JP4288028B2 - Planar area specifying method and image measuring apparatus - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image measurement method for measuring the position and orientation of a plane from an image, and specifying an area occupied by the measured plane in the image, and an image measurement apparatus that implements the image measurement method.
[0002]
[Prior art]
Conventionally, a method for measuring three-dimensional geometric information of a plane (planar spatial parameters such as plane orientation and distance) from an optical flow or a parallax vector of an image has been provided (JP 2000-357235 A). .
[0003]
According to the method described in the publication, motion information obtained from two images obtained by photographing the subject at different times t0 and t1 depending on the camera moving toward the subject, or the left eye From parallax information obtained from two images obtained by photographing the subject at the same time t by the camera and the right eye camera, the spatial parameters of the plane in the image (plane orientation, plane arrival time, plane up to plane Can be measured.
[0004]
[Problems to be solved by the invention]
However, in the method for measuring the spatial parameter of the plane, the plane (spatial parameter) can be measured, but it cannot be specified in which region in the image the measured plane exists.
[0005]
An object of the present invention is to make it possible to specify an area occupied by a plane (a plane specified by a spatial parameter) measured by a method for measuring a plane spatial parameter in an image.
[0006]
[Means for Solving the Problems]
In order to solve the above-mentioned problem, the present invention is a plane area specifying method, which detects local motion in an image, votes the motion information in a voting space constituted by plane space parameters, By obtaining a spatial parameter and searching for a local region based on the data of the voting result, the region occupied by the plane specified by the spatial parameter in the image is specified.
[0007]
According to the present invention, it is possible to specify an area occupied by a plane (a plane specified by a spatial parameter (such as plane orientation n and normalized time tc)) measured by a plane spatial parameter measurement method. become.
[0008]
When the above planar area specifying method is implemented by a computer system, the image measuring apparatus can be specified as follows.
[0009]
A plane parameter voting for voting a local motion detection unit that detects local motion at an arbitrary position in an image and a voting space composed of the spatial parameters of the plane based on the voting table. A voting data storage unit for extracting voting data information, a voting data storage unit for storing voting data information, and the voting data storage unit based on the peak data obtained from the peak extracting unit. A plane area extraction unit that searches and identifies a local area included in the plane area.
[0010]
According to the present invention, the area occupied by the plane (peak data, plane specified by the plane parameters (plane orientation n and normalization time tc, etc.)) measured by the plane spatial parameter measurement method is specified. It becomes possible. That is, it is possible to specify a local area (local area constituting the plane area) included in the plane area specified by the peak data and the plane space parameters (plane orientation n and normalization time tc).
[0011]
In order to solve the above-mentioned problem, the present invention is a plane area specifying method that detects local parallax in an image and votes the parallax information on a voting space constituted by plane space parameters. The plane space parameter is obtained, and the local area is searched based on the voting result data, whereby the area occupied by the plane specified by the space parameter in the image is specified.
[0012]
According to the present invention, it is possible to specify a region occupied by a plane (a plane specified by a spatial parameter (such as plane direction n and shortest distance dc)) measured by the plane parameter measurement method. Become.
[0013]
When the above planar area specifying method is implemented by a computer system, the image measuring apparatus can be specified as follows.
[0014]
Planar parameter voting for voting a local parallax detection unit that detects local parallax at an arbitrary position in an image and a voting space composed of the spatial parameters of the plane based on a voting table. A voting data storage unit for extracting voting data information, a voting data storage unit for storing voting data information, and the voting data storage unit based on the peak data obtained from the peak extracting unit. A plane area extraction unit that searches and identifies a local area included in the plane area.
[0015]
According to the present invention, the area occupied by a plane (peak data, a plane specified by a spatial parameter (such as plane orientation n and shortest distance dc)) measured by the plane parameter measurement method is specified in the image. Is possible. That is, it is possible to specify a local area (local area constituting the plane area) included in the plane area specified by the peak data and the plane space parameters (plane orientation n and normalization time tc).
[0016]
The present invention can also be specified as follows. A planar area specifying method, in which a plurality of local area images are cut out from captured images obtained by shooting a subject at different times by a camera moving toward the subject, and a local area image is extracted for each of the cut out local area images. A local motion at an arbitrary position in the region image is detected and obtained based on the detected motion information for each of a plurality of elements constituting a voting space (for example, the planar parameter voting space 101 in the embodiment). Vote the voting value, and associate the voting value with the specified information of the voting element of the voting value and the specified information of the local region image that is the basis of the voting value (based on the voting value or local motion detection) (For example, the voting strength table 106 in the embodiment), the peak data is extracted from the voting result, and the case is extracted based on the extracted peak data. From the obtained information, search for designation information of the local area image contributing to the appearance of the extracted peak data among the plurality of local area images, and take a local area image designated by the searched designation information Identifies as the middle planar area.
[0017]
According to the present invention, the area occupied by the plane (peak data, plane specified by the plane parameters (plane orientation n and normalization time tc, etc.)) measured by the plane spatial parameter measurement method is specified. It becomes possible.
[0018]
The present invention can also be specified as follows. A method for specifying a planar area, in which a plurality of local area images are obtained from captured images obtained by photographing a subject at the same time by a plurality of cameras (for example, two cameras) that move toward the subject or are stationary. A plurality of array elements that form a voting space (for example, the planar parameter voting space 101 in the embodiment) by detecting a local parallax at an arbitrary position in the local region image for each cut out local region image Voting values obtained based on the detected parallax information are voted for each, and the voting value and the designated information of the voting target element of the voting value and the basis of the voting value (the basis for detecting the voting value or local motion) The local area image designation information is stored in association with each other (for example, the voting strength table 106 in the embodiment), and peak data is extracted from the voting result, Based on the extracted peak data, the local information that has contributed to the appearance of the extracted peak data among the plurality of local area images is searched from the stored information, and the search is performed. The local area image specified by the specification information is specified as a planar area in the captured image.
[0019]
According to the present invention, the area occupied by the plane (the peak data, the plane specified by the plane spatial parameters (such as the plane orientation n and the shortest distance dc)) measured by the plane spatial parameter measurement method is specified. It becomes possible to do.
[0020]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, a system configuration for carrying out the planar area specifying method of the present invention will be described with reference to the drawings. FIG. 1 is an external view of a computer system for carrying out a planar area specifying method (including a planar spatial parameter measuring method) according to the present invention. FIG. 2 is a hardware configuration diagram of the computer system shown in FIG.
[0021]
The computer system 300 includes a central processing unit (CPU) 311 and a RAM 312, a magnetic disk controller 313, a floppy disk driver 314, an MO driver 315, a mouse controller 316, and a keyboard controller 317 that are connected to the CPU 311 via a bus or the like. , A display controller 318, a communication board 319, an image input board 320, and the like.
[0022]
The magnetic disk controller 313 is for accessing a magnetic disk 321 built in the main body 301 (see FIG. 1). The floppy disk driver 314 and the MO driver 315 are for loading the floppy disk 322 and the MO 323, respectively, and for accessing the loaded floppy disk 322 and the MO 323.
[0023]
A mouse controller 316 and a keyboard controller 317 transmit the operations of the mouse 304 and the keyboard 303 to the inside of the computer, respectively.
[0024]
The display controller 318 is a controller that displays an image on the CRT display 302 in accordance with a program executed by the CPU 311.
[0025]
Further, the communication board 319 is connected to a communication network such as a LAN or the Internet via the communication line 400. The communication board 319 serves to receive image data via the communication network, for example. I'm in charge.
[0026]
Further, the image input board is connected to an external camera CA (electronic still camera, video camera, or the like), and plays a role of capturing image data obtained by photographing with the camera CA into the computer. . Although only one camera is shown in FIG. 2, two cameras are connected to the image input board 320, and the same camera is used from different directions corresponding to the parallax of human eyes, for example. It is also possible to input two images obtained by photographing the subject at the same time.
[0027]
In addition, a camera CA and a removable recording medium such as a flash memory or an IC card are mounted on a moving body such as an aircraft, a bipedal walking robot, or a vehicle, and an image captured by the camera CA is recorded on the recording medium and stored in a computer. An image may be read from the recording medium by a connected reading device or the like and input into the computer. Further, an image may be read from such a recording medium and input into the computer via a wireless or wired communication line.
[0028]
In the computer system 300, a program stored in the floppy disk 322 or the MO 323 is installed, or a program transmitted via the communication line 400 is installed, and the planar area specifying method (planar space parameter of the present invention described later) is installed. (Including the measurement method). The function for implementing the planar area specifying method is realized by executing a predetermined program in the computer system 300. These functions are shown in various functional block diagrams and flowcharts to be described later.
[0029]
Next, an outline of the first embodiment of the plane area specifying method (including the plane space parameter measuring method) will be described.
[0030]
In the plane area specifying method of the present embodiment, a plane (plane specified by plane orientation n and standardized time tc) measured by a plane spatial parameter measurement method (JP 2000-357235 A) is an image. It is intended to specify the area occupied in the middle (in the captured image).
[0031]
As shown in FIG. 3, the plane space parameter is measured by a camera CA (aircraft in this embodiment) moving toward a certain subject (in this embodiment, a runway is illustrated as a subject). For two images G0 and G1 (one image G0 is shown in FIG. 8) obtained by photographing the subject at different times t0 and t1. By performing the processing shown in S100 to S112 in FIG. 12, the plane spatial parameters (plane orientation n and normalized time tc, etc.) in the image are to be measured.
[0032]
In order to facilitate understanding of the planar area identification method (including the measurement method of the spatial parameter of the plane), first, it is used in the planar area identification method. _ij The τ table and the plane parameter voting space will be described.
[0033]
_ij The τ table 100 is for deciding to vote the calculation strength of the local velocity τ in the local region number i to the element number j of the spatial parameter (the element number j is a total vote consisting of the azimuth and arrival time parameters). Serial number in space). In this embodiment, as shown in FIG. 4 and FIG. _ij The table that associates τ with _ij It is called the τ table. _ij The method for creating the τ table is described in detail in FIG. 111 of JP-A-2000-357235 and the description thereof. _ij Only functional blocks used to create the τ table are shown (see FIGS. 6 and 7).
[0034]
As shown in FIGS. 9 and 10, the planar parameter voting space 101 is a cylindrical voting space in which a plurality of circles 101b including a plurality of array elements 101a overlap each other (see FIG. 8). It is formed by being executed. A serial number j (1, 2,..., Last) is assigned to each element 101a of the array. Each element 101a of the array corresponds to a vote value storage area (area secured in the RAM) to which the serial number j is assigned (see FIG. 11).
[0035]
As shown in FIG. 8, the planar region specifying method according to the present embodiment sequentially extracts local regions from an input image and performs local motion detection for each of them in the first stage. Based on the detection result, voting is performed on the voting space constituted by the plane space parameters, and the voting values are accumulated. Here, a three-dimensional array of plane orientation (two-dimensional) and normalized arrival time up to the plane was taken as a spatial parameter representing the plane. By detecting the peak from the voting space, the plane direction n and the standardized arrival time tc in the image are obtained. The plane detection method so far has been reported so far (Japanese Patent Laid-Open No. 2000-357235, J83-D-II, No.12, pp.2786-2797, etc.).
[0036]
In the second stage, on the basis of the azimuth obtained by peak detection and the standardized arrival time, voting data recorded in advance is searched, and a local motion detection region contributing to voting for the peak is obtained. A portion including these local regions is a region where the plane exists in the image. In this way, it is possible to identify and extract a planar area (area occupied by a plane in the image) in the image. As described above, according to the present embodiment, it is possible to determine in which region in the image the plane in which the direction / distance is detected exists in the image, in units of local regions that perform speed detection / parallax detection. It can be easily obtained.
[0037]
Next, an example in which the computer system 300 implements the planar region specifying method (including the planar spatial parameter measuring method) according to the present embodiment will be described in detail with reference to FIGS. In the following description, it is assumed that the moving direction v of the camera CA is known, and the plane azimuth n and the arrival time tc to the plane are considered as the plane spatial parameters. Also, _ij It is assumed that the τ table is created in advance before the processing of S100 and subsequent steps is started.
[0038]
The following processing is realized by executing a predetermined program. When the predetermined program is executed, first, i = 0 is set as an initial value to a local area number i that designates each local area (or local area image) (S100). The local area scanning unit 102 scans the inside of the image, and from each of the two images at time t0 and time t1, an image portion (hereinafter referred to as a local area) corresponding to a preset local area (i-th local area). Each image) is cut out (S101). FIG. 13 shows an example of a cut-out local area image. The local motion detection unit 103 detects local motion at an arbitrary position in the image (S102). Specifically, the local motion detection unit 103 calculates a motion vector (local velocity) from these two extracted local region images (i-th local region image).
[0039]
As a local motion detection method, (a) a pole conversion + 1-dimensional correlation + inverse pole conversion method (see JP-A-6-44364) or (b) block matching used in MPEG motion estimation. Can be used. By this local motion detection process, a local motion detection plane (local velocity map) corresponding to the input image is obtained. FIG. 14 shows an example of a local motion detection plane (local velocity map).
[0040]
The local motion detection plane is composed of coordinate values indicating velocity and intensity information thereof. The intensity information is the reaction intensity in (a), the correlation value in the case of using correlation calculation in (b), and the sum of absolute differences in (b). When the distance calculation such as is used, the distance value is meant. The coordinates at which the reaction intensity or the correlation value is maximum (the minimum coordinate in the case of distance) indicate the speed of the local region. Such intensity information (reaction intensity, correlation value, distance value) can be used directly / indirectly as a vote value for the plane parameter voting space.
[0041]
Next, j = 0 is set as an initial value to an element number j that designates an element (and a vote value storage area) of each array in the plane parameter voting space (S103).
[0042]
The plane parameter scanning unit 104 _ij From the τ table, the local area number i and the element number j at that time are associated with each other. _ij τ is read (output) (S104).
[0043]
The plane parameter voting unit 105 votes the motion information detected by the local motion detection unit 103 to a voting space constituted by plane space parameters based on the voting table. That is, the plane parameter voting unit 105 uses the calculated value (also referred to as intensity, hereinafter referred to as voting value) based on the exercise information calculated in S102. _ij Voting is performed on the element number j of the plane parameter voting space 101 according to the τ table (S106). Specifically, it was read in S104 _ij The vote value for τ (calculated in S102) is added to the vote value storage area (area indicated by j) secured in the RAM.
[0044]
The vote value may be voted as it is (without performing any operation on the vote value), or the vote value is voted 10 times or 1/10 times (intensity calculation) as necessary. (S105). Further, the vote value may be binarized and voted, or 0 may be voted when the vote value is smaller than a certain value.
[0045]
At the time of voting, the calculated value (strength) as the voting value is stored in the voting strength table 106 (for example, the voting strength table 106 on the left side of FIG. 15) in association with the local region number i and the element number j at the time of voting. (S107). The voting strength table 106 corresponds to a voting data storage unit that stores voting data information. The strength table 106 associates the voting value (voting strength) with the voting value designating element j of the voting value and the local region image designating information i based on the voting value (that is, i at the time of voting). It is a table.
[0046]
The processing from S104 to S107 is performed on the element numbers of all the arrays constituting the planar parameter voting space 101 (S108, S109). Further, the above-described processing from S101 to S109 is performed on all local area images in the image (S110, S111). In the present embodiment, there are a total of 11 local area images from i = 0th to i = 10th indicated by circular areas in the image (see FIG. 8). Processing is performed on these 11 local region images.
[0047]
After scanning all the images, the peak extraction unit 107 performs peak extraction in the planar parameter voting space 101 (S112). That is, the peak extraction unit 107 extracts peak data from the voting result. For example, the peak appears in a cross section of the planar parameter space 101 as shown in FIG.
The coordinate position (n, tc) of the extracted peak in the plane parameter voting space 101 indicates the azimuth n and arrival time tc of the plane (the ground portion below the horizontal line in the image shown in FIG. 8) in the image. ing. As a result, a plane is detected. That is, the plane direction n and the arrival time tc are measured as plane space parameters. The measured coordinate position (n, tc) corresponds to the element number jp and can also be represented by the element number jp (S113).
[0048]
In the next stage, the plane area detection unit 108 searches the voting data storage unit based on the peak data obtained from the peak extraction unit 107 and identifies a local area included in the plane area. Hereinafter, the process of the plane area | region detection part 108 is demonstrated.
[0049]
First, the voting strength table 106 is searched based on the detected element number jp of the plane space parameter (S115), and the local area number i voted for the element jp and the calculated value (intensity) as the voting value are obtained. Check (S116). That is, the local area number i associated with the element number jp and its voting value are read from the voting strength table 106.
[0050]
Then, using the read voting value, a determination process as a planar area (for example, determination whether or not the read voting value is equal to or greater than a certain threshold value) is executed (S116), and the i-th local area image is planar It is determined whether or not the region is included in (the plane specified by the azimuth n and the arrival time tc). For example, the symbol “1” or the like indicating the same group is stored in the table in association with the i number whose voting value is equal to or greater than the threshold (see FIG. 16).
[0051]
By repeating this region determination (S118, S119), the planar region detecting unit 108 determines the local region number i finally obtained (for example, the local number in which the symbol “1” indicating the same group is associated with the i number). A portion covered by the region i) is identified as a region of a plane in the image (a plane identified by the azimuth n and the arrival time tc) and extracted (S120).
That is, the local area specified by the finally obtained local area number i is specified as the area occupied by the plane specified by the plane space parameters (direction n and arrival time tc) in the captured image. Further, it is assumed that the local region specified by the finally obtained local region number i is specified as the local region included in the plane region specified by the plane spatial parameters (direction n and arrival time tc). You can also.
Further, based on jp as the extracted peak data, search the voting intensity table 106 for the designation information i of the local area image that contributed to the appearance of the extracted peak data among the plurality of local area images, It can also be understood that the local area image designated by the searched designation information i is specified as a planar area in the captured image.
[0052]
The voting strength table 106 is as shown on the right side of FIG. _ij It may be shared with the τ table. In this way, a memory for storing the table can be saved. It is also effective to record not only the voting intensity but also an attribute such as whether or not the detection speed is a peak on the local motion detection plane and use the information for region determination. In this way, it is possible to specify the region where the plane exists more accurately.
[0053]
In the above embodiment, it was created in advance. _ij Although an example in which voting is performed using the τ table has been described, the present invention is not limited to this, and a voting strength table is created simultaneously with voting while sequentially calculating voting element numbers. _ij Voting may be performed using the τ table.
[0054]
In the above embodiment, an example in which a total of 11 local area images corresponding to a circular area are used as local area images has been described (see FIG. 8). The present invention is not limited to this. For example, as shown in FIG. 19, as a local region image, the entire image is divided into a grid shape and the divided rectangular regions (in addition, triangles, pentagons, ellipses). A local region image corresponding to any graphic region may be used. The number of local area images can be set to an appropriate number as necessary.
[0055]
Further, in the above-described embodiment, the example in which the area of one plane included in the image is specified has been described. The present invention is not limited to this, and the areas of each of the plurality of planes can be specified as follows.
[0056]
For example, as shown in the lower part of FIG. 19, consider an image including three planes obtained by photographing a cube. In this case, three peaks are extracted in S112. The coordinate positions (n1, tc1), (n2, tc2), (n3, tc3) of the three extracted peaks in the plane parameter voting space 101 are the azimuth n and arrival time tc of each of the three planes present in the image. Show. These measured coordinate positions (n1, tc1), (n2, tc2), (n3, tc3) correspond to the element numbers jp1, jp2, and jp3, and can also be represented by the element numbers jp1, jp2, and jp3. (S113).
[0057]
Next, the processing from S114 to S119 is performed for each jp (jp1, jp2, jp3). At this time, a different threshold is used for each jp in the determination process of S116. Then, the symbols “1”, “2”, “3” and the like indicating the same group are stored in the table in association with the i-number whose voting value is equal to or greater than the threshold (see FIG. 16).
[0058]
In this way, the first plane in the image covers the portion covered by the finally obtained local region number i (for example, the local region i in which the symbol “1” indicating the same group is associated with the i number). It can be specified and extracted as a region of (a plane specified by the direction n1 and the arrival time tc1). The region of the second plane (the plane specified by the azimuth n2 and the arrival time tc2) and the region of the third plane (the plane specified by the azimuth n3 and the arrival time tc3) can be similarly specified and extracted.
[0059]
Next, an outline of the second embodiment of the plane area specifying method (including the plane space parameter measuring method) will be described.
[0060]
In the plane area specifying method of the present embodiment, a plane (a plane specified by the plane orientation n and the shortest distance dc) measured by a plane space parameter measurement method (Japanese Patent Laid-Open No. 2000-357235, etc.) is included in the image. This is intended to specify the area occupied by (in the captured image).
[0061]
As shown in FIG. 3, the plane space parameter is measured by photographing the subject at the same time using a left-eye camera CA and a right-eye camera CA (in this embodiment, they are mounted on an aircraft). The two images G0 and G1 obtained in this way (one image G0 is shown in FIG. 8) are subjected to the processing shown in S200 to S212 in FIG. A parameter (such as the plane direction n and the shortest distance dc to the plane) is to be measured.
[0062]
Used in planar area identification method _ij The σ table has been described in the first embodiment. _ij corresponds to the τ table (ie shown in FIG. _ij τ table _ij the value of τ _ij The one replaced by σ _ij As the plane parameter voting space, the same one as described in the first embodiment is used.
[0063]
Next, an example in which the plane area specifying method (including the plane space parameter measurement method) according to the present embodiment is implemented in the computer system 300 will be described in detail with reference to FIGS. 17 and 18. In the following description, it is assumed that the moving direction v from the left eye camera CA to the right eye camera CA is known, and the plane space parameter is the plane direction n and the shortest distance dc to the plane. In addition, for deciding to vote the calculated intensity of the local parallax σ in the local region i for the element number j of the spatial parameter _ij It is assumed that the σ table is created in advance before the processing of S200 and subsequent steps is started.
[0064]
The following processing is realized by executing a predetermined program. When the predetermined program is executed, first, i = 0 is set as an initial value to a local area number i that designates each local area (or local area image) (S200). The local area scanning unit 202 corresponds to a preset local area (i-th local area) from two images obtained at the same time from the left eye camera CA and the right eye camera while scanning the image. Each image portion to be performed (hereinafter referred to as a local region image) is cut out (S201). FIG. 13 shows an example of a cut-out local area image. The local parallax detection unit 203 detects local parallax at an arbitrary position in the image (S202). Specifically, the local parallax detection unit 203 calculates a displacement vector (local parallax) from these two cut out local area images (i-th local area image).
[0065]
Next, j = 0 is set as an initial value to an element number j that designates an element (and vote value storage area) of each array in the plane parameter voting space (S203).
[0066]
The plane parameter scanning unit 204 includes: _ij From the σ table, the local area number i and the element number j at that time are associated with each other. _ij σ is read (output) (S204).
[0067]
The plane parameter voting unit 205 votes the parallax information detected by the local parallax detection unit 203 to a voting space composed of plane space parameters based on the voting table. That is, the plane parameter voting unit 205 uses the calculated value (also referred to as intensity, hereinafter referred to as voting value) based on the parallax information calculated in S202. _ij Voting is performed on the element number j of the plane parameter voting space according to the σ table (S206). Specifically, it was read in S204. _ij The vote value for σ (calculated in S202) is added to the vote value storage area (area indicated by j) secured in the RAM.
[0068]
The vote value may be voted as it is (without performing any operation on the vote value), or the vote value is voted 10 times or 1/10 times (intensity calculation) as necessary. (S205). Further, the vote value may be binarized and voted, or 0 may be voted when the vote value is smaller than a certain value.
[0069]
At the time of voting, the calculated value (intensity) as the voting value is stored in the voting intensity table 206 (for example, the voting intensity table 106 on the left side of FIG. 15) in association with the local region number i and element number j at the time of voting. (S207). The voting strength table 206 corresponds to a voting data storage unit that stores voting data information. The strength table 206 associates the voting value (voting strength) with the voting value voting destination element designation information j and the local region image designation information i based on the voting value (that is, i at the time of voting). It is a table.
[0070]
The above processing from S204 to S207 is performed for the element numbers of all the arrays constituting the planar parameter voting space 101 (S208, S209). Further, the processing from S201 to S209 is performed on all local region images in the image (S210, S211). In the present embodiment, since there are a total of 11 local region images from i = 0th to i = 10th indicated by a circular region in the image (see FIG. 8), S201 to S209 Processing is performed on these 11 local region images.
[0071]
After scanning all images, the peak extraction unit 207 performs peak extraction in the planar parameter voting space 101 (S212). That is, the peak extraction unit 207 extracts peak data from the vote result. For example, the peak appears in a cross section of the planar parameter space 101 as shown in FIG.
The coordinate position (n, dc) of the extracted peak in the plane parameter voting space 101 indicates the azimuth n and the shortest distance dc of the plane (the ground portion below the horizontal line in the image shown in FIG. 8) in the image. ing. As a result, a plane is detected. That is, the plane direction n and the shortest distance dc are measured as plane space parameters. The measured coordinate position (n, dc) corresponds to the element number jp and can also be represented by the element number jp (S213).
[0072]
In the next stage, the plane area detection unit 208 searches the voting data storage unit based on the peak data obtained from the peak extraction unit 207 and identifies a local area included in the plane area. Hereinafter, the process of the plane area detection unit 208 will be described.
[0073]
First, the voting strength table 206 is searched based on the detected element number jp of the planar space parameter (S215), and the local area number i voted for the element jp and the calculated value (intensity) as the voting value are obtained. Check (S216). That is, the local area number i associated with the element number jp and its voting value are read from the voting strength table 206.
[0074]
Then, a determination process as a planar area (for example, determination of whether or not the read voting value is equal to or greater than a certain threshold value) is executed using the read vote value (S216), and the i-th local area image is planar. It is determined whether the region is included in (the plane specified by the azimuth n and the shortest distance dc). For example, the symbol “1” or the like indicating the same group is stored in the table in association with the i number whose voting value is equal to or greater than the threshold (see FIG. 16).
[0075]
By repeating this region determination (S218, S219), the planar region detection unit 208 determines the local region number i finally obtained (for example, the local number in which the symbol “1” indicating the same group is associated with the i number). The part covered by the area i) is specified as an area of a plane in the image (a plane specified by the azimuth n and the shortest distance dc) and extracted (S220).
In other words, the local area specified by the finally obtained local area number i is specified as the area occupied by the plane specified by the plane space parameters (direction n and shortest distance dc) in the captured image. Further, it is assumed that the local area specified by the finally obtained local area number i is specified as the local area included in the plane area specified by the plane spatial parameters (direction n and shortest distance dc). You can also.
Further, based on jp as the extracted peak data, search the voting intensity table 106 for the designation information i of the local area image that contributed to the appearance of the extracted peak data among the plurality of local area images, It can also be understood that the local area image designated by the searched designation information i is specified as a planar area in the captured image.
[0076]
In the above embodiment, an example in which a total of 11 local area images corresponding to a circular area are used as local area images has been described (see FIG. 8). The present invention is not limited to this. For example, as shown in FIG. 19, as a local region image, the entire image is divided into a grid shape and the divided rectangular regions (in addition, triangles, pentagons, ellipses). A local region image corresponding to any graphic region may be used. The number of local area images can be set to an appropriate number as necessary.
[0077]
Further, in the above-described embodiment, the example in which the area of one plane included in the image is specified has been described. The present invention is not limited to this, and the areas of each of the plurality of planes can be specified as follows.
[0078]
For example, as shown in the lower part of FIG. 19, consider an image including three planes obtained by photographing a cube. In this case, three peaks are extracted in S212. The coordinate positions (n1, tc1), (n2, tc2), (n3, tc3) of the three extracted peaks in the plane parameter voting space 101 are the azimuth n and arrival time tc of each of the three planes present in the image. Show. These measured coordinate positions (n1, tc1), (n2, tc2), (n3, tc3) correspond to the element numbers jp1, jp2, and jp3, and can also be represented by the element numbers jp1, jp2, and jp3. (S213).
[0079]
Next, the processing from S214 to S219 is performed for each jp (jp1, jp2, jp3). At this time, a different threshold is used for each jp in the determination process of S216. Then, the symbols “1”, “2”, “3” and the like indicating the same group are stored in the table in association with the i-number whose voting value is equal to or greater than the threshold (see FIG. 16).
[0080]
In this way, the first plane in the image covers the portion covered by the finally obtained local region number i (for example, the local region i in which the symbol “1” indicating the same group is associated with the i number). It can be specified and extracted as a region of (a plane specified by the direction n1 and the arrival time tc1). The region of the second plane (the plane specified by the azimuth n2 and the arrival time tc2) and the region of the third plane (the plane specified by the azimuth n3 and the arrival time tc3) can be similarly specified and extracted.
[0081]
[Others] The present invention can be specified as follows.
(Supplementary note 1) Local motion is detected in an image, the motion information is voted on a voting space composed of plane spatial parameters to obtain a plane spatial parameter, and a local region is searched based on the data of the voting result. A plane area specifying method for specifying an area occupied in the image by the plane specified by the spatial parameter.
(Supplementary Note 2) A local motion detection unit that detects local motion at an arbitrary position in an image, a plane parameter voting unit that votes the motion information based on a voting table into a voting space composed of plane spatial parameters, Based on the peak data obtained from the peak extraction unit that extracts peak data from the voting result, the voting data storage unit that stores the voted data information, and the peak extraction unit, the voting data storage unit is searched. An image measurement apparatus comprising: a planar area extraction unit that identifies a local area included in a planar area.
(Appendix 3) Detecting local parallax in an image, voting the parallax information to a voting space composed of plane spatial parameters to obtain a plane spatial parameter, and searching for a local region based on voting result data A plane area specifying method for specifying an area occupied in the image by the plane specified by the spatial parameter.
(Additional remark 4) The local parallax detection part which detects the local parallax of the arbitrary positions in an image, The plane parameter voting part which votes the parallax information to the voting space comprised by a plane spatial parameter based on a voting table, Based on the peak data obtained from the peak extraction unit that extracts peak data from the voting result, the voting data storage unit that stores the voted data information, and the peak extraction unit, the voting data storage unit is searched. An image measurement apparatus comprising: a planar area extraction unit that identifies a local area included in a planar area.
(Supplementary Note 5) A plurality of local region images are cut out from captured images obtained by shooting the subject at different times depending on the camera moving toward the subject, and each local region image thus cut out is included in the local region image. A local motion at an arbitrary position is detected, a vote value obtained on the basis of the detected motion information is voted for each of a plurality of array elements constituting the voting space, and the vote value is assigned to a vote destination element of the vote value. Corresponding and storing the designation information and the designation information of the local region image that is the basis of the voting value, extracting peak data from the voting result, and from the stored information based on the extracted peak data The local area image that has contributed to the appearance of the extracted peak is searched for in the plurality of local area images, and the local area image specified by the searched specification information is captured. A plane area specifying method for specifying as a plane area in a shadow image.
(Supplementary Note 6) A plurality of local region images are cut out from captured images obtained by shooting the subject at the same time by a plurality of cameras moving toward or stationary at the same time, and for each of the cut out local region images, A local parallax at an arbitrary position in the local area image is detected, and a vote value obtained based on the detected parallax information is voted for each of a plurality of array elements constituting the voting space. The designation information of the voting destination element and the designation information of the local area image based on the voting value are stored in association with each other, the peak data is extracted from the voting result, and the storage is performed based on the extracted peak data. The specified information of the local area image that contributed to the appearance of the extracted peak data is searched from the plurality of local area images, and specified by the searched specification information. A plane area specifying method for specifying a local area image to be performed as a plane area in a captured image.
[0082]
The present invention can be implemented in various other forms without departing from the spirit or main features thereof. For this reason, said embodiment is only an illustration in all the points, and is not interpreted limitedly.
[0083]
【The invention's effect】
As described above, according to the present invention, it is possible to specify the area occupied by the plane (plane specified by the spatial parameter) measured by the plane spatial parameter measurement method. In addition, since it is possible to specify in which region in the image the plane exists, it is possible to cut out the plane from the image. As a result, the avoidance operation can be appropriately performed by estimating the size of the moving obstacle. A plane detected by being divided into a plurality of planes by overlapping or the like can also be extracted. It is also possible to reconstruct the three-dimensional spatial structure from the image by extracting a plurality of planar regions.
[Brief description of the drawings]
FIG. 1 is an external view of a computer system for carrying out first and second embodiments of the present invention.
2 is a hardware configuration diagram of the computer system shown in FIG. 1. FIG.
FIG. 3 is an example of a subject photographed in the first and second embodiments of the present invention.
FIG. 4 is a functional block diagram for implementing the first embodiment of the present invention in a computer system.
FIG. 5 is used in the first embodiment of the present invention. _ij It is a figure for demonstrating the example of (tau) table.
FIG. 6 is used in the first embodiment of the present invention. _ij It is a functional block diagram used for preparation of a τ table.
FIG. 7 is used in the first embodiment of the present invention. _ij It is a functional block diagram used for preparation of a τ table.
FIG. 8 is a diagram for mainly explaining a planar parameter voting space and a local region image used in the first and second embodiments of the present invention.
FIG. 9 is a diagram for explaining the structure (cross-sectional structure) of a planar parameter voting space used in the first and second embodiments of the present invention.
FIG. 10 is a diagram for explaining peaks appearing in a planar parameter voting space used in the first and second embodiments of the present invention.
FIG. 11 is a diagram for explaining a relationship between a voting value storage area secured in a RAM used in the first and second embodiments of the present invention and voting.
12 is a flowchart for explaining an operation when the first embodiment of the present invention is implemented by the functional blocks of FIG. 4;
FIG. 13 is an example of a local region image cut out in the first and second embodiments of the present invention.
FIG. 14 is an example of a local motion detection plane (local velocity map) obtained in the first embodiment of the present invention.
FIG. 15 is an example of a voting strength table used in the first and second embodiments of the present invention.
FIG. 16 is an example of a table for grouping local area images used in the first and second embodiments of the present invention.
FIG. 17 is a functional block diagram for implementing the second embodiment of the present invention in a computer system.
FIG. 18 is a flowchart for explaining an operation when the second embodiment of the present invention is implemented by the functional blocks of FIG. 17;
FIG. 19 is a diagram for explaining a modification of the present invention.
[Explanation of symbols]
100 _ij τ table
101 planar parameter voting space
101a Array elements
102 Local area scanning unit
103 Local motion detector
104 Planar parameter scanning unit
105 Planar parameter voting section
106 Voting strength table
107 Peak extractor
108 Planar area detector
202 Local area scanning unit
203 Local parallax detection unit
204 Planar parameter scanning unit
205 Planar parameter voting section
206 Voting Strength Table
207 Peak extractor
208 Planar area detector
300 computer system
301 body
311 CPU
312 RAM
313 Magnetic disk controller
314 floppy disk driver
315 MO driver
316 Mouse controller
317 Keyboard controller
318 Display controller
319 Communication board
320 Image input board
321 magnetic disk
322 floppy disk
323 MO
400 communication line
CA camera
n Planar orientation
tc Normalization time
dc shortest distance
G0, G1 images

Claims (5)

By detecting local motion in the image, voting the motion information to a voting space composed of spatial parameters of the plane, obtaining spatial parameters of multiple planes, and searching for local regions based on the data of the voting results A plane area specifying method for specifying an area occupied by a plurality of planes specified by the spatial parameter in an image. A local motion detector that detects local motion at an arbitrary position in the image;
A plane parameter voting unit for voting the motion information to a voting space composed of plane space parameters based on a voting table;
A peak extraction unit that extracts a plurality of peak data from the voting results;
A voting data storage for storing voting data information;
Wherein based on the plurality of peak data obtained from the peak extraction unit, an image measuring device and a plane area extracting unit searches the voting data storage unit specifying each local regions included in the plurality of planar regions. By detecting local parallax in an image, voting the parallax information to a voting space composed of plane spatial parameters, obtaining spatial parameters of multiple planes, and searching for local regions based on the data of voting results A plane area specifying method for specifying an area occupied by a plurality of planes specified by the spatial parameter in an image. A local parallax detector that detects local parallax at an arbitrary position in the image;
A plane parameter voting unit for voting the parallax information to a voting space composed of plane space parameters based on a voting table;
A peak extraction unit that extracts a plurality of peak data from the voting results;
A voting data storage for storing voting data information;
Wherein based on the plurality of peak data obtained from the peak extraction unit, an image measuring device and a plane area extracting unit searches the voting data storage unit specifying each local regions included in the plurality of planar regions. A plurality of local region images are cut out from captured images obtained by shooting the subject at different times depending on the camera moving toward the subject, and an arbitrary position in the local region image is extracted for each of the cut out local region images. A local motion is detected, and a vote value obtained based on the detected motion information is voted for each element of a plurality of arrays constituting a voting space composed of plane space parameters, and the vote value is voted to Store in association with the designation information of the element and the designation information of the local region image that is the basis of the vote value, extract a plurality of peak data from the vote result, based on the extracted plurality of peak data, from the stored information, the designation information of the local region image that has contributed to the emergence of the peaks extracted explore the multiple designations are the search among the plurality of local region image A plurality of local region image plane region specifying method for specifying a plurality of planar regions in the captured image specified by broadcast.

Images