repeated.
`If, on the other hand, no ambient intensity change or
`difference is detected, control passes directly to box 24,
`where a second comparison is done between the ?rst
`and second visual images. If a suf?cient difference exists
`between the two images, an alarm is indicated (box 25).
`Otherwise, control passes to box 26 and the process
`repeats as described above. ,
`It is the unique method used to compare old and new
`images which enables the present invention to operate
`35
`independently of changes in ambient light intensity.
`This comparison is best understood with reference to
`FIGS. 3A and 3B.
`FIG. 3A represents a ?rst digital representation of a
`?rst visual image. The representation comprises a plu
`rality Of pixels X(1,1), X03), X03), . . . , X(Z4Q,256), where
`each pixel has an associated gray scale indicative of
`brightness or light intensity. For example, an individual
`pixel may have a gray scale, G, ranging from 0 to 63,
`where 0 indicates black and 63 indicates white, or from
`0 to 255, where 0 indicates black and 255 indicates white
`depending on the interface board used.
`Similarly, FIG. 3B represents a second digital repre
`sentation of a second visual image, such as that of an
`image obtained at a later point in time than the ?rst
`image mentioned previously. The second representation
`comprises a plurality Of pixels Y(],1), Yul), Y(1,3), . .
`. ,
`“240356), where each pixel has an associated gray scale
`as discussed above.
`55
`The comparison method of the invention functions by
`comparing digital representations of corresponding
`pixels of the ?rst and second images. For example,
`X(3,1) and Y(3,1) are corresponding pixels. There are
`three variables which factor into the comparison pro
`cess as follows:
`1. The total number of comparisons of corresponding .
`pixels from each image. (Determined by J, the num
`ber of pixels to jump over when making comparisons
`from each image).
`2. The number of comparisons, N, which must yield a
`difference in gray scale in order to signify an alarm
`condition.
`
`6
`3. The magnitude, K, ofthe gray scale difference neces
`sary within a given pixel-pair to signify an alarm
`condition.
`It can be readily appreciated that a comparison of
`each pixel of the ?rst image with its corresponding pixel
`of the second image would be extremely time consum
`ing if all 61,440 pixels of each image were compared.
`Obviously, comparison processing time is related to the
`number of pixels compared. For example, in a software
`driven embodiment of the invention, a comparison of
`20,480 pixels (skipping every third pixel) takes approxi
`mately 0.71 seconds of processing time, whereas com
`paring 8,777 pixels (skipping every seventh pixel) takes
`approximately 0.43 seconds.
`Fortunately, it is unnecessary to compare each and
`every pixel to achieve a system which functions inde
`pendently of changes in ambient light intensity. More
`over, the number of comparisons needed, the percent
`age of those comparisons which must show a difference,
`the spatial distribution of the pixels compared, and the
`magnitude of the gray difference between pixels being
`compared are variables dependent upon whether the
`comparison is being done to determine a change in
`ambient light intensity, or to determine an alarm condi
`tion.
`The ?rst comparison of the method determines
`whether a difference in ambient light intensity exists
`between the ?rst and second visual images. In a pre
`ferred embodiment, experiments indicate that, for an
`image comprising 61,440 pixels, as few as 1,536 pixels
`need be compared (skipping every 40th pixel, i.e.,
`J =40) and yet still achieve accurate and reliable results.
`In other words, it is only necessary to compare approxi
`mately 2% to about 4% of the total image, assuming
`that the compared pixels are distributed throughout the
`image. Of course, more pixels could be compared, but
`this would increase processing time. The processing
`time required to compare 1,536 pixels is less than 0.3
`seconds.
`Experiments also indicate that, for J =25-40, N may
`be in the approximate range of N=90-110 to achieve
`accurate results. In other words, approximately 3% to
`about 8% of the compared pixel-pairs must exceed the
`predetermined K value (magnitude of gray difference)
`in order to indicate a difference in ambient light inten
`sity.
`Finally, in making the ambient light comparison,
`' experiments indicate that a low K value is preferred
`(e.g., K=2 or 3) since the change in ambient light may
`50
`be very small. This is the magnitude, K, of the gray
`scale difference within a given pixel-pair comparison
`necessary to signify an alarm condition. For example, if
`one pixel has a gray scale of 21 and its corresponding
`pixel has a gray scale of 27, then the gray scale differ
`ence is said to be 6 (27-21). The value of K selected
`affects the sensitivity of the system. The lower the value
`of K, the more sensitive the system. In a system such as
`the preferred embodiment where each pixel has a gray
`scale range from 0 to 63, a K value of 2 to 3 is a differ
`ence equivalent to 3%—5% of the total gray scale.
`Once again, experimental data suggest that values of
`K=2 or 3 ensure reliable operation in nearly all ambient
`light conditions. This is not to say that other values of K
`will not be suitable, only that values of K=2 or 3 are
`preferred. In fact, other values of K have been proven
`to achieve suitable results, depending upon the ambient
`light conditions. Obviously, the method will still work
`
`40
`
`65
`
`Page 11 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`5,151,945
`7
`satisfactorily with larger K values, but will simply be
`less sensitive to changes in ambient light intensity.
`Also, it is noted that, in a preferred embodiment, the
`pixels compared are uniformly distributed throughout
`the images (i.e., every 25th pixel, every 40th pixel, etc.).
`This is not to imply that uniform distribution is abso
`lutely necessary, although it is preferred. In checking
`for ambient light differences, it is necessary, however,
`that compared pixels be distributed widely throughout
`the images.
`Summarizing, then, in a preferred embodiment, the
`?rst comparison to determine if a difference in ambient
`light intensity exists between the ?rst and second im
`ages is preferably made with K=2 or 3, N=90—l 10, and
`J=25-40. These ranges are intended to be guidelines
`and approximations, and it is not intended or implied
`that other combinations of J, K and N will not work
`satisfactorily, only that satisfactory results have been
`obtained when the variables are selected within these
`ranges.
`.
`In the second comparison, or so-called common sur
`veillance mode, a different range of the variables are
`utilized to determine if an alarm condition exists. It has
`been determined experimentally that N=2 or 3
`achieves satisfactory results when J =3-7. While other
`25
`values of N may also work, it has been found that N=l
`often results in false alarms, and high values of N result
`in low surveillance sensitivities which may not detect
`small moving objects within the visual ?eld. Similarly,
`although larger values of J may work satisfactorily, as J
`increases small moving objects may not be detected.
`In a preferred embodiment, the K value in the second
`comparison is determined by the average gray scale
`value of the ambient (AGA) according to Table I be-_
`low. The average gray scale value of the ambient is
`determined by adding all of the individual gray scales
`values (GS) of each considered pixel and then dividing
`this total by the number of pixels considered.
`TABLE I
`4
`5
`6
`3
`7
`9
`l2
`4
`l4
`l5
`l6
`l3
`l6
`l6
`l6
`l6
`24
`25
`26
`23
`l8
`l8
`l9
`l8
`34
`35
`36
`33
`2O ' 2O
`21
`21
`43
`44 45
`46
`23
`24
`24
`24-
`53
`54
`55
`56
`26
`27'
`27
`27
`63
`32
`
`2
`1
`AGA
`2
`I
`K
`l2
`AGA ll
`l6
`K
`l6
`22
`AGA 21
`l8
`K
`18
`32
`AGA 31
`20
`K
`20
`42
`AGA 41
`23
`K
`23
`52
`AGA 51
`K
`26 26
`AGA 61
`62
`K
`30
`31
`
`7
`l4
`l7
`l7
`27
`l9
`37
`21
`47
`24
`57
`28
`
`8
`l6
`18
`l7
`28
`l9
`38
`22
`48
`25
`58
`28
`
`9
`l6
`19_
`17
`29
`l9
`39
`22
`49
`25
`59
`29
`
`10
`l6
`20
`17
`30
`l9
`40
`22
`50
`25
`60
`29
`
`5
`
`30
`
`35
`
`45
`
`50
`
`8
`Moreover, it is not necessary to actually make all the
`planned comparisons if, for example, a sequence of
`early comparisons indicate a problem. For instance, if
`2,458 comparisons are to be made, but only 90 compari
`sons indicating a difference are required to signal an
`alarm, then the processing can stop as soon as the 90th
`comparison indicating a difference is reached. This may
`occur at any time (i.e., on the 2,450th comparison or
`even on the 90th comparison, etc.). This manner of
`processing ensures reliability by preventing false alarms
`while minimizing processing time.
`A First Physical Embodiment For Implementing The
`Method
`In a ?rst embodiment, the method of the invention
`may be implemented using a software driven system of
`a personal computer, such as an IBM-PC or equivalent.
`The software necessary to implement the system is
`included in the micro?che appendix. Also required in a
`video imaging interface for converting the analog video
`signal to digital signals for processing by the computer.
`In a preferred software driven embodiment, a Model
`DT2803 “Frame Grabber” was used as the video inter
`face (available from Data Translation, Inc., 100 Locke
`Drive, Marlborough, Massachusetts 01752-1192). Of
`course, any commercially available equivalent video
`imaging interface could be used in lieu of the DT2803.
`The DT2803 Frame Grabber is a single-board, micro- -
`processor-based video imaging interface, suitable for
`use with the IBM personal computer series (IBM
`PC/AT/XT) and functionally IBM-compatible per
`sonal computers. This video interface provides real»
`time 6-bit digitization of an RS-l70/RS-330/NTSC or
`CCIR/PAL compatible input signal. The DT2803
`plugs into the PC backplane, and includes a video imag
`ing input analog to digital converter and look-up tables,
`a 64 kilobyte frame-store memory, a video imaging
`output digital to analog converter and look-up tables,
`and microprocessor and control logic. For a more com
`plete description of the capabilities and operation of the
`video imaging interface, the reader is referred to the
`User Manual for. the DT2803 Low-Cost Frame Grab
`ber, available as Document UM-03286A, copyright
`1985, by Data Translation, Inc. This document is incor
`porated herein by reference as representative of the
`general state of the art with respect to video imaging
`interfaces.
`.
`The software included in the micro?che appendix is
`self-executing. After booting up the computer and load
`ing the software, the user merely types the word
`“ALARMS” on the keyboard and then follows the
`self-explanatory menu driven instructions for setting the
`system sensitivity. If desired, the user can omit setting
`sensitivity levels, in which case the levels are automati
`cally set by the software.
`In the software driven embodiment, video camera 11
`provides analog video signals to analog-to-digital con
`verter 14 (DT2803), and the digital signals are then
`processed by the computer as previously described.
`When an abnormal or alarm condition is detected by the
`system, the software sounds an audible alarm through
`the internal speaker of the computer. With minor soft
`ware modi?cation, the system can also be programmed
`to sound an external alarm. For this purpose, an addi
`tional output port interface board is required, such as
`DT2801, also available from Data Translation, Inc.
`
`Although K is determined automatically in the pre
`ferred embodiment, a less sophisticated embodiment is
`possible, where K is determined manually and by trial
`and error by the system operator. For example, if sur
`veillance of an indoor room having a constant ambient
`light intensity is desired, the operator may manually set
`the sensitivity level and experiment with persons or
`objects moving in and out of the visual ?eld until satis
`factory results are obtained.
`Obviously, increasing the number of pixels compared
`is one way of increasing the reliability of the system, but
`at the cost of increasing processing time. Another
`method of improving reliability without necessarily
`increasing processing time, is in the selection of the
`number of comparisons which must indicate a differ
`ence in gray scale in order to signify an alarm condition.
`
`55
`
`65
`
`Page 12 of 14
`
`SAMSUNG TECHWIN AMERICA 1006
`
`

`
`5,151,945
`10
`Commercially available equivalent interfaces are also
`in ambient light intensity exists between said ?rst
`suitable.
`and second visual images, wherein said difference
`in ambient light intensity is de?ned to exist when a
`?rst percentage of said ?rst predetermined number
`of ?rst remotely displaced comparisons result in
`pixels having a difference in gray scale of a ?rst
`predetermined amount; and,
`d. selectively making a second predetermined number
`of second comparisons of corresponding near
`neighboring pixels from said ?rst and second digi
`tal representations if and only if no signi?cant dif
`ference in ambient light intensity exists between
`said ?rst and second visual images,
`e. and indicating an alarm condition when a second
`percentage of said second predetermined number
`of second near neighboring comparisons result in
`pixels having a difference in gray scale of a second
`predetermined amount.
`2. A method as described in claim 1 wherein said
`second p