JPH0697162B2 - Distance measurement method for three-dimensional objects - Google Patents

Description

TECHNICAL FIELD OF THE INVENTION The present invention relates to a distance measuring method for a three-dimensional object for obtaining three-dimensional position information of the object.

[Technical Background of the Invention and Problems Thereof] For example, in a nuclear power plant, it is considered to inspect and monitor each device and equipment by an autonomous vehicle or a robot. In general, ultrasonic waves, light (such as a television camera), and laser are applied to the visual system of the above-mentioned autonomous vehicle or robot. When ultrasonic waves are used in the visual system, they are widely used because the distance to the object can be easily measured.However, because ultrasonic waves have directivity, obstacles existing in a specific direction Mainly detection.

In order to obtain three-dimensional position information in the forward direction of the autonomous vehicle or robot, mechanical two-dimensional scanning is required, which causes a decrease in information input speed, and when ultrasonic waves are used. However, the detection resolution of the object is poor, and highly accurate three-dimensional position information cannot be obtained. Further, although the above-mentioned television camera is widely used as a two-dimensional image input device, in the case of a television camera, it is difficult to detect the distance to an obstacle and three-dimensional position information cannot be obtained. Therefore, it is considered to obtain three-dimensional position information by using two TV cameras, but in this case also, the image captured by the TV camera is limited in the features of obstacles, the nature of the environment, etc. However, it has not been possible to obtain highly accurate three-dimensional position information.

[Object of the Invention] An object of the present invention is to provide a distance measuring method for a three-dimensional object capable of obtaining three-dimensional position information of an unspecified number of objects with high accuracy.

[Summary of the Invention] A distance measuring method for a three-dimensional object according to the present invention moves an imaging unit in parallel from an initial position to a final position with respect to an object to be imaged, and the imaging unit measures the distance from the initial position to the final position. Grasp the component surface of the imaged object for each predetermined movement position as an image of uniform brightness within the plane, and divide the grasped component surface of the imaged object into regions of the same brightness. All the planes divided into regions at the position and the final position are stored, and the image pickup unit is obtained from the picked-up image for each predetermined moving position from the initial position to the final position through a plurality of surfaces obtained by dividing the region at the initial position. The cross-section data of a predetermined width parallel to the moving direction is extracted and created and stored as a cross-section image, the cross-section image is divided into areas with the same brightness, and the area-divided area is shared by the area-divided areas. Each component element surface is associated with the surface stored at the initial position and the final position included in the same plane as the same surface, and obtained at the initial position and the final position of each associated component surface of the object. The average value of the expansion / contraction ratio in the moving direction of the image capturing unit on each component surface is determined as the expansion / contraction coefficient of each component surface, and the expansion / contraction coefficient of each component surface determines the image in the initial position and the image in the final position. The coordinates of the pixels on all the imaging screens that make up each component surface are associated, and all of the component surfaces are identified from the associated coordinates of all the pixels and the initial position and the final position of the imaging unit. The visual sense at the initial position and the final position of the point on the target object corresponding to the pixels of is determined, and from this determined vision, the imaged target corresponding to all the pixels on each surface that is a constituent element of the plurality of target objects. It is characterized by comprising calculating the distance of the points above.

[Examples of the Invention] Examples of the present invention will be described below with reference to the drawings. In FIG. 1 showing a schematic configuration of a three-dimensional image generation apparatus to which the distance measuring method for a three-dimensional object of the present invention is applied,
Reference numeral 1 denotes an image pickup unit including an area type image sensor and its lens system. The image pickup unit 1 can be moved by a moving mechanism 2. The moving mechanism 2 includes a pulse motor 3 and a shaft 4 connected to the pulse motor 3. The image pickup unit 1 moves along the shaft 4 in the direction of the arrow in FIG. 1 and picks up images at the respective positions P _{0 to} P _M as shown in FIG.

The image pickup unit 1 has a function of converting the optical information of the outside world into an electric signal of a two-dimensional image, and by supplying a control clock from the control circuit 5, the image signal is converted into an analog-digital conversion circuit ( Hereinafter, it will be referred to as an A / D conversion circuit) 7. The pulse motor 3 is controlled by the control circuit 5 via the driver 8. The conversion memory circuit 6 is composed of the A / D conversion circuit 7, the address counter 9 and the image memory 10. The image signal digitized by the A / D conversion circuit 7 is generated by the address counter 9. It is configured to be stored in the image memory 10 by the address signal.

The image data of the image memory 10 input at the initial position P ₀ in FIG. 2 is divided into areas by the image processing circuit 11, and as a result, a different number (area number) is assigned to each area. The divided images are stored in the first image memory 13 of the storage circuit 12. The image processing circuit 11 obtains the vertical dimension and the position (the vertical direction here is the direction orthogonal to the moving direction of the image pickup unit 1) of each area of the data stored in the first image memory 13, and the details will be described later. Common cross-section data to be determined.

Next, move the image pickup unit 1 to the position of P ₁ and store the image in the image memory.
Enter in 10. From the image memory 10, section data at the same position as the extracted section data is retrieved and stored in the image memory 14. At this time, the sectional data of P ₁ is stored after the sectional data of P ₀ . This processing is performed at the position of the imaging unit 1.
The processes from P ₀ to P _{M are} performed, and cross-sectional data is generated in the image memory 14. Here, the image (original image) in P _M is stored in the image memory 10.

Next, the image in P _M of the image memory 10 is the image processing circuit.
The area is divided by 11 and, as a result, a different number (area number) is assigned to each area, and the area is divided into a memory circuit.
Twelve third image memories 15 are stored. Image memory
The image of the cross-section data (cross-section image) generated in 14 is divided into regions by the image processing circuit, and the regions are numbered. This area number connects each area of the area-divided image at P ₀ (image memory 13) and the area-divided image at P _M (image memory 15) (details will be described later).

Correspondence of pixel coordinates from the area-divided image stored in the third image memory 13 by the arithmetic circuit 16 when the imaging unit 1 is at the initial position P ₀ and the area-divided image at the final position P _{M by the} imaging unit 1. It is configured so that the distance to the target object is calculated. The image processing circuit 11 and the arithmetic circuit 16
Are both controlled by the control circuit 5.

Next, the operation of the three-dimensional image generation apparatus to which the distance measuring method for a three-dimensional object of the present invention configured as described above is applied will be described. First, in FIG. 1, when the image pickup unit 1 is at the initial position P ₀ at the left end of the moving mechanism 2, the control circuit 5
Outputs a two-dimensional image signal output start signal and a drive pulse (hereinafter referred to as a clock) from the image pickup unit 1. The image pickup unit 1 outputs the image signal of the image picked up by this signal to the A / D conversion circuit 7. The output start signal from the control circuit 5 functions as an initialization of the address counter 9,
The clock is the counter input of the address counter 9 and A /
It functions as a sampling signal for the D conversion circuit 7 and generates an address signal for the image memory 10. That is, the A / D conversion circuit 7 A / D-converts the image signal from the image pickup unit 1 to digitize it by the sampling signal, and writes the image signal digitized by the address signal in the image memory 10. The strength of the image signal from the image pickup unit 1 is determined by the brightness of the imaged target. Further, the writing of the digitized image signal into the image memory is performed by representing the brightness at each position of the captured image by a numerical value. The image memory 10 stores image signals as shown in FIG. 3, for example. FIG. 3 is a diagram showing the existence of an object to be imaged, and reference numerals 20, 21, and 22 in the drawing show rectangular parallelepipeds of the objects existing around.

Next, the control circuit 5 outputs the following command signal to the image processing circuit 11 based on the data stored in the image memory 10. That is, the image stored in the image memory 10 is divided into areas, and area-divided images with area numbers are generated. This will be explained using the image of FIG. 3 described above. First, the image shown in FIG. 3 (image with contrast)
The edges are smoothed to remove the fluctuation due to the noise of the image signal, and each of the component surface of the rectangular parallelepiped 20, 21, 22 is grasped as an image of uniform brightness. This operation removes high-frequency changes in the luminance distribution in the same direction, and is not necessary in the case of an image having a uniform luminance distribution.

In the case of FIG. 3, the rectangular parallelepiped 20, 21, 22 Represent the same brightness (density), and are generally defined by the illumination direction and the surface orientation, and are images with contrast. The smoothed image has a brightness difference, that is, Area division is performed based on the minimum value of the difference between the image data, and a cluster of pixels having the same brightness is found. FIG. 4 is a result of area division of the image of FIG. 3, and shows eight areas (to in the figure) + background area (in the figure).
Indicated by. As a result, the nine areas are numbered 1, 2, ... 9 as shown in FIG. Becomes The area-divided image shown in FIG.
Generated to 13. Next, the vertical dimension and position of each area of the area-divided image shown in FIG. 4 are obtained for each area number, and stored in the memory inside the image processing circuit 11 as a bit pattern as shown in FIG.

Here, for the image shown in FIG. 4, the horizontal direction is the x-axis and the vertical direction is the y-axis. Further, this coordinate axis is related to the moving direction of the image pickup unit 1, and an axis parallel to the moving direction is the x axis. The bit pattern in FIG. 5 detects the position on the y-axis where each region exists for each region in FIG. 4, sets the region where each region exists as “1”, and sets each region number as the horizontal axis. The y-coordinate position of each area in FIG. 4 for each area number is shown in
It is shown as an axis.

The procedure for obtaining the cross-section data f (y ₁ ), f (y ₂ ) ... f (y ₅ ) to be extracted from this bit pattern data is shown below. Regarding the data of the bit pattern shown in FIG. 6 (a), y = 0 first changed to "1": The value y of y near the center of the area number 9 (range of "1"): y ₁ was obtained, and y Other areas including ₁ and "1" in area 9 are erased. In FIG. 6A, the area including y ₁ is only the area number 9. Next, similar processing is performed to obtain y ₂ and y ₃ . At this time, the areas 3 and 6 become "1" at the same time, but in this case, they are obtained based on a predetermined procedure, for example, from the smallest area number. Similarly, y ₄ and y ₅ are obtained as shown in FIGS. 6 (b) and 6 (c).

The y = y _i thus obtained becomes the coordinates of the y-axis passing through each area of the area-divided image (FIG. 4) stored in the image memory 13, and using this value, the imaging unit 1 sets the initial position P When the cross section data f (x, y _i ) is extracted from the image in the image memory 10 at the time ₀ , the initial address of the image memory 14 (the storage location of the cross section data at P ₀ in the figure is extracted as shown in FIG. 7). Is stored at the initial address, and the storage location of the cross-section data at P ₁ in the figure is stored at the next address, ...).

Next, the control circuit 5 outputs a signal to the driver 8 to move the imaging unit 1 to the next position P ₁ parallel to the object to be imaged. As a result, the driver 8 outputs a signal to the pulse motor 3 to move the image pickup unit 1 to the position P ₁ .
Then, perform the operation performed at the initial position P ₀ described above,
The image signal is stored in the image memory 10, the cross-section data f (x, y _i ) is taken out based on the extraction position y ₁ , y ₂ ... Y ₅ of the cross-section data already determined, and the predetermined address of the image memory 14 is obtained. (Following the storage location of the section data of P ₀ ).

Next, the control circuit 5 outputs a signal to the driver 8 to move the image pickup unit 1 to the next position P ₂ . As a result, the driver 8 outputs a signal to the pulse motor 3 and the imaging unit 1
To position P ₂ . Then, the same operation is performed. Then, the imaging unit 1 is sequentially moved to perform processing up to the final position P _M. As a result, cross-sectional images as shown in FIGS. 7, 8 and 9 are generated.

Next, the control circuit 5 sends the final position P to the image processing circuit 11.
The image of _M (stored in the image memory 10) is instructed to perform the above-described image processing (smoothing and area division performed on the image of P ₀ ) and store the image in the image memory 15. As a result, the final position P is stored in the image memory 15.
The area-divided image of the image at _M is stored. The area-divided image at this time does not necessarily match the correspondence between the area number and the area (pattern in FIG. 4) of the area-divided image generated by the imaging unit 1 at the initial position P ₀ . That is, the area number And may differ from the above correspondence. For this reason, the information for connecting each area of the area-divided image of the image memory 13 to each area of the area-divided image of the image memory 15 is the cross-sectional image of the image memory 14.

That is, the cross-sectional image of FIG. 9 shows each image obtained by the movement of the image pickup unit 1 by a cross section at a certain position of the y-axis, and in this cross-sectional image, a certain area having the same brightness is included in the image pickup unit. It shows how to move with the movement of 1. For example, the movement of the divided image area at the position y = 5 is shown in the sectional image of f (x, y ₅ ) in FIG. 9B, and each area shown in the sectional image is Which region of the original image corresponds to is determined by associating regions of the same brightness.

The procedure is shown below. The control circuit 5 instructs the image processing circuit 11 to perform area division for each of the component images recognized as the same plane for each cross-sectional image of the image memory 14, for example, from the position where the same luminance is continuous and the continuous position. As a result, the ninth
The sectional image shown in f (x, y ₅ ) of FIG. 10 (b) is shown in FIG. 10 (b).
The area is divided as follows. In the area number of Fig. 10 (b) Is included in the region number of the image at P ₀ (Fig. 10 (a)), Is included in a certain area of the area-divided image at P _M. That is, the areas that commonly include the area number of the cross-sectional image are processed as the same area. Similar processing is performed for f (x,
Repeat for other area numbers in the y ₅ ) cross-sectional image. By performing these processes on all the cross-sectional images, the areas of the image memories 13 and 15 can be associated with each other.

Of the area-divided images obtained in this way, the initial position
Using the region-divided image obtained at P ₀ and the region-divided image obtained at the final position P _M , arithmetic processing is performed to calculate the distance to the object and obtain three-dimensional position information. That is, the initial position
The image captured at P ₀ and the image captured at the final position P _M are information about the same object captured at different positions (in other words, with parallax). By associating these two data The three-dimensional position information is obtained.

The processing will be described below. First, the control circuit 5 outputs an instruction signal for causing the arithmetic circuit 16 to calculate the three-dimensional position information. The arithmetic circuit 16 extracts from the image memory 13 and the image memory 15 each region (component surface) of the associated region-divided images, that is, the region-divided image at the initial position P ₀ and the multi-valued image at the final position P _M , and extracts each of them. The expansion / contraction coefficient is calculated for each area (component surface).

This calculation operation will be described with reference to FIG. Generated at the initial position P ₀ and the final position P _M, retrieves the set of region surface which is compatible with (indicated by a region number Fig. 10 (b)), determine the scale factor of the region (component surface). The expansion / contraction coefficient represents the degree of change in the component surface due to the difference in parallax, with the initial position P ₀ as a reference. In the case of FIG. 11, the image b at the final position P _M extends in the moving direction of the image pickup unit 1 with respect to the image a at the initial position P ₀ . The length shown in the figure for this coefficient
K ₁ , K ₂ ...... K _N and the length L ₁ , L ₂ ...... L _N ratio, that is, L ₁ / K ₁ ,
L ₂ / K ₂ …… L _N / K _N is calculated and the average value is calculated. Is the expansion / contraction coefficient A ₀ of the area. This coefficient is calculated for all the generated regions and used to associate all the pixels of the region j at the initial position P ₀ with the pixels forming the region j at the final position P _M. That is, the position of the pixel in the direction is calculated using the expansion / contraction coefficient A ₀ with the left edge of the region at the initial position P ₀ and the left edge of the region at the final position P _M as references. As a result, it is possible to associate each pixel unit as shown in FIG.

Next, the three-dimensional position information is calculated from the associated coordinates for each pixel unit. In FIG. 13, the imaging center of the imaging unit 1 is assumed to move in parallel, and the initial position P ₀ and the final position P _M
The image captured in step 1 is one-dimensional, and the center pixel of the image is the origin of the coordinate system. One point Q ₀ of the object is obtained as coordinate values X ₀ and X _{m at the} initial position P ₀ and the final position P _M , respectively. Now, the movement amount P _M −P ₀ of the image pickup unit 1 is 1, the distance between the lens of the image pickup unit 1 and the image pickup surface is Y ₀ , the viewing angle when viewing Q ₀ from the initial position P ₀ is θ ₀ , and the final position P _M If the viewing angle when Q ₀ is viewed from is θ _M , the viewing angles θ ₀ and θ _M are calculated by the following equations.

θ ₀ = tan ^-1 Y ₀ / X ₀ (1) θ _M = tan ^-1 Y ₀ / X _M (2) If the distance to one point Q ₀ on the object is YQ ₀ , then YQ ₀ = { sin θ ₀ · cos θ _M / sin (θ _M −θ ₀ )} · 1 ...
(3) The above calculation is performed on all the area surfaces of the initial position P ₀ and the final position P _M to obtain distance information regarding the surfaces (all points) forming the object to be imaged. These pieces of distance information are associated with the pixels stored in the image memory 13 and additionally stored. With this, the distance to the object to be imaged can be calculated. From the above processing results, it is possible to obtain three-dimensional position information regarding the entire object to be imaged.

According to the three-dimensional image generation apparatus according to the above-described embodiments, the information of the outside world is imaged while moving the image pickup apparatus 1 including an area image sensor at a predetermined interval to give parallax.
Then, the image at the initial position P _{0 and} the image at the final position P _M are associated with each other, the distance to the object is obtained, and the three-dimensional position information can be obtained. Therefore, if such a position is applied to a visual system such as an autonomous vehicle or a robot, it is possible to judge whether or not the vehicle can pass even if there is an unspecified obstacle ahead. Becomes

In the present embodiment, the image pickup unit 1 is composed of an area type image sensor or the like, but the present invention is not limited to this, and the same effect can be obtained even if it is composed of a linear type image sensor. Further, instead of the moving mechanism of the image pickup unit, the image pickup unit and the image memory may be paired and a large number of them may be arranged so that a plurality of images can be input at a time. It may be configured such that the different portion is placed on the turntable.

[Effects of the Invention] As described above, in the present invention, the component surface of the object to be imaged captured by the imaging unit at the initial position is detected for each same brightness, and the component surface is traced as the imaging unit moves. Then
The component surface grasped at the initial position and the component surface grasped at the final position are associated with each other, and based on this, the distance to the object to be imaged is calculated to obtain the three-dimensional position information. This makes it possible to obtain three-dimensional position information of a large number of unspecified objects with high accuracy, and when this is applied to an automated vehicle or robot, for example, there are unspecified obstacles. It is possible to determine whether or not the vehicle can be driven even when the vehicle is running, and it is possible to recognize the object to be imaged by storing information such as the shape and size of the object to be imaged.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram showing an embodiment of a three-dimensional image generation apparatus to which a distance measuring method for a three-dimensional object of the present invention is applied, and FIG. 2 is a diagram for explaining movement of an image pickup unit of the present invention. FIGS. 3 and 4 are explanatory diagrams for explaining the region-divided image, FIGS. 5 and 6 are bit pattern diagrams of the region-divided image, and FIGS. 7, 8 and 9 are Cross-sectional image view when the imaging unit is sequentially moved, FIG. 10 is a region-divided diagram of the region-divided image, FIGS. 11 and 12 are explanatory diagrams for explaining the arithmetic operation, and FIG. 13 is a three-dimensional position. It is an explanatory view for calculating information. 1 ... Imaging unit, 2 ... Moving mechanism 3 ... Pulse motor, 4 ... Shaft 5 ... Control circuit, 6 ... Conversion memory circuit 7 ... A / D conversion circuit, 8 ... Driver 9 ... Address Counter, 10 ... Image memory 11 ... Image processing circuit, 12 ... Storage circuit 13,14,15 ... Image memory, 16 ... Arithmetic circuit

Claims (1)

[Claims] 1. A component of an object to be imaged is moved in parallel from an initial position to a final position with respect to the object to be imaged, and the imaging unit moves the image pickup object at each predetermined moving position from the initial position to the final position. Grasp a surface as an image of uniform brightness within the surface, divide the grasped component surface of the object to be imaged into areas of the same brightness, and store all surfaces that have been divided into areas at the initial and final positions. Then, cross-sectional data of a predetermined width parallel to the moving direction of the image pickup unit is extracted from the captured image at each predetermined moving position from the initial position to the final position through a plurality of surfaces obtained by dividing the area at the initial position. Then, it is created and stored as a cross-sectional image, this cross-sectional image is divided into areas for each same brightness, and the surface stored at the initial position and the final position that includes the area-divided area in common from this area-divided cross-sectional image is displayed. same Corresponding each component surface as a surface, in each component surface of the expansion and contraction ratio in the moving direction of the imaging unit of the image obtained at the initial position and the final position of each component surface of the corresponding object The average value is obtained as the expansion / contraction coefficient of each component surface, and by the expansion / contraction coefficient of each component surface, the pixels on all the imaging screens that make up each component surface in the image at the initial position and the image at the final position The coordinates are associated with each other, and from the associated coordinates of all the images and the initial position and the final position of the imaging unit, the initial position and the final position of the point on the target object corresponding to all the pixels of each component surface. It is characterized in that a visual angle at a position is obtained, and the distances of points on the object to be imaged corresponding to all pixels of each surface, which is a constituent element of the plurality of objects, are calculated from the obtained visual angle. Three-dimensional Distance measuring method in elephant products.