JPS6214014A - Range measurement - Google Patents

Abstract

PURPOSE:To enable the recognition of the three-dimensional arrangement of an object by using a TV camera, summing a plurality of image signals from some different positions with signs alternately inverted and processing the data thus obtained. CONSTITUTION:When the push button 23 of a controller 20 is pressed, a motor 34 displaces a camera 1 to produce a pulse 102. The pulse 102 operates an A/D converter 2, a sign inverter 3 and the like and image data with signs alternately inverted are stored in an image memory 5. The data stored in the memory 5 for every horizontal scanning line are transmitted to a high speed Fourier transformer 6, in which a horizontal direction spatial frequency spectrum is obtained, and the spectrum of the whole screen is calculated by an adder 4. Thus, after a smoothing is conducted on the spectrum by operating a constant data memory 22, a multiplier 11, an interpolator 12 and the like, the spectrum is led to another high speed Fourier transformer 13 to be restored to an original image (point). This procedure is performed on every scanning line and the image of the object being measured is completed on another image memory 14. Therefore, the three-dimensional arrangement of a given object apart from the camera by a given range can be recognized.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a method for measuring the distance of a three-dimensional arrangement of objects, and particularly to a method for recognizing a three-dimensional structure using a television camera.

[Background of the invention]

Conventionally, the following methods have been known as three-dimensional structure recognition methods using images. First, it is used in autofocus cameras, etc., which compares images at two different positions while moving a reflector, and then selects the angle of the reflector that best matches the two images. This is a method to determine the distance between. With this method, only the distance to a specific point such as the center of the field of view is known; in order to know the three-dimensional structure of the entire field of view, the field of view must be scanned by changing the angle of the measurement system or by changing the point at which the two images are compared. There was a need.

Another method being considered is a method in which two images obtained from two viewpoints are processed by a computer to reproduce and recognize a three-dimensional world. This method finds corresponding points in two images and calculates the distance and positional relationship using the principle of triangular quantities. I need.

In addition, unless an advanced and complex recognition algorithm is incorporated to determine if there is an object in the image whose distance is to be determined and where in the image it is, it will be necessary to compare the two images at each point in the entire field of view, which will take a long calculation time. Is required.

[Purpose of the invention]

An object of the present invention is to provide a method that can recognize the three-dimensional arrangement of objects by combining relatively simple known algorithms.

[Summary of the invention]

The present invention creates a single image by adding image signals from many different positions while changing the sign alternately, calculates the distance to the subject from the spatial frequency of this image, and calculates the distance to the subject from the spatial frequency of this image. From the image data of , an image of an object at a distance corresponding to that frequency is obtained.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to FIG. The apparatus of this embodiment includes a television camera 1. Guide dedicated 31. Feeding rod 32. Feed gear 33. A television camera moving mechanism composed of an electric motor 34 and an A/D converter 2. Code converter 3. Adder 49 Image memory 5° Fast Fourier transformer 6. Adder 7.
Peak detector 8, multiplier 92 display! +! ! l O, multiplier 11. Interpolator 12. Fast Fourier inverse transformer 139, image memory 14, display CRT 15. Constant data memory 2
2° timing pulse generator 21. and a controller 20.

The guide rod is fixed horizontally, that is, in the direction of the horizontal scanning line of the television camera, and the television camera 1 can be moved left and right along this guide rod without changing the direction of the lens. The television camera is equipped with a male-threaded attachment 32.
This has a feed gear 3 with a female thread on the inside and a gear on the outside.
3 is fitting.

This gear meshes with a gear attached to the rotation of the electric motor 34, and although the feed gear rotates, it does not move in the direction of the rotation axis, so when the electric motor rotates, the feed gear rotates and the feed rod moves. The TV camera moves horizontally. The electric motor is equipped with a position signal generator 35 that generates a pulse 102 at every constant rotation speed.

The television camera 1 uses a highly sensitive image pickup tube so that a sufficiently bright image can be obtained even when the lens is stopped down to increase the depth of field so that the image does not blur due to distance.

A video signal (luminance signal) from a television camera 1 is sent to an A/D converter 2 and digitized. The code converter 3 is capable of adding a negative sign to the digital value of the video signal, and inverts the sign for each image. The adder 4 reads out the contents of the image memory 5, adds the sent video signal thereto, and writes the contents into the image memory again.

The fast Fourier transformer 6 calculates the spatial frequency spectrum of the luminance signal for each horizontal scanning line sent from the image memory 5. The adder 7 calculates the sum of the spectra of all horizontal scanning lines, and the spatial frequency spectrum of the entire image memory is obtained. The peak detector 8 determines the peak frequency of the spectrum. It compares the amplitude value data of each frequency sequentially sent from the adder 7 with the previous one, and when the new amplitude value is smaller, the frequency value is determined. Output. This frequency value is held until the new amplitude value is greater than the previous one.

In this way, the peak frequency is determined and sent to the multiplier 9. The multiplier 9 outputs the product of the peak frequency value and a constant (described later). The obtained value is displayed on the display 10.

The multiplier 11 extracts a specific frequency component from the spectrum of each horizontal scanning line obtained by Fourier transform and corrects the value, and stores the amplitude and constant data of each frequency of the spatial frequency spectrum of the image in the memory 22. data (
The interpolator 12, which performs multiplication with (described later), performs spectrum smoothing by interpolating values between extracted specific frequencies. The obtained spectrum is converted back into an image signal by the fast Fourier inverse transformer 13 and stored in the image memory 14. The stored image data is displayed on the CRT 15.

The timing pulse generator 21 is a device for capturing one image data (A/D converter 2, code conversion W#3° adder 49
A timing pulse is generated so that the image memory 5) operates in synchronization. The controller 20 also only generates a start signal so that the data processing systems (numerals 5 to 14) operate in an orderly manner, and does not have a judgment function or arithmetic function.

Next, the operation of the device will be explained. The measurement is initiated by pressing the button 23 on the controller 20. When the button 23 is pressed, a clear signal 103 is sent to the image memory 5 and past data is erased. At the same time, a starting signal 101 is sent to the electric motor 34.

When the electric motor 34 starts, the television camera 1 starts moving.
Further, a pulse is generated at each constant rotation speed of the electric motor.

Each time a pulse occurs, the A/D converter 2. Code converter 3.
The adder 4 operates, and image data with alternating signs are stored on the one-image memory 5. At this time, the television camera moves horizontally (in the scanning line direction), and the pulse generation interval is constant, so the image data on the image memory is an alternating array of aerial images seen from many positions that are moved at equal intervals little by little. It will exist with a different sign. Also, the number of pulses generated during the entire movement of the television camera is adjusted so that it is an even number. At this time, the data in the image memory is as follows.

- For example, if the subject is close-range 1IIL as shown in Figure 2
Assume that there is a sphere at 0 and a triangular pyramid at far distance LT. When the television camera moves, the image of a nearby object moves significantly, so that images of a sphere with a coarse pitch and a triangular pyramid with a fine pitch are created on the image memory, as shown in FIG. The background image, which can be considered to be at infinity, does not move even if the camera moves horizontally because the direction of the lens remains unchanged, and since the number of positive and negative changes is the same, they cancel each other out and are erased.

If the horizontal angle of view of the television camera is 09, the horizontal width of the image is W, and the distance the television camera moves while one pulse is generated is α, then the pitch of the spherical image is a (the distance between the positive image and the negative image). ) is Lc After the television camera has finished moving and the image has been photographed, data processing begins. Data for each horizontal scanning line is sent from the image memory 5 to the fast Fourier transformer 6 according to a signal from the controller 20, and a horizontal spatial frequency spectrum is obtained.The spectrum of the entire screen is calculated by the adder 4 and sent to the peak detector 8. Since the transmitted images in the 0-image memory are a large number of identical images lined up at equal intervals, they have frequency components according to their pitches, as shown in Figure 4. In other words, the spatial frequency peak of the spherical image is The denominator 2 comes from the fact that there are alternating positive and negative images.

Since e and W are constants of the television camera, and α is a preset value of 2Wα, by setting □ as the θ constant of the multiplier 9, the peak frequency f,...! ? are sent to the multiplier, so the distances Lc and LT to the object are displayed on the display. As described above, according to the present invention, distances can be determined no matter where a plurality of objects are located on the screen.

If the brightness distribution of the image of the object is P(x, y), then the sum of 8 images obtained by shifting the luminance distribution by a pitch a while changing the sign alternately is Pab(X+y) with the X axis in the horizontal direction. =
-1)'P(x-n a, y) (3) This is the image P of the object P (x, y) in the image memory
Ob (xe y), and when it is Fourier transformed (represented by F), F(Pab(x, y))=F(2:(-1)'P(x-
n a,y)=E(1)"e"""F(P(x,y))
(4) becomes. Fourier transform of the original image F(P(x,y)
), (f)=E(-1)"e""" (5)
When calculated, the result is as shown in FIG. 5. In this embodiment, the number of pulses generated from the television camera moving mechanism is 20.

That is, in the above equation, N=20.

It can be seen that it also has harmonics that are odd multiples of that value. Therefore, in the observed image, there is a large peak at odd-numbered harmonics, and extremely small values in other parts. Since the pitch a is a function of the distance to the object as shown in equation (2), the Fourier transforms of images of objects at different distances have peaks at different positions. Therefore, only the Fourier transform of the image of the object located at the peak of the spatial frequency corresponding to the distance can be obtained. To do this, create the following function RL(J) and multiply it with the spatial frequency spectrum.

eL eL , 2Wα 2Wα=O Other than the above (6) An example of RL(J) is shown in Figure 6.Here, the peak extraction width was determined so that the absolute value of the coefficient of the real part was 1 or less, so 0 The numbers .96 and 1.04 appeared. R, (j) depends on L, but since it is a function determined by the television camera and the shooting pitch α, various L (Ll, L□, L,

), and the data obtained by discretizing the calculation is stored in the constant data memory 22 in advance.

Now, the signal from the controller is sent to the image memory 5. Fast Fourier transformer6. When sent to the constant data memory 22, the spatial frequency spectrum of a certain scanning line of the image and the function R4 (assumed to be performed for the L process first) are sent to the multiplier 11.
The product of both is created.

As a result, as shown in FIG. 7, the spectrum of the peaked portion is restored. At this time, the spectrum of the object at a position other than L1 disappears because its peak position does not match RL (x.

y) and stored in the image memory 14. Since the above procedure is performed for all scanning lines, the distance f is stored in the image memory.
The image of the object at position iL1 is completed. This is CRT
15.

Next, the same processing is performed using the function read from the constant data memory 22 as RL2, and an image of the object at a distance L2 is displayed, and this is repeated for each type of L.

As described above, according to this embodiment, as shown in FIG. 8, the image of an object at a certain distance from the television camera can be displayed separately from objects located at other positions. In other words, the three-dimensional arrangement of objects can be recognized. Furthermore, as explained above, there is no need for judgment in the data processing process, and the distance to the object and its three-dimensional arrangement can be determined using a simple algorithm.

Another embodiment of the present invention is shown in FIG. 9. In this embodiment, the computer 24 performs data processing. Also, the same effect is achieved by moving the reflector 16 instead of moving the television camera 1.The reflector 16 is equipped with a reflector moving mechanism similar to the Temebi camera moving mechanism of the previous embodiment.1 Reflector 16 is 45 for fixed TV camera 1
It moves back and forth with respect to the television camera 1 while maintaining an angle of . Since the subject is photographed by being reflected on the reflecting mirror, by moving the reflecting mirror back and forth, the same image as in the previous embodiment can be obtained by moving the television camera in the horizontal direction.

The electric motor 34 starts rotating in response to the measurement start signal, and the reflecting mirror starts moving.The electric motor 34 has a constant rotation speed, so that the moving speed of the reflecting mirror is constant. The captured image is constantly converted into a digital signal by the A/D converter 2 and sent to the computer. A computer captures images at regular intervals using an internal clock. This allows images to be obtained from positions moved in the horizontal direction at regular pitches.

FIG. 10 shows a flowchart of the program contained within the computer, and the processing content is the same as that performed by the hardware in the previous embodiment. First, erase the old image integration value, set the number of shots counter n to 0, and then capture the image data G at a predetermined time. times) repeat.

Thereafter, the obtained image data is Fourier transformed for each scanning line to obtain a spatial frequency spectrum F(j'+y-).

If only the distance to the object is needed to be known (this is due to human input to the computer), then the resulting spatial frequency spectra are summed and the sum of all scanlines is taken to the 0th order of this total scanline. The frequencies giving the maximum amplitude of the spectrum are detected, and distances are calculated from these values using the relationship L=(2Wα/θ)J and printed.

On the other hand, when displaying an image of an object at each distance, calculate the functions R, , (f) using equations (5) and (6), and calculate the spatial frequency spectrum F Cf,y of each scanning line. −)
Make the product of This result is interpolated, smoothed, and then inversely Fourier transformed back into a luminance signal. This process is repeated for all scanning lines, and after completion is displayed on a CRT. After that, repeat the same process for the fire distance of 0.1.The value of distance may be determined in advance,
It may also be input into a computer at the processing stage.

As described above, according to this embodiment, the distances between multiple objects in an image can be determined using a simple machine configuration and a simple algorithm, and the three-dimensional arrangement of objects can be recognized.

Since the data processing algorithm of the present invention is simple and does not involve complicated judgments, the processing system can be configured with analog circuits.

FIG. 11 is an example in which the present invention is implemented using an analog circuit.

In this embodiment, instead of moving a TV camera or a mirror, a plurality of TV cameras 1 are lined up in the horizontal direction (scanning line direction), and the sum and difference of the outputs are taken alternately, which is the same as moving the cameras. It's working. To accomplish this, the output of a single sweep controller 56 is used as a sweep signal so that all television cameras scan the screen in synchronization. The outputs of the arranged television cameras are alternately connected to the positive input terminal and negative input terminal of the differential amplifier 51. This output passes through a differentiating circuit 52, a parallel narrowband filter 53, and a detection circuit, and then is input to an integrating circuit 55, and its output is displayed on the CRT 15. The CRT scan lines also receive sweep signals from the same controller 56 so that they are swept in synchronization with those of the television camera.

The passing frequency of the filter 53 is selected to be 2τα, where the scanning time of one scanning line is τ, and to be an odd multiple thereof.

Further, L is the distance from the television camera to the position to be viewed, and α is the pitch of the television camera.

When an object is photographed by this television camera group 1, the output from the differential amplifier 51 is the same as in the previous embodiment.

As shown in Figure 3, images viewed from different positions are superimposed with alternating signs. The scanning line output of a television camera has a spatial frequency that is a temporal frequency, so when this output passes through a filter, it becomes Only spatial frequency signals are obtained. Therefore, since the detected signal is displayed on the CRT as a luminance signal, only the image of the object at a distant position is displayed. Furthermore, in order to obtain an object image at a position at a different distance, in addition to changing the passing frequency of the filter, the sweep time τ of the scanning line may also be changed. In order to prevent the spectrum from being removed, a differentiating circuit 52 is inserted to emphasize the high frequency part, and an integrating circuit 55 is used to restore the original spectrum.

According to this embodiment, since there is no interpolation operation as performed in the previous embodiment, the resolution is somewhat degraded, but it has the effect that an image of an object at a specific distance can be obtained in real time.

The method of superimposing images viewed from different positions by alternating the signs is shown in Fig. 12, in addition to the method of moving a television camera or reflector as described above and the method of lining up a large number of television cameras. A method using a semi-transparent reflecting mirror 57 is also possible.

At this time, the reflectance of each reflecting mirror is r, = -, that is, 1° 1/2.
1/3, 1/4, etc., the brightness of the image from each reflecting mirror is equal.0 The reflected light from the reflecting mirrors is alternately reflected by two cameras, one for "sum" and one for "difference". The output of the camera is sent to the television camera 1, the difference between the two is created by the differential amplifier 51, and the output is sent to the image memory in the CPU via the A/D converter 2.

The subsequent data processing is the same as in the embodiment shown in FIG. 10, and the scanning line control of the television camera is the same as in the embodiment shown in FIG.

According to this embodiment, images viewed from different positions can be superimposed with alternating signs by using only two television cameras without any moving parts, and the A/D converter 2. By processing the data using the computer 24 or the like, there is an effect that the three-dimensional arrangement of objects can be recognized.

〔Effect of the invention〕

According to the present invention, it is possible to determine the distances from a television camera to multiple objects in an image using relatively simple equipment or a combination of known algorithms, and to determine the distances of objects at a certain distance from the television camera. This has the effect that images can be displayed separately from objects in other positions, that is, the three-dimensional arrangement of objects can be recognized.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 is a block diagram of an embodiment of the present invention.
Figure 3 is an explanatory diagram of the contents of the image memory, Figure 4 is a schematic diagram of the spectrum obtained by frequency analysis, Figure 5 is for quantitatively explaining the principle. FIG. 6 is an explanatory diagram showing the functional form used in data processing, FIG. 7 is an explanatory diagram of the data processing process,
Fig. 8 is an explanatory diagram of an output image, Fig. 9 is a block diagram of another embodiment of the present invention, Fig. 10 is a flow chart of the program used, and Fig. 11 is a block diagram of another embodiment. 12 are explanatory diagrams showing yet another embodiment. 1... Television camera, 2... A/D converter, 3... Code converter, 4... Adder, 5... Image memory, 6...
- Fast Fourier transformer, 7... Adder, 8... Peak detector, 1o... Display device, 11... Multiplier, 12
...Interpolator, 13...Fast Fourier inverse transformer, 14.
...Image memory, 15...CRT, 20...Controller, 21...Timing pulse generator, 22...
Constant data memory. 34...Electric motor, 35...Position compensation signal generator, 16...
-Reflector, 24...Computer, 51...Differential amplifier, 53...Filter, 57...Semi-transparent reflector.

Claims (1)

[Claims] 1. In a distance measurement method using a television camera, an image is obtained by alternately inverting the signs of multiple image signals from different positions and summing them, and the distance to the subject is calculated from the magnitude of the spatial frequency of this image. A distance measurement method characterized by calculating.