JPH03277082A - Controller - Google Patents

Abstract

PURPOSE:To reduce the scale of a calculation program and to decrease the processing time by providing 1st, 2nd arithmetic means and a drive means driving a controlled system and approximating an output membership function to control the controlled system. CONSTITUTION:Presume that a rule (a) in fuzzy deduction is considered for data A, B, then a probability Fx with a condition of a Big is obtained with respect to an input data (x) from a membership function defined by the data A at first. Similarly, a probability Fy with a condition of a Small is obtained with respect to an input data (y) from a membership function defined by the data B. Then a probability of a Middle of a membership function being an output C of the rule is obtained based on the probabilities Fx, Fy. Then a gravity center M of an area surrounded by an external size of the output membership function C obtained finally and the coordinate axes is obtained and its x-coordinate is the speed and the direction of a focus motor.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a control device, and particularly to a control device suitable for application to control using fuzzy inference.

(Prior art) In recent years, in the field of various types of control, in order to perform control that is more natural and adapted to the situation, ambiguous parts are incorporated into control to perform control that is more natural and adapted to the situation. So-called fuzzy inference is used to obtain

In such a control system, the degree of compliance with preset conditions is determined using a predefined membership function for multiple pieces of control information, and control is determined based on the results and a predefined output function. It calculates quantities.

That is, a plurality of pieces of control information are detected, each piece of detected information is applied to a membership function prepared in advance, and checked against a preset rule. Then, the degree of compliance with the rule is quantitatively determined, the result is substituted into the output membership function, an output membership function is generated that takes each situation into consideration, and the constraint quantity is calculated. Specifically, a control output that reflects various ambiguities is obtained by calculating the center of gravity of the outer shape of the output membership function.

In this way, control using fuzzy inference is comprised of steps such as expansion of membership functions, rule (condition) determination, and defuzzification process for calculating control variables.

On the other hand, all kinds of objects can be controlled that include such ambiguity, but in the field of imaging devices such as video cameras, for example, the shooting conditions are extremely diverse, such as the focus adjustment device, and it is difficult to accurately In a system that requires the operator's intuitive judgment in order to perform accurate control, it is extremely effective to perform human control that incorporates the above-mentioned ambiguity.

Also, of course, this kind of control is not only effective for imaging devices, but can be applied to devices in all fields regardless of the type of control target, but miniaturization and cost reduction are important in all fields. This is an essential condition, and the demand is particularly strong in the field of small electronic consumer devices such as home appliances.

(Problems to be solved by the invention) However, according to the conventional control system described above,
When fuzzy inference is used for the control, a relatively large-scale system is required for the defuzzification process of expanding membership functions, determining rules (conditions), and calculating control variables. This is because the defuzzification process of expanding the membership function and calculating the control amount requires sophisticated and complicated calculation processing.

Specifically, in the process of defuzzification, the centroid of the outline of the output membership function is obtained to obtain the final result, but the outline of this output membership function changes depending on a number of rule compliance conditions, and it is usually Since the shape is extremely complicated, calculations for determining the center of gravity of the outer shape are also complicated and cannot be performed easily.

In particular, with the miniaturization of various devices in recent years, there has been an increasing need to use single-chip microcontrollers to perform fuzzy inference on software. The constraints are thick (,
Therefore, when attempting to accurately perform the above-mentioned center of gravity calculation, a major problem arises.

(Means for Solving the Problems) The present invention has been made to solve the above-mentioned problems, and is characterized by a control that calculates input information and outputs the H amount of the controlled object. a first calculation means that quantitatively calculates the degree to which the input information conforms to preset conditions based on a predetermined function and generates an output function for controlling the control object; a second calculating means for performing a gravity center calculation on the output function output from the first calculating means; and a driving means for driving the controlled object based on the output of the second calculating means, The function is located in the control device as an approximation function that is subjected to approximation processing so as to be vertically symmetrical with respect to the center value of the input data value.

The present invention is also characterized by a control device that calculates input information and outputs a control amount of a controlled object, the control device calculating the degree to which the input information conforms to preset conditions based on a predetermined function. a first calculation means for quantitatively calculating and generating an output function for controlling the controlled object; and approximation processing for the outer shape of the function so that it is vertically symmetrical with respect to the center value of the input data value. an approximation processing means for performing a centroid calculation on the approximate function; a second calculation means for performing a centroid calculation on the outer shape of the output function output from the first calculation means; and a drive means for driving the controlled object based on the output.

Further, the present invention is characterized by a control device that calculates input information and outputs a control amount of a controlled object, which performs predetermined approximation processing to determine the degree to which the input information conforms to preset conditions. a first calculation means for calculating an output membership function for controlling the controlled object based on the input membership function obtained by the calculation; and an output membership function output from the first calculation means. The control device is characterized by comprising: a second calculating means for calculating a center of gravity; and a driving means for driving the controlled object based on the output of the second calculating means.

(For I) In this way, by approximating the output membership function for controlling the controlled object, the calculation can be greatly simplified without reducing the accuracy of the fuzzy inference output. It becomes possible to reduce the size of the calculation program capacity and shorten the processing time.

(Embodiment) Hereinafter, an embodiment of the control device according to the present invention will be described in detail with reference to each figure, taking as an example a case where the control device of the present invention is applied to an automatic focusing device such as a video camera.

First, an outline of the automatic focus adjustment device targeted in this embodiment will be explained.

There are various types of automatic focusing devices for cameras, but in devices such as video cameras and electronic still cameras that have an imaging means that photoelectrically converts the subject image to obtain a video signal, A method is known in which the focus is adjusted by detecting the sharpness and fineness of the image and controlling the position of the focusing lens so that the sharpness and fineness are maximized.

A method has been proposed for controlling the focusing lens driving speed in such a focusing device by detecting the blur width of the edge portion of the subject image and changing the lens driving speed according to this detected value ( For example, Japanese Patent Application No. 1-77063 filed by the present applicant).

However, in this method of detecting focus by extracting a signal component corresponding to the focus state from the video signal, the signal component fluctuates depending on the subject and its environment, making it difficult to perform accurate and natural focus adjustment. Tend.

Fuzzy reasoning, which has recently been used in various control fields, can be effectively applied to such controls that contain ambiguity.

In such a case, for example, there is a method that processes the high-frequency signal components in the video signal and the detected blur width of the subject using fuzzy inference, determines the speed and direction of the focus morph for driving the focusing lens, and performs automatic focus adjustment. Conceivable.

Specifically, for example, membership functions are defined for the detected blur width of the subject and the high-frequency component signal values of the video signal, and the speed of the focusing lens is calculated by performing adaptation calculations for each rule based on these functions. It is something to control.

FIG. 2 is a block diagram showing an embodiment of the automatic focus adjustment device according to the present invention, and FIG. 3 is a flowchart for explaining the control operation of the logical control device 11 that controls the entire system in FIG.

In Fig. 2, 1 is a photographing lens system, and a focusing lens group (hereinafter referred to as focusing lens) la
and a zoom lens group (hereinafter referred to as a zoom lens) lb that performs zooming. The focusing lens 1a is
It is driven in the optical axis direction by a focus motor 16 via a focus drive circuit 15 according to a command from a logic control device 11, which will be described later. Similarly, the zoom lens 1b is driven in the optical axis direction by a zoom motor 20 via a zoom drive circuit 19 according to a command from a logic control device 11, which will be described later. 2
is an aperture (iris) that controls the amount of incident light, and drives the iris drive circuit 13 via the iris control circuit 12 so that the video signal level is always constant.
It is activated by operating the .

The incident light is from the focusing lens 1a and the zoom lens 1b.
, is incident through the aperture 2, and solid-state image sensor 3 such as CCD.
The signal is imaged on the imaging surface of , converted into an electrical signal, amplified to a predetermined level by the preamplifier 4 , and then supplied to the process circuit 5 . The process circuit 5 performs predetermined signal processing on the input imaging signal, converts it into a standardized television signal, and outputs the signal to a video output terminal, an electronic viewfinder (EVF), or a VTR integrated camera (not shown). In this case, the signal is output to the recording signal processing circuit of the VTR, etc.

The subject discrimination filter 6 is a full range filter that is set to be able to determine the magnitude of the contrast of the subject from the video signal output from the preamplifier 4, and cannot accurately determine the focus on the subject and control the focus morph drive speed. In the case of extremely low-brightness or high-brightness objects, the control may be canceled or reset, or as described later,
Information for controlling the setting of membership functions for determining focus determination and focus morph drive speed is output.

The bandpass filter 7 extracts high frequency components for focus detection from the video signal.

Furthermore, the output of the preamplifier 4 is led to a blur width detection circuit 8, which calculates the blur width of the subject on the imaging surface. This blur width detection circuit 8 uses information corresponding to the width of the edge portion of the subject image, and the closer the in-focus state is, the smaller the width of the edge portion of the subject becomes. Therefore, focusing is performed so that this edge width is minimized. Focus adjustment can be performed by driving the lens 1a.

This itself is proposed by the present applicant, for example, in Japanese Patent Application Laid-open No. 62-10
3616, JP-A No. 63-128878, and the like.

The output of the whole range filter, high frequency component signal, and blur width signal described above are determined by the focus detection area (rangefinding frame) within the imaging plane.
passes through a gate circuit 9 for setting the peak detection circuit 10,
A peak value or an integral value within a limited distance measurement frame for distance measurement within the screen is obtained, and is supplied to a logic control device 11 constituted by, for example, a microcomputer, which comprehensively controls the operation of the entire device. . In addition to this, the logic control device 11 includes a focusing lens encoder 18, an aperture encoder 17, and a zoom encoder 19.
Detected values are also supplied and used for various controls.

Further, it is desirable to set a distance measurement area (focusing area) according to the depth of field and the degree of focus. Therefore, the logic control device 11 detects the depth of field from various encoder information provided in the optical system, and sets a focusing area suitable for the depth of field (sends the area switching signal to the gate circuit 9). For example, if the depth is deep or the degree of focus is low, set the area wide,
Conversely, when the depth is less than t or when the degree of focus is high, the area is set narrow.

The logic control device 11 uses time-series changes in this information.
Then, the driving speed, direction, stop, restart, etc. of the focusing lens 1a are determined, and the focus driving circuit 15
A control signal corresponding to the determination result is output to drive the focus motor 16 to move the focusing lens 1a.

Next, the flow of control operations performed by the logic control device 11 will be explained using FIG.

In FIG. 3, 3tep 1 is the subject discrimination filter 6, bandpass filter 7, and blur width detection circuit 8 in FIG.
This is a routine that A/D converts 8 analog signals for each field and calculates the desired value for the focusing operation. Calculate the difference value.

5tep 2 is a routine that determines the operation mode of the focusing control flow and shifts the control to a predetermined routine.

A block 20 consisting of 5tep3 and 5tep4 is one of the operation modes, in which it is determined whether or not the focus motor 16 should be restarted after focusing based on the change in the input signal, and the block 20 is set to a direction 1 focus determination mode or a zoom start mode, which will be described later. This routine is called restart mode.

A block 21 consisting of 5tep8 and 5tep9 is a routine that performs processing using fuzzy inference to actually control the speed of the focus motor 16 and determine focus using fuzzy inference, and is the center of the focus control flow. This is called inference mode. This will be explained in detail later.

Block 22 consisting of 5 tep 12 to 5 tep 14
is the direction and focus determination mode, and when the restart mode 20 is determined to restart, the camera shifts to this mode. Here, it is reconfirmed whether or not the focus is in focus, and the restart direction of the focus motor 16 is determined. When it is determined that the object is in focus, the control mode is shifted to the inference mode.

Block 23 consisting of 5 steps 18 is called a zoom start mode, and is a routine for restarting the focus motor 16 when the restart conditions are satisfied during zooming to the telephoto side, restart mode 2o, direction 1 focus judgment mode 2
2, control is transferred to inference mode 21.

5tep 5~5tep 7.5tepl O, 5te
pl 1, 5tep15-5tep17.5tep1
9 is a routine for setting a control mode based on the results of the four control routines described above.

5tep20 is a routine for actually driving the focus motor 16 according to the drive speed and direction of the focus motor 16 determined from the above results.

With the above configuration, data such as the high frequency component detection value of the video signal, the blur width detection value of the subject image, the high frequency component difference value in the video signal, and the difference value of the blur width signal of the subject are captured in step 1. Subsequently, in step 52, the control mode is determined, a control routine corresponding to the focus control mode is selected, and control is transferred.

Restart routine 2 if restart mode is selected
0, and in step 3, it is determined whether or not a zoom operation is being performed. If it is determined that a zoom operation is being performed, the process advances to step 7, and the restart is not determined, and control is resumed. Set zoom startup mode to mode,
Perform zoom operation. After setting the control mode, 5 steps
The process advances to step 20 to drive the focus motor.

If it is determined in step 3 that no zoom operation has been performed, the process proceeds to step 4, where it is determined whether or not the focus is out of focus based on changes in the input signal data, that is, whether or not to restart the camera. be done. If it is determined that the focus state is out of focus and a restart is necessary, the process proceeds to step 5, where the restart mode 20 is set as the control mode, and automatic focus adjustment is restarted.

Further, if a restart determination has not been made in 5tep4, the control mode is set to the direction focus determination mode in 5tep5, and the process proceeds to 5tep20, where the drive of the focus motor 16 is controlled.

If it is determined in 5 step 2 that the fuzzy inference mode is set as the control mode, the process proceeds to fuzzy inference mode 201, and in 5 step 8, the driving speed for driving the focusing lens 1 to the in-focus point is set. Then, the focusing lens is driven, and the focus is determined at 5 steps 9. If it is determined that the focus is in focus, the focus is determined at 5 steps 9.
Proceed to step o, set the restart mode 200 as the control mode, and drive the focusing lens in 5tep20.

Also, if it is determined that the focus is out of focus at 5tep9, 5tep9
Proceed to pl 1, set the fuzzy inference mode as the control mode, and then set the focusing lens drive speed based on the prediction of the in-focus point by fuzzy inference.
At 20, the focus motor is driven.

If it is determined at 5 tep 2 that the mode is the direction 9 focus determination mode 202, the restart direction of the focusing lens is determined at 5 tep 2. This two-direction focus determination mode is set only when the restart mode is set in step 5 of the restart mode 20.

After the restart direction of the focusing lens is determined in 5tep12, it is determined in 5tep13 whether or not a zoom operation is being performed. If the zoom operation is in progress, the process advances to 5tep15 and zooms to control mode. The control mode is set, and the process advances to step 520 to drive and control the focus motor 16.

In addition, if it is determined that the in-focus state is reached in 5tep14,
In step 17, the fuzzy inference mode is set as the control mode, and in step 520, the focus motor 16 is driven.

Also, if it is determined in 5tep14 that it is out of focus,
At 5tep16, the direction 1 focus determination mode is continued, and the process returns to 5tep20 to drive the focus motor 16.

If the zoom startup mode shown in block 23 is set at 5tep 2, the zoom lens is driven in the zoom startup routine at 5step 18, and then at 5tep 2, the zoom lens is driven.
After setting the fuzzy inference mode 22 as the control mode in step p19, the process advances to step 520 and drive control of the focus motor 16 is performed.

After each mode has been set and the focus motor has been driven, the process returns to step 1, the input data is updated, and the above-described control flow is repeated.

Next, step 8 in FIG. 3, which constitutes the main part of the present invention.
The focus motor 16 according to the fuzzy inference mode shown in
The control operation will be explained using FIG. 4, FIG. 5, FIG. 6, and FIG. 7.

FIG. 4 is a control flow of the focus motor based on fuzzy reasoning.

5tepl O1 is a data input routine of a membership function used for rule determination of 5tepl O2.

This is similar to step 1 in FIG. 3, in which the high frequency component detection value in the video signal, the subject blur width detection value, and the difference value therebetween are input.

5tepl O2, 5tepl O3, 5tepl O
4 uses these data to perform fuzzy inference and finally determine the speed and direction of the focus motor.

Specifically, fuzzy inference is performed as shown in FIG. In other words, if a rule such as a) is considered for certain data A and B, first, from the membership function defined for data A, Bi
The probability Fx of the condition g is found, and similarly for data B, the probability Fy that the input data y is Small is found.
is required. Next, the probability called Middle of the membership function C, which is the output of the rule, is calculated using Fx
, Fy.

In other words, since the inputs of the rule are Fx and Fy, Fx and Fy
By applying the smaller value of C to the membership function of C, the probability line of Mi dd 1 e is compressed as shown in Figure 5(c), and the outer shape and coordinate axis of the finally obtained output membership function C are The center of gravity value M of the enclosed area is determined, and its X coordinate becomes the speed and direction of the focus motor 16.

If we return to Figure 2 and consider what has been said above, we find that 5
The membership calculation of tepl O3 is performed based on the membership function set in advance in the logic control unit, based on each condition of the input data (for example, the Big,
Routine to find the probability of Sma11 etc.), 5tepl
02 is determined by a routine that calculates the logical sum or logical product of the probabilities of each membership function, 5tepl O
The output calculation in step 4 is a routine that calculates the probability of the output membership function from the logical sum or logical product of each membership function calculated in 5tepl O2, calculates the center of gravity of the outer shape, and determines the speed and direction of the focus motor 16. You can think about it.

For convenience of explanation, the above is based on two input membership functions, 1
Although a determination method using one rule including six output membership functions has been shown, in this example, in reality, as shown in FIG. The speed and direction of the focus motor 16 are determined by the ship function and 13 rules as shown in FIG.

In this way, when there are multiple rules, the output membership function consists of the outer shape that overlaps all the outer shapes at the time when all the conditions have been judged, and calculates the speed and direction of the focus motor that will be the output from the center of gravity. There is.

In FIG. 7, (a) is the membership function of the detected blur width ES of the edge part of the object, and S (Small
), M (Middle), and B (Big).

Figure (b) shows the membership function of the difference value dES of the blur width ES value, that is, the amount of change, and the NB (Negative
B i g), P B (Positive B
ig), NS (Negative Sma
l1), PS (Positive Small
membership function that evaluates l), 20 (zero)
is a membership function that indicates the probability of a range near the 0 point.

FIG. 6(c) shows the membership function of the high frequency component detection value FV in the video signal, and the probability in each region is determined for the two functions S (Small) B (Big).

Figure (d) shows the membership function for determining the difference value dFV of the detected high frequency component value FV, that is, the state of the amount of change.
As in b), its value is evaluated by a function representing the probability of each region of NB, N520, PS, and PB.

Figure (e) shows the membership function of the value PFMS corresponding to the delay time after the output of the speed command when reversing the focus morph and the result appears, and is in the negative range N (
Negat i ve), positive range P (Positi
ve), the values of three areas in the range ZO near O are evaluated.

Figure (f) is the membership function of the rotation direction FMDIR of the focus motor, and similarly to Figure (e), the negative range N
(Negative), range near 0 20. The probabilities of three ranges of positive range P (Positive) are evaluated.

(g) of the same figure shows the membership function of the speed FMS of the focus motor 16 as an output, and in the negative (N) direction, NB (NegativeBig), NM (Nega
tive Middle), NS (Negative
The three speed regions of e Small 1) are PB (Positive Big),
PM (Positive Middle), PS
Three types of velocity regions (Positive Small) are each set with a range ZO in the vicinity of O, and the probability of each region is determined.

Then, for these six membership functions and one output membership function, focus motor speed FM is determined according to each of the 13 rules shown in FIG.
The probability under each condition of S is set, and the probability data obtained by each membership function is compared with the output membership function shown in FIG. 6(f). Then, the center of gravity set by the probability value of each of these membership functions is determined in FIG. 6(f), and the focus motor is driven using the focus morph speed EMS corresponding to the position of the center of gravity.

As a result, in contrast to conventional binary control, all the probability data calculated in each membership function for each setting condition is taken into consideration, and by weighting them to obtain an output that becomes the center of gravity, it is possible to The most suitable natural focus morph control can be performed.

The above is an example of focus morph control using fuzzy inference, and the above-mentioned membership functions are set in advance based on development research and experience.

However, it is known that the dynamic range of the high-frequency signal component used here varies greatly depending on the subject. Furthermore, even if the object blur width signal is normalized by the contrast so that it is not affected by the contrast, there is actually a problem that the S/N ratio deteriorates due to the low contrast and the signal is influenced by noise, and the dynamic range fluctuates. That is, a subject with low contrast has a small dynamic range, and a subject with high contrast has a wide dynamic range.

Therefore, if the membership function is fixed, even if the probability determination is appropriate for one subject, it may be inappropriate for another subject.

Therefore, in the present invention, the object discrimination filter shown at 6 in FIG. Focus motor control is performed using membership functions.

This will be explained in more detail below using figures.

FIG. 8 is a control flow of the focus motor using fuzzy reasoning according to the present invention. 5tep 201 is ~5t
The control operation itself shown in ep204 is 5t in FIG.
epl O1 performs the same operation as the routine shown in ~5tep104.

Here, a mode determination routine shown in step 205 is further added. In this routine, the output of the object discrimination filter 6 shown in FIG. is detected, and when the value is below a certain threshold, it is determined that the subject is a low contrast object, and the membership function of the detected blur width value ES and high frequency component value FV is adjusted to make each detected value as large as possible, as shown in Fig. 9 (a). (In order to increase the probability of detection, each area of the function is shifted to the Small side, as shown by the arrow in the figure. This also reduces the deterioration of detection accuracy due to low contrast, even for objects with low contrast. It is possible to reliably determine the state of the subject and make a focus determination without being affected by the effects of the camera, and to set the drive speed of the focus motor with high accuracy.

Contrary to the above, when the output of the subject discrimination filter 6 is above a certain threshold, it is determined that the subject is a high contrast subject, and the membership function of the detected blur width and high frequency component values is adjusted as much as possible, as shown in FIG. 9(b). In order to increase the probability of determining each detected value to be small, each region of the function is shifted toward the Big side as indicated by the arrow in the figure. This can prevent erroneous determination of the in-focus state due to high brightness.

If the subject is other than the above, that is, if the subject is determined to be a normal subject, a membership function set in an intermediate region as shown by dotted lines in FIGS. 9(a) and 9(b) is used.

By the way, fuzzy inference is originally suitable for control that involves ambiguity that is difficult to express in binary terms, but in situations where the input detection data itself changes depending on various conditions, the reliability of the calculation itself may be affected. Therefore, the present invention solves this problem by changing the membership function settings according to the contrast of the subject, and this invention solves this problem by changing the membership function settings according to the contrast of the subject. Natural control can be exercised.

As described above, in focus motor control using fuzzy inference, by changing the input membership function depending on the contrast of the subject, optimal focus motor control can always be performed regardless of the subject, and the features of fuzzy control can be maximized. This makes it possible to achieve the same comfortable focusing operation.

According to the above-described embodiment, the input membership function is corrected as a result of object discrimination, but the present invention is not limited to this, and for example, the output membership function may be corrected.

Also, instead of changing the membership function settings, the rules may be changed. In other words, for a low-contrast subject, set the membership function conditions for the detected blur width value and high-frequency component value to Middle for Big, and
e is lowered to Small, and the output membership judgment is also lowered one step at a time to slow down the focus motor speed as much as possible to prevent hunting, and in the case of high contrast subjects, On the contrary,
Judgment of membership of blur width detection value and high frequency component value
Set the mall to Middle and the Middle to Big, raise the output membership judgment by one step, and make the mountain climbing operation as fast as possible, so that the focusing lens stops midway with a blur. This makes it possible to prevent such malfunctions.

Furthermore, in the above-described embodiment, a method was described in which the contrast of the subject is determined by the subject discrimination filter 6, and the settings of the membership function or rule of fuzzy inference are changed especially for low-luminance subjects and high-luminance subjects.
Especially when the condition of the subject is poor, the accuracy of focus judgment and focusing lens drive speed control will also decrease.
It is also possible to use an algorithm that cancels these control operations or resets them to the initial state and performs control again without changing the membership functions or rules.

In addition, in the above embodiment, the membership function of the high frequency component detection value, the blur width detection value, or the speed of the focusing motor that is the output, which changes depending on the contrast of the subject image, is made variable. Even more detailed and highly accurate control can be achieved by making the membership function settings variable for elements whose characteristics change greatly depending on the external environment.

According to the embodiment described above, the speed of the focus motor 16 is determined by six input membership functions as shown in FIG. 6, one output membership function, and 13 rules as shown in FIG. If there are multiple rules like this, the output membership function is
It consists of the outer shape obtained by superimposing all the outer shapes at the time when all the conditions have been determined, and the speed and direction of the focus motor which is the output from the center of gravity are determined.

However, the outline of the input membership function is usually the first
As shown in FIG. 6(a) and FIG. 6, it has a complicated shape. If an attempt is made to collate this shape ni input information and perform a predetermined calculation using software, the scale of the calculation program will be enormous, the processing time will be long, a large amount of memory capacity will be required, and calculation accuracy will be reduced. Problems such as difficulty in increasing the memory size arise, making it difficult to realize a system that is high speed, highly accurate, and requires only a small memory area. In particular, when control is performed using a one-chip microcomputer for control, the effects of these problems become noticeable, making it difficult to perform high-speed and highly accurate output calculations.

The present invention solves this problem, simplifies the computation of membership functions that are inherently complex, reduces processing time (and reduces program capacity), and reduces the occupied memory area. This method enables control by software-based fuzzy inference without reducing accuracy and calculation accuracy, and each calculation by this fuzzy inference can be greatly simplified by approximation processing of input membership functions as described later. It is something that can be done.

Among the input membership functions shown in Fig. 6, the input data are dES (difference value of object blur width signal), dEV (
(difference value of video signal high frequency component), PFMS (focus morph speed value delayed by several fields), FMDI
Looking at the membership function R (a guideline value for past focus morph speed values), the input data value (8bit
= 256 levels) are expressed in a nearly symmetrical form with respect to the central value.

For example, if we take a membership function representing dES as an example, we consider the membership function to be symmetrical with respect to the center value of the input data value, as shown in Figure 1(a), and the membership function in Figure 1(b)
As shown in , replace the data values in the half of the center value of the membership function with the data values in the right half.

In other words, NB becomes PB, NS becomes ps, and 20 on the left half
(hereinafter referred to as zOL) is replaced with 20 (hereinafter referred to as ZO8) in the right half. Furthermore, in order to determine the difference between the two, the left half NB, NS, and ZOL are marked with a (-) sign.
Add a (+) sign to PB, PS, and ZO.

To express these values a little more concretely, A in the same figure (a)
The point is the value of point A° in the same figure (b), and similarly the value of point B, 0
The points also have the values of point B° and point C° in FIG. 2(b), respectively.

By the above method, the input data value (8 bits = 256 steps) can be represented in half, ie, 128 steps, as shown in FIG. 2B, and the capacity of the microcomputer program can be reduced.

Conversely, if 256 steps are used as the data value in the right half without reducing the capacity of the microcomputer program, the result will be as shown in Figure (C) compared to Figures (a) and (b). The precision of input data values can be doubled, and membership functions with higher precision can be obtained.

Although this embodiment has been described using the dES membership function as an example, the present invention is also applicable to other membership functions as long as they can be approximated symmetrically with respect to the center value.

As explained above, by symmetrically approximating the input membership function and creating a new membership function, it is possible to reduce the program capacity of the microcomputer, or to increase the membership function with the same program capacity of the conventional microcomputer. The precision of the function can be increased.

Further, according to the above embodiment, the input membership function is approximated vertically symmetrically with respect to its center value, thereby reducing the calculation program capacity in a microcomputer, etc., but the membership function itself is not limited to the microcomputer. In consideration of compatibility with advanced systems that have no limitations on program capacity, etc., various membership functions that have not been subjected to approximation processing are stored in ROM, and advanced systems can perform fuzzy inference without performing approximation processing. In a normal system or a system particularly requiring high-speed processing and reduction in program capacity, the membership-related data read from the ROM may be subjected to approximation processing and used.

(Effects of the Invention) As described above, according to the control device of the present invention,
In a system that performs control using fuzzy inference, by approximating the input membership function during fuzzy inference into an appropriate shape, calculations can be simplified, the calculation program can be simplified and its scale reduced, and the calculation It is possible to increase the processing speed and reduce the program and data storage area.

As a result, a general-purpose one-chip microcomputer, for example, can be used as the control microcomputer, and control using fuzzy inference by software can be easily realized on the general-purpose microcomputer.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining membership function approximation means in the present invention; FIG. 2 is a block diagram showing the configuration when the control device in the present invention is applied to an automatic focus adjustment device such as a video camera; Fig. 3 is a flowchart for explaining the control in the automatic focusing device of the present invention, Fig. 4 is a flowchart showing the speed control procedure of the focus motor by fuzzy inference, and Fig. 5 is the basics of fuzzy inference rules and membership functions. FIG. 6 is a diagram for explaining operations using membership functions used in fuzzy inference in the present invention. FIG. 7 is a diagram showing rules for fuzzy inference in the present invention. FIG. 8 is a flowchart for explaining focus motor speed control in consideration of object mode discrimination in the present invention, and FIG. 9 is a diagram showing control of membership function setting conditions based on mode discrimination in the present invention. et al (: iEs Kezujin Kametsuba Shippu;! dES dFV FMS FMD I R Output Kameshibaship l!l Number 8 Figure ES Edge Radiation dES : Edge Radiation Difference FV High Frequency Component dFV High l1li!1 Component Difference PFMS
・Focus morph reverse speed FMD I R + focus motor rotation direction FMS: Focus morph motor B
Rolling speed S: Sma l + M + Middle B BIg 70: Zero (ES = B + and fdEs = ZO1O) nfFMs = Z
Set it to O1. +i'S: Po5itive Small112,
(FMDIR=Pl and fEs=B+ and fdEs=P
When Bl, (FMS=PSl. 3, (PFMS=P) and (ES=B+ and (dE
When s=NBl, (FMS=NS). (NS+NegativeSmal114, (FMDI
When R=Nl and (ES=Bl and (dEs=PBl), f
Let FMs=Psl. 5, fPFMs=N) and (ES=Bl and fdE
When s=NBl. (FMS=PSl. (PM:Po5itive Middle+6, fF
MDIR = Pl and (FV = Bl and -) + dFV = PB
) and (dEs#PB), then fFMs=PMl. 7, (PFMS=P) and (FV=Bl and -) (dFV
=NB), then (FMS=NM). (NM: Negative Middle+8, fF
MDIR=N)t'(FV=Bl and -)fdFV=PB
) and (dEs*PBl +1) fFMs=NMl
shall be. 9. When fPFMs=N+and (FV=Bl (1), TFMS-PMl is set. (PB: Po5itive Biglo, (F
MDIR=Pl and (FV=Sl and -) (dFV=Ps
1 and fdEssPBl (FMS=PB). +1. (PFMS=Pl and (FV=Sl and fdFV
=Nsl, then fFMs=NBl. (NB: Negative Bigl+2.fFMD
I R=Nl and (FV=S+ and (dFV=Psl and (dEs#pBl) (FMS=NB). 3, (PFMS=Nl and (FV=Sl and [dFV!
NS) (FMS=PBl).

Claims (3)

[Claims] (1) A control device that calculates input information and outputs a controlled variable of a controlled object, the control device quantitatively calculating the degree to which the input information conforms to preset conditions based on a predetermined function. a first calculation means for generating an output function for controlling an object; a second calculation means for performing a gravity center calculation on the output function output from the first calculation means; and the second calculation means. driving means for driving the controlled object based on the output of the controller, and the function is an approximate function that has been subjected to approximation processing so as to be vertically symmetrical with respect to the center value of the input data value. control device. (2) A control device that calculates input information and outputs a control amount of a controlled object, the control device quantitatively calculating the degree to which the input information conforms to preset conditions based on a predetermined function. a first calculation means for generating an output function for controlling an object; and approximating the outer shape of the function so that it is vertically symmetrical with respect to the center value of the input data value, and determining the center of gravity of the approximate function. approximation processing means for performing calculations; second calculation means for performing gravity center calculations on the outline of the output function output from the first calculation means; and based on the output of the second calculation means, the control target A control device comprising: a drive means for driving; and a control device. (3) A control device that calculates input information and outputs a controlled variable of a controlled object, which calculates the degree to which the input information conforms to preset conditions using an input membership function that has been subjected to a predetermined approximation process. a first calculation means that calculates based on the output membership function and generates an output membership function for controlling the controlled object; and a second calculation means that performs a gravity center calculation on the output membership function output from the first calculation means. A control device comprising a calculation means and a drive means for driving the controlled object based on the output of the second calculation means.