JPH1068621A - Distance detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect the distance to an object, e.g. an automobile, accurately in order to prevent crash even under disadvantageous condition where the inner pattern of an image can not be detected because the object is shaded by bright backlight. SOLUTION: A pair of image data ID1, ID2 representative of the pattern of an image in the field of view are generated in a pair of image sensors, 21, 22 in an image sensor means 20. A plurality of fields of view are set side by side in the field of view of the image sensor by extracting a pair of window parts WD1, WD2 from the pair of image data for a field of view means 30. An apparent distance corresponding to the parallax of a pair of images in the field of view is then calculated for each field of view from the pair of window parts by a distance measuring means 40. Subsequently, a plurality of probable distance candidates are selected from the frequency distribution of apparent distance by a distance determining means 50. A range EA where an object 1 having outline substantially enveloping the distance candidates is then selected and the distance of the object 1 is determined from the enveloped distance candidates an outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a so-called passive or passive distance detecting apparatus for detecting an image to be detected by an image sensor and detecting a distance to the object to prevent a collision of an automobile.

[0002]

2. Description of the Related Art The above-mentioned passive distance detecting apparatus receives images to be detected through a pair of image sensors through optical means such as a lens through optical paths different from each other, and utilizes the so-called parallax of the images having different optical paths. It is more suitable for detecting a long distance than a so-called active distance detection device that detects the distance to the target by applying ultrasonic waves or infrared rays to the target and detecting the distance from the time reflected from it. It has been used widely in autofocus cameras because of its excellent characteristics, such as extremely good detection directivity and low susceptibility to interfering sound waves and light. The potential is also being considered for applications.

As is well known, a distance detecting device using this image sensor detects a distance using the principle of triangulation, and a parallax on a sensor of an image to be detected received by a pair of image sensors. If σ is detected, the base length, which is the distance between a pair of lenses of the optical means for forming an image on the image, is b, and the focal length of the lens is f, the target distance d is d = bf / σ. It can be calculated with a simple formula.
In an actual application, it is customary to use the parallax σ directly as an index of the distance without having to calculate the value of the distance d from this formula.

However, in an autofocus camera, a detection target can be specified by catching it with its viewfinder, whereas in an application for preventing collision of an automobile, the detection target is not always located in front of an image sensor and is specified by a driver. You can't put a burden. However, even if the target exists in a direction forming an arbitrary angle θ from the front of the image sensor, fortunately, if the parallax σ is detected, it is known that the distance d can be calculated by the same formula as above. A method of setting a plurality of visual fields within the field of view of the sensor, providing a plurality of image sensor pairs, detecting a distance for each of the visual fields, and employing a most probable distance value is adopted.

Japanese Patent Application No. 7-167321 filed by the present applicant proposes a distance detecting device of this type. In this device, a distance to be detected from a visual field in which an object should be detected or a frequency distribution of detected distance values within a detection range can be obtained. The most probable distance range is selected from the result of calculating the average value while selecting a likely distance range and sequentially moving the area selected to include this distance range within the field of view. In a wide field of view including a plurality of image sensor pairs, there may be a plurality of detection targets having different distances from each other. In such a case, a plurality of likely distance ranges are selected. You can also find the distance of

[0006]

However, there is a limit to the accuracy of image detection by an image sensor, so that it may not be possible to accurately detect a distance under bad conditions. As a disadvantageous condition for detection, there is a case where the internal pattern of the image cannot be detected with sufficient accuracy because the target vehicle is receiving bright backlight such as sunset. In this case, since the entire object becomes a shadow that is very dark compared to the brightness of the backlight, it is very difficult to detect the internal pattern of the video. On the other hand, when detecting a distance from an image, the image received by the pair of image sensors is mutually shifted, and the shift value when the pattern matches or shows the highest correlation is the above-described parallax σ. If the internal pattern is poor, the accuracy of pattern matching and correlation test decreases, and the accuracy of detecting the parallax σ as a distance index also decreases.

However, even if the internal pattern of the image of the target cannot be detected, the boundary between the target and the backlight can be detected, so that the distance information cannot be obtained at all. However, according to the results of the experiment, the internal pattern information is poor. It is inevitable that the detection accuracy of the distance is reduced only by the information on the outline of the image. In order to improve the detection accuracy of the internal pattern, it is necessary to use a high-quality and very expensive image sensor capable of obtaining video data with a high accuracy of, for example, 8 bits or more for the image sensor, but of course there is a practical disadvantage.

[0008] Based on this situation, the object of the present invention is to provide a particularly high image sensor even under disadvantageous conditions in which the internal pattern of an image to be detected is very scarce and only information on the contour pattern can be obtained. It is an object of the present invention to provide a distance detecting device that can accurately detect a target distance without using an accurate one.

[0009]

According to the distance detecting apparatus of the present invention, a pair of image data representing a pattern of an image in the field of view is generated by an image sensor having a pair of image sensors, and the field of view is set by a field of view setting means. A plurality of visual fields to be detected in the field of view of each image sensor are set side by side, and the parallax of the video pair in the visual field within the visual field from the pair of window parts extracted corresponding to the visual field from the pair of video data by the distance calculating means. A corresponding apparent distance is calculated for each visual field, and a plurality of detection distance candidates are selected from a plurality of apparent distance frequency distributions by the distance determining means, and an existence range of an object having a contour substantially enclosing them is selected. The above-mentioned problem is solved by determining and outputting the target distance from the distance candidates belonging to the existence range.

The present invention focuses on the fact that even when the accuracy or reliability of the detection distance is considerably low, the distribution of the visual field whose distance has been detected almost accurately tends to converge within a certain range within the field of view. Yes, the distance determination means first selects a plurality of probable distance candidates from the frequency distribution of detected distances, and the existence of an object with a contour that almost encompasses the range where the corresponding fields of view are gathered in the field of view By selecting a range, the distance of the target can be determined almost accurately from the distance candidates within this range.

The image sensor means as described above uses an integrated circuit device incorporating an image sensor and its related circuits, and is used as an image detection module together with optical means for forming an image in the field of view on the image sensor. It is convenient to put together. It is desirable that the field-of-view setting means be set so that a plurality of fields of view slightly overlap each other so that images in the field of view of the image sensor can be captured without omission. Further, it is advantageous to use an integrated circuit device incorporating a plurality of distance calculation circuits as the distance calculation means to calculate the distances for a plurality of fields of view simultaneously and in parallel, in order to increase the distance calculation speed as much as possible. While a hardware circuit is advantageous for the distance calculation means, the distance determination means is advantageously loaded as software on a processor such as a microcomputer, preferably together with the view setting means. The distance determination means may include a plurality of distance candidates, for example, two to several detection distances from the frequency distribution so as to extract as useful information as possible from the plurality of detection distances calculated by the distance calculation means. Is advantageously selected.

In particular, in the case of automobile collision prevention, it is necessary to automatically find an unspecified target, so that it is desirable to set a wide field of view for capturing the target. For this purpose, a plurality of image sensors are provided in the image sensor means. The image is captured within the detection range in which the fields of view are arranged in a two-dimensional matrix by combining multiple fields of view and multiple image sensor pairs set for each image sensor pair by field of view setting means, and used as distance calculation means. It is desirable to calculate the apparent distance of the image for each field of view forming the matrix. Further, in this case, a two-dimensional map showing the distribution state of the distance candidates selected by the distance determining means within the detection range is created, and a rectangular contour is formed so as to circumscribe the pattern in which the distance candidates are distributed in this map. It is reasonable to select an area as the existence range of the object.

In order to set the existence range correctly and accurately determine the distance of the object, the apparent size of the object when the object exists at a distance position corresponding to the distance candidate selected by the distance determination means. After confirming that the existence range of the object selected as described above is approximately the same size as this frame, the distance of the object is determined from the distance candidates belonging to this existence range. Is desirable. Furthermore, in order to ensure the accuracy of the determined distance, it is most reasonable to adopt the average value of all distance candidates belonging to the target existence range as the target detection distance.

However, distance candidates may be discretely distributed in the detection range. If the degree of discreteness is large, it becomes difficult to match the existence range and the size of the frame as described above. Create a map weighted by taking into account the aggregation state of the candidates, set a frame representing the apparent size when the target exists at the distance position corresponding to the distance candidate, and set the It is advantageous to select a target existence range by fitting a frame to the map so that the total of the weights assigned is maximized.

[0015]

Preferred embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 1 shows an example of the configuration of a distance detection device according to the present invention together with a target whose distance is to be detected.
FIG. 2 shows an operation example of the distance determining means constituting the device of the present invention, and FIG. 3 shows a different operation example of the distance determining means. In the embodiment described below, it is assumed that the distance detection device is for preventing collision of an automobile.

In the example shown in the upper part of FIG. 1, the detection target 1 which is an automobile has received a backlight such as a bright sunset, and therefore has a dark shadow as shown by hatching and should be recognized in a normal state. The internal patterns of the name plate, bumper, tail lamp, window, etc. are unclear, and only the outline of the boundary with the backlight is prominent. The optical means 10 shown below captures the image of the object 1 by a pair of lenses 11 and 12 and gives the image to the image sensor means 20 through different optical paths L1 and L2. And the distance b between the optical axes of 12 and 12 is the so-called base line length for detecting the distance d of the object 1 based on the principle of triangulation.

In order to prevent the collision of a car, it is necessary to catch the object 1 in an unspecified direction with a wide field of view. Therefore, the image sensor means 20 incorporates a plurality of pairs of image sensors 21 and 22, for example, a few to a dozen pairs. Is good. In this application, it is preferable from experience that the lenses 11 and 12 of the optical means 10 and the image sensors 21 and 22 are arranged in a vertical direction, but for convenience of illustration, they are shown in a horizontal arrangement in FIG. I have.
The object 1 is in the direction of the angle θ from the front of the optical means 10 and its image is formed on the image sensors 21 and 22 located at the position of the short focal length f between the lenses 11 and 12. Each sigma ₁ the deviation from the reference position of the image sensor 21 and 22 of the center of the _picture, and sigma _2, and when the σ = σ ₁ + σ _2, target 1
Is d = bf / σ regardless of the angle θ as described above.
Can be calculated from the following equation, but σ is usually used as an index of the distance d.

It is convenient to use a CCD for the image sensors 21 and 22. In the example shown in the figure, a switching circuit 23 for receiving a detection signal from the CCD, an amplifier circuit 24, and an AD converter 25 are incorporated in the image sensor means 20. Image sensor 21
For example, the A / D converter 25 outputs 8-bit precision video data representing a video pattern received by the video converter 22. The image sensor means 20 is an integrated circuit chip in which these related circuits are built together with a plurality of pairs of image sensors 21 _and 22, and is mounted together with the optical means 10 in the form of a module housed in a small package and mounted on an automobile. It is convenient.

The processor 70 shown in the lower part of FIG. 1 is, for example, a microcomputer, and is shown in a form incorporating the visual field setting means 30, the distance calculating means 40 and the distance determining means 50 for convenience of illustration. Preferably, the means 40 is configured as a hardware integrated circuit device separate from the processor 70. Memory shown at the top of processor 70
71 receives the output from the aforementioned AD converter 25 of the image sensor means 20 and stores it as a pair of video data ID1 and ID2. Each image data is a set of digital sensor data representing the intensity of light received by each of several hundred optical sensors of the image sensors 21 and 22. The field-of-view setting means 30 extracts the window portions WD1 and WD2 from these video data ID1 and ID2 as shown by thin lines in the figure, thereby providing a plurality of fields of view for capturing the object 1 within the field of view of the image sensors 21 and 22. In general, dozens are set side by side, and a plurality of fields of view set for each image sensor and a detection range DA in which the fields of view are arranged in a two-dimensional matrix by a plurality of pairs of image sensors 21 and 22 are set as shown in FIG. Is set so as to show the outline overlapping with the object 1. Note that the fields of view set for each pair of image sensors are preferably slightly overlapped with each other so that images in the field of view can be captured without omission.

The distance calculating means 40 is preferably a hardware integrated circuit device as described above, and a so-called gate array is suitable for this. In the example of FIG.
From the distance calculation means 40, each unit calculation circuit
Reference numeral 41 denotes a pair of window portions WD1 and WD extracted from the pair of video data ID1 and ID2 from the visual field setting means 30, as indicated by the thin lines in the figure.
2, the index σ of the apparent distance corresponding to the parallax of the image pair in each field of view constituting the matrix of the detection range DA is calculated. Since a plurality of unit calculation circuits 41 perform this calculation in parallel, the calculation can be completed in a short time.
This calculation result is preferably stored in the memory 72 of the processor 70.

Advantageously, the distance determining means 50 is preferably loaded in software in the processor 70 together with the field setting means 30 described above. This distance determination means 50 is the above-described detection range DA which is a calculation result of the above-described distance calculation means 40.
Within the two-dimensional matrix, the apparent distances for each field of view are read from the memory 72, a frequency distribution of distances is first created, and then a plurality of candidates for the most probable detection distances are selected. After selecting the existence range of the target 1 having the enclosing contour, the distance of the target 1 is determined and output from the distance candidates belonging thereto. Hereinafter, an operation example of the distance determining means 50 will be described with reference to FIGS. FIG. 2 (a) is a diagram showing a frequency distribution of apparent distances, FIG. 2 (b) and FIGS. 3 (a) and 3 (b) are schematic diagrams showing a procedure for selecting an existence range, and FIGS. 2 (c) and 3 (b). (c) is a flowchart showing an operation example of the distance determination means.

In the frequency distribution diagram of FIG. 2A, the abscissa indicates the index σ of the apparent distance calculated by the distance calculating means 40, and the ordinate indicates the frequency of occurrence n. Although there is some unevenness in the distribution, in the example shown in the figure, there is a range A with a high frequency n at the center, and σ within this range A is selected as a most probable distance candidate. The width of the range A is determined so as to include a plurality of distance candidates, for example, two to several detection distances from the frequency distribution so as to extract as useful information as possible from the plurality of distances calculated by the distance calculation means 40, or It is preferable that the frequency included therein is determined so as to be several to several ten percent of the total frequency.

FIG. 2B is a map showing how the distance candidates are selected and distributed in the detection range DA in this manner. For simplicity, 1 is added to the field of view corresponding to the distance candidates. , 0 is added to the other fields of view. In the map, the existence range EA of the target 1 is indicated by a thin-line frame. In the embodiment of FIG. 2, a rectangular outline circumscribing the distribution pattern 1 in FIG. Is selected as the existence range EA. However, in order to ensure the accuracy of the detection distance, it is desirable to confirm that this selection is appropriate. For this reason, in the embodiment of FIG. 2, a frame FM representing the size of the target 1 located at a position separated by a distance corresponding to the distance candidate is set as shown in the figure, and the width and height of the existence range EA substantially coincide with it. Make sure you do. Existence range
After the EA is determined in this way, it is reasonable to calculate the total average value of the distance candidates belonging to the EA and use it as the detection distance of the target 1.

FIG. 2C is a flowchart showing the above operation of the distance determining means 50 which is software. First step S51
Then, a frequency distribution of the distance is created as shown in FIG.
At 52, a distance candidate within the range A is selected. In step S53, a matrix of 0s and 1s corresponding to the map of the distance candidates in FIG. 2 (b) is created. In step S54, an existence range EA having a rectangular outline is selected. In step S55, a frame corresponding to the distance candidates is selected. Set FM. In step S56, it is determined whether or not the existence range EA is almost the same as the frame FM. If so, the existence range EA is determined. Therefore, in step S57, the total average value σ _av of the distance candidates belonging to the existence range EA is determined. It is calculated as the distance index σ, but if not, a value σ _f that clearly indicates that the distance detection has failed is inserted into the distance index σ in step S58. In step S59, the average value σ _av or σ _f indicating failure in detection is output as the distance index σ, and the operation as the distance determination means 50 is completed.

However, as shown in FIG. 2B, the distance candidates are not always distributed collectively in the detection range DA, but may be distributed considerably discretely. It is difficult to determine the EA by matching it with the frame FM. The embodiment shown in FIG. 3 can be suitable in such a case, and even when the distance detection fails in the embodiment of FIG. FIG. 3A is a map corresponding to FIG. 2B, and shows an example in which the distance candidates indicated by 1 are discretely distributed over almost the entire detection range DA. In the embodiment of FIG. 3, even in such a case, by using a point where a difference appears in how the visual fields corresponding to the distance candidates are present in the existence range EA and in other cases, weights are added to each distance candidate in consideration of the aggregation state. Make a map with, and set the above-mentioned frame FM corresponding to the distance candidate,
The existence range EA is selected by fitting the frame FM to the map such that the sum of the weights of the distance candidates in the frame FM is maximized.

FIG. 3B shows an example of a map in which each distance candidate is weighted, corresponding to FIG. 3A. In the example shown in the figure, a weight of 1 is first given to each distance candidate itself. If there are distance candidates adjacent to the upper, lower, left and right sides, a weight of 2 is added to each distance candidate, and there are four adjacent distance candidates in oblique directions. Are also weighted by 1 for each. The result of applying the frame FM shown below to this map in the manner described above is shown by the existence range EA. In the example of the figure, this existence range
The sum of the weights of the distance candidates in the EA is 51, and as can be easily understood, if the position of the existence range EA is slightly shifted in the detection range DA, the sum of the weights decreases. Can be uniquely determined in most cases.

FIG. 3C shows an operation example of the distance determining means 50 in this embodiment. Steps S51 and S52 of the first
Is the same as that of FIG. 2 (c), but the next step S53a
Then, a map A shown in FIG. 3A is created, and furthermore, a step S53b
The map B shown in FIG. In step S61, the flow control flag F is set to 0, and in step S55, the flow control flag F is set to correspond to the distance candidate in the map.
(b) Set the frame FM. In step S62, it is checked whether or not the flag F is 0. Since this is the case at first, the operation proceeds to step S63, in which the flag F is set to 1, and in step S54, the frame FM is added to the map B as described above. The existence range EA shown in FIG. 3B is set by fitting in the manner described above, and the size of the frame FM is stored in preparation for the subsequent operation.

In the next step S57, the set existence range is set.
After calculating the average value σ _av of the distance candidates belonging to the EA, the flow returns to step S55, and the frame FM is set again so as to correspond to the average value σ _av . In the subsequent step S62, it is determined whether or not the flag F is 0. Since the flag F is set to 1 this time, the determination always comes out as no, and the operation proceeds to step S64. In this step S64, the new frame FM reset in step S55
Is the same as that stored previously in step S54. If this determination is negative, the flow returns to step S54 to re-select using the frame FM for which the existence range EA has been reset. In most cases, however, the determination result is clear. Then, the average value σ _av calculated in step S57 is output as the distance index σ, and the operation of the distance determination means 50 in the embodiment of FIG. 3 is completed.

[0029]

As described above, in the present invention, the visual field in which the distance in the field of view of the image sensor is almost accurately detected even when the reliability of the detection distance is considerably low because the image pattern to be detected has a poor detail structure. Focusing on the point that the distribution is likely to be gathered in a certain range, the distance determination means selects the most probable distance candidates from the frequency distribution of the detected distance, and determines the range where they are gathered in the field of view Since the existence range of the object with the outline that almost wraps around was selected, and the distance of the object was determined from the distance candidates within this existence range,
(a) The distance of the object is almost certainly and accurately even under bad conditions where the object such as a car receives a bright backlight and becomes a dark shadow, so that the image can only be captured from the boundary between the shadow and the backlight. (B) Since the existence range is determined at the same time as the detection of the distance of the object, the field of view and the visual field to capture the object are set to the direction and width suitable for the existence range, so that the detection of the object can be performed. The possibility of failure can be reduced and the distance detection accuracy can be increased. The distance detecting device according to the present invention having such features is particularly suitable for preventing automobile collision.

The embodiment in which a plurality of unit calculation circuits are incorporated in the distance calculation means by using an integrated circuit device in the distance calculation means has the effect of shortening the distance detection time by simultaneously proceeding distance calculation for a plurality of visual fields within the field of view. The embodiment in which the distance determining means and the visual field setting means are loaded into the processor in the form of software has the effect that their operation details can be easily and at any time changed to suit the usage of the apparatus.

A plurality of pairs of image sensors are incorporated in the image sensor means, and each field of view within a two-dimensional matrix detection range is obtained by combining these image sensor pairs and a plurality of fields of view set for each image sensor pair by the field of view setting means. The embodiment in which the distance calculation means calculates the distance for each case, particularly in the case of car collision prevention, captures an image in a wide field of view, so that objects existing in unspecified directions can be leaked without burdening the driver. There is an effect that can be found automatically.

The embodiment in which a plurality of distance candidates are selected from the frequency distribution of detected distances by the distance determination means is advantageous in that useful information can be extracted as much as possible from the plurality of apparent distances calculated by the distance calculation means. is there. The embodiment in which the two-dimensional map showing the distribution state within the detection range of the distance candidate is created by the distance determination means and the existence area having a rectangular outline circumscribing the distribution pattern is selected, the object existence area can be easily and reasonably determined. In the embodiment in which a frame indicating the apparent size when the target is located at a position corresponding to the distance candidate is set and the existence range substantially matches with the embodiment, the existence range is correctly selected. The distance can be accurately detected.

Further, the distance determining means creates a weighted map for each distance candidate in consideration of the aggregation state thereof, and a frame representing the apparent size when the object is located at a position corresponding to the distance candidate. In this embodiment, the range of the target is selected by setting the frame so that the sum of the weights assigned to the distance candidates on the map is maximized in the frame. For this reason, even under disadvantageous conditions in which the distance candidates are widely and discretely distributed over almost the entire detection range, there is an effect that the target existence range can be reliably and rationally selected and the distance can be accurately determined.

[Brief description of the drawings]

FIG. 1 is a configuration circuit diagram showing a configuration example of a distance detection device of the present invention and a detection target.

FIG. 2 shows an operation example of a distance determination means used in the present invention,
Figure (a) is a diagram showing the frequency distribution of apparent distances.
(b) is a schematic diagram showing how to set the existence range, and FIG.
6 is a flowchart illustrating an operation example of a distance determination unit.

FIGS. 3A and 3B show different operation examples of the distance determination means used in the present invention. FIG. 3A is a schematic diagram showing an example of distribution of distance candidates within a detection range, and FIG. 3B is a diagram showing weighted distance candidates. FIG. 3C is a schematic diagram showing a distribution example and a procedure for selecting an existence range, and FIG. 4C is a flowchart showing an operation example of the distance determination means.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Detection object 10 Optical means for forming an image on an image sensor 11, 12 A pair of lenses of optical means 20 Image sensor means 21, 22 Image sensor pair 30 Field of view setting means 40 Distance calculating means 41 Constructing distance calculating means Unit calculation circuit 50 Distance determination means 70 Processor or microcomputer A Selection range of distance candidate b Base line length in distance detection d Distance to be detected DA Detection range FM Frame corresponding to the existence range f Lens focal length EA Existence range ID1, ID2 A pair of video data n Occurrence frequency in the frequency distribution of distance candidates WD1, WD2 Window portion in video data σ Distance index θ Angle in the direction in which the target exists

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Akihiko Okabe 1-1, Tanabe-Nitta, Kawasaki-ku, Kawasaki, Kanagawa Prefecture Inside Fuji Electric Co., Ltd. No. 1 Fuji Electric Co., Ltd.

Claims (5)

[Claims] 1. An image sensor means comprising a pair of image sensors and emitting a pair of image data representing a pattern of an image in the field of view, and a field of view in which a plurality of fields of view to be detected are set in the field of view of the image sensor. Setting means; distance calculating means for calculating, for each visual field, an apparent distance corresponding to the parallax of the pair of images in the visual field from a pair of window portions extracted corresponding to the visual field from the video data; A distance determining means for selecting a plurality of detection distance candidates from the frequency distribution of the distance, selecting an existence range of an object having a contour substantially enclosing them, and determining and outputting the distance of the object from the distance candidates belonging thereto; and A distance detection device comprising: 2. The apparatus according to claim 1, wherein a plurality of image sensor pairs are incorporated in the image sensor means, and a plurality of visual fields and a plurality of image sensor pairs set for each image sensor pair by the visual field setting means are provided. Capture images in the two-dimensional matrix-like detection range that is combined,
A distance detecting device, wherein an apparent distance is calculated for each field of view constituting a matrix within the detection range by distance calculating means. 3. The apparatus according to claim 2, wherein a two-dimensional map showing a distribution state of the distance candidates selected by the distance determination means within a detection range is created, and the distribution pattern of the distance candidates in the map is circumscribed. A distance detecting device, wherein a rectangular area is selected as a target existence range. 4. The apparatus according to claim 1, wherein when an object is present at a distance position corresponding to the distance candidate selected by the distance determination means, a frame representing the apparent size of the object is set, and the selected object is set. A distance detection device for determining the target distance from distance candidates within the existence range after confirming that the existence range is substantially the same size as the frame. 5. An apparatus according to claim 1, wherein a total average value of distance candidates within the range of the target selected by the distance determining means is determined as the detection distance of the target. apparatus.