JP2767923B2 - Fuzzy control circuit - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention will be described in the following order.

A Industrial application field B Outline of the invention C Conventional technology (Fig. 7) D Problems to be solved by the invention (Fig. 7) E Means for solving the problems (Fig. 3) F function ( Fig. 3) G embodiment (G1) First embodiment (Fig. 1 to Fig. 6) (G2) Other embodiments H Effect of the invention A Industrial application field The present invention relates to a fuzzy control circuit, for example, This is suitable for application to aperture control of a television camera.

B. Outline of the Invention According to the present invention, in a fuzzy control circuit, high-precision control data can be obtained with a simple configuration by quantizing the membership functions so that sampling points are interchanged.

C Conventional Art Conventionally, some television cameras have been designed so that not only the aperture can be automatically adjusted but also the manual adjustment can be made (Japanese Patent Application Nos. 63-208825 and 63-213098).
No., Japanese Patent Application No. 63-215850).

That is, as shown in FIG. 7, in this type of television camera, the iris motor 2 drives the wing-shaped aperture.

The aperture is such that the opening amount thereof can be detected by the Hall element 3, and the iris motor 2 is driven based on the detection result.

That is, the comparison result between the output voltage V _K of the Hall element 3 and the reference voltage V _REF output from the reference voltage generation circuit 7 in accordance with the operation amount of the aperture control is obtained via the resistors 4, 5 and 6. The comparison result is output to the inverting input terminal of the operational amplifier circuit 10.

The operational amplifier circuit 10 includes an integrating circuit having a resistor 11 and a capacitor 12, and outputs an output signal of the integrating circuit to the iris motor 2.
Is output to the drive coil 13.

At this time, the operational amplifier circuit 10 is configured to input the reference voltage of the reference power supply 16 to the non-inverting input terminal via the resistor 14, and to connect the braking coil 18 to the inverting and non-inverting input terminals, respectively. ing.

On the other hand, the operational amplifier circuit 20 is configured by an integrating circuit having a capacitor 22 and a resistor 24, similarly to the operational amplifier circuit 10, and receives an output signal of the operational amplifier circuit 10 via an input resistor 21 at an inverting input terminal. Has been made.

Further, the operational amplifier circuit 20, like the operational amplifier circuit 10,
The reference voltage of the reference voltage 16 is input to the non-inverting input terminal via the resistor 26, and the output signal is output to the remaining input terminal of the drive coil 13.

Thus, at the aperture, the output voltage of the Hall element 3
As V _K coincides with the reference voltage V _REF, while being braked by the brake coil 18, adapted to be driven by the drive coil 13, thus setting a desired value determined the amount of opening of the aperture at the reference voltage V _REF Have been made to gain.

D Problems to be Solved by the Invention By the way, in this type of manual adjustment diaphragm, variations in the iris motor 2, the hall element 3 and the diaphragm, changes due to temperature characteristics, rattling, etc. cannot be avoided, and modeling is difficult. There is a characteristic.

Furthermore, while the electrical conversion characteristics of the image sensor become non-linear, the aperture can be adjusted only to the linear type. On the contrary, when the signal level of the image signal changes, the visual brightness becomes logarithmic. Change.

Therefore, in the conventional manual adjustment, it is difficult to make a quick and smooth response in accordance with the user's operation feeling, and the adjustment accuracy is still insufficient.

As one method of solving this problem, a method of controlling the aperture by applying fuzzy control can be considered.

However, in order to generate control data with high accuracy by applying the fuzzy inference, it is necessary to use a membership function with high resolution.

However, when such a high-resolution membership function is used, there is a problem in that the number of data expressing the membership function is increased, the memory circuit is correspondingly enlarged, and the configuration of the fuzzy control circuit is enlarged.

The present invention has been made in view of the above points, and has as its object to propose a fuzzy control circuit capable of obtaining high-precision control data with a simple configuration.

Means for Solving E Problem In order to solve such a problem, in the present invention, a plurality of membership functions NL, N
M, NS, ZR, PS, PM, PL data (D _{NL1 to} D _NL32 ), (D
_{NM1 ~D NM32), (D NS1} ~D NS32), (D ZR1 ~D ZR32),
_{(D PS1 ~D PS32), (} D PM1 ~D PM32), (D PL1 ~D PL32)
In the fuzzy control circuit 54 that infers the control data D _CONT by using a plurality of membership functions NL, NM, NS,
ZR, PS, PM, PL data _{(D NL1 ~D NL32), (} D NM1 ~D
_{NM32), (D NS1 ~D NS32} ), (D ZR1 ~D ZR32), (D PS1 ~
_{D PS32), (D PM1 ~D} PM32), (D PL1 ~D PL32) , the data a plurality of the membership functions NL, NM, NS, ZR, PS, PM, and PL quantizes (D _NL1 ~D _NL32) , (D _NM1- D _NM32 ),
(D _{NS1 to} D _NS32 ), (D _{ZR1 to} D _ZR32 ), (D _PS1 to
(D _PS32 ), (D _{PM1 to} D _PM32 ), and (D _{PL1 to} D _PL32 ) to obtain the first and second membership functions (NL and NM) rising up adjacent to each other, (NM and NS), (NM NS and Z
R), (ZR and PS), (PS and PM), and (PM and PL), the first membership function NL, NM, NS, ZR, PS,
Between the PM sampling point and the following sampling point, a second
Set sampling points for the membership functions NM, NS, ZR, PS, PM, and PL.

F action First and second membership functions (NL and NM), (NM and NS), (NS and ZR) rising up next to each other,
(ZR and PS), (PS and PM), (PM and PL)
If the sampling points of the second membership functions NM, NS, ZR, PS, PM, PL are set between the sampling points of the first membership functions NL, NM, NS, ZR, PS, PM and subsequent sampling points, , The adjacent membership functions NL, NM,
With the inference results obtained from NS, ZR, PS, PM, and PL, the accuracy of the inference results can be improved by complementing the overall inference results.

G Example Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

(G1) First Embodiment (G1-1) Configuration of the Embodiment In FIG. 1 in which the same reference numerals are assigned to parts corresponding to FIG. 7, reference numeral 30 denotes a television camera as a whole, and a lens 32 the focused light through the aperture 33 guided on the imaging surface of the solid-state imaging device 34, thereby obtaining an image signal S _S and forms an image of a subject on the imaging surface.

Analog digital conversion circuit (A / D) 35
Receives the output signal of the Hall element 3 via the amplifier circuit 36, and outputs the output digital signal (hereinafter, referred to as aperture amount data) to the control circuit 38.

In contrast, the analog digital conversion circuit (A / D) 4
0 is the reference voltage V output from the variable resistor 42 for aperture adjustment
Upon receiving _REF , the signal is converted into a digital signal and output to the control circuit 38.

The control circuit 38 creates control data D _CONT based on the aperture amount data and the reference voltage V _REF , and transmits the control data D _CONT via a digital / analog conversion circuit (D / A) 44 and a low-pass filter circuit (LPF) 46. Thus, the aperture 33 is controlled to an opening determined by the reference voltage V _REF .

At this time, the control circuit 38 creates control data D _CONT using fuzzy inference.

In other words, the fuzzy inference has a feature that it can be reliably controlled with a simple configuration even for an object that cannot be modeled.

Therefore, it is possible to reliably obtain a desired aperture with a simple configuration by applying to the aperture control which is difficult to model.

Further, depending on the setting of the membership function, the brightness can be controlled quickly and smoothly with sufficient accuracy in accordance with the operational feeling of the user, thereby improving the usability of the television camera as compared with the conventional case. be able to.

(G1-2) Configuration of Control Circuit 38 The control circuit 38 receives the aperture amount data and the data of the reference voltage V _{REF from} the subtraction circuit 50, and thereby calculates the deviation (hereinafter referred to as deviation data) θ of the aperture 33 with respect to the control target. It has been made to detect.

Further, the control circuit 38 receives the deviation data θ from the differentiating circuit 52, and thereby generates differential data Δθ of the deviation data θ.

The control circuit 38 supplies the deviation data θ and the differential data Δθ of the deviation data θ to the fuzzy inference unit 54, and the fuzzy inference unit 54 fuzzily infers the control data D _CONT .

(G1-2-1) Membership Function The fuzzy inference unit 54 has a read-only memory circuit, and the read-only memory circuit stores the deviation data θ, the differential data Δθ of the deviation data θ, and the data of the membership function relating to the control data D _CONT. Are stored to form a table.

That is, as shown in FIG. 2, in the membership function relating to the deviation data θ, the value of the deviation data θ is normalized from the value −1 to 1 from the largest value on the negative side to the largest value on the positive side. Are assigned to the seven membership functions (NL, NM, NS, ZR, PS, PM, and PL), which are expressed on the horizontal axis.

That is, in the first membership function NL,
The second membership function NM indicates a state in which the value of the deviation data θ is the largest on the negative side, whereas the second membership function NM is a state in which the value is slightly closer to the positive side.

On the other hand, the third membership function NS is
Represents a state in which the deviation data θ is closer to the positive side from the membership function NM of FIG.
Indicates a state in which the value of the deviation data θ is almost 0.

Further, the fifth and sixth membership functions PS and PM
Represents a state in which the value of the deviation data θ is sequentially increased from the state of almost 0 to the positive side, and the seventh membership function PL
Represents the state of the largest positive side.

Similarly, in the membership function regarding the differential data Δθ of the deviation data θ, the value of the differential data Δθ from the largest value on the negative side to the largest value on the positive side is normalized from the value −1 to 1, and this is transformed into a horizontal value. 7 for the axis
Two membership functions (NL, NM, NS, ZR, PS, PM and
PL) is assigned.

That is, in the first membership function NL,
The second membership function NM represents a state in which the value of the differential data Δθ is the largest on the negative side, whereas the second membership function NM is a state in which the value is slightly closer to the positive side.

On the other hand, the third membership function NS represents a state closer to the positive side, while the fourth membership function ZR represents a state of almost zero.

Further, the fifth and sixth membership functions PS and PM
Represents the state that increases sequentially from the almost zero state to the positive side, while the seventh membership function PL represents the state that increases to the positive side.

Similarly, in the membership function relating to the control data D _CONT , the control data D _CONT is normalized from a value of −1 to 1, and the resulting membership function is represented on the horizontal axis by seven member functions (NL, NM, NS, ZR, PS, PM and PL) are assigned.

That is, the first, second and third membership functions
In NL, NM, and NS, the control data D _CONT indicates a state in which the diaphragm 33 is narrowed, and indicates that the change speed is large, medium, and small, respectively.

On the other hand, the fourth membership function ZR
The fifth, sixth, and seventh membership functions PS, PM, and PL represent a state in which the diaphragm 33 can be opened by the control data D _CONT. In the case of, it represents a large case.

Each membership function (PL, PM, PS, ZR, NS, NM and
NL) is set to a function that rises into a triangular shape shifted by a predetermined amount with respect to the horizontal axis, except for the membership functions on both sides (that is, except for the membership functions NL and PL).

In addition, each member function (PL, PM, PS, ZR, NS,
NM and NL) are such that the triangular portion rises sharply as the central membership function rises, and becomes smaller as the distance of the rising portion or near the center.

Therefore, the deviation data θ, the differential data Δ of the deviation data θ
When the θ and the control data D _CONT are small, that is, when the user slightly changes the aperture of the aperture 33, when the user slowly changes the aperture 33, the aperture of the aperture 33 is slightly smaller than the adjustment target. The more the distance is shifted, the higher the accuracy of the result of the fuzzy inference can be obtained, whereby the aperture 33 can be adjusted with a higher degree of accuracy in accordance with the operational feeling of the user.

Further, conversely, when the deviation data θ, the differential data Δθ of the deviation data θ and the control data D _CONT are large,
In other words, when the user greatly changes the aperture of the diaphragm 33, when the user changes the aperture 33 quickly, or when the aperture of the diaphragm 33 is largely deviated from the adjustment target, the diaphragm 33 is controlled at a higher speed. As a result, the diaphragm 33 can be adjusted at a high speed at a speed according to the operation feeling of the user.

In particular, in the fuzzy inference, the accuracy of the detected data (in this case, the deviation data θ and the differential data Δθ of the deviation data θ, the analog digital conversion circuits 35 and 40
(Depending on the resolution) is high, it is possible to output control data with high accuracy.

Therefore, the analog-to-digital conversion circuit 35 with low resolution
And 40, highly accurate control data D _CONT can be obtained.

In addition, since the control data D _CONT with high accuracy can be obtained by using the analog digital conversion circuits 35 and 40 with low resolution, high accuracy can be obtained even with respect to variations in the aperture 33, the iris motor 2 and the Hall element 3. Control data
D _CONT can be obtained, and thus the aperture 33 can be controlled with high accuracy.

Further, since the center member function rises sharply and the distance of the rising portion is set to be small, the diaphragm 33 can be controlled smoothly even when the diaphragm 33 is narrowed down.

Actually, in this type of diaphragm, the amount of incident light changes in proportion to the area of the opening, so that the more the diaphragm is narrowed, the more precise the control becomes.

However, in the conventional control method, since the control is performed linearly, high-precision control cannot always be performed in a state where the aperture is stopped down.Thus, in this respect, the usability can be remarkably improved as compared with the conventional control method. it can.

In addition, each member function (NL, NM, NS, ZR, PS, PM
And PL), the triangular portion is selected so as to be left and right asymmetric with respect to the center membership function ZR, whereby the incident light amount of the image sensor 34 is not affected by the operation amount of the aperture adjustment operation element. The operation amount is changed to a linear shape so that the operation amount and the change in visual brightness match.

Thus, the aperture 33 can be adjusted with high accuracy and high speed in accordance with the operational feeling of the user.

Therefore, the usability can be remarkably improved in comparison with the related art.

As shown in FIG. 3, the deviation data θ, the differential data Δθ of the deviation data θ, and the membership function relating to the control data D _CONT are respectively divided into 32 at equal intervals on the horizontal axis with 5-bit data, and the divided horizontal axis coordinates Is used as a sampling point and quantized into 8-bit data.

At this time, in the first, third, fifth, and seventh membership functions (NL, NS, PS, and PL), data ( _DNL1 , _{......, D NLn, D NLn +} 1, ......, D NL32),
(D _NS1 ,…, D _NSn , D _{NSn + 1} ,…, D _NS32 ),
(D _PS1 ,…, D _PSn , D _{PSn + 1} ,…, D _PS32 ),
The sampling points of (D _PL1 ,..., D _PLn , D _{PLn + 1} ,..., D _PL32 ) are set to coincide with each other, whereby each data (D _{NL1 to} D _NL 32), (D _{NS1 to} D _{NS1 to} D _NS32 ), (D _PS1 ~
D _PS32), are made to represent the shape of the (at D _PL1 ~D _PL32), respectively the membership functions (NL, NS, PS and PL).

On the other hand, in the second, fourth, and sixth membership functions (NM, ZR, and PM), data (D _NM1 ,...,...) Representing the membership functions (NM, ZR, and PM) are displayed.
D _NMn , D _{NMn + 1} ,…, D _NM32 ), (D _ZR1 ,…, D _ZRn , D
_{ZRn + 1} ,…, D _ZR32 ), (D _PM1 ,…, D _PMn , D _{PMn + 1} ,
_.. , _DPM32 ) are the first, third, and fifth sampling points.
And data (D _{NL1 to} D _NL32 ) representing the seventh membership function (NL, NS, PS and PL), (D _{NS1 to} D _NS32 ),
The sampling points (D _{PS1 to} D _PS32 ) and (D _{PL1 to} D _PL32 ) are set so as to be shifted by 1/2 sampling pitch, whereby each data (D _{NM1 to} D _NM32 ), (D
_ZR1 ~D _ZR32), are made to represent the shape of the (D _PM1 ~D _PM32) in the membership function (NM, ZR and PM).

By doing so, the membership function that rises adjacently (in FIG. 3, the membership function PS
And data between PM, PM and PL), each data (D _PS1 ~
D _PS32 ) and (D _{PM1 to} D _PM32 ), (D _{PM1 to} D _PM32 ) and (D
_{PL1 to DPL32} ) can be set so that the sampling points are _interchanged .

Thus, in the fuzzy inference, even if one inference result obtained from the membership function is missing, the entire inference result can be supplemented by the inference result obtained from the membership function adjacent to the membership function. By expressing each membership function with data such that sampling points are interchanged,
The accuracy of the inference result can be improved accordingly.

Actually, when expressing each membership function without exchanging the sampling points in this way, in order to obtain the inference result with the same accuracy as when the sampling points are exchanged, the horizontal axis should be divided into 64, which is twice the division of 32. Must represent the membership function.

Thus, by quantizing the membership functions so that the sampling points are exchanged with each other, highly accurate control data D _CONT can be obtained with a simple configuration.

Incidentally, in this embodiment, data (D _{NL1 to} D _NL32 ) and (D _NM1 to D _NM1 ) representing each membership function are represented.
_{D NM32), (D NS1 ~D} NS32), (D ZR1 ~D ZR32), (D PS1
-D _PS32 ), (D _{PM1 -D PM32} ), and (D _{PL1 -D PL32} ), the horizontal axis is set to common address data and stored in the read-only memory circuit. In the inference using the fourth and sixth membership functions (NM, ZR and PM), normalized deviation data θ and deviation data θ
And the control data D _CONT are shifted by 1/2 sampling pitch to detect the values of the second, fourth and sixth membership functions (NM, ZR and PM).

Thus, the configuration of the read-only memory circuit is simplified, and highly accurate control data is obtained.

(G1-2-2) Rules for Fuzzy Inference The fuzzy inference unit 54 uses a predetermined function when performing fuzzy inference using the deviation data θ, the differential data Δθ of the deviation data θ, and the membership function for the control data D _CONT . Inference is performed according to rules R1 to R9.

That is, the technique for controlling the aperture 33 based on the deviation data θ and the differential data Δθ of the deviation data θ is based on the following rules RH1 to RH9 expressed in natural sentences including ambiguous words.
It is represented by

Rule RH1 If the deviation data θ is a large positive value, the diaphragm 33 is opened at a high speed.

Rule RH2 If the deviation data θ is a large negative value, stop the aperture 33 at a high speed.

Rule RH3 If the deviation data θ is a positive intermediate value and the differential data Δθ of the deviation data θ is almost 0, the diaphragm 33 is opened at a medium speed.

Rule RH4 If the deviation data θ is a small positive value and the differential data Δθ of the deviation data θ is a small positive value, the diaphragm 33 is opened at a small speed.

Rule RH5 If the deviation data θ is a small positive value and the differential data Δθ of the deviation data θ is a small negative value, the control of the diaphragm 33 is stopped.

Rule RH6 If the deviation data θ is a medium negative value and the differential data Δθ of the deviation data θ is almost 0, the diaphragm 33 is squeezed at a medium speed.

Rule RH7 If the deviation data θ is a small negative value and the differential data Δθ of the deviation data θ is a small negative value, the diaphragm 33 is squeezed at a small speed.

Rule RH8 If the deviation data θ is a small negative value and the differential data Δθ of the deviation data Δ is a small positive value, the control of the diaphragm 33 is stopped.

Rule RH9 If deviation data θ is almost 0 and deviation data θ
When the differential data Δθ is almost 0, the control of the diaphragm 33 is stopped.

Therefore, the rules RH1 to RH9 can be represented using the following rules R1 and R9 using symbols.

Rule R1 IF θ = PL THEN D _CONT = PL Rule R2 IF θ = NL THEN D _CONT = NL Rule R3 IF θ = PM AND Δθ = ZR THEN D _CONT = PM Rule R4 IF θ = PS AND Δθ = PS THEN D _CONT = PS rule R5 IF θ = PS AND Δθ = NS THEN D _CONT = ZR rule R6 IF θ = NM AND Δθ = ZR THEN D _CONT = NM rule R7 IF θ = NS AND Δθ = NS THEN D _CONT = NS rule R8 IF θ = NS AND Δθ = PS THEN D _CONT = ZR Rule R9 IF θ = ZR AND Δθ = ZR THEN D _CONT = ZR Thus, the fuzzy inference unit 54 is adapted to perform fuzzy inference according to the above rules R1 to R9. ing.

(G1-2-3) Fuzzy inference processing The fuzzy inference unit 54 uses the mamdami technique to perform fuzzy inference on the control data D _CONT .

That is, the fuzzy inference unit 54 detects the deviation data θ and the differential data Δθ of the deviation data θ, normalizes the detection result, and then detects the value of the corresponding membership function by referring to the table.

Here, the fuzzy inference unit 54 calculates, for example, a value as the normalized deviation data θ and the differential data Δθ of the deviation data θ.
When 0.6 and 0.1 are obtained, first, according to the condition of the rule R1, from the membership function (PL) on the deviation data θ,
Find the value (in this case the value 0 is obtained).

Further, the fuzzy inference unit 54 truncates the membership function PL relating to the control data D _{CONT which is the} consequent part of the rule R1 with the detected value (0) of the membership function.

As a result, the control data D
It is possible to obtain a membership function obtained by truncating the membership function PL relating to _CONT by a value 0 (that is, the membership function PL is an inference result of the rule R1).

Similarly, the fuzzy inference unit 54 detects a value from the membership function (NL) relating to the deviation data θ according to the condition of the rule R2 (in this case, a value 0 is obtained), and detects the value (0) of the detected membership function. And the membership function NL for the control data D _{CONT which is the} consequent part of the rule R2.
Cut off the head.

Subsequently, the fuzzy inference unit 54 detects a value from the deviation data θ and the membership functions (PM) and (ZR) relating to the differential data Δθ of the deviation data θ according to the condition of the rule R3 (for example, in this case, the values 0.7 and 0.8, respectively). Is obtained),
According to the condition of “AND”, a smaller value (0.7) is selected from the detected values (0.7 and 0.8).

Further, the fuzzy inference unit 54 uses the selected value (0.7) to set the membership function P for the control data D _CONT of rule R3.
M is truncated, and as a result, as shown in FIG. 4, a trapezoidal inference result obtained by truncating the triangular membership function PM at a value of 0.7 (that is, an inference result of rule R3) can be obtained.

Similarly, the fuzzy inference unit 54 sequentially outputs the deviation data θ and the differential data Δ of the deviation data θ according to the conditions of the rules R3 to R9.
The value is obtained from the membership function for θ, and each rule R3
After selecting the smaller value in accordance with the "AND" condition of .about.R9, the membership function for the corresponding control data D _{CONT is} truncated using the selected value.

Thus, as shown in FIG. 4, an inference result formed by the truncated group of membership functions can be obtained.

(G1-2-4) De-Fuzzy Processing When the fuzzy inference unit 54 obtains the inference result formed by the truncated membership function group, it executes the processing procedure shown in FIG. 5 to detect the center of gravity of the inference result. As a result, the inference result is subjected to a defocusing process, and a definite value of the inference result is detected.

Here, in the membership function relating to the control data D _CONT , data is stored by dividing the horizontal axis from value −1 to 1 into 32. At this time, the coordinate on the horizontal axis is represented by 5-bit data. Has been made to represent.

Accordingly, as shown in FIG. 6, in the fuzzy inference unit 54, as an inference result of the control data D _CONT (FIG. 6 (A)), the horizontal axis is set to 5-bit address data, and the value 0 is assigned to each address. An inference result (FIG. 6 (B)) in which the data from to is stored is obtained.

The fuzzy inference unit 54 stores the data of the inference result in the memory circuit, moves from step SP1 to step SP2, and sequentially accumulates the data of the inference result.

At this time, the fuzzy inference unit 54 sequentially stores the accumulated addition result in the memory circuit in accordance with the address of the addition data (FIG. 6 (C)). When the addition processing is completed, the process proceeds to step SP3, where the final address is stored. (In this case, the value is 2.5) is halved.

Further, the fuzzy inference unit 54 moves to step SP4,
After detecting the address at which the addition result of the value closest to the division result (in this case, the value is 1.25) is obtained (in this case, address 4 is detected, and the data at the address is hereinafter referred to as the centroid vicinity data). Subsequently, the process proceeds to step SP5, and subsequently detects an address at which an addition result of a value close to the division result is obtained (in this case, address 5 is detected, and this address is hereinafter referred to as an address of adjacent data).

The fuzzy inference unit 54 subsequently proceeds to step SP6 and loads the addition result of the data near the center of gravity and the adjacent data from the memory circuit (in this case, the addition result of the values 1.1 and 1.8 of the addresses 4 and 5 is detected, respectively). ), The addition result is linearly interpolated, and an address at which an addition result having a value equal to the division result is obtained is detected.

That is, in this case, the fuzzy inference unit 54 calculates After obtaining the value 0.21, the value 0.21 is added to the address 4 as expressed by the following equation 4 + 0.21 = 4.21 (2).
The address of 21 centroids is detected.

In this way, the address of the center of gravity can be detected with high precision equal to or less than the resolution of the membership function, and thereby a highly accurate definite value can be detected.

The fuzzy inference unit 54 creates control data T _CONT before normalization from the detected address of the center of gravity, outputs a drive signal to the iris motor 2 based on the control data D _CONT, and proceeds to step SP7. The processing procedure ends.

Thus, since the control data D _CONT can be created with a high precision equal to or less than the resolution of the membership function, even if the resolution of the membership function is set to a low resolution, it is possible to obtain high-precision control data, and as a whole, a simpler With this configuration, a television camera with high aperture control accuracy can be obtained.

Further, the process of dividing the cumulative addition value by half only requires shifting the cumulative addition data by one bit.

Therefore, in the defocus processing of this embodiment, a definite value can be detected substantially only by the cumulative addition processing and the arithmetic processing of the equations (1) and (2).

On the other hand, the position of the center of gravity is Therefore, if the arithmetic processing is performed as defined, since the membership function is divided into 32 in this case,
It requires 32 multiplications, 64 additions, and one division.

Therefore, according to this embodiment, a fixed value can be detected much more easily than in the case of performing arithmetic processing as defined, and a television camera having a simple configuration can be obtained as a whole. .

(G1-3) Operation of Embodiment In the above configuration, the result of detecting the opening amount of the stop 33 output from the Hall element 3 is based on an analog digital conversion circuit.
After being converted into aperture amount data via 35, the control circuit 38
Is input to

On the other hand, the reference voltage V _REF output from the variable resistor 42 for aperture adjustment is changed in voltage in response to the operation amount of the user, and is converted into a digital signal through the analog digital conversion circuit 40. The data is input to the control circuit 38.

In the control circuit 38, the aperture amount data and the reference voltage V _REF
Is subtracted by the subtraction circuit 50 and converted into deviation data θ.

The deviation data θ is directly input to the fuzzy inference unit 54, converted into differential data Δθ via the differentiating circuit 52, and then input to the fuzzy friction unit 54.

In the fuzzy inference unit 54, the deviation data θ and the differential data Δθ are normalized and then used for fuzzy inference according to the conditions of the rules R1 to R9.
The inference result of _CONT is obtained.

The inference result of the control data D _CONT is subjected to a de-fuzzy process to obtain a final value, and the iris motor 2 is driven based on the final value, thereby controlling the aperture 33 to the brightness desired by the user.

(G1-4) Effects of the embodiment According to the above configuration, the membership functions (NL and NM), (NM and NS), (NM and NS), (NS and
ZR), (ZR and PS), (PS and PM), (PM and PL)
Quantized by shifting by two sampling pitches, the first
Between successive sampling points of the membership function NL, NM, NS, ZR, PS, PM of the second membership function NM,
By setting the sampling points for NS, ZR, PS, PM, and PL, the adjacent membership functions NL, NM, NS, ZR, P
The entire inference result can be supplemented by the inference result obtained from S, PM, and PL, and thus, the accuracy of the inference result can be improved and highly accurate control data D _CONT can be obtained.

(G2) Other Embodiments In the above embodiments, the case where the aperture of the stop is detected using the Hall element has been described. However, the method of detecting the aperture is not limited to this, and various detection methods can be widely used. Can be applied.

Further, in the above-described embodiment, the case where the fuzzy inference is performed on the basis of the deviation data θ and the differential data Δθ has been described. However, the present invention is not limited to this. For example, the fuzzy inference may be performed with reference to the signal level of the imaging signal. May be.

Further, in the above-described embodiment, the membership functions (NL and NM) rising adjacent to each other, (NM and N
(S), (NS and ZR), (ZR and PS), (PS and PM), (P
M and PL are shifted by 1/2 sampling pitch and quantized. However, the present invention is not limited to this. For example, three successively rising membership functions (NL, NM and NS), (NM , NS and ZR), (NS, ZR and
(PS), (ZR, PS and PM), and (PS, PM and PL) are shifted by 1/3 sampling pitch and quantized, and between successive sampling points of the first membership function, the following second and second sampling points are performed. The sampling point of the third membership function may be set.

Furthermore, in the above-described embodiment, the case where the fuzzy inference is performed using the Mumdami method has been described, but the present invention is not limited to this, and various inference methods can be widely applied.

Further, in the above-described embodiment, the case where the present invention is applied to the case where the aperture of the television camera is manually adjusted has been described. However, the present invention is not limited to the television camera, but may be applied to various imaging devices such as an electronic still camera. This can be widely applied to manual adjustment of the aperture.

H Advantageous Effects of the Invention As described above, according to the present invention, by setting the sampling points of the subsequently rising membership function between successive sampling points of one membership function, highly accurate control data can be obtained with a simple configuration. A fuzzy control circuit that can be obtained can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a television camera according to an embodiment of the present invention, FIG. 2 is a schematic diagram showing a membership function, FIG. 3 is a schematic diagram showing its data, and FIG. FIG. 5 is a flow chart showing a procedure of a degassing process, FIG. 6 is a schematic diagram used for the description, FIG.
FIG. 1 is a block diagram showing a conventional television camera. 1, 30 TV camera, 2 iris motor, 3 Hall element, 33 aperture, 35, 36 analog digital conversion circuit, 38 control circuit, 50 subtraction circuit, 52 ... Differentiator circuit, 54 ... Fuzzy inference unit, 44 ... Digital analog conversion circuit.

────────────────────────────────────────────────── ─── Continuation of the front page (58) Field surveyed (Int. Cl. ⁶ , DB name) G05B 13/00-13/02 G06F 9/44 554 JICST file (JOIS)

Claims (1)

(57) [Claims] In a fuzzy control circuit for inferring control data by using data of a plurality of membership functions stored in a memory means, the data of the plurality of membership functions is obtained by quantizing the plurality of membership functions. In obtaining data, the sampling point of the second membership function is set between the sampling point of the first membership function and the subsequent sampling point in the first and second membership functions rising adjacent to each other. A fuzzy control circuit characterized by the following.