TW201005416A - Fuzzy control device for zoom and autofocus aperture of video camera lens - Google Patents

Abstract

This invention relates to a fuzzy control device for zooming and autofocus aperture of video camera lens, which uses the technique of image processing and the fuzzy control method to track a target and to control the focus, aperture, and zoom of a video camera lens. In respect of autofocus control of the present invention, the spectrum of the image is captured and used to design an input membership function for controlling the focus of the video camera lens; in respect of automatic control of aperture of the lens, the input membership function is designed by using the standard deviation and grayscale mean value of the target for controlling the aperture of the video camera lens; in addition, the input membership function is designed by using the reliability of the image captured with the target dimension, the change of movement velocity and the target's size so as to achieve the control of autofocus of the video camera lens.

Description

201005416 IX. Description of the invention: [Technical field to which the invention pertains] The present invention relates to an automatic mismatch control device for a face-to-face lens. The hair-sharing method __ green is used to track the moving target and control the focus, aperture and zoom of the camera lens. [Prior Art]

With the advancement of computer technology, the application and development of computer vision combined with cameras has become more and more extensive. For all defense, meteorology, medicine, industrial production and manufacturing, transportation, agriculture, fishery, and security monitoring materials, we can see the use of cameras. Application examples. At present, the processing of motion pictures is the main research project of computer vision, and the visual ship tracking or unmarking system is the research theme of the lion. In the related research of real-time viewing and tracking, it is the main application project to complete accurate guidance and tracking; in the related research of camera lens control, 'mainly study how to control the focus, aperture and zoom of the lens, clear and Large and small targets _ like, for subsequent processing, analysis, identification and tracking control applications. The visual feeding suit chasing _ the first step is to practice the county, and the quality of the photographic gamma is related to the success or failure of the smashing of the results. Therefore, the image quality is very important. However, due to the factors affecting the quality of the image couch, in addition to the ambient light source and the target target speed, the lens control of the camera is also the image quality of the god. Therefore, in the case of Na Na, if the shirt U is very poor, the target image is small, and when the environment is not changed, the result of the incorrect secret image processing may occur, and the tracking fails. Therefore, by controlling the autofocus of the camera lens, auto-tuning of the iris, and auto-zooming 201005416, and appropriately scaling the size of the object to obtain a clear image can be used for subsequent processing. In order to achieve the purpose of instant tracking, the lens control and image processing must have the function of instant and fast calculation. The camera lens is controlled to control the focus, aperture and zoom of the camera lens in order to obtain a clear and appropriately sized target image. For example, if the target is very small in the image and moves at a high speed, the trustworthiness of the target image information and the outline of the target will be reduced. 'There is too much noise, and the ambient light source when the image is taken changes too much. It may cause the target to exceed the identified range, resulting in tracking failure. Therefore, the present invention hopes to track the image of the object and can be controlled within an appropriate size range. Therefore, the factors considered in the image are the size of the target, the moving speed and the change of the moving direction, and the background of the ring at that time. [Explanation] ❹ The invention of the case of the lion mirror control, when the image is captured by the camera (10), 'first judge whether the target to be tracked is in the captured image, and its household _ = shirt Such as 疋 No β is clearly identifiable, the irregular movement of the secret moving object, the change of the light of the environmental background during the tracking, the change of the lighting condition If the photo-advancement of the post-advancement control and _ unexpectedly, the fortune of the fortune is turned over, and the silk touch is what the animal is and cannot track the target. Therefore, the image acquisition control is important, = the case is self-defeating, the vehicle is self-contained, and the automatic is based on the above purpose, in order to have a better method of image acquisition. The mail image _ spectrum is the input attribution function of Wei 201005416. The average and standard deviation of the image grayscale values are used as the input attribution function for the lens aperture control. The image's trustworthiness is determined by the size of the target, the change of the moving speed, and the change of the contour of the image. These losers touch the simple _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Average) Defuzzification is used as the output of the lens focus, aperture and zoom fuzzy controller, and finally the converted fruit is converted into the actual mirror focus, fine and motor control dust. [Fuzzy autofocus fuzzy controller] The correct step for the sharp image of the line, the opposite - the demotion lens and the r scene object image - the Gauss lens recording law (G_ian _ to indicate: 7

~ Γ-"_ "' FL λ DI rh A U1 (1) where aFL is the focal length of the lens—th), DI is the distance between the target and the center of the mirror, and A is the distance between the center of the mirror and the imaging plane. Also known as the image width fine = her), also known as the camera constant (camera_ (four), in general, the hole. = image plane shift A also changes, the age can make the image blur, so must = the length of the focal length FL, In order to obtain a clear image. To judge 撷 = frequency =: like the Fu Li Ye conversion life blur ^ The more stable the machine is, the opposite is the opposite of the image, the difference between the image and the value of the image is the difference between the two sides of the image. If , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , The input attribution function of the focus control is established. The invention adopts the method of discretizing the domain and the language term attribution function of the input and output variables to design the fuzzy controller. The discretization of the input and output variables of the focus control The comparison table is shown in Table 1, where the input variable is the Fourier spectrum of the image /, the domain is [0~256], and the output variable is the control voltage & of the focus motor. The field of view is the input of the focus control of Table 1. Discrete table of output variable domain

Discrete level image FFT spectrum / s focus motor control voltage F / -3 〇 ^ <32 -8 ^ V/K -6 -2 32 parts <64 -6^Vf<A -1 64^fs<96 -4 ^Vf<-2 0 96^/^160 Γ -2^Vf^2 +1 160</,^192 2<Vf^4 +2 192 </<224 4<Vf^6 +3 224</ The input attribute function of ^256 6<Vf^S 〇 image Fourier transform spectrum language term is shown in Table 2. The definition function FFi(Lqw) of the image of the image of the Fourier transform surface of the image indicates the result of the image-rich leaf transformation. The spectrum component has fewer high-frequency components and more low-frequency components, which means that the captured image is blurred. The high-frequency components and low-frequency components of the spectral distribution are moderately distributed, indicating that the captured image is clear. Milk Qiuyou said that the spectrum has a high frequency component, and the image of Na is very clear. The output attribution function of the focus motor control voltage language term is shown in Table 3, in which the focus motor control voltage language term attribution function %(Dee·) is defined to indicate that the shortened focus 201005416 is from '/^(Medium) means moderate 'every coffee (10) off) Indicates the growth focal length

Table 2 image FFT linguistic item entry attribute function

Voltage language item output attribution letter

After determining the input and output attribution functions, the formula of the tiling Then. describes the image of the rich conversion result and the mirror focus control. The S fuzzy control rule is as follows: Define its i?Fi - IF fs is FFx and Afs is plus TiffiN ^ .g ^ Also: IF Ais ^ and from is minus ^ ^ ❹ Know: IF A is FF2 and ^ is pks TH£n ^ is ^ ^:IF fs is FF2 and^ is ^ TiffiN F / .§ ^ ^F5:IF fs is FF3 TfffiN vf is FL2 where the input variable /, is the rich conversion spectrum signal of the image, (4) the difference between the Fourier transform spectrum signals of the captured image at 6 and /2 , household ^ =,) / (6) research 'sand 2' item for the image of the Fourier transform spectrum signal is a function to read the scales her thorough round, Wei control focus, "Da / 3⁄4 '/3⁄4 ' 73⁄4 is the focus motor control The output of the voltage belongs to the function. In the fuzzy control rule that controls the focus motor, for example, the fuzzy control rule is known, 201005416 Ο If the miscellaneous Fu Li decorated fine spectrum 峨 < is FFi, the wire has more low frequency components. The composition is less, and when it is positive, its fuzzy round-off attribution function is ton, and the table does not need to be protected first or The age of sugar control, the growth of the material or the direction of the focal length is positive, and its mechanical scale control Wei surface loss does not need to change the magnitude and polarity of the voltage. For example, the fuzzy control gauge W' its output control is the same as knowing, just The amount of change of the heart is small. For example, the fuzzy control rule ^, if the shadow of the Fourier transform frequency axis U is the research, and the slave is negative, then the fuzzy output attribution function is ~ silk autumn increase silk weaving The direction is opposite to the direction of the positive focus. It is necessary to change the amount of control that was originally increased or decreased, that is, the size and polarity of the wheel must be changed, shortening or increasing the distance in the opposite direction to the original, and the corresponding focus. The light output of the motor control is, for example, the same as the output control of the module, but the amount of change of the microphone is small. Other fuzzy control rules are analogous. After determining the fuzzy control rule of the lens focus control, the fuzzy inference operation is performed. , ', 疋 模糊 模糊 健 Jianxian, its algorithm is expressed as follows··

Kf =^, URF2 u^3 uRfs (7) After obtaining the fuzzy relation rule of input and output, the present invention performs the synthetic calculus with the maximal nasal method, as follows: _ : = max{min{A(/i)5/U(^ )}} (3) =中/</;) is the attribution function of the Fourier transform spectrum of the image of the input attribution function 'generation_ is the fuzzy relation rule base'. After the fuzzy inference results are obtained, the weighted average is selected, and the fuzzy inference is obtained. The method is as follows: (4)

The result of the looseness is unambiguous, as the actual turn-off signal of the focus control unit: ° J FI,

ZRLP (4) is the result of the deblurring rotation, which indicates the control number, and ^ is the degree of attribution of the attribution function of the control letter of the age of the wire. Finally, the deblurred wheel:: turn: Mo: The output control voltage of the focus motor (as shown in equation (5): , *

Vf=-FI (5) [Fuzzy controller for lens aperture] The grayscale statistical graph of the digital image implies the image--some characteristics and information such as: grayscale (four) average value is the county level, the average value is higher It represents a more image of the party. Conversely, a lower average value represents a darker image, so the average value of the order value is proportional to the brightness of the image. The standard deviation of the grayscale value is the grayscale (10)_thin, and the high_nor_image indicates that the variation is high, and vice versa, the variation is low. Therefore, the fuzzy controller that uses the gray scale statistical map to design the lens aperture, if the average recording of the grayscale value of the image is 'or the light source is too bright or the aperture of the lens is too large, the aperture must be reduced; otherwise, the surface will be enlarged. Because the average value and standard deviation of the Wei image surface values are used as the two input function of the light __, the set of attribution functions is assumed to be G, • and horse respectively to determine the size of the lens aperture. The input variable _ is defined as the grayscale average of the image, as shown in the following equation: 201005416 - NxM (6) In equation (6), "^ is the total number of pixels in the image. If # = Λ/ of an image, the total pixel is additionally Define the wheeled variable % as the standard deviation of the image grayscale value. As shown: σ" [255 CTn=yi(gg„)2P(g) ^ ρω=ψ ❹, Ν (8) where (g) is the grayscale value of the image, which is the gray level of the pixel in the image. The value is the number of points of g, so Pfe) is the probability that the gray scale value of the image is g. The gray scale value of the image of the present invention is [〇~255]. The discretization of the input and output variables of the horse aperture control is as follows: Table 4 shows that the field of the image grayscale mean watt is [0~255], the standard deviation of the image grayscale value is [〇~128], and the field of the output variable aperture motor control voltage is [~8~~β] 〇",, Table 4 Aperture Control Input and Output Variables Discretization Comparison Table Discrete Position Image Grayscale Average 艮 Image Grayscale Value Standard σ„ Aperture Motor Control Voltage ^ -3 〇^g„<64 0^σ„<16 -2 —---- 64$ watts <96 16 鑫σ„<32 -1 96 彡 <112 32^σΛ<48 -4^ Vm<.2 0 —-------------- 112^^^144 48^ σ„ ^80 -2^Vm^2 +1 —----- 144<g„^ 160 80<σ„^96 2<Vl +2 --- 160<g„^192 96<σ„^112 4<Vm^6 +3 —--- 192<g„^255 112<σ„^128 6<Vi^8 The input attribute function G of the image brightness language item is as shown in Table 5. Among them, Table 12 201005416: like the Lai Ying County resignation (four) loses people to touch the money Dingcheng seven brightness 3 words touch pure, to miscellaneous _), ten know tears, ^ table 0 shot (ZR), Q indicates The outline, & indicates that G is very bright (PB). ❹ Table 5 y image input function of gray level mean language term

::ί, the value of the standard deviation language entry input attribute function such as table ❹ == order = difference round, function ::: two _, =: = standard deviation - means too low too high _, horse means after high (four) . (ZR)4 is not a peach S) 'Private indicates the round-in attribution function of the difference language item 0.6 0.5 0~ +1 ~+Γ +3 ~ ~Γ 0 0 ~0~ ~~〇~ 0.6 0 0 0 丄0.5 0 b 0-6 1 0.6 0 0 0.7 1 0.7 0 0.6 0.8 1 The round-off assignment function of the upper term is shown in Table 7. The 1 record of the ° item is defined as seven language items ugly 1_) ^6 〇~0~ J^6(PM) The aperture motor control voltage language will control the aperture motor control voltage 201005416, which means that the aperture opening is very small s), indicating moderate (four), /) _ f2 table is not small _, indicating shame indicates big _. 5 τ大(PS) 'Win means too large (PM)' After determining the input and rotation attribution function, use Zhuang style to describe the relationship between the job ash and the county _.THEN... The fuzzy control rule is as follows: the fuzzy control rule of the brother's aperture ^•IF gn isGiand σ j ❹ In the formula (9) /=1,2,.··7, Feng 2,..·7, (1) 2 "7: Coffee (4) (9) Time, "the average gray level of the image, the stalk door (four), 2, ..49, the input variable watt is the image _ side _ input _: ^ = touch difficult standard deviation , input attribution function '匕 is the aperture of the aperture fuzzy controller = gray: the standard deviation of the motor, is the aperture motor control fine output ^ ' to control the aperture to discrete scale +2 0.6 +3 ❹ 0 0.5 0

0J 1 into the attribution ^ ^ Invented the design of 2 groups of 7x7. A lose 1 genera function ' and designed 7 output attribution function key K $ as such 'by 2 sets of input attribution function, domain IF...THEN.&quot The fuzzy control rules of the model input are 4, ^ 3, and the fuzzy control rules are common, and are composed of R1, and IR2, ... and ir49, for example. : IF gn is G1 and σ„ is El THEN Vm is IRy 201005416

Rjr2 · IF gn is Gx and ση is E2 THEN VIR is IR6

Rm : IF gn is Gx and is THEN VIR is /J?6 i?iR47: IF gn is Gn and σ„ is Es THEN Vm is IR2 Riras : IF g; is G7 and ση is E6 THEN VIR is ir2

Rir49 : IF is Gn and σ„ is Εη THEN VIR is a fuzzy control rule that controls the aperture motor, such as the fuzzy control rule. If the grayscale average of the captured image is low, the image brightness is very dark (Gi), and The standard deviation of the grayscale value of the image is very low (when the time is five, the fuzzy output attribution function is Han?, indicating that the aperture must be opened to a large opening, and the output of the corresponding aperture motor control voltage is the factory' For example, if the fuzzy control rule and JR*9' have a high grayscale average value, the image brightness is very bright (Gy), and the grayscale value of the image is very high (stone), then the fuzzy output is The attribution function is / see, indicating that the aperture must be closed to a small opening, the corresponding aperture motor controls the output of the cake and other fuzzy control rules are analogous. The fuzzy control rules for the entire lens aperture control are shown in Table 8. Table 8 Fuzzy control rules for camera aperture state Ei e3 e4 E5 e6 Εη Gi 1R7 IRi IRe IR6 13⁄4 TR1 〇2 1K6 IKi IK5 IRs IRs IR4 IR^ (3⁄4 IRe IRs IRs IR4 IR4 TR, IR3— 〇α 1K6 IRs IR4 IR4 IR4 13⁄4' TR, g5 IRs 1K 5 1K4 IR4 IR3 IRs IR, g6 IRs IR4 1R3 IR3 IR3 TR. IR, Gy IR4 IR3 1R2 1R2 IR2 ir2 IR) After determining the fuzzy control rules of the lens aperture control, the fuzzy inference algorithm is used to determine the fuzzy relation rule base. The algorithm is expressed as follows: _ 15 201005416 ^mlR — VJ RIR2 U....^«48 ^ and 49 (j〇) After finding the fuzzy relation rule of the input and the round, the maximum-minimum calculus is used. The synthesis is calculated as follows: =max {min{(//(gn)η μ(σ„)), M(Rm!R) }} (11) Μ&1)&;Μσ") is the input attribution function, which represents the gray level average g of the image and the standard deviation of the grayscale value of the shirt image as the entire fuzzy relation rule base and (4) the attribution function. Ο Find the model inference _, and then use the weighted average to make the fuzzy inference derived from the fuzzy inference, as the actual output signal of the fine control, the method is as follows: 7 uIRi ς^λ (12: ΣΜη ίί /Λ/ In the formula () 4·«; ^ unambiguous output, indicating the actual circle control letter of the control aperture size m wire _ _ silk _ bribe, is the synthesis of the paste inference _ number The degree of attribution. The margin of the _ the lens used is F22c, the position of the silky initial recording. Finally, the output of the deblurred output is converted into the output control voltage PiR of the actual aperture motor, as shown in the following equation. : (13) [Fuzzle Zoom Blur Controller] The zoom blur controller of the size camera lens whose input variable & is defined as the target d, and the input attribute set is set to , and the input variable heart is defined as the target 16 201005416 The change of the moving speed of the object, and set its input attribute set to easy, the rounding variable c« is defined as the trustworthiness of the contour of the target obtained by the image processing, and the input attribute set is set to q, Using the above The set of belonging functions determines the size of the zoom target' to complete the zoom control of the lens. The present invention defines the input variable a as the horizontal and vertical length d) of the target projected to the x-axis and the less-axis at time, /y(g The diagonal line of the equation is as follows: an = (0 + ^(0 (14) Ο

The input variable is defined as the relationship of the moving speed of the target at time and G, as shown in equation (15): (15) (16) b _

MO in equations (15) and (16) where ' and b indicate the moving speed of the moving target at time & 匕! and Α(ί„-ι) indicate the focal length of the lens at time 4 and U, ^ is before and after The focal length ratio of the two images. The input variable c" is defined as the reliability of the target image. Let a characteristic value that is not the contour of the target, defined as the horizontal width of the target contour 々仏) and the horizontal width & 〇 «) plus the vertical width / / 匕), then the ?? As follows: (〇^c„<i> (17) where (18) y a '“Ο ” 4 (〇+冬(〇

Mn=c„-X„+{\-cn)Mn_x 17 (19) 201005416 In equation (17), 'not the estimated contour of the target at time ,, < and respectively at time ^ and 4 ι, the outline of the object stored in the memory buffer, and assume that its initial value = 0.5, that is, the initial value of the object's outline is 0.5. Therefore, when the stone is closer, the q increases. The discretization comparison table of the zoom control input variable is shown in Table 9, wherein the field of the input variable size A is [0~800], and the unit is pixel (pixel).

The subject of the change of the moving speed of the target is [04], and the domain of the trustworthiness of the target image is [〇~1]. The discretization comparison table of the zoom control output variable is shown in Table 1〇, where the field of the output variable zoom motor control voltage & is _8]. Table 9 Input Variable Theory of Zoom Control Discretization Comparison Table Discrete Position Target Size an Target Movement Speed Change Target Image Trustworthiness c„ -3 Ο$α„<30 〇^ό„<〇 .ι 0 guest c„<0.1 -2 30^α„<60 〇·1^ό„<〇.2 0.1^c„<0.2 -1 60^α„<80 〇.2^ό „<〇.4 0.2^c„<0.4 0 80^α„^160 〇·4^ό„^〇.6 0.4 ^c„ ^0.6 +1 160<α„^180 °·6<6„ ^〇.8 0.6 <c„ ^0.8 +2 "180<fl„S210 0·8<δ„^〇.9 0.8 <c„ ^0.9 +3 *- 一—. 210<α„^800 ° ·9<όΛ^ι 〇.9<c„^l Table 10 zoom control output variable field discretization comparison table discrete step zoom, motor control Ray r -3 -8^Fk-6 -2 -6g^< .4 -1 _4 outer <-2 0 +1 2<V/^4 +2 4<^/S6 +3 201005416 The language entry input attribution function of the zoom control target size is as shown in Table 11, which will The target size is defined as three language item attribution functions, which are small (Small), medium (Medium), and A3 (Large). Table 11 Language item of target size input attribution function 4 discrete level -3 -2 -1 0 +1 +2 +3 Cold (Small) 1 0.9 0.7 0 0 0 0 ^(Medium) 0 0 0.6 1 0.6 0 0 j3 (Large) 0 0 0 0 0.7 0.9 1 The language entry of the target speed of the zoom control changes the input function as shown in Table 12, in which the change relationship of the target moving speed is defined as the three language item attribution function. It means that the speed changes slowly (Slow), indicating that the speed changes moderately (Medium), indicating that the speed changes quickly (Fast) 〇 Table 12 The moving object speed change of the language item input attribution function is discrete level -3 -2 -1 〇 +1 +2 +3 5i(Slow) 1 0.8 0.5 〇〇〇〇^(Medium) 〇〇0.6 1 0.6 〇〇53(Fast) 〇〇〇〇0.5 0.8 1 Zoom Control Target Outline Confidence The language item input attribution function ❹ G is as shown in Table 13, in which the trustworthiness of the outline of the target is defined as two language item attribution functions, respectively, q for low trust (Bad) and C2 for high trust (Good) 〇 Table 13 object outline contour trustworthiness language entry input attribution function g散级级 -3 -2 -1 〇+1 +2 +3 Ci(Bad) 1 0.9 0.6 0.5 0.2 〇〇C2(Good) 〇〇0.2 0.5 0.6 0.9 1 Lens zoom motor control voltage language item output attribution function as table 14, which defines the lens zoom motor control voltage language term attribution function R indicates that the target is very 19 201005416 small (NB), must be enlarged a lot 'A table is not a target too small teacher ^, must be enlarged, & indicates the target Small _ 'to be enlarged' 匕 indicates that the target size is moderate (zr)\ indicates that the target is large (PS) and must be reduced. F6 indicates that the target is too large and wide, and must be reduced, and A indicates that the target is large (PB). It is necessary to reduce the number of reports. Table 14 Zoom motor control term system function laCNS) le(PM) -ϋ.5 0,8 1 Included 2 == attribution function, then fine... delete the mode _ mirror =::::= Change and external degree and indication: The relationship between the job and the fuzzy control rule is as follows: 1, 2, 3 is called ~SQ (4) (2〇) a A^-gdtpB 蛀,~'2'/== 1,2,...7, the round-in variable «is a function of moving the target variable at time ^, and when the time is at the time of the heart, the print 2 is the input of the moving target size. The change of 7 consecutive degrees 'material target shift can be #•任A, C is the day* is the target shape rim in the coffee 4

It is the change of the coffee: the two things of the device: the input attribute function of the round fineness. K reaches the number of control touch wheels_2, tree rhyme, [is the fuzzy type II description of the zoom horse, the design of the three groups of 3 coffee 2 20 201005416 input attribution function, and designed 7 output attribution The function, therefore, consists of three sets of input attribute functions...then...the fuzzy implied description of the fuzzy control rules consists of 18 'composed by the brother lKl8, for example: and zl and z2 Rz3 Η7 α» is Αχ and bn is B\ and c„ is C\ THEN Vz is 73 ^ is and bn is B2 and cn is Q THEN Vz is Y2 117 ««is and bn is B3 and c„ is Q THEN Vz is Yl ^zl6 and zl7 and zl8 ❹ ❹ is A3 and bn is Bx and c„ is C2 THEN Vz is γ5 〇n ls K is B2 and c„ is C2 THEN Vz is Y6

Jp . anlsA2mdbnisB3mdcnisC2THENVz isy7 ^ variable ",, motor fuzzy control rules, such as fuzzy control rules and a, if the moving speed, and the trustworthiness is low (6), Z is the model change: the horse two function is K, indicating The target must be slowly enlarged 'the corresponding sample...^The output of the crane is again, for example, _ control fineness. If the output movement speed changes rapidly (6) and the trustworthiness is high (Q), then Modularity, the relative function of A ' indicates that the target must be quickly reduced, and the output of the zoom motor control voltage must be reduced. The output of the control voltage of the zoom motor is other than the fuzzy control of the entire lens zoom control. As shown in Table 15. Main 1 C==秘秘...· State Bi b2 b3 Cl Ai r3 r2 Yi a2 >5 Ya Y3 —-- ~~ «1 — -JL__ Ye Yi C2 A\ r3 r2 Yl A2 r3 r4 Ys L'. a3 ys~ r6 Yi After the fuzzy control rule of the lens control of the lens of the camera, the fuzzy push 21 201005416 is used to calculate the fine Wei riding. The operation silk is as follows:

Kmz = R2l u Rz2 (J ......u Rzl7 u After obtaining the fuzzy relational rules of input and output, the calculus is synthesized by the maximal and minimum algorithms, as shown in the following equation: ^, ( =α„)η μφ„)η μ {〇η)) . Rmz , =m^{^man)nMibn)^M^cn)),M{Rm) }} (22) ❺ In the formula (22), the ground, and the second is the input attribution function. , respectively, represents the target size ~, the change relationship of the target moving speed \ and the target shape contour trust degree c", is the attribution function of the entire fuzzy relation rule base. After obtaining the fuzzy inference silk, the present invention is weighted and flat touched The fuzzy result obtained by the inference is unambiguous, and as the actual control of the Wei lion, the method is as follows: γ Σ~ is ~~J^T (23) The output of the solution is defuzzified in (23) As a result, it is indicated that the actual wheeling signal of the control zoom motor is at z., which is the appropriate degree of the result of the synthetic fuzzy inference, and the degree of attribution of the function of the generalized fuzzy inference result is 1. The camera of the present invention is used in the experiment. The length of the zoom lens ranges from 1' = 7.5 to 120 mm. Finally, the deblurred output is converted into the actual zoom motor. The output control voltage Γζ is as follows: (24) [Embodiment] The present invention analyzes the frequency domain signal of the captured image by using Fourier transform, and the average value and freshness of the image calculated by 22 201005416 ^ To set up the paste controller ^ Lai gang _ silk fine. _, will take the image of the axis target object 'after the image processing operation, you can get the target size, the speed of change, Ke trust, the target wheel scale three The variables, as the three input attribution functions of the fuzzy logic operation, are re-controlled, and the length of the focal length of the lens and the size of the object are zoomed to complete the automatic zoom tracking control of the lens. The following is a description of the specific technical features of the case and the specific technical features of the case (4). Figure 1 shows the control flow of the lens tracking control system. First, the moving object image obtained by the camera is imaged. After processing the program, the resulting image information will be processed to control the touch and the light, and the image will be moved to the appropriate size and clear image of the target. The purpose of tracking and lens control. Figure 2 is the control strategy of the lens tracking control system. Firstly, the camera system ❹ = positive 'the initial setting of the money recording system, the age is checked and the money starts to start photography = test 'search, target Object, and immediately determine whether the target exists in the image of the image of the image, if not 'continue to the next image; if it is, then enter the tracking and lens control system, when the target moves, tracking, And judge that the chasing object is consistent with the target jersey. If the two match, Na holds the lens control of the camera and controls the focus, aperture and zoom to zoom the target and adjust the target. Control to the appropriate size; if not, re-enter the photo detection mode and determine if the target exists? The lens control of the present invention comprises three parts, the first part is lens focus control 23 201005416 system, the second part is lens aperture control, and the third part is lens zoom control. The first part is the lens focus control. Firstly, adjust the focal length of the lens to too long, so that the target object is smeared. The difficulty of controlling the wheel shape is shown in Figure 3. In the front part of the figure, the front part indicates that the focus control is _4 volt. This battery age makes the focus motor Reverse shortening the focal length' The rear part of the b segment indicates that the focus motor is zero after the focus is completed. Fig. 3 is an enlarged wave paste of the & section in Fig. 3, and Fig. 5 is an enlarged waveform diagram of Fig. 3 tb. Then adjust the focal length of the lens to a short distance, so that the image of the target is blurred. After the focus control is completed, the target image is clearly visible. The output waveform of the focus control is shown in Fig. 6. Fig. 7 is the enlargement of the e segment in Fig. 6. Waveform diagram. The first part is the control of the lens aperture. Firstly, the aperture is adjusted to be moderate, so that the average value of the captured image is 112~Μ4, and the variation is BO4~_. The brightness is 4 fox. The aperture control experiment is divided into two experiments. At the first test, the illumination of the experimental environment is suddenly dimmed, so that the brightness of the experiment is 235 [like, the average value of the object f is 78.5, and the number of faces is ancestor 6, The detailed statistics are shown in Figure 8. After that, the control experiment of the lens aperture is started. After the aperture control is completed, the average value of the captured object image is 118 2, the variation number is period 8, and the gray scale statistical chart is as shown in FIG. 9 'The aperture waveform of the aperture control is as shown in FIG. As shown in the figure, before the d segment, the aperture control voltage is _36 volts. This voltage will cause the optical motor to reverse and increase the aperture. The front section of the e section indicates that the fine control voltage is +3.2 volts, indicating that the fine opening is too large. The aperture is reversed to reduce the aperture, and the rear of the g segment indicates that the aperture motor voltage is zero after the aperture control is completed. When the second aperture control experiment was carried out, the environmental background of the first completion of the aperture control experiment was taken as the reference. When the experiment was performed, the experimental environment was rayed bright, so that the brightness of the experiment 24 201005416 was 621 Lux, and the captured target image was taken. The average is 182 6, and the variance is 1302.8. The grayscale statistics are shown in Figure 11. After that, the lens aperture control experiment is started. After the aperture control is completed, the average value of the target image acquired is 128·5, the variation number is 2836.6, and the gray scale statistical graph is shown in Fig. 12, and the output voltage waveform of the aperture control is as shown in the figure. 13 shows the front part of the h segment indicates that the aperture control voltage is +36 volts. This voltage will cause the aperture motor to reverse and reduce the aperture. The front of the i segment indicates the aperture control voltage -1.5 volts, indicating that the aperture is reduced too much. Re-inverting and increasing the aperture, the rear of the k-segment indicates that the aperture motor voltage is zero after the aperture control is completed. The third part is the lens zoom control, and the zoom control is divided into two experiments. In the first experiment, the target was first adjusted to a too small state, and the image size of the captured object was 36.2 pixels. After that, the lens zoom control experiment was started. After the zoom control was completed, the captured target image size was 126.8 pixels. The target has a zoom control voltage output waveform from too small to as shown in Figure 14. The front part of the figure indicates that the zoom control voltage is -8.4 volts. This voltage causes the zoom motor to reverse the magnification of the target until the target Zoomed to the appropriate size '1 segment' rear section indicates that the zoom motor power is zero after the zoom is completed. When the zoom control experiment is performed for the second time, the target is first adjusted to an excessive state, and the target image size of the couch is 196.4 pixels. Then, the control experiment of the lens zoom is started. After the zoom control is completed, the image size of the captured object is 1465 pixels. The output voltage waveform of the target is controlled by an excessively large to moderate zoom. As shown in Figure 15, the front part of the figure indicates that the zoom control voltage is +9.6 volts. This voltage causes the zoom motor to reverse the target down until the target It is reduced to the appropriate size, and the rear of the segment indicates that the zoom motor voltage is zero after the zoom is completed. Figure 15 shows the control voltage waveform of the zoom control that is too large to moderate 25 201005416. In summary, the experimental results of the lens control show that the lens control algorithm proposed in the present invention is effective, and at the same time, the lens control can be continuously performed to obtain a clear and appropriately sized target image. As mentioned above, the present invention combines the new reliance, advancement and practicality, and meets the requirements of the invention patent case, and submits an application according to law. [Simplified illustration of the diagram] ❹ Figure 1 Control flow chart of the lens tracking control system Figure 2 Strategy flow chart of the lens tracking control system Figure 3 Output voltage waveform of the focal length too long to complete the focus control Figure 4 is the amplified voltage waveform of section a of Figure 3 FIG. 5 is an enlarged voltage waveform diagram of the b segment of FIG. 3. FIG. 6 is a short-to-short focal length waveform to complete the focus control. FIG. 7 is an enlarged voltage waveform of the c-segment of FIG. 6. FIG. 8 is a dark image of the aperture control experiment. Gray scale statistical chart Figure 9 Gray scale statistical image of the target image controlled by the aperture Figure 10 Voltage waveform for the dark image to the completion of the aperture control Figure 11 Gray scale statistics of the image during the aperture control experiment Figure I2 Complete the aperture Gray scale statistical chart of the controlled object image FIG. 13 is a control voltage pattern for the image control to complete the aperture control. FIG.

Claims (1)

201005416 X. The scope of application for patents: 1. A kind of camera lens from the surface of the focus and the face control of Wei, the fuzzy control method is used to design the controller of the lens; the spectrum of the image is used as the input of the lens focus control. The attribution function; the image gray level (four) mean value and standard deviation are used as the input attribution function of the lens aperture control; the lens zoom control is used as the lens size control according to the size of the target object, the change of the moving speed and the change of the contour of the image. The transfer attribute function, these round-in attribution functions use fuzzy inference calculations to determine the 〇=, fine and listen to the model ship, and use the max-minimum calculus as the fuzzy synthesis inference algorithm, and use the weighting Average (four) Soul (10) avemg__ The naval ship is difficult, and the thinning of the system is the rotation of the controller. Finally, the output of the de-turning is converted into the actual copper focus, aperture and zoom motor control voltage. 2. If the application is fine! According to the fuzzy control device for the rotation of the lens and the change of the lens, the image of the moving object obtained by the lens is processed by the image processing program, and the obtained image information is processed to control the focus, aperture and Wei of the camera. The size and clear image of the moving target are taken, and the feeding motor of the tracking device is controlled to achieve the purpose of tracking and lens control. 3. For example, the camera lens autofocus aperture and the fuzzy control of the zoom as described in the patent application scope, the fuzzy controller of Wei's automatic difficulty, the method of discriminating the domain of the input and output variables and the discretization of the language term attribution function , incoming fuzzy controller. 4. For example, the camera lens autofocus aperture and zoom 27 as described in the application for patents 〗 〖2010, the film is controlled by the woven fabric, the automatic light __ turn (four) device, the side with gray-scale system to set Wei _ die difficult, silk If the average value of the scorpion is high, it means that the light source is too bright or the aperture of the lens is too large, and the size must be reduced; otherwise, the aperture must be enlarged. 5·For example, the camera lens autofocus aperture and the zoom control of the zooming method described in the first paragraph of the patent. The tracking is difficult to control, the input variable & is defined as the target size, and the input is attributed to The function set is set to A, the input variable is defined as the target object speed, and JL sets its input attribute to and . The input variable G is defined as the trustworthiness of the target shape of the image processing. Degree 'and its input attribute set is set to &, the above-mentioned input attribution function set is used to determine the size of the zoom target, in order to complete the zoom control of the lens. 28