JP2005114581A - Distance measurement method - Google Patents

Abstract

[Purpose] In distance measurement based on the principle of triangulation, a reliability judgment is performed to detect a defect caused by an evaluation function representing a correlation between two photosensor array outputs not being symmetrical in the vicinity of the minimum value. Provide a distance measurement method that can.
The reliability is evaluated by estimating the true minimum value f (x) and comparing it with a predetermined value H. For the predetermined value H, the random variation (a%) of the photosensors, the average number of bits of the photosensor output digitized by the A / D converter (n), and per photosensor array applied to the calculation of the evaluation function Determined in consideration of the number of photosensors (m). Further, in order to deal with a case where the S / N ratio of the photosensor output is low, such as a low-contrast subject, a shift value i that gives the estimated true relative displacement and the minimum value of the evaluation function discretely obtained for each photosensor pitch. Reliability evaluation based on the positional relationship with ₀ is also performed.
[Selection] Figure 1

Description

  The present invention relates to a distance measuring method used for an autofocus device such as a camera.

FIG. 3 is a distance measurement principle diagram based on the principle of triangulation. The light of the subject 55 is formed as subject images 56 and 57 on the photosensor arrays 53 and 54 through the lenses 51 and 52. Points P ₆ and P ₇ are intersections of light beams (optical axes 58 and 59) passing through the center points P ₂ and P ₃ of the lenses 51 and 52 from the infinity of the front and the photosensor arrays 53 and 54. The distance between the points P ₆ and P ₇ is B, and the distance between the photosensor arrays 53 and 54 and the lenses 51 and 52 is f (substantially equal to the focal length of the lenses 51 and 52). Further, the deviation of the subject images 56 and 57 from the optical axes 58 and 59 is defined as X1 and X2, and the length obtained by adding these X1 and X2 is defined as X. Reference numeral 61 denotes a measurement system.
Here, since the triangle P ₁ P ₂ P ₄ and the triangle P ₂ P ₅ P ₆ and the triangle P ₁ P ₄ P ₃ and the triangle P ₃ P ₇ P ₈ are similar to each other, the distance d to the subject 55 is expressed by the following equation: Is required.

(Equation 5)
d = B · f / (X1 + X2) = B · f / X
X is a relative transition of two images based on the case where the subject 55 is at infinity, that is, when the two subject images 56 and 57 are at the intersections of the optical axes 58 and 59 of the lenses 51 and 52 and the photosensor array. . Since B and f are constants, the distance d can be obtained by detecting X.
In order to increase the distance measurement accuracy, it is necessary to accurately obtain the relative displacement X of the two images. However, since it cannot be obtained directly beyond the accuracy and resolution determined by the physical arrangement (pitch) of the photosensor, the sensor pitch is calculated by first calculating the relative displacement with the accuracy of the photosensor pitch and then performing interpolation calculation. A technique for obtaining the above accuracy is disclosed in Patent Document 1 and the like. The prior art will be described below.

  FIG. 4 is a configuration diagram showing a general example of a distance detecting device, FIG. 5 is a diagram for explaining a first conventional example of an interpolation method, and FIG. 6 is a diagram for explaining a second conventional example of an interpolation method. FIG. 4, 53 and 54 are the same photosensor arrays as shown in FIG. 3, and 62 and 63 are A / D converters. Analog outputs from the sensors in the photosensor arrays 53 and 54 are converted into digital signals by A / D converters 62 and 63. An arithmetic processing unit 64 (hereinafter referred to as a microcomputer) such as a microcomputer obtains a relative displacement X between two images, and controls the imaging lens using this X. As described above, such a distance detection apparatus generally forms a system using a microcomputer. Now, the output values of the A / D converted photosensor arrays 53 and 54 are respectively L (1), L (2),..., L (M1), R (1), R (2),. .., R (M2), and the evaluation function f (i) indicating the degree of mismatch between the images with respect to the relative displacement of the two images of X = i · p (p is the photosensor pitch) is, for example, the expression (1) or (2 ).

(Equation 6)
f (i) ≡Σ | L (j) −R (j + i) | (1)
j

(Equation 7)
f (i) ≡Σ | L (j−s) −R (j + t) |
j
s + t = i (2)

If the two images on the photosensor arrays 53 and 54 are relatively shifted by the i ₀ photosensor pitch, f (i ₀ ) = 0. However, since it is rare that X is an integral multiple of the optical sensor pitch, f (i ₀ )> 0 is usually satisfied.
In the predetermined range of i, the minimum value of f (i) can be easily obtained by a microcomputer. However, as described above, this is usually only a first order approximation, so that highly accurate detection is performed. For this purpose, interpolation calculation is required. A first conventional example relating to this interpolation method is shown in FIG.

In the following description, including FIG. 5 and the embodiment of the invention, i ₀ is i giving the minimum value of f (i) for each photosensor pitch,

(Equation 8)
f (i ₀ -1)> f (i ₀ +1) (3)
Suppose that The condition of the expression (3) is for determining that the true relative displacement X is between i ₀ and i ₀ +1. When f (i ₀ −1) <f (i ₀ +1), i ₀ The positions of −2 and i ₀ −1 and i ₀ +2 and i ₀ +1 may be interchanged, and so on.
As shown in FIG. 5, the first conventional example of the interpolation method starts with a straight line L0 connecting the point (i ₀ -1, f (i ₀ -1)) and the point (i ₀ , f (i ₀ )). obtains the inclination -B1, also determine the points _{(i 0 + 1, f (} i 0 +1)) and the point _{(i 0 + 2, f (} i 0 +2)) the slope B2 to the straight line connecting the. Then the point _{(i 0, f (i 0} )) is the street slope - {B1 + K1 (B1- B2)} pull a line L2 is a point _{(i 0 + 1, f (} i 0 +1)) is the street slope The straight line L3 which is B1 + K1 (B1-B2) is drawn. The value of the abscissa X at the intersection of the straight lines L2 and L3 is finally set as the interpolated relative displacement.

The interpolation method according to the first conventional example is formulated as follows. That is, the function values f (i ₀ −1) and f (i ₀ +1) are compared, and a) f (i ₀ −1)> f (i ₀ +1), b) f (i ₀ −1) = F (i ₀ +1), c) f (i ₀ −1) <f (i ₀ +1) is determined. In the case of a), the following formula (I), b) In the case of x = i ₀ , c), x given by the following equation (II) or a value obtained by multiplying this by the photosensor pitch is set as the relative displacement amount.

(Equation 9)
x = i ₀ + 1 / 2− {f (i ₀ +1) −f (i ₀ )} / A
A = 2 {B1 + K1 (B1-B2)}
B1 = f (i ₀ −1) −f (i ₀ )
B2 = f (i ₀ +2) −f (i ₀ +1)
(A constant of K1 = 0 to 0.5) (4)

(Equation 10)
x = i ₀ −1 / 2 + {f (i ₀ −1) −f (i ₀ )} / C
C = 2 {D1 + K1 (D1-D2)}
D1 = f (i ₀ +1) −f (i ₀ )
D2 = f (i ₀ −2) −f (i ₀ −1)
(A constant of K1 = 0 to 0.5) (5)

This first conventional example was invented as an improvement of the second conventional example described below. That is, the second embodiment of the interpolation method is to draw a straight line L0 connecting the points _{(i 0 -1, f (i} 0 -1)) and the point _{(i 0, f (i 0} )) as shown in FIG. 6 , drawn to pass this straight line L4 absolute values for the slope -B1 the straight line L0 is the sign reversed with inclination of reverse + B1 point _{(i 0 + 1, f (} i 0 +1)), the straight line L0 and L4 The value of the abscissa X of the intersection point is finally set as the interpolated relative displacement. Here, the concept of interpolation calculation is that the graph of f (x) is symmetrical in the vicinity of x = x ₀ that gives the true minimum value of f (x), and is a graph (linear line) expressed by a linear expression. ), The second conventional example has the following problems. In both the first and second conventional examples, in addition to the above, f (x) decreases monotonously in the range of x <x ₀ and monotonically increases in the range of x> x ₀ in the vicinity of x = x _0. The assumption is also made.

Ie to a problem inherent above, the point _{(i 0, f (i 0} )) of the inclination -B1 through the straight line and the point _{(i 0 + 1, f (} i 0 +1)) of straight intersection of slope + B1 through It is only in the case of f (i ₀ ) = f (i ₀ +1) that the abscissa coincides with x = x ₀ giving the true minimum value of (x). Usually, since f (i ₀ ) and f (i ₀ +1) are not equal, in the second conventional example, basically, an error cannot be avoided. In contrast, in the first conventional example, the point _{(i 0, f (i 0} )) and point _{(i 0 + 1, f (} i 0 +1)) The absolute values of the two straight lines of slope through respective first By adding a correction term K1 (B1-B2) to B1 of the conventional example 2 and correcting it to B1 + K1 (B1-B2) (a constant of K1 = 0 to 0.5), f (i ₀ ) and f Even when (i ₀ +1) is not equal, the abscissa of the intersection of the two straight lines is an interpolation value close to x = x ₀ giving the true minimum value.

On the other hand, a method for evaluating the reliability of the interpolation calculation and determining whether the interpolation calculation gives a true relative displacement has been proposed. This is because a subject with mixed perspective, a subject including a repetitive pattern, or the like may cause so-called false focusing. When f (i ₀ ) is larger than a predetermined value, the degree of coincidence of images is low and the reliability is low. A method for determining is known (for example, see Patent Document 2).
Japanese Patent Publication No. 7-54371 Japanese Patent Laid-Open No. 2002-214521 (pages 4, 6 and 4)

As described above, it is rare that the relative displacement X is an integral multiple of the optical sensor pitch. Therefore, when the reliability is determined by f (i ₀ ), it is necessary to leave a margin corresponding to the determination value. There is. For example, as shown in FIG. 7, when f (i ₀ ) is almost equal to f (i ₀ +1), the difference between the true minimum value and f (i ₀ ) is considered to be large. It is necessary to prevent misjudgment that the degree of matching is low. The true minimum value of the evaluation function shown in FIG. 7 is, for example, a straight line connecting point (i ₀ −1, f (i ₀ −1)) and point (i ₀ , f (i ₀ )) as shown in FIG. It is estimated as the preparative point _{(i 0 + 1, f (} i 0 +1)) and the point _{(i 0 + 2, f (} i 0 +2)) the connecting intersection in the height (y-coordinate value) fmin, fmin is sufficiently small image It can be determined that the degree of coincidence is high. On the other hand, although the value of f (i ₀ ) itself is not so small, the determination value for f (i ₀ ) is somewhat large in order to prevent erroneous determination that the image matching degree is low in this case. I have to. Since the determination value is thus large, f (i ₀ −1) is almost equal to f (i ₀ +1), and f (i ₀ ) is a true minimum value, contrary to the case of FIG. For cases that are considered to be almost equal, there is a problem that reliability may not be judged correctly, for example, it may be judged that the degree of coincidence is high despite the fact that the degree of coincidence of images is actually low. There is.

Further, in the interpolation methods shown in the first conventional example and the second conventional example, f (x) in the vicinity of x = x ₀ that gives the true minimum value of f (x) is x = x _0. Is assumed to be symmetrical. Based on this, there is a true relative displacement between i ₀ and i ₀ +1 when f (i ₀ −1)> f (i ₀ +1) and the interpolation method described in the background art, and f ( When i ₀ −1) <f (i ₀ +1), it is determined that a true relative displacement exists between i ₀ −1 and i ₀ . That is, the conventional distance measurement method is largely based on this left-right symmetry, but this right-left symmetry may be lost. A case where the left-right symmetry is broken will be described below.
Normally, the outputs of the two photosensor arrays are equal except for relative displacement. When j is a photosensor number in the photosensor array, Img (j) is a function representing an image signal, and z is a relative displacement, outputs L (j) and R (j) of the two photosensor arrays are L (j ) = Img (j + z), R (j) = Img (j). Thus, for example, | L (j) −R (j + i) | in the expression (1) becomes | Img (j + z) −Img (j + i) |, and the graph of f (i) in the expression (1) represents i It turns out that it becomes symmetrical with respect to = z. Conversely, if L (j) and R (j) cannot be represented by the same function Img (j), the evaluation function may not be symmetrical.

  The first case where L (j) and R (j) cannot be represented by the same function Img (j) is a so-called mixed perspective where images of a plurality of subjects having different distances are simultaneously received by the photosensor array. In this case, L (j) and R (j) are expressed as L (j) = Img1 (j + z1) + Img2 (j + z2), R (j) = Img1 (j) + Img2 (j). It is assumed that there are two subjects with different distances and the relative displacements are z1 and z2, respectively. In this case, | (Img1 (j + z1) + Img2 (j + z2)) − (Img1 (j + i) + Img2 (j + i)) | = | (Img1 (j + z1) −Img1 (j + i)) + (Img2 (j + z2) −Img2 (J + i)) | needs to be calculated, but in order to grasp the whole, this is replaced by | Img1 (j + z1) -Img1 (j + i) | + | Img2 (j + z2) -Img2 (j + i) | Consider it. In other words, this is equivalent to obtaining the evaluation functions for the two subjects independently and adding them together at the end. An example thereof is shown in FIG. 8, but the graphs Img1 and Img2 are not symmetrical although the graphs Img1 and Img2 are symmetrical. It is to be noted that the substitution with respect to the above absolute value formula is not correct only in the interval between the relative displacements z1 and z2 giving the minimum values of the two graphs Img1 and Img2, and the above substitution is used to grasp the overall tendency. no problem.

A second case where L (j) and R (j) cannot be expressed by the same function Img (j) is a case where the photosensor cannot follow a sudden change in a subject image such as a point light source. That is, the frequency component of the subject image is too high, and the subject image cannot be reproduced by sampling determined by the photosensor pitch of the photosensor array, and L (j) and R (j) have different function forms. is there. Examples of this case are shown in FIGS. As shown in FIG. 9, different photosensor outputs are obtained by the two photosensor arrays LEFT and RIGHT for the same point light source, and the relative displacement of the two images is sequentially changed as shown in FIGS. 10 (a) to 10 (c). When the evaluation function is obtained by shifting, the graph becomes asymmetrical as shown in FIG.
In addition to the above, a low-contrast subject has a small difference between a normal signal and noise, that is, a so-called SN ratio is small, so that L (j) and R (j) are unlikely to have the same shape and the left-right symmetry may be lost. is there. Furthermore, in a subject with a repetitive pattern such as a striped pattern, moire occurs due to the relationship between the repetitive pitch of the subject image and the photosensor sensor pitch, or a pattern such as gradation is superimposed on the repetitive pattern, and the above Img1 (j) is repeated. The pattern Img2 (j) may be a combination of patterns such as moire and gradation, which may lead to a collapse of the symmetry of the evaluation function.

The loss of left-right symmetry of the evaluation function in this way also means that the basis for the above-described interpolation method is lost. Further, when the right-left symmetry is greatly broken, as shown in FIG. 12, (i ₀ -1, f (i ₀ -1)) and _{(i 0, f (i 0} )) linearly and _{(i 0 + 1, f (} i 0 +1)) and _(i 0 + 2, intersections of a straight line connecting the _{f (i} 0 +2)) connecting the no longer exists between i ₀ and i ₀ +1, and if f (i ₀ −1)> f (i ₀ +1), then there is a true relative displacement between i ₀ and i ₀ +1, and f (i _{In the case of 0} -1) <f (i ₀ +1), the above judgment that a true relative displacement exists between i ₀ -1 and i ₀ also leads to doubt, and the collapse of symmetry is This causes a problem in the reliability of the result of the interpolation calculation.
Accordingly, the present invention has been made to solve the above-described problems, and the object of the present invention is to measure the distance that can deal with a problem subject by detecting that there is a problem in the above-described reliability. It is to provide a method.

Accordingly, in order to solve the above-described problem, the invention according to claim 1 projects the light beams of different portions of the light beams radiated or reflected from the subject onto at least two photosensor arrays, respectively. In order to obtain the distance from the relative displacement of the subject image formed on the array to the subject,
A scale f (i) indicating a degree of inconsistency between the two images with respect to a displacement amount i (integer) times the sensor pitch of the photosensor array, a value i ₀ giving the minimum value, and a function value f ( i ₀₎ and before and after _i 0 _-1 of the _{i 0, i 0 -2, i} 0 + 1, i 0 +2 to the corresponding function value _{f (i 0 -1), f} (i 0 -2), f ( i ₀ +1) and f (i ₀ +2) are obtained respectively, and the relative displacement amount is obtained by interpolation calculation from these values, and the function g (x0, x1, x2, x3) indicating the reliability of the interpolation calculation is obtained. Is compared with the function value f (i ₀ −1) and f (i ₀ +1),
A) f (i ₀ −1)> f (i ₀ +1)
B) f (i ₀ −1) = f (i ₀ +1)
C) f (i ₀ −1) <f (i ₀ +1)
To determine which relationship
In the case of a), the function g (x0, x1, x2) is assumed as x0 = f (i ₀ −1), x1 = f (i ₀ ), x2 = f (i ₀ +1), x3 = f (i ₀ +2). , X3)
B), g (x0, x1, x2, x3) = f (i ₀ )
In the case of c), x0 = f (i ₀ +1), x1 = f (i ₀ ), x2 = f (i ₀ −1), x3 = f (i ₀ −2) and the function g (x0, x1, x2) , X3), and when the value of the function g (x0, x1, x2, x3) exceeds a predetermined value H, it is determined that the reliability of the interpolation calculation is low. .

  The invention according to claim 2 is characterized in that, in the invention according to claim 1, the function g (x0, x1, x2, x3) is expressed by the following equation (I).

(Equation 11)
g (x0, x1, x2, x3) = A1 / A2
A1 = 2x0 · x2-x0 · x3 + 2x1 · x3-3x1 · x2
B1 = x0−x1−x2 + x3 (I)
The invention according to claim 3 is the invention according to claim 1, wherein the function g (x0, x1, x2, x3) is expressed by the following equation (II).

(Equation 12)
g (x0, x1, x2, x3) = (2x1-x0 + x2-k1 (x3-x2)) / 2
Constant of k1 = 0 to 0.5 (II)
The invention according to claim 4 is the invention according to claim 2 or 3, wherein the output of the photo sensor is digitized by an n-bit AD converter, and the digitized photo sensor output is m data per photo sensor array. Is used to calculate the scale f (i) indicating the degree of inconsistency between the two images, and when the average random variation of the photosensor output is a%, the predetermined value H is expressed by the following formula (III): It is characterized by that.

(Equation 13)
H = k2 · (a / 100) · 2 ⁿ · m
Constant of k2 = 0.1 to 1.0 (III)
The invention according to claim 5 is that, in the invention according to claim 1, the function g (x0, x1, x2, x3) is expressed by the following equation (IV), and the predetermined value H is 0: Features.

(Equation 14)
g (x0, x1, x2, x3) = x1 + x3-2x2 (IV)

  The distance measuring method of the present invention relates to an evaluation function indicating the correlation between two object images calculated when measuring the distance to the object based on the principle of triangulation, and estimates the true minimum value to estimate the true minimum It is possible to provide a distance measurement method capable of improving the reliability evaluation accuracy of the distance measurement result by evaluating the value itself or the relative displacement that gives the true minimum value.

In the present invention, not only the true relative displacement x is obtained by interpolation calculation, but also the true minimum value f (x) is estimated and compared with a predetermined value H to perform reliability evaluation. For the predetermined value H, the random variation (a%) of the photosensors, the average number of bits of the photosensor output digitized by the A / D converter (n), and per photosensor array applied to the calculation of the evaluation function It may be determined in consideration of the number of photosensors (m). Further, in order to deal with a case where the S / N ratio of the photosensor output is low, such as a low-contrast subject, it is preferable to perform reliability evaluation based on the positional relationship between the estimated true relative displacement and i ₀ described in the background art section. .
Details will be described below with reference to examples.

In the first embodiment, when f (i ₀ −1)> f (i ₀ +1) is satisfied and the true relative displacement is determined to be between i ₀ and (i ₀ +1), as shown in FIG. point two _{(i 0 -1, f (i} 0 -1)) and the point _{(i 0, f (i 0} )) connecting the straight line and the point _{(i 0 + 1, f (} i 0 +1)) and the point _{(i 0} The height (y coordinate value) fmin of the intersection connecting +2, f (i ₀ +2)) is obtained and compared with a predetermined value H. The condition of f (i ₀ −1)> f (i ₀ +1) is for judging that the true relative displacement X is between i ₀ and i ₀ +1, and f (i ₀ −1) <f In the case of (i ₀ +1), the positions of i ₀ -2 and i ₀ −1 and i ₀ +2 and i ₀ +1 may be interchanged and the same may be performed. Note that when f (i ₀ −1) = f (i ₀ +1), interpolation is not performed and f (i ₀ ) is a true minimum value.
Although details of the calculation are omitted, when f (i ₀ −1)> f (i ₀ +1), x0 = f (i ₀ −1), x1 = f (i ₀ ), x2 = f (i ₀ +1) ), X3 = f (i ₀ +2), fmin = g (x0, x1, x2, x3) is given by the following equation.

(Equation 15)
g (x0, x1, x2, x3) = A1 / A2
A1 = 2x0 · x2-x0 · x3 + 2x1 · x3-3x1 · x2
B1 = x0−x1−x2 + x3 (I)
As described above, when f (i ₀ −1) <f (i ₀ +1), the positions of i ₀ −2 and i ₀ −1 and i ₀ +2 and i ₀ +1 are interchanged, and x0 = f (i ₀ +1), x1 = f (i ₀ ), x2 = f (i ₀ −1), and x3 = f (i ₀ −2). As described above, when f (i ₀ −1) = f (i ₀ +1), fmin = g (x0, x1, x2, x3) = f (i ₀ ) is set.

  The predetermined value H is determined by random variations (a%) of photosensors, the average number of bits (n) of the photosensor output digitized by the A / D converter, and one photosensor array applied to the calculation of the evaluation function Taking into account the number of photosensors (m), it may be determined as follows: K1 is a constant for adjustment.

(Equation 16)
H = k2 · (a / 100) · 2 ⁿ · m
Constant of k2 = 0.1 to 1.0 (III)
Equation (III) is based on the idea that the true minimum value should be zero except for random variations. If the obtained fmin = g (x0, x1, x2, x3) is larger than the predetermined value H, it is determined that the correlation is low and the reliability is low.

The second embodiment estimates the true minimum value based on the concept of the first conventional example related to the interpolation calculation described in the background art section. That is, fmin shown in FIG. 5 is obtained and compared with the predetermined value H. Although details of the calculation are omitted, as in the first embodiment, when f (i ₀ −1)> f (i ₀ +1), x0 = f (i ₀ −1), x1 = f (i ₀ ), If x2 = f (i ₀ +1), x3 = f (i ₀ +2) and f (i ₀ −1) <f (i ₀ +1), then x0 = f (i ₀ +1), x1 = f ( If i ₀ ), x2 = f (i ₀ −1), and x3 = f (i ₀ −2), then fmin = g (x0, x1, x2, x3) is given by the following equation.

(Equation 17)
g (x0, x1, x2, x3) = (2x1-x0 + x2-k1 (x3-x2)) / 2
Constant of k1 = 0 to 0.5 (II)
When f (i ₀ −1) = f (i ₀ +1), fmin = g (x0, x1, x2, x3) = f (i ₀ ) is set as in the first embodiment. Further, when k1 = 0, this corresponds to the second embodiment relating to the interpolation calculation described above.
The method for determining the predetermined value H is the same as in the first embodiment.

A low-contrast subject has almost no change in the above-described L (j) and R (j), and thus takes almost the same value. Therefore, | L (j) −R (j + i) | (J−s) −R (j + t) | also becomes a value almost close to zero, and eventually the value of the evaluation function itself becomes small. If x0, x1, x2, and x3 in the formulas (I) and (II) are close to zero, the true value estimated by the formulas (I) and (II) even if the left-right symmetry of the evaluation function is lost. May be less than the predetermined value H. In this case, the left-right symmetry is broken and the place where it should be determined that there is a problem in reliability as described above is erroneously determined as no problem. In such a case, it is effective to evaluate the collapse of the symmetry of the direct evaluation function. However, since the S / N ratio of the evaluation function for a low-contrast object is small, it is preferable to detect only those whose symmetry is clearly broken. That is, as shown in FIG. 12, even though f (i ₀ −1)> f (i ₀ +1) holds and the true relative displacement is between i ₀ and (i ₀ +1), the point ( _{i 0 -1, f (i 0} -1)) and the point _{(i 0, f (i 0} )) connecting the straight line and the point _{(i 0 + 1, f (} i 0 +1)) and the point _(i 0 + 2, f This is to detect a case where there is no intersection point connecting (i ₀ +2)) between i ₀ and (i ₀ +1). As in the first embodiment, the condition of f (i ₀ −1)> f (i ₀ +1) is for determining that the true relative displacement X should be between i ₀ and i ₀ +1. In the case of i ₀ -1) <f (i ₀ +1), the positions of i ₀ -2 and i ₀ -1 and i ₀ +2 and i ₀ +1 may be interchanged, and so on. When f (i ₀ −1) = f (i ₀ +1), the following determination is not performed.

Although details of the calculation are omitted, as in the first and second embodiments, when f (i ₀ −1)> f (i ₀ +1), x0 = f (i ₀ −1), x1 = f (i ₀ ), X2 = f (i ₀ +1), x3 = f (i ₀ +2), and if f (i ₀ −1) <f (i ₀ +1), then x0 = f (i ₀ +1), x1 = If f (i ₀ ), x2 = f (i ₀ −1), x3 = f (i ₀ −2) and g (x0, x1, x2, x3) in the following equation is positive, Since the intersection is out of the range that should be (in the case of FIG. 12, the intersection exists to the left of i ₀ ), it is determined that there is a problem in reliability.

(Equation 18)
g (x0, x1, x2, x3) = x1 + x3-2x2 (IV)

FIG. 1 shows a flowchart relating to the reliability determination executed by the microcomputer 64 of FIG. 4 according to the present invention as a fourth embodiment.
In the flowchart shown in FIG. 1, in S1 to S3, the microcomputer 64 takes in the data obtained by digitally converting the analog outputs of the photosensor arrays 53 and 54 by the A / D converters 62 and 63, and the evaluation function f (i) indicating the degree of inconsistency. And i = i ₀ giving the minimum value is obtained. Next, in S4, the four flags FLG1, FLG2, FLG3 and EQF are initialized to zero. FLG1, FLG2, and FLG3 are flags that are set to “1” when it is determined that there is a problem in reliability by performing the determinations described in the first, second, and third embodiments, respectively. The flag EQF is “1” when f (i ₀ −1) = f (i ₀ +1), and is “0” otherwise. In S5, the magnitudes of f (i ₀ −1) and f (i ₀ +1) are compared. If it is determined that they are equal, the process branches to S6, EQF is set to “1”, and the subroutine SINRAISEI is called. On the other hand, S5 with _{f (i 0 -1)> f} (i 0 +1) or if it is determined that _{f (i 0 -1)> f} (i 0 +1) at S5, each branch to step S7 or S8 x0 , X1, x2, and x3 are set and the subroutine SINRAISEI is called. When the reliability is determined by the subroutine SINRAISEI and the values of the flags FLG1, FLG2, and FLG3 are determined according to the result, the operation is terminated.

FIG. 2 shows a flowchart regarding the subroutine SINRAISEI. First, in S9, it is determined whether EQF = 0, that is, f (i ₀ −1) = f (i ₀ +1) is satisfied, and if it is determined that it is satisfied, the process branches to S10. If f (i ₀ −1) = f (i ₀ +1) holds, it is determined that there is no point in performing the interpolation process. In S10, f (i ₀ ) is regarded as the true minimum value and the predetermined value H Compare the magnitude relationship of. If it is determined that f (i ₀ )> H, FLG1 and FLG2 are set to “1”, and the subroutine processing is terminated. If it is determined in step S10 that f (i ₀ ) ≦ H, the subroutine processing is terminated as it is (FLG1 and FLG2 remain “0”).
If it is determined in S9 that EQF is not 0, that is, f (i ₀ −1) and f (i ₀ +1) are not equal, the process branches to S12. In S12, two estimated values of g1 and g2 are calculated regarding the true minimum value, and g3 for judging the collapse of the symmetry of the evaluation function described in the third embodiment is calculated. G1, g2, and g3 correspond to g (x0, x1, x2, and x3) described in the first, second, and third embodiments, respectively.

  In step S13, g1 is compared with a predetermined value H. If g1> H, the process branches to step S14, and FLG1 is set to "1" and the process proceeds to step S15. If g1 ≦ H is determined in S13, the process branches to S15 as it is. In S15, g2 is compared with a predetermined value H. If g2> H, the process branches to S16 to set FLG2 to “1” and proceeds to S17. If g2 ≦ H is determined in S15, the process branches to S17 as it is. In S17, g3 is compared with a predetermined value “0”. If g3> 0, the process branches to S18 to set FLG3 to “1” and the subroutine processing is terminated. If g1 ≦ 0 is determined in S17, the subroutine processing is terminated as it is.

It is a flowchart for demonstrating embodiment of this invention. It is a flowchart for demonstrating subroutine SINRAISEI. It is a distance measurement principle diagram based on the principle of triangulation. It is a block diagram which shows the general example of a distance detection apparatus. FIG. 10 is a diagram for explaining a first conventional example relating to an interpolation method. It is a figure for demonstrating the 2nd prior art example regarding the interpolation method. f _{(i 0)} is a diagram for explaining a case almost equal to _{f (i} 0 +1). It is a figure for demonstrating the case where an evaluation function is not left-right symmetric. It is a figure for demonstrating the example from which the output of two photosensor arrays is not equal with respect to a point light source. FIG. 10 is a diagram for comparison by changing the relative position of the photosensor array output shown in FIG. 9. It is a figure for demonstrating the evaluation function obtained from the photosensor array output shown in FIG. It is a figure for demonstrating Example 3 of this invention.

Explanation of symbols

51, 52 Lens 53, 54 Photo sensor array 55 Subject 56, 57 Subject image 58, 59 Optical axis 62, 63 A / D converter 64 Arithmetic processing device (microcomputer)
FLG1, FLG2, FLG2, EQF flag fmin Estimated true minimum value H Reference value for reliability evaluation k1, k2 constant

Claims (5)

Of the light beams emitted or reflected from the subject, different portions of the light beam are projected onto at least two photosensor arrays, and the distance from the relative displacement of the subject image formed on each sensor array to the subject is determined. To seek
A scale f (i) indicating a degree of inconsistency between the two images with respect to a displacement amount i (integer) times the sensor pitch of the photosensor array, a value i ₀ giving the minimum value, and a function value f ( i ₀₎ and before and after _i 0 _-1 of the _{i 0, i 0 -2, i} 0 + 1, i 0 +2 to the corresponding function value _{f (i 0 -1), f} (i 0 -2), f ( i ₀ +1) and f (i ₀ +2) are obtained respectively, and the relative displacement amount is obtained by interpolation calculation from these values, and the function g (x0, x1, x2, x3) indicating the reliability of the interpolation calculation is obtained. Is compared with the function value f (i ₀ −1) and f (i ₀ +1),
A) f (i ₀ −1)> f (i ₀ +1)
B) f (i ₀ −1) = f (i ₀ +1)
C) f (i ₀ −1) <f (i ₀ +1)
To determine which relationship
In the case of a), the function g (x0, x1, x2) is assumed as x0 = f (i ₀ −1), x1 = f (i ₀ ), x2 = f (i ₀ +1), x3 = f (i ₀ +2). , X3)
B), g (x0, x1, x2, x3) = f (i ₀ )
In the case of c), x0 = f (i ₀ +1), x1 = f (i ₀ ), x2 = f (i ₀ −1), x3 = f (i ₀ −2) and the function g (x0, x1, x2) , X3), and when the value of the function g (x0, x1, x2, x3) exceeds a predetermined value H, it is determined that the reliability of the interpolation calculation is low. The distance measurement method according to claim 1, wherein the function g (x0, x1, x2, x3) is expressed by the following equation (I).
(Equation 1)
g (x0, x1, x2, x3) = A1 / A2
A1 = 2x0 · x2-x0 · x3 + 2x1 · x3-3x1 · x2
B1 = x0−x1−x2 + x3 (I) The distance measurement method according to claim 1, wherein the function g (x0, x1, x2, x3) is expressed by the following equation (II).
(Equation 2)
g (x0, x1, x2, x3) = (2x1-x0 + x2-k1 (x3-x2)) / 2
Constant of k1 = 0 to 0.5 (II) The scale of f (i) indicating the degree of inconsistency between the two images by digitizing the output of the photosensor by an n-bit AD converter and using the digitized photosensor output using m data per photosensor array. The distance measuring device according to claim 2, wherein the predetermined value H is expressed by the following equation (III) when the average random variation of the photosensor output is a%.
(Equation 3)
H = k2 · (a / 100) · 2 ⁿ · m
Constant of k2 = 0.1 to 1.0 (III) The distance measuring method according to claim 1, wherein the function g (x0, x1, x2, x3) is expressed by the following equation (IV), and the predetermined value H is 0.
(Equation 4)
g (x0, x1, x2, x3) = x1 + x3-2x2 (IV)

Images