JPH06214285A - Camera TTL metering device - Google Patents

Abstract

(57) [Abstract] [Purpose] Corrects photometric errors at high speed and with high accuracy even if the lens configuration changes due to replacement of the taking lens or zooming. [Structure] Part of the subject light that has passed through the photographing lens 2 is re-imaged on the photometric sensor 10 by the photometric reimaging lens 9. The image on the photometric sensor 10 is converted into an electric signal and sent to the photometric calculation device 12, and at the same time, a lens CP that stores each lens data provided in the taking lens 2 is stored.
The lens data is transferred from U13. Photometric calculation device 1
2 uses lens data and a neural network
The brightness information of the photometric sensor 10 is corrected to derive an optimum exposure control value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a TTL photometric device for a camera in which the taking lens can be exchanged, which performs photometry using light that has passed through the taking lens.

[0002]

2. Description of the Related Art TT of a conventional camera with interchangeable lenses
In the L photometric device, it is desirable that the amount of light reaching the light receiving element of the TTL photometric device is ideally proportional to the amount of light reaching the film surface. However, there is a problem in that the output of the light receiving element causes a photometric error due to the difference in the lens configuration of the attached photographing lens.

Therefore, in the conventional apparatus, a simple approximate expression or conditional expression using parameters such as the lens open F value and focal length is used so that the photometric error in all the photographic lenses mounted can be reduced. It is designed to correct the photometric error of the taking lens.

[0004]

However, when the photometric error of multi-division photometry is corrected by the conventional method, the above-mentioned approximate expression and conditional expression must be prepared for each divided photometric element. There is a problem that the scale of the TTL photometer increases as the number increases.

Further, when these correction processes are performed by programming by the CPU, there is a problem that the scale of programming becomes large and the operation time also increases.

The present invention has been made in view of the above circumstances, and is a TTL of a camera for performing multi-division photometry with a large number of divisions.
It is an object of the present invention to provide a TTL photometric device for a camera, which is capable of correcting a photometric error at high speed and with high accuracy even if the lens configuration of the photometric device changes due to replacement of a taking lens or zooming.

[0007]

TTL of the camera of the present invention
The photometric device is a TTL photometric device of a camera having a taking lens 2 as a replaceable taking lens, and the photometric sensor 10 as a light receiving means for detecting the brightness information of the taking screen.
And a lens CP as a storage unit provided in the taking lens 2 for storing a plurality of lens data of the taking lens 2.
A photometric calculation device 12 as a correction unit that forms a predetermined neural network using U13 and lens data and position information of the photometric sensor 10 as input, and corrects the brightness information detected by the photometric sensor 10 based on the output of this neural network.
It has and.

The lens data can be data including information on a plurality of apertures, F-numbers of lenses in the taking lens, or focal lengths.

[0009]

In the TTL photometric device of the camera of the present invention, the photometric calculation device which receives a plurality of lens data of the taking lens 2 stored in the lens CPU 13 and position information on the photographic screen where each pixel of the photometric sensor 10 performs photometry is input. Since the brightness information on the photographic screen detected by the photometric sensor 10 is corrected by the output of the predetermined neural network formed in 12, a large number of photometric sensors can be used even if the lens configuration changes due to replacement of the photographic lens 2 or zooming. On the other hand, the photometric error can be corrected at high speed and with high accuracy.

The lens data is used as the information of a plurality of apertures, the F value,
Alternatively, by using the data including the focal length, the photometric error can be corrected at high speed and with high accuracy even for vignetting of the taking lens 2.

[0011]

Embodiments of the present invention will be described below with reference to the drawings.

1 to 7 relate to an embodiment of the present invention. FIG. 1 is a block diagram showing a configuration of a main part of a camera provided with a TTL photometric device for the camera, and FIG. 2 is a photometric calculation device 12 shown in FIG.
A configuration diagram showing the configuration of the neural network formed in the
FIG. 3 is an explanatory view for explaining a divided shape of an image pickup surface of the two-dimensional sensor in FIG. 1, and FIG. 4 is an explanatory view for explaining a state of a light flux when the taking lens in FIG. 1 is represented by three aperture images, 5 is an explanatory view for explaining a sigmoid function used for calculation in the photometric calculation device 12 of FIG. 1, FIG. 6 is a flowchart for realizing the neural net of FIG. 2 by programming, and FIG. 7 is a neural network calculation of FIG. 6 is a flowchart showing a flow of a neural net calculation subroutine for performing.

As shown in FIG. 1, the T of the camera of this embodiment is
An interchangeable taking lens 2 is detachably attached to a camera body 1 equipped with a TL photometric device. The light from the subject that has passed through the taking lens 2 is reflected by the quick return mirror 4 and imaged on the focusing screen 5. Then, the subject image on the focusing screen 5 passes through the pentaprism 7 via the condenser lens 6 and is observed by the eyepiece lens 8, and the subject image is two-dimensionally displayed by a CCD or the like by the photometry re-imaging lens 9. An image is re-formed on the photometric sensor 10 including a sensor.

The image on the photometric sensor 10 is converted into an electric signal and sent to the photometric calculation unit 12 composed of a CPU or the like. Further, the photometric calculation device 12 is provided in the photographing lens 2, and has a lens CPU 13 that stores lens data such as a focal length, an F value (including zooming), and aperture information representative of lens characteristics described later. Then, these lens data are transferred. Then, the photometric calculation device 12
Uses these lens data to correct the brightness information of the photometric sensor 10 by a neural network described later, and derives an optimum exposure control value based on the corrected brightness information output of the photometric sensor 10.

That is, the TTL photometry device of the camera of this embodiment is composed of the photometry reimaging lens 9, the photometry sensor 10, the photometry calculation device 12, and the lens CPU 13.

On the other hand, the photometric sensor 10 on which the subject image on the focusing screen 5 is re-formed is divided into a plurality of pixels 10a as shown in FIG. 3, and each pixel 10a of the photometric sensor 10 performs photometry. The x-axis and the y-axis are set as the position information corresponding to the shooting screen, and the input values x (−5 to +5) and y (−3 to +3) of the neural network described later are set.

Next, in FIG. 4, the taking lens 2 is attached to the film 1
It is a figure which represented by the three aperture images seen from the 1st surface side, and which explains the cause of a photometry error. In FIG. 4, Po,
Pf and Pb are openings 31o, 31f and 31b, respectively.
Shows the distance from the film surface, and Do, Df, Db are
The diameters of the openings 31o, 31f, 31b are shown.

In FIG. 4, these three openings 31
Only the light fluxes that have passed through o, 31f, and 31b reach the two-dimensional sensor 8. For example, the light flux reaching the central portion A of the focal plane is limited only by the opening 31o at the position Po in the center, but in the upper part B of the focal plane, the upper side of the luminous flux reaching the rear side Pb is It is limited by the opening 31b at the position, and the lower side of the light flux is limited by the opening at the front Pf position.

Therefore, so-called vignetting occurs at the point B, and the amount of light reaching it becomes smaller than the amount of light reaching the point A.

Thus, the photometric error of the photometric sensor 10 is
The three openings 31o, 31f, 31b of the taking lens 2,
It is determined by the position on the photographic screen where the photometric sensor 10 measures light. In this embodiment, these three openings 31o, 31
The lens data of f and 31b are stored in the lens CPU 13 (see FIG. 2) in advance, and this information is used to correct the photometric error of the calling photometric sensor 10 as needed.

If the number of these apertures is three, it is possible to reproduce the vignetting of most photographing lenses, but it is determined according to the required accuracy.

Further, the amounts of these photometric errors are measured in advance in a number of taking lenses and camera bodies to be used, or calculated by computer simulation, and used for learning the neural network shown in FIG. The teacher value.

FIG. 2 shows the photometric sensor 10 in this embodiment.
2 shows a configuration of a neural network that outputs a correction value H of 1. The neural network is configured by an input layer that receives nine inputs, two intermediate layers, and an output layer that outputs a correction value. There is.

I [1] to I [9] are inputs to the neural network, and the three apertures 31o and 31 of the taking lens 2 are used.
Values Po, Do, Pf, D indicating the position and diameter of f, 31b
The input values are f, Pb, Db and the coordinate positions x, y on the photographic screen where each pixel of the photometric sensor 10 performs photometry and the constant 1. The intermediate layer is two layers, the intermediate layer 1 weights the input values I [1] to I [9], and the intermediate layer 2 weights the output of the intermediate layer 1 with the weight Wn of the connection optimized by learning in advance. The added result is converted by the sigmoid function shown in FIG. 5 and used as the input value to the next neuron layer.

That is, in the intermediate layer 1, the weights W1 [n, 1], W1 [n, 2], ..., W1 for the input value I [n].
[N, 10] is weighted added and converted by a sigmoid function,
It is used as the input value of the middle layer 2. Further, in the middle layer 2, weights W2 [1], W2 [2],
..., W2 [11] is weighted and added, and the value converted by the sigmoid function is used as the input value of the output layer.

Further, in the output layer, the output of the intermediate layer 2 (-
1 to 11) is corrected to the range of the actual correction value, the result of weighted addition of W3 [1] and W3 [2] to the output of the intermediate layer 2 and I [9] is corrected without passing through the sigmoid function. The value H is output as it is.

The connection weight W of this neural network is
The photometric calculation device 12 is optimized by a computer or the like in advance using a learning method such as an error back-propagation algorithm, which is a known method.
It is installed inside.

FIG. 6 shows a flowchart for realizing the neural network of this embodiment by programming a microcomputer or the like.

In FIG. 6, in step s1, lens data such as values Po, Do, Pf, Df, Pb, Db indicating the positions and diameters of the three openings 31o, 31f, 31b are input from the taking lens. At step s2, the lens parameters and the coordinate positions x and y on the photographic screen where each pixel 10a of the photometric sensor 10 performs photometry and the constant 1 are set to the input values I [1] to I [9] of the neural network.

Step s3 is a neural network operation subroutine which will be described later. At step s3, I [1]
I [9] is input to obtain a correction value H [x, y] for each coordinate position x, y on the photographing screen where each pixel 10a of the photometric sensor 10 performs photometry as an output.

At step s4, the output S of the photometric sensor
Get [x, y]. In step s5, the sensor output S [x, y] is corrected by the correction value H [x, y], and S '
Calculate [x, y]. In step s6, S '[x,
y] is used to perform the exposure calculation. In step s7, lens data is input. In step s8, step s
It is determined whether the lens data input in 7 has changed, and if the lens data that is a parameter has changed,
Returning to step s2, the correction value H [x, y] is recalculated, and if there is no change in the parameter, it returns to step s4 and repeats exposure calculation.

FIG. 7 shows the flow of the neural network operation subroutine for performing the neural network operation in s3.

The neural network operation subroutine is step s21, in which the input values I [1] to I [9] and the weights W1 [n, 1] to W1 of the coupling between the input layer and the intermediate layer 1 are input.
[N, 9] integration, that is, Σ (W1 [i, n] · I [i]) n = 1 to 10 is performed, and the intermediate layer 1 is processed through the sigmoid function f shown in FIG.
The output N1 [n] of is calculated. Here, in the addition symbol Σ, addition is performed for i = 1 to 9.

Next, in step s22, step s2
Outputs N1 [1] to N1 [10] of the intermediate layer 1 obtained in 1
And the weights W2 [1] to W of the coupling between the intermediate layer 1 and the intermediate layer 2
The result of adding W2 [11] to the integration result with 2 [10], that is, Σ (W2 [i] · N1 [i]) + W2 [11], is passed through the sigmoid function f to obtain the output N2 of the intermediate layer 2. Calculate Here, in the addition symbol Σ, the addition is i = 1 to 1
Do about 0.

In step s23, the result of adding W3 [2] to the product of the weight W3 [1] of the coupling between the intermediate layer 2 and the output layer to the output N2 of the intermediate layer 2 obtained in step s22, that is, W3 [1 ] · N2 + W3 [2] is taken as the output N3, and this N3 is the result of the function, that is, H [x,
y] to end this routine.

As described above, according to this embodiment, the photometric sensor 10 detects the brightness information of the photographing screen, and the lens CPU
A plurality of lens data of the photographing lens 2 stored in 13 and position information on the photographing screen where each pixel of the photometric sensor 10 performs photometry is input, and the photometric sensor is output by a predetermined neural network formed in the photometric calculation device 12. Since the brightness information detected by 10 is corrected, the photometric error can be corrected at high speed and with high accuracy even if the lens configuration changes due to replacement of the taking lens 2, zooming, or the like.

In the present embodiment, the neural network is realized by programming the microcomputer, but the present invention is not limited to this, and in order to execute the neural network at a higher speed, for example, a neural network is formed by a dedicated arithmetic circuit dedicated to the operation of the neural network. You may realize the net.

[0038]

As described above, according to the TTL photometric device for a camera of the present invention, a predetermined neural network which receives a plurality of lens data of the photographing lens stored in the storage means and the position information of the light receiving means as input. Since the luminance information of the light receiving means is corrected by the output of the light receiving means, even if the lens configuration is changed due to replacement of the photographing lens or zooming, the photometric error in the various light receiving means can be corrected at high speed and with high accuracy. The effect is that you can do it.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a configuration of a main part of a camera including a TTL photometric device for a camera according to an embodiment.

FIG. 2 is a configuration diagram showing a configuration of a neural network formed in the photometric calculation device 12 of FIG.

FIG. 3 is an explanatory diagram illustrating a divided shape of an imaging surface of the two-dimensional sensor in FIG.

FIG. 4 is an explanatory diagram illustrating a state of a light flux when the taking lens in FIG. 1 is represented by three aperture images.

5 is an explanatory diagram illustrating a sigmoid function used for calculation in the photometric calculation device 12 of FIG.

FIG. 6 is a flowchart for implementing the neural network shown in FIG. 2 by programming.

FIG. 7 is a flowchart showing a flow of a neural network operation subroutine for performing the neural network operation of FIG.

[Explanation of symbols]

 2 Photographic lens 9 Re-imaging lens for photometry 10 Two-dimensional sensor 12 Photometric calculation device l3 Lens CPU

Claims (2)

[Claims] 1. A TTL photometric device for a camera having an interchangeable taking lens, comprising: a light receiving means for detecting luminance information of a taking screen; and a plurality of lens data of the taking lens stored in the taking lens. A correction that corrects the brightness information detected by the light receiving means by the output of the neural network by forming a predetermined neural network with the storage means provided and the position information of the plurality of lens data and the light receiving means as input A TTL photometric device for a camera, comprising: 2. The TTL photometric device for a camera according to claim 1, wherein the plurality of lens data includes information on a plurality of apertures, an F value of a lens in the taking lens, or a focal length.