WO2012077287A1 - 姿勢状態推定装置および姿勢状態推定方法 - Google Patents
姿勢状態推定装置および姿勢状態推定方法 Download PDFInfo
- Publication number
- WO2012077287A1 WO2012077287A1 PCT/JP2011/006499 JP2011006499W WO2012077287A1 WO 2012077287 A1 WO2012077287 A1 WO 2012077287A1 JP 2011006499 W JP2011006499 W JP 2011006499W WO 2012077287 A1 WO2012077287 A1 WO 2012077287A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- posture state
- likelihood map
- likelihood
- map
- region
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/77—Determining position or orientation of objects or cameras using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Definitions
- the present invention relates to a posture state estimation device and a posture state estimation method for estimating a posture state of an object based on image data obtained by imaging an object having a plurality of regions connected by joints.
- Behavior analysis includes, for example, abnormal behavior detection on the street, purchase behavior analysis in a store, work efficiency improvement support in a factory, and form guidance in sports.
- Non-Patent Document 1 describes a technique for estimating the posture state of a person based on image data captured by a single-eye camera.
- the technology described in Non-Patent Document 1 detects a silhouette (outer shape) of a person from image data, and extracts a ShapeContext histogram, which is one of shape feature quantities, from the silhouette.
- the discriminator is configured for each posture of motion to be identified, using the variance-covariance matrix of the extracted histogram as an input. Thereby, the prior art can estimate the posture state of the person regardless of the position and the direction of the person.
- An object of the present invention is to provide a posture state estimation device and a posture state estimation method capable of accurately estimating the posture state of an object having a joint.
- the posture state estimation device is a posture state estimation device that estimates a posture state of the object based on image data obtained by photographing an object having a plurality of regions connected by joints, A likelihood map generation unit configured to generate a likelihood map indicating a distribution of likelihood that each part is positioned for at least two or more parts, and the likelihood map previously associated with the posture state When the degree of coincidence between the learning likelihood map and the estimated likelihood map which is the likelihood map generated based on the image data is high, the posture state associated with the learning likelihood map is the object state And a posture state estimation unit for estimating the posture state.
- the posture state estimation method is a posture state estimation method for estimating a posture state of the object based on image data obtained by photographing an object having a plurality of regions connected by joints. Generating a likelihood map indicating a distribution of likelihood that each part is positioned for at least two or more parts, and learning likelihood map which is the likelihood map previously associated with the posture state Determining the degree of coincidence with the estimated likelihood map that is the likelihood map generated based on the image data, and when the degree of coincidence is high, the posture state associated with the learning likelihood map Estimating as the posture state of the object.
- the posture state of an object having a joint can be estimated with high accuracy.
- position state estimation apparatus based on Embodiment 1 of this invention which concerns on Embodiment 1 of this invention Diagram for explaining image data in the first embodiment
- a flowchart showing an example of the operation of the posture state estimation apparatus according to the first embodiment Flow chart showing an example of estimation phase processing in the first embodiment Diagram for explaining the omega shape in the first embodiment
- compatible table in this Embodiment 1. A diagram showing an example of the content of the part / region correspondence table according to the first embodiment
- a diagram showing an example of a case where it is determined that the designated posture is in Embodiment 1.
- Block diagram showing an example of the configuration of the posture state estimation apparatus according to Embodiment 2 of the present invention A flowchart showing an example of the operation of the posture state estimation device according to the second embodiment Flow chart showing an example of learning phase processing in the second embodiment
- position state estimation apparatus based on Embodiment 3 of this invention The figure for demonstrating the relationship between the posture of the person based on this Embodiment 3, and the brightness of each site
- Embodiment 1 The first embodiment of the present invention is an example in which the present invention is applied to an apparatus for estimating whether the posture state of a photographed person matches the posture state designated by the user.
- site refers to a group of parts of the human body divided by joints. That is, the sites are, for example, the head, shoulders, upper right arm, right forearm, upper left arm, left forearm, right upper knee, right lower knee, left upper knee, and left lower knee.
- site region is a region that can be occupied by a site in the image, that is, the movable range of the site.
- the "posture state” to be estimated refers to the posture of two or more parts to be focused (hereinafter referred to as "target parts"). Further, the “posture” is represented by information such as the position of a joint connecting the target site in a two-dimensional coordinate system or a three-dimensional coordinate system, or the length of each related site and the angle between the sites I assume. Therefore, “posture state estimation” means estimating a posture state by estimating these pieces of information.
- the above-mentioned position, length, and angle may be expressed by relative values based on a predetermined body part of a person, or may be expressed by absolute values in a two-dimensional coordinate system or a three-dimensional coordinate system. .
- a group of a plurality of pixels corresponding to a predetermined size may be regarded as one pixel, and the same processing may be performed. Thereby, processing can be performed at high speed.
- the value of the pixel serving as the center of gravity of the plurality of pixels may be used as the value of the plurality of pixels, or the average value of the values of the plurality of pixels is used for the plurality of pixels. It may be used as a value.
- FIG. 1 is a block diagram showing an example of the configuration of a posture state estimation apparatus according to Embodiment 1 of the present invention. For simplicity of explanation, peripheral devices of the posture state estimation device are also illustrated.
- posture state estimation apparatus 100 includes posture state management unit 110, posture state designation unit 120, image data acquisition unit 130, region region estimation unit 140, likelihood map generation unit 150, and posture state estimation unit 160. .
- the posture state management unit 110 previously stores, for each posture state, identification information of the posture state, identification information of two or more target parts specified for the posture state, and a likelihood map in association with each other.
- the likelihood map is a distribution of likelihood (likelihood) that each target portion is positioned on the image, and the details will be described later.
- the posture state only the posture state in which the information is stored in the posture state management unit 110 is referred to as “the posture state”.
- the likelihood map associated with the posture state and stored in advance in the posture state management unit 110 is referred to as a “learning likelihood map”.
- Posture state designation unit 120 receives a designation of a posture state to be a target of estimation from the user via an input device (not shown) such as a keyboard. Specifically, the posture state designation unit 120 creates a list of posture states with reference to, for example, the posture state management unit 110, and receives the designation of the posture states by displaying the list as options. For example, the posture state designation unit 120 receives designation of a posture state that "the right arm is bent”. Then, posture state designation unit 120 outputs identification information of the specified posture state to part region estimation unit 140 and posture state management unit 110.
- identification information of the designated posture state and the designated posture will be collectively referred to as "designated posture”.
- posture state designation unit 120 outputs two or more focused parts associated with the designated posture to part region estimation unit 140.
- the posture state designation unit 120 outputs the “upper right arm” and the “right forearm” for the posture state “the right arm is bent”.
- the target portion associated with the designated posture is referred to as a “designated portion”.
- the image data acquisition unit 130 acquires image data of an image captured by the single-eye camera 200 installed in a predetermined three-dimensional coordinate space by wired communication or wireless communication, and outputs the image data to the part region estimation unit 140.
- the image data is described as including only one image. However, the image data may include images of a plurality of people or may not include human images.
- FIG. 2 is a diagram for explaining image data.
- a three-dimensional coordinate system 410 having an origin O as a position where the position of the single-eye camera 200 is projected on the ground is set.
- the vertical direction is the Y axis
- the Y axis and the direction orthogonal to the optical axis 411 of the single-eye camera 200 are the X axis
- the direction orthogonal to the X axis and the Y axis is the Z axis.
- the installation angle of the monocular camera 200 is represented by, for example, an angle ⁇ between the Y axis and the optical axis 411. Then, the single-eye camera 200 performs focusing by focusing on a certain plane 412 included in the range of the angle of view ⁇ of the single-eye camera 200. Image data of the image captured in this manner is transmitted to the posture state estimation device 100. Hereinafter, image data to be subjected to estimation of the posture state is referred to as “estimated image data”.
- the part region estimation unit 140 of FIG. 1 estimates the part region of the designated portion input from the posture state designation unit 120 based on the estimated image data input from the image data acquisition unit 130. Specifically, part region estimation section 140 estimates the position and orientation of a reference part of a person from estimated image data. Then, part region estimation section 140 estimates a part region of each designated portion based on the estimated position and orientation of the reference portion.
- part region estimation unit 140 outputs estimated image data, a designated posture, and information indicating a part region for each designated portion (hereinafter referred to as “part region data”) to likelihood map generation unit 150.
- the likelihood map generation unit 150 generates a likelihood map from the estimated image data input from the part region estimation unit 140. At this time, the likelihood map generation unit 150 lowers the likelihood that the designated region corresponding to the region will be located for regions other than the region indicated by the region data input from the region estimation unit 140. Generate a likelihood map. In such a likelihood map, only the likelihood of the movable range of the target part of the designated posture (for example, "upper right arm” and “right forearm” in the case of the posture state "right arm is bent") is high It becomes such information. Then, likelihood map generation section 150 outputs the generated likelihood map to posture state estimation section 160 together with the designated posture input from part region estimation section 140.
- the likelihood map generated based on the estimated image data is referred to as “estimated likelihood map”.
- Posture state estimation unit 160 obtains a learning likelihood map associated with the designated posture input from likelihood map generation unit 150 from posture state management unit 110.
- the posture state management unit 110 may output the learning likelihood map to the posture state estimation unit 160 in response to an instruction from the posture state estimation unit 160.
- posture state estimation unit 160 estimates the specified posture when the degree of coincidence between the acquired learning likelihood map and the estimated likelihood map input from likelihood map generation unit 150 is high. It is estimated as the posture state of the person included in the image data. That is, posture state estimation unit 160 estimates that the person photographed by single-eye camera 200 has taken the posture state specified by the user. Then, the posture state estimation unit 160 transmits information to the information output device 300 such as a display device by wired communication or wireless communication, and notifies the user of the estimation result.
- the information output device 300 such as a display device by wired communication or wireless communication
- the posture state estimation device 100 is a computer including storage media such as a CPU (central processing unit) and a RAM (random access memory). That is, the posture state estimation device 100 operates when the CPU executes a control program to be stored.
- a CPU central processing unit
- RAM random access memory
- Such a posture state estimation apparatus 100 uses a likelihood map indicating the distribution of likelihood for each part, for example, even when the right arm is accommodated in the outer shape of the trunk on the image, "the right arm is bent" It can be determined whether or not it is in the following posture state. That is, the posture state estimation device 100 can estimate the posture state of a person with high accuracy as compared with the prior art.
- posture state estimation apparatus 100 estimates a part region which is a movable region of a designated part and lowers the likelihood value for regions other than the part region, the accuracy of the likelihood map can be improved.
- FIG. 3 is a flowchart showing an example of the operation of the posture state estimation device 100.
- part region estimation section 140 determines whether or not there is an instruction for posture state estimation. For example, when a new posture state is specified in posture state designation unit 120 or when new estimated image data is input to image data acquisition unit 130, part region estimation unit 140 Judge that there was an instruction. Part region estimation unit 140 proceeds to step S4000 when there is a posture state estimation instruction (S3000: YES). If there is no posture state estimation instruction (S3000: NO), part region estimation section 140 proceeds to step S5000.
- a posture state estimation instruction S3000: YES
- step S4000 posture state estimation apparatus 100 executes estimation phase processing for estimating posture state, and proceeds to step S5000. Details of the estimation phase process will be described later.
- step S5000 part region estimation unit 140 determines whether an instruction to end processing has been issued by user operation or the like. When there is no instruction to end the process (S5000: NO), part region estimation section 140 returns to step S3000, and continues waiting for an instruction for posture state estimation. In addition, part area estimation unit 140 ends a series of processing when there is an instruction to end processing (S5000: YES).
- FIG. 4 is a flowchart showing an example of the estimation phase process (step S4000 in FIG. 3).
- posture state designation unit 120 receives designation of a posture state from the user, acquires a designated posture, and acquires a designated part corresponding to the designated posture.
- part region estimation unit 140 acquires estimated image data from monocular camera 200 via image data acquisition unit 130.
- the estimated image data may be still image data constituting moving image data continuously captured in real time by the single-eye camera 200, or image data captured and stored in advance.
- step S4300 part region estimation section 140 performs processing for estimating the position and orientation of the reference part (hereinafter referred to as “reference part estimation processing").
- the reference part estimation process is roughly divided into a first process of estimating a person's shoulder joint position and a second process of estimating a person's torso direction.
- the part region estimation unit 140 detects an omega shape from the estimated image data, and estimates a shoulder joint position based on the omega shape.
- FIG. 5 is a diagram for explaining the omega shape.
- the omega ( ⁇ ) shape is a characteristic edge shape of a region including the head and shoulders of a person, and among human bodies, the probability of being most stably photographed when using a surveillance camera or the like is high. It is a shape. In addition, the head and the shoulder have little change in the relative position with the human torso. Therefore, the part region estimation unit 140 detects the omega shape first to detect the positions of the head and shoulders of the person, and estimates the part region of the other part with reference to these, thereby achieving high accuracy of the part region. Estimate.
- Omega shapes can be detected, for example, using a detector made by Real AdaBoost, etc., using a sufficient number of sample images.
- a feature amount used for the detector for example, a histogram of gradient (HoG) feature amount, a Sparse feature amount, a Haar feature amount, or the like can be used.
- HoG histogram of gradient
- Sparse feature amount a Sparse feature amount
- Haar feature amount or the like.
- Boosting method for example, in addition to the Boosting method, it is also possible to use an SVM (support vector machine), a neural network, or the like.
- Part region estimation section 140 first detects omega shape 421 from image 420 of the estimated image data.
- a pixel (a pixel of an edge portion) constituting the omega shape 421 is a digital signal "1" and the other pixels are digital signals "0".
- a relatively small rectangular area including the omega shape 421 is determined as the omega area 422.
- the lower side of the omega region 422 is referred to as a reference line 423.
- the part region estimation unit 140 removes noise included in the omega region 422. Specifically, part region estimation unit 140 corrects digital signal “1” present in the region surrounded by omega shape 421 among the pixels of omega region 422 as digital signal “0” as noise. Do. This correction is possible, for example, by performing a so-called closing process.
- the closing process is a process of enlarging or reducing the image area at a predetermined pixel ratio or a predetermined ratio. This correction can improve the accuracy of the distance histogram described later.
- part region estimation section 140 obtains the vertical distance from reference line 423 to omega shape 421 for each position of reference line 423.
- FIG. 6 is a diagram for explaining the vertical distance from the reference line 423 to the omega shape 421.
- the part region estimation unit 140 treats the direction of the reference line 423 as the X axis and handles the vertical direction of the reference line 423 as the Y axis.
- the part region estimation unit 140 sets, for example, the number of pixels from the left end of the reference line 423 as an X coordinate value.
- part region estimation section 140 acquires the number of pixels in the Y-axis direction from reference line 423 to the pixels forming omega shape 421, that is, the vertical distance to omega shape 421 as vertical distance d (X).
- the pixels forming the omega shape 421 are, for example, those closest to the reference line 423 among the pixels of the digital signal “1”.
- part region estimation section 140 generates a distance histogram in which data of n (n is a positive integer) vertical distances d (X) are associated with X coordinates.
- FIG. 7 is a diagram showing an example of the distance histogram generated by the part region estimation unit 140 based on the omega region 422 shown in FIG.
- part region estimation section 140 generates distance histogram 430 showing the distribution of vertical distance d (X) in the XY coordinate system, using vertical distance d (X) as the value of the Y axis.
- the distance histogram 430 swells in a shape corresponding to the shoulder, and of those, it protrudes in a range corresponding to the center of the head.
- part region estimation section 140 applies a predetermined threshold Th to generated distance histogram 430 to perform binarization processing. Specifically, part region estimation unit 140 replaces the Y coordinate value of the X coordinate at which vertical distance d (X) is equal to or greater than threshold value Th to “1”, and vertical distance d (X) is less than threshold value Th. Replace the Y-coordinate value of X-coordinate, which is The threshold value Th is set with a high probability in the omega region 422 to be larger than the vertical distance d (X) at the upper end of the shoulder and smaller than the vertical distance d (X) at the upper end of the head.
- the binarization processing is not limited to this, and may be another method such as, for example, so-called Otsu binarization (Otsu method).
- FIG. 8 shows an example of the result of binarizing the distance histogram 430 shown in FIG.
- the range 441 of “1” indicates the range of the X coordinate of the image area of the central portion of the head (hereinafter referred to as “head area”). Further, the entire range 442 including the range 441 where “1” is to be shown indicates the range of the X coordinate of the image area of the shoulder (hereinafter referred to as “the shoulder area”). Therefore, part region estimation unit 140 extracts the range in the X axis direction of omega region 422 in image 420 of the estimated image data as the X axis direction range of the shoulder region, and the X axis direction of range 441 which becomes "1". The range is extracted as an X-axis direction range of the head region.
- the part region estimation unit 140 calculates various parameters indicating the position and the orientation of the reference part based on the extracted shoulder region and head region.
- FIG. 9 is a diagram for explaining various parameters indicating the reference part.
- the part region estimation unit 140 uses H (xh, yh), RSE (x_rse), RD (symbols in parentheses as parameters indicating the position of the reference part). It is assumed that x_rd), RS (x_rs, y_rs), RSU (y_rsu), and LS are used.
- H is the center of gravity of the head.
- RSE is the position of the end of the right shoulder.
- RD is the distance in the X-axis direction from the center of gravity of the head to the end of the right shoulder.
- RS is the position of the right shoulder joint (hereinafter referred to as “right shoulder position”).
- RSU is at the top of the right shoulder.
- LS is the position of the left shoulder joint (hereinafter referred to as "left shoulder position").
- Part region estimation section 140 calculates the value of each parameter, for example, as follows.
- the part region estimation unit 140 determines the right shoulder region based on whether or not a person (body) is facing the monocular camera 200 from the shoulder region extracted based on the result of the binarization processing. .
- the part region estimation unit 140 determines whether the person is facing the monocular camera 200 based on whether the skin color component of the color information of the head region is equal to or more than a predetermined threshold.
- a predetermined threshold it is assumed that a person is facing the monocular camera 200, and the left shoulder region toward the image is determined to be the right shoulder region.
- part region estimation section 140 calculates the barycentric position of the right shoulder region as right shoulder position RS (x_rs, y_rs). Further, the part region estimation unit 140 calculates the barycentric position H (xh, yh) of the head, and the distance in the Y axis direction between the barycentric position H (xh, yh) and the omega shape 421 (hereinafter referred to as “head The right shoulder position RS (x_rs, y_rs) may be calculated using the height ⁇ h ′ ′).
- the part region estimation unit 140 sets a value that is a predetermined ratio to the head height ⁇ h as a distance from the center of gravity H of the head to the right shoulder position RS in the X axis direction. It should be (xh-x_rs).
- the part region estimation unit 140 may set a position lower than the shoulder height by a value ⁇ h / 2 that is half the head height ⁇ h as the Y coordinate y_rs of the right shoulder position RS.
- part region estimation section 140 calculates a point at which the inclination of the edge of omega shape 421 (that is, the change rate of the distance histogram) exceeds a threshold as position RSE (x_rse) of the end of the right shoulder. Then, part region estimation section 140 calculates distance RD (x_rd) in the X-axis direction between center-of-gravity position H of the head and position RSE of the end of the right shoulder.
- part region estimation unit 140 similarly calculates the left shoulder position LS.
- the calculation method of each parameter is not limited to the above-mentioned example.
- the part region estimation unit 140 holds a reference part correspondence table in advance.
- the reference part correspondence table is a combination of the center of gravity H of the head, the right shoulder position RS, and the left shoulder position LS (hereinafter referred to as “reference part position”), and the orientation of the body estimated from the position of the reference part This table is described in association with (hereinafter referred to as "direction of reference part”). That is, the reference part table is a table describing the relative positional relationship of each part.
- part is an omega-shaped part which shows a human head and a shoulder part as mentioned above. Therefore, the orientation of the reference part is the orientation of the human body (body).
- the part region estimation unit 140 derives the orientation of the reference part corresponding to the position of the reference part calculated from the estimated image data from the reference part correspondence table.
- part region estimation unit 140 sets the center of gravity position H of the head as an origin, and the length between the center of gravity position H of head and right shoulder position RS or left shoulder position LS is 1 Use the normalized values to derive the quasi-site orientation.
- the right shoulder position RS and the left shoulder position LS may be described in the reference part correspondence table.
- a line passing through the center of gravity H of the head and the right shoulder position RS or the left shoulder position LS and a vertical straight line passing through the center of gravity H of the head (hereinafter referred to as "head vertical line" The angle formed by) may be described.
- the reference part correspondence table describes the distance between the center of gravity H of the head and the left shoulder LS when the distance between the center of gravity H of the head and the right shoulder position RS is 1. It is good.
- the part region estimation unit 140 derives the direction of the reference part by calculating parameters corresponding to the parameters described in the reference part correspondence table.
- FIG. 10 is a diagram showing an example of the content of the reference portion correspondence table.
- the reference part correspondence table 450 describes the projection angle 452, the coordinates 453 of the left shoulder position LS, the coordinates 454 of the center of gravity H of the head, and the direction 455 of the reference part in association with the identifier 451.
- Each coordinate is expressed, for example, using a predetermined two-dimensional coordinate system parallel to the two-dimensional coordinate system of the screen, with the right shoulder position RS as the origin.
- the projection angle 452 is, for example, an angle of the predetermined two-dimensional coordinate system (that is, the installation angle ⁇ shown in FIG. 2) with respect to the XZ plane of the three-dimensional coordinate system 410 described in FIG.
- the orientation 455 of the reference portion is represented by, for example, a rotation angle with respect to each of the XYZ axes of the three-dimensional coordinate system 410 described in FIG.
- part region estimation section 140 estimates the position and orientation of the reference part. This is the end of the description of the reference part estimation process.
- step S4400 in FIG. 4 the part region estimation unit 140 estimates the part region for each designated part based on the estimated position and orientation of the reference part (hereinafter referred to as "part region estimation processing"). Do.
- the part region estimation unit 140 holds a part region correspondence table in advance.
- the part region correspondence table is a table in which the position and the direction of the reference part are described in association with the part regions of the other parts.
- the part region estimation unit 140 derives a part region of the designated part corresponding to the position and the orientation of the reference part estimated from the estimated image data from the part region correspondence table.
- the part region is defined, for example, by the pixel position of the image of the estimated image data. Therefore, the part region estimation unit 140 determines whether each pixel belongs to a part region of any designated portion with respect to all the pixels of the entire image of the estimated image data.
- FIG. 11 is a diagram showing an example of the content of the part / region correspondence table.
- the part region correspondence table 460 associates the position 463 of the head and shoulder area (reference part), the direction 464 of the head and shoulder area (reference part), and the area 465 of each part in correspondence with the identifier 461. Describe.
- Each position and area are represented, for example, by values of a two-dimensional coordinate system of an image.
- the projection angle 462 is, for example, an angle of the predetermined two-dimensional coordinate system (that is, the installation angle ⁇ shown in FIG. 2) with respect to the XZ plane of the three-dimensional coordinate system 410 described in FIG.
- the position 463 of the head and shoulder area is, for example, the right shoulder position RS.
- the orientation 464 of the head and shoulder region is represented by, for example, a rotation angle with respect to each of the XYZ axes of the three-dimensional coordinate system 410 described in FIG.
- the direction 464 of the head and shoulder area may not necessarily be described in the part area correspondence table 460.
- the region 465 of each portion is represented by, for example, the center coordinates and the radius of the circle when the region is approximated by a circle.
- the part region estimation unit 140 may not necessarily use the part region correspondence table 460 when obtaining a part region.
- the part region estimation unit 140 uses the various other body constraint information from the orientation of the reference part derived from the reference part correspondence table 450 (see FIG. 10) to set each part region, for example, from the reference part. It may be calculated dynamically in the order of connection.
- the body constraint information is information including constraints regarding the position of each part.
- part region estimation unit 140 sends information indicating whether or not it is a part region of a designated portion for each pixel to the likelihood map generation unit 150 as portion region data for all pixels of the entire image of the estimated image data. Output.
- the part region data may have, for example, a structure in which pixel information Kij indicating whether or not the part region of any designated part corresponds to all pixel positions (i, j) of the estimated image data.
- each element of the pixel information Kij takes “1” when it belongs to the part region of the corresponding designated part, and takes “0” when it does not belong to it.
- k1 corresponds to the region of the right upper arm
- k2 corresponds to the region of the right forearm.
- the part area data may indicate which part area of the designated part corresponds to each part area preset in the image, or the coordinates of the outer edge of the part area may be indicated for each designated part. good.
- the part area correspondence table describes the part area corresponding to the normalized reference part. Further, other information such as the right shoulder position RS and the left shoulder position LS may be described in the region region data, as in the case of the reference region correspondence table described above.
- the part region estimation unit 140 derives a part region of each designated part by calculating parameters corresponding to the parameters described in the part region correspondence table.
- FIG. 12 is a diagram showing an example of the content of part region data. Here, in order to simplify the description, the positions of the respective parts in the upright state are illustrated together.
- the part region data indicates the part region 471 of the upper right arm which is the designated part and the part region 472 of the right forearm which is the designated part. As described above, these region regions 471 and 472 are estimated based on the position and orientation of the previously estimated reference region 473.
- part region estimation section 140 estimates the part region of each designated portion. This is the end of the description of the part region estimation process.
- step S4500 in FIG. 4 the likelihood map generation unit 150 generates an estimated likelihood map by calculating the likelihood value for the portion region for each designated portion (hereinafter, “estimated likelihood map generation processing” To do).
- the likelihood map generation unit 150 determines, from the estimated image data, an image feature suitable for representing the position and direction of the designated portion for each pixel in the portion region of the designated portion, and the designated portion The likelihood value indicating the likelihood of being located is calculated. Then, the likelihood map generation unit 150 generates an estimated likelihood map indicating the distribution of likelihood values of each pixel using the likelihood value calculated from the estimated image data.
- the likelihood value may be a value normalized to be in the range of 0 to 1, or may be a real number including a positive integer or a negative number.
- the SIFT feature quantity is a 128-dimensional vector and is a value calculated for each pixel.
- the SIFT feature amount is particularly effective for detecting a portion that can rotate in various directions, such as an arm, because it is not affected by the scale change, rotation, and translation of an object to be detected. That is, the SIFT feature quantity is suitable for the present embodiment in which the posture state is defined by the relative joint position and angle of two or more target parts.
- the classifier Hk is generated by the AdaBoost algorithm. That is, until the strong classifier Hk can determine with a desired accuracy whether or not it is the upper right arm and the right forearm with respect to a plurality of learning images prepared for each part in advance. Learning is repeated, and a plurality of weak classifiers are generated by cascading.
- the likelihood map generation unit 150 calculates the image feature quantity for each designated part and each pixel, the image feature quantity is input to the strong classifier Hk, and the output of each weak classifier constituting the strong classifier Hk is used. On the other hand, for each weak classifier, a sum of values obtained by multiplying the reliability ⁇ obtained in advance is calculated. Then, the likelihood map generation unit 150 subtracts the predetermined threshold value Th from the calculated sum to calculate the likelihood value ck for each designated portion and for each pixel.
- c1 represents the likelihood value of the upper right arm
- c2 represents the likelihood value of the right forearm.
- the likelihood map generation unit 150 determines, for each pixel, whether the pixel is included in any part region, and if it is included, the likelihood value is calculated using the classifier of that part, For example, the likelihood value of that part may be set to zero. In other words, the likelihood map generation unit 150 determines the determinant (K ij) of the pixel information output from the part region estimation unit 140 and the determinant (C ij) of the likelihood value of each pixel calculated independently of the part region. The result of integration of and may be used as a final estimated likelihood map.
- FIG. 13 is a diagram illustrating an example of the estimated likelihood map. Here, only the likelihood value of one designated part (for example, the upper right arm) of the estimated likelihood map is shown, and the higher the likelihood value, the darker the shaded area is. As shown in FIG. 13, the estimated likelihood map 478 represents the distribution of the likelihood that the designated part is located.
- the likelihood map generator 150 generates an estimated likelihood map. This completes the description of the estimated likelihood map generation process.
- step S4600 posture state estimation unit 160 acquires a learning likelihood map corresponding to the specified posture from posture state management unit 110. Then, the posture state estimation unit 160 performs a matching degree determination process of determining whether or not the learning likelihood map and the estimated likelihood map match, based on whether the matching degree is equal to or higher than a predetermined level. Do.
- Posture state estimation section 160 first binarizes the estimated likelihood map and the learning likelihood map using predetermined threshold values. Specifically, the posture state estimation unit 160 sets the likelihood value of each pixel and each designated part to the digital signal “0” when it is equal to or more than a predetermined threshold, and a digital signal when it is less than the predetermined threshold. Convert to "1".
- FIG. 14 is a diagram illustrating an example of a state after the estimated likelihood map illustrated in FIG. 13 is binarized.
- the pixels of the digital signal “1” are represented in gray, and the pixels of the digital signal “0” are represented in white.
- the estimated likelihood map 479 after binarization represents the distribution of a portion with high likelihood of the designated part being located.
- posture state estimation unit 160 takes the product of likelihood values binarized for each pixel and for each designated part between the estimated likelihood map and the learning likelihood map, and designates all pixels and all designations.
- the sum of the values for the site is taken as the evaluation value.
- posture state estimation section 160 superimposes the estimated likelihood map and the learning likelihood map in a predetermined positional relationship, multiplies the likelihood value information after binarization for each pixel, and multiplies them. Calculate the sum of all pixels and designated parts.
- Posture state estimation unit 160 shifts the positional relationship of superposition between the estimated likelihood map and the learning likelihood map by movement and rotation, and performs the above-described arithmetic processing on each positional relationship. Then, posture state estimation unit 160 obtains the maximum value among the obtained evaluation values as a final evaluation value representing the degree of coincidence, and when this evaluation value is equal to or greater than a predetermined threshold value, a learning likelihood map and It is determined that the estimated likelihood map matches.
- a threshold value an appropriate value is set in advance by learning or the like.
- Posture state estimation section 160 may not necessarily binarize the estimated likelihood map and the learning likelihood map. In this case, posture state estimation section 160 can more accurately determine the degree of coincidence between the learning likelihood map and the estimated likelihood map. When binarization is performed, the posture state estimation unit 160 can determine the degree of coincidence at high speed.
- posture state estimation section 160 determines the degree of coincidence between the estimated likelihood map and the learning likelihood map. This is the end of the description of the matching degree determination process.
- posture and state estimation section 160 matches the learning likelihood map and the estimated likelihood map (S4600: YES), processing proceeds to step S4700. Further, if the learning likelihood map and the estimated likelihood map do not match (S4600: NO), posture state estimation section 160 proceeds to step S4800.
- step S4700 posture state estimation unit 160 notifies the user that the posture of the person included in the target image data is the designated posture via information output device 300, and returns to the processing of FIG.
- posture state estimation unit 160 notifies the user that the posture of the person included in the target image data is not the designated posture via information output device 300, and returns to the process of FIG. In the case where it is not possible to determine the posture state, such as when a person is not detected from the target image data, the posture state estimation unit 160 may notify that effect.
- the notification in steps S4700 and S4800 can be performed according to the presence or absence of an output such as a character display, an image display, an audio output, and a vibration output, or a difference in output content.
- an output such as a character display, an image display, an audio output, and a vibration output, or a difference in output content.
- posture state estimation apparatus 100 can estimate a part region and generate an estimated likelihood map indicating the distribution of the likelihood for each designated part.
- the posture state estimation device 100 can estimate the posture state by comparing the generated estimated likelihood map with the learning likelihood map associated with the designated posture.
- FIG. 15 is a diagram illustrating an example of the case where it is determined that the designated posture is set.
- the site region 482 to be compared is a range that covers the whole body.
- the designated site is the whole body area of the person 491, and the site area 492 to be compared is a range including the whole body.
- the learning likelihood map of the part region 482 based on the learning image 480 shown in FIG. 15A and the estimation likelihood map of the part region 492 based on the estimated image 490 match when relatively moved and rotated. . Therefore, posture state estimation apparatus 100 can determine that the person included in learning image 480 is in the posture state of "upright posture".
- FIG. 16 is a view showing another example in the case where it is determined that the designated posture is taken.
- the part region 482 to be compared is a range including the upper right arm and the right forearm.
- the right arm is bent is designated in the estimated image 490 in the estimation phase.
- the designated part is the upper right arm and the right forearm of the person 491
- the part region 492 to be compared is a range including the upper right arm and the right forearm.
- the learning likelihood map of part region 482 based on learning image 480 shown in FIG. 16A matches the estimation likelihood map of part region 492 based on estimated image 490 when relatively moved and rotated. . Therefore, posture state estimation apparatus 100 can determine that the person included in learning image 480 is in the posture state that “the right arm is bent”.
- posture state estimation apparatus 100 can accurately perform posture state estimation as described above.
- posture state estimation apparatus 100 has a posture that “the right arm is bent” even if the postures of parts other than the right arm are different as shown in FIG. 16 (B) to FIG. 16 (E). It can be determined that it is in the state. That is, the posture state estimation apparatus 100 can extract the posture focusing on only the designated part regardless of the postures of other parts.
- posture state estimation apparatus 100 uses a likelihood map indicating the distribution of likelihoods for each part, it is possible to estimate a person's posture state with high accuracy.
- posture state estimation apparatus 100 estimates a part region and generates an estimated likelihood map in which the likelihood value is lowered for regions other than part region, the accuracy of the likelihood map is improved and posture state estimation is performed. It can be done with higher accuracy.
- the posture state estimation apparatus 100 estimates only a certain posture state that is specifically designated. However, among the plurality of posture states, which posture state corresponds to You may make it estimate. In this case, for example, the posture state estimation device 100 may treat all posture states in which the learning likelihood map corresponding to the posture state management unit 110 is stored as the designated posture. Further, the posture state estimation apparatus 100 may handle all posture states as specified postures when execution of posture state estimation is instructed without specifying any posture state.
- image data used for posture state estimation may be data of an image captured by a stereo camera or a plurality of cameras.
- the posture state estimation apparatus 100 may use image data captured by one camera and position information of an object obtained from installation parameters of the stereo camera.
- the posture state estimation apparatus 100 includes image data captured by one of the cameras and position information of an object obtained from installation parameters of each camera. You may use.
- part region estimation unit 140 may not perform the above-described reference part estimation processing.
- part region estimation unit 140 may hold body direction information.
- the method of estimation of the part region performed by the part region estimation unit 140 is not limited to the above-described example.
- the part area estimation unit 140 extracts an edge part (hereinafter simply referred to as an “edge”) of the image from the estimated image data, and based on the range of Y coordinate values of the area surrounded by the edge, each part area You may estimate Specifically, for example, in the region surrounded by the edge, the part region estimation unit 140 estimates a region from the position with the highest Y coordinate value to 20% as the region of the head.
- the region estimation unit 140 may set the region of 15% to 65% as the region of the trunk, the region of 55% to 85% as the region of knee, and the region of 75% to 100%. It is estimated as the region under the knee.
- the part region estimation unit 140 extracts a moving body by taking a background difference between the images, and extracts the entire region including the extracted region. , It may be a candidate of the region of each part. As a result, it is possible to speed up the process when estimating the part region.
- posture state estimation apparatus 100 estimates the position of a portion one by one in order of proximity to the reference portion, and repeats the processing of estimating the portion region of the next portion based on the estimated position, A part region may be estimated.
- the posture state estimation unit 160 sets the installation angle ⁇ of the monocular camera 200.
- the corresponding learning likelihood map may be a comparison target.
- posture state estimation apparatus 100 does not necessarily have to perform part region estimation.
- the likelihood map generator 150 uniformly calculates the likelihood value for all the regions of the image.
- the type of likelihood map handled by the posture state estimation apparatus 100 is not limited to the likelihood map generated by the above-described example.
- the estimated likelihood map and the learning likelihood map may be generated by extracting parallel lines from edges.
- the likelihood map generation unit 150 is provided in advance with, for example, a correspondence table in which the length of the shoulder joint and the standard thickness value of each part are associated.
- the likelihood map generation unit 150 searches a set of parallel lines separated by a distance corresponding to the standard thickness of the part in the part region, while rotating the determination direction by 360 degrees. Then, when there is a corresponding set of parallel lines, likelihood map generation section 150 repeats the process of voting for each pixel in the area surrounded by those parallel lines, and finally the final pixel Generate an estimated likelihood map based on the number of votes.
- the estimated likelihood map and the learning likelihood map will include the direction of parallel lines and the number of votes (hereinafter referred to as "the likelihood value of the direction") for each pixel and for each designated part.
- the likelihood value for each pixel and for each designated part is an eight-dimensional value corresponding to eight directions.
- the distance and angle of the parallel line to be voted may be different for each part.
- the likelihood map generation unit 150 determines, for each designated part, for example, the direction with the highest likelihood value of the direction as the main edge direction of the designated part. At this time, even if posture state estimation unit 160 takes the sum value of the likelihood values of all pixels for each direction and determines that the direction in which the sum value is the highest is the direction in which the direction likelihood value is the highest. good.
- posture state estimation section 160 superimposes the estimated likelihood map and the learning likelihood map so that the main edge directions of the designated parts coincide with each other, and calculates the degree of coincidence.
- the subsequent processing is the same as the method already described in the present embodiment.
- the method in which the direction of the edge is taken into consideration can add constraints to the positional relationship of superposition between the estimated likelihood map and the learning likelihood map, thereby reducing the processing load.
- the posture state estimation unit 160 calculates the coincidence of the angles formed by the edge directions of the designated parts among a plurality of designated parts with an evaluation value representing the coincidence of the estimated likelihood map and the learning likelihood map. Do. Then, when the evaluation value is within the predetermined range, posture state estimation unit 160 determines that the posture of the person included in the target image data is the designated posture.
- the method of determining the degree of coincidence using only the edge direction can eliminate the processing of repeatedly calculating a plurality of evaluation values while rotating the image, thereby further reducing the processing load. .
- the second embodiment of the present invention is an example in which the generation of a learning likelihood map is performed together in the posture state estimation apparatus.
- the posture state estimation device according to the present embodiment performs learning phase processing for generating a learning likelihood map, in addition to estimation phase processing for estimating the posture state.
- FIG. 17 is a block diagram showing an example of the configuration of the posture state estimation apparatus according to Embodiment 2 of the present invention, and corresponds to FIG. 1 of Embodiment 1.
- the same parts as those in FIG. 1 are denoted by the same reference numerals, and the description thereof is omitted.
- the posture / state estimation apparatus 100a has a likelihood map generation unit 150a different from that of the first embodiment.
- the image data acquisition unit 130 and the part region estimation unit 140 in the present embodiment perform the same process as the process for estimated image data on image data input in the learning phase (hereinafter referred to as “learned image data”). Perform and estimate part regions.
- posture state designation unit 120 in the present embodiment also receives designation of the posture state and the part in the learning phase, and outputs the designated posture and the specified part.
- the likelihood map generation unit 150a performs the same processing as processing on estimated image data on learning image data, and reduces the likelihood of the designated part corresponding to the part region being located. Generate a map. However, the likelihood map generation unit 150 a stores the likelihood map generated from the learning image data in the posture state management unit 110 as a learning likelihood map, in association with the designated posture and the designated part. Further, the likelihood map generation unit 150a does not output the likelihood map generated from the learning image data to the posture state estimation unit 160.
- Such posture state estimation apparatus 100a receives learning image data input and designation of posture state and part, generates a learning likelihood map, and estimates posture state for target image data using the generated learning likelihood map It can be performed.
- FIG. 18 is a flowchart showing an example of the operation of the posture state estimation device 100a, which corresponds to FIG. 3 of the first embodiment. The same steps as in FIG. 3 are assigned the same step numbers, and the explanation thereof is omitted.
- step S1000a part region estimation section 140 determines whether or not there is an instruction for posture state learning.
- Posture state learning is, in other words, generation of a learning likelihood map.
- part region estimation unit 140 specifies a new posture state in posture state designation unit 120, or new estimated image data is input to image data acquisition unit 130.
- Switching between the learning phase and the estimation phase is performed, for example, by receiving a predetermined operation from the user via an input device (not shown) such as a keyboard.
- an instruction for posture state learning S1000a: YES
- part region estimation section 140 proceeds to step S2000a.
- part region estimation section 140 proceeds to the processing of steps S3000 to S5000 described in the first embodiment.
- step S2000a posture state estimation apparatus 100 executes learning phase processing for learning posture state, and proceeds to the processing of steps S3000 to S5000 described in the first embodiment.
- FIG. 19 is a flowchart showing an example of the learning phase process (step S2000a in FIG. 18).
- the posture state designation unit 120 receives designation of a posture state from the user, acquires a designated posture, and acquires a designated part corresponding to the designated posture. In addition, it is necessary to perform designation from outside of the designated part at least once corresponding to the designated posture.
- the posture state designation unit 120 stores a set of the designated posture and the designated part, and automatically determines the designated part from the designated posture after the second time, and omits acceptance of designation of the part. Also good.
- part region estimation unit 140 acquires learning image data from monocular camera 200 via image data acquisition unit 130.
- the learning image data may be still image data constituting moving image data continuously captured in real time by the single-eye camera 200, or image data captured and stored in advance.
- the learning image data may not be data of an image obtained by photographing an actual person, but may be data of an image generated by computer graphics (CG) software on a computer.
- CG computer graphics
- the image data of the motion capture software can simultaneously acquire three-dimensional posture information of a person, convenience in generating a learning likelihood map can be improved.
- step S2300a part region estimation section 140 performs processing similar to the reference part estimation processing described in Embodiment 1 on the learning image data to estimate a reference part.
- step S2400a part region estimation section 140 performs processing similar to the part region estimation processing described in the first embodiment on the learning image data, and estimates a part region for each designated portion.
- likelihood map generation section 150a performs the same process as the estimated likelihood map generation process described in the first embodiment on the learning image data, and sets likelihood values for part regions for each designated part. To generate a learning likelihood map.
- step S2600a the likelihood map generation unit 150a causes the posture state management unit 110 to store the generated learning likelihood map in association with the designated part and the designated posture, and returns to the processing of FIG.
- the likelihood map generation unit 150a In the case where the same posture state is specified for a plurality of learning image data, that is, when there are a plurality of learning likelihood maps having the same specified posture, the likelihood map generation unit 150a generates likelihood values.
- the posture state management unit 110 may store a learning likelihood map composed of the average value of
- the posture state estimation device 100a can generate and store a learning likelihood map in response to input of learning image data and designation of a posture state and a part.
- a learning likelihood map is generated based on a part region 482 shown in FIG. 16A of the first embodiment, and the posture state that “right arm is bent”, “upper right arm” and “right It is stored in association with a part called "forearm".
- posture state estimation apparatus 100a since posture state estimation apparatus 100a according to the present embodiment generates a learning likelihood map indicating the distribution of likelihood for each part for each posture state, and uses the generated learning likelihood map, It is possible to estimate the posture of a person with high accuracy.
- the estimated image 490 in FIGS. 16B to 16E is not in the posture state that “the right arm is bent”. It will judge. In order to prevent this in the prior art, it is necessary to prepare learning images and generate a learning likelihood map for all the contours shown in FIGS. 16 (B) to 16 (E). Such exhaustive learning takes time and effort. In addition, as the number of learning likelihood maps stored increases, it also takes time to determine the degree of coincidence. On the other hand, in the case of using the posture state estimation apparatus 100a according to the present embodiment, as described above, learning can be performed on the learning image 480 shown in FIG. 16A, and the number of learning likelihood maps can also be suppressed. .
- posture state estimation apparatus 100a uses an estimated likelihood map generated by another method such as the estimated likelihood map based on the edge direction described in the first embodiment, the learning likelihood is calculated by the corresponding method. It shall generate a degree map.
- the application of the present invention is not limited to the estimation of the posture state of the person described in the first embodiment and the second embodiment.
- the present invention can also be applied to posture state estimation of various objects having a plurality of parts connected by joints, such as robots.
- the unevenness map is a map in which an image is divided into unevenness of the surface of the subject shown in the image.
- FIG. 20 is a block diagram showing a main configuration of a posture state estimation apparatus according to Embodiment 3 of the present invention, which corresponds to posture state estimation apparatus 100 in FIG. 1 of Embodiment 1.
- the same components as in FIG. 1 will be assigned the same reference numerals as in FIG.
- the posture state estimation apparatus 100b of FIG. 20 further includes an unevenness map estimation unit 145b in addition to the configuration of FIG.
- the unevenness map estimation unit 145 b generates an unevenness map of each part. More specifically, the unevenness map estimation unit 145 b receives the estimated likelihood map and estimated image data from the likelihood map generation unit 150. Then, the unevenness map estimation unit 145 b generates an unevenness map based on the input information, and outputs the generated unevenness map to the posture state estimation unit 160 b. Details of the method of generating the unevenness map will be described later. Hereinafter, the unevenness map generated from the estimated image data is referred to as "estimated unevenness map".
- posture state estimation unit 160b holds in advance, for each posture state, a concavo-convex map (hereinafter referred to as “learned concavo-convex map”) learned from the reference model in the posture state.
- the posture state estimation unit 160b estimates the posture state of the subject based on the degree of coincidence between the estimated unevenness map and the learning unevenness map in addition to the degree of coincidence between the estimated likelihood map and the learning likelihood map. That is, in addition to the operation of the first embodiment, posture state estimation unit 160b further performs matching between the estimated unevenness map and the learning unevenness map.
- the unevenness map estimation unit 145 b estimates the orientation of the surface of the part from the brightness information of the part on the image.
- the brightness is, for example, the level of brightness
- the brightness information is information indicating the brightness or the level of brightness.
- FIG. 21 is a diagram for explaining the relationship between the posture of a person and the brightness of each part.
- first posture shown in FIG. 21A and the second posture shown in FIG. 21B are different postures, it is assumed that the silhouettes when viewed from the front are the same as shown in FIG. 21C. In this case, it is not possible to correctly estimate whether the posture of the target person is the first posture or the second posture only from the area information including the edge of the front image.
- the length of the left leg is shorter than the length of the left arm, so it can be inferred that the left knee may be bent.
- the left knee may be bent or extended.
- the posture state estimation device 100b estimates the region of the region using brightness information in addition to the region information, since it corresponds to the posture in which the joint position (division of each region) can not be identified only from such region information. Do.
- FIG. 21D is a diagram showing the brightness of each portion when the first posture is photographed from the front when natural light from above is used as a light source, as a density.
- FIG. 21E is a diagram showing the brightness of each portion when the second posture is photographed from the front, when natural light from above is used as a light source, as a density.
- the higher the density the lower the brightness (darker).
- five levels of "-2, -1, 0, 1, 2" are defined as the brightness in order from the darker level.
- the level "0" is, for example, the level of brightness of the surface perpendicular to the ground.
- the brightness level of each area of the image becomes brighter as the area of the surface facing upward, and conversely, becomes darker as the area of the surface pointing downward.
- the area of the head, torso, and left arm is at level “0”, and the area of the right leg is somewhat dark It becomes level "-1".
- the upper right arm In the first posture, the upper right arm is lowered vertically and the right forearm extends forward, so the area of the upper right arm is at level "0" and the area of the right forearm is as shown in D of FIG. It becomes level "2".
- the upper right arm In the second posture, the upper right arm is pulled backward and the right forearm is directed downward, so that the area of the upper right arm becomes level "-2" as shown in FIG. 21E. The area is level "2".
- the entire left foot extends forward, so the regions above the left knee and below the left knee become level “1” as shown in FIG. 21D.
- the area of the left thigh becomes level “2” and the left knee The area is level "-2”.
- each part can be regarded as a plane of the same brightness. Therefore, the position of the part can be estimated from the brightness information of the part on the image.
- FIG. 22 steps in common with FIG. 4 of the first embodiment are assigned the same step numbers as in FIG. 4 and descriptions thereof will be omitted.
- the unevenness map estimation unit 145 b performs estimated unevenness map generation processing.
- the estimated unevenness map generation process is a process of generating an estimated unevenness map from the estimated image data acquired in S4200 and the estimated likelihood map generated in S4500.
- FIG. 23 is a diagram showing a processing flow of the unevenness map generation processing (step S4510 b in FIG. 22).
- pk is binary information, and the value of pk takes, for example, either 0 indicating that there is no possibility of being part k or 1 indicating that there is a possibility of part k.
- the unevenness map estimation unit 145b selects one part to be processed. For example, when the right arm is to be subjected to the unevenness map generation processing, the unevenness map estimation unit 145b first selects the right forearm that is most distant from the main part.
- the unevenness map estimation unit 145b acquires a region of the part selected in S6100b (hereinafter referred to as a part likelihood region) from the estimated likelihood map generated in S4500.
- a region likelihood region pixels in which the likelihood of the right forearm on the estimated likelihood map exceeds a predetermined threshold value are extracted as a region likelihood region of the right forearm.
- the unevenness map estimation unit 145b extracts brightness information of the region likelihood region extracted in S6200b from the estimated image data acquired in S4200.
- the brightness information may be extracted, for example, by converting only the luminance (brightness of pixels) from the RGB value of each pixel constituting the estimated image data into a gray scale (black and white gradation) image. it can.
- the unevenness map estimation unit 145b groups the brightness information of the part likelihood area obtained in S6300b using the brightness threshold.
- the unevenness map estimation unit 145 b may set the threshold of brightness as a preset fixed value or may set it dynamically.
- an example of a method of dynamically setting the threshold will be described.
- FIG. 24 is a diagram for describing a method of area classification using the body constraint of the right forearm. In order to simplify the explanation, it will be described that the torso has only the right arm.
- a head-shoulder area and a torso area 501b connected thereto are estimated based on the estimated right shoulder position 500b.
- the region where the right upper arm and the right forearm can exist is as in the region 502b
- the region where only the right forearm can exist is as the region 503b.
- the regions 502b and 503b can be calculated from, for example, the region-region correspondence table shown in FIG.
- the unevenness map estimation unit 145b first extracts the luminance value (brightness information) of the pixels present in the region from the region 503b in which only the right forearm can exist among the region likelihood regions for the right forearm.
- the unevenness map estimation unit 145 b removes n pieces from the smallest ones and n pieces from the largest ones from the extracted data of the luminance value. Furthermore, the unevenness map estimation unit 145 b treats the minimum value and the maximum value of the data (the number of data is m ⁇ 2 n) after removing the 2n pieces of data as the threshold value of the brightness information of the right forearm The upper limit value and the lower limit value of the range of luminance values).
- a is a value set in advance.
- the unevenness map estimation unit 145 b determines that the right forearm, for example, of the unevenness vector Oij of the pixel that satisfies this threshold (that is, within the range of the brightness value treated as the right forearm) in the part likelihood region of the right forearm. Is set to a value (for example, 1) indicating that it may be the right forearm.
- the unevenness map estimation unit 145 b sets the threshold value of the luminance value using only the brightness information of the part likelihood area in which only the right forearm exists due to the physical restriction. As a result, the unevenness map estimation unit 145b can specify a pixel having the brightness information of the right forearm without being affected by other parts.
- the unevenness map estimation unit 145 b extracts the luminance value (brightness information) of the pixel from the region 502 b in which only the upper right arm and the right forearm can exist in the region likelihood region of the right forearm.
- the unevenness map estimation unit 145 b deletes, from the extracted data of the brightness value, the one that fits the threshold value of the brightness information of the right forearm obtained in the previous step. Then, assuming that b% of the total number p of the remaining luminance value data is q, the unevenness map estimation unit 145 b generates q pieces from the smallest and q pieces from the largest from the extracted luminance value data. except for.
- the unevenness map estimation unit 145 b treats the minimum value and the maximum value of the data (the number of data is p ⁇ 2 q) after removing these 2 q pieces of data as the threshold value of the brightness information of the upper right arm (the right forearm The upper limit value and the lower limit value of the range of luminance values).
- the value of b is a value set in advance.
- the unevenness map estimation unit 145 b determines, for example, the upper right arm of the unevenness vector Oij of the pixel that satisfies this threshold (that is, within the range of the brightness value treated as the upper right arm) in the region likelihood region of the right forearm.
- this threshold that is, within the range of the brightness value treated as the upper right arm
- a value for example, 1 indicating that there is a possibility of being the upper right arm is set.
- the unevenness map estimation unit 145b removes the threshold value by excluding data in the range of the brightness value treated as the right forearm from the data of the brightness information of the region likelihood area in which only the right upper arm and the right forearm exist due to the physical restriction.
- the unevenness map estimation unit 145b can specify the pixel having the brightness information of the upper right arm without being influenced by the other part, and can specify the pixel having the brightness information of the upper right arm with high accuracy.
- the unevenness map estimation unit 145 b sequentially sets the threshold of the brightness information using the brightness information of the area where only the site exists, from the site distant from the main site, and the brightness information for each site Group to estimate the area.
- the unevenness map estimation unit 145b extracts, for example, luminance information of pixels present in the area 502b in which only the upper right arm and the right forearm can exist among the part likelihood areas of the right forearm and the upper right arm. A process of classifying into two groups of the forearm and the upper right arm may be performed. Then, the unevenness map estimation unit 145 b sets the above-described threshold value using, for example, Otsu's binarization. Thereby, the unevenness map estimation unit 145b can set the threshold value of the brightness information of the upper right arm and the right forearm even when there is no region likelihood area of the right forearm in the area 503b where only the right forearm can exist.
- the unevenness map estimation unit 145 b may set, for example, the same value as the right forearm with respect to the brightness information of the upper right arm. Thereby, the unevenness map estimation unit 145 b can set the brightness information of the upper right arm even when the directions of the surfaces of the upper right arm and the right forearm are similar (when extending straight).
- the unevenness map estimation unit 145 b determines whether or not all of the parts to be subjected to the unevenness map generation process have been processed. For example, in the case of generating the estimated unevenness map also for the left arm, the unevenness map estimation unit 145 b returns to S6100 b, and performs the same process as the right arm for the left arm.
- the unevenness map estimation unit 145 b outputs the generated estimated unevenness map to the posture state estimation unit 160 b.
- posture state estimation section 160 b performs matching between the learning likelihood map and the estimated likelihood map, and thereafter performs matching between the learning unevenness map and the estimated unevenness map. Then, posture state estimation section 160b determines, as in the first embodiment, whether the estimated likelihood map matches any of the learning likelihood maps.
- posture state estimation unit 160b determines the degree of coincidence of likelihood for each pixel between the estimated concavo-convex map and the learning concavo-convex map. evaluate. For example, the posture state estimation unit 160b counts the pixels having the same identifier for all the pixels, and determines that the matching degree with the estimated unevenness map is high with respect to the learning unevenness map having the largest count value. Note that, as in the case of the likelihood map, the posture state estimation unit 160b may perform the matching process after performing the scaling process on the image area when the sizes are different.
- the posture state estimation apparatus 100b since the posture state estimation apparatus 100b according to the present embodiment generates a concavo-convex map and uses the matching of the concavo-convex map in combination, the accuracy of posture estimation can be further improved.
- the present embodiment may be applied to posture state estimation apparatus 100a according to the second embodiment. That is, similar to the generation of the learning likelihood map, the learning unevenness map may be generated.
- the posture state estimation device and the posture state estimation method according to the present invention are useful as a posture state estimation device and a posture state estimation method capable of estimating the posture state of an object having a joint with high accuracy.
- Attitude state estimation device 110 Attitude state management unit 120 Attitude state designation unit 130 Image data acquisition unit 140 Part region estimation unit 145b Irregularity map estimation unit 150, 150a Likelihood map generation unit 160, 160b Attitude state estimation unit 200 Single-lens camera 300 information output device
Abstract
Description
本発明の実施の形態1は、本発明を、撮影された人の姿勢状態が、ユーザが指定した姿勢状態に一致しているか否かを推定する装置に適用した例である。
本発明の実施の形態2は、姿勢状態推定装置において、学習尤度マップの生成を併せて行うようにした例である。本実施の形態に係る姿勢状態推定装置は、姿勢状態の推定を行う推定フェーズ処理の他に、学習尤度マップを生成する学習フェーズ処理を行う。
図20は、本発明の実施の形態3に係る姿勢状態推定装置の要部構成を示すブロック図であり、実施の形態1の図1の姿勢状態推定装置100に対応するものである。なお、図20において、図1と共通する構成部分には、図1と同一の符号を付して説明を省略する。
110 姿勢状態管理部
120 姿勢状態指定部
130 画像データ取得部
140 部位領域推定部
145b 凹凸マップ推定部
150、150a 尤度マップ生成部
160、160b 姿勢状態推定部
200 単眼カメラ
300 情報出力装置
Claims (11)
- 関節により接続された複数の部位を有する物体を撮影した画像データに基づいて前記物体の姿勢状態の推定を行う姿勢状態推定装置であって、
前記画像データから、少なくとも2つ以上の前記部位について、各部位が位置することの尤もらしさの分布を示す尤度マップを生成する尤度マップ生成部と、
前記姿勢状態に予め対応付けられた前記尤度マップである学習尤度マップと、前記画像データに基づいて生成された前記尤度マップである推定尤度マップとの一致度が高いとき、当該学習尤度マップと対応付けられた前記姿勢状態を、前記物体の姿勢状態として推定する姿勢状態推定部と、を有する、
姿勢状態推定装置。 - 前記尤度マップは、少なくとも、前記部位が位置することの画素毎の尤度値を前記部位毎に示す情報であり、
前記姿勢状態推定部は、
対応する前記画素毎および前記部位毎の前記尤度値の一致度がより高いほど、学習尤度マップと前記推定尤度マップとの一致度がより高いと判定する、
請求項1記載の姿勢状態推定装置。 - 前記姿勢状態推定部は、
前記推定尤度マップを拡大、縮小、平行移動、もしくは回転、またはこれらの組み合わせにより変換した情報と、前記学習尤度マップとの一致度が高いとき、学習尤度マップと前記推定尤度マップとの一致度が高いと判定する、
請求項1記載の姿勢状態推定装置。 - 前記少なくとも2つの部位について、前記画像データにおける各部位の可動範囲を、その部位の部位領域として推定する部位領域推定部、を更に有し、
前記尤度マップ生成部は、
前記部位領域以外の領域については、前記部位領域に対応する前記部位が位置することの尤もらしさを低くした前記推定尤度マップを生成する、
請求項1記載の姿勢状態推定装置。 - 前記物体は人であり、
前記部位領域推定部は、
前記画像データから前記人の頭部および肩部の位置および向きを検出し、これらの位置および向きから、前記部位領域を推定する、
請求項4記載の姿勢状態推定装置。 - 前記推定の対象となる前記姿勢状態の指定を受け付ける姿勢状態指定部、を更に有し、
前記姿勢状態推定部は、
前記推定の対象として指定された前記姿勢状態に対応付けられた前記学習尤度マップと前記推定尤度マップとの一致度が高いとき、前記物体の姿勢状態が指定された前記姿勢状態である旨の通知を行う、
請求項1記載の姿勢状態推定装置。 - 前記姿勢状態指定部は、
前記学習尤度マップの生成の指示と、前記生成の対象となる前記姿勢状態の指定とを受け付け、
前記尤度マップ生成部は、
前記学習尤度マップの生成が指示されたとき、所定の画像に基づいて前記学習尤度マップを生成し、
生成された前記学習尤度マップを、指定された前記姿勢状態に対応付けて格納する姿勢状態管理部、を更に有する、
請求項6記載の姿勢状態推定装置。 - 前記姿勢状態指定部は、
2つ以上の前記部位の指定を更に受け付け、
指定された前記2つ以上の前記部位について、前記所定の画像における各部位の可動範囲を、その部位の部位領域として推定する部位領域推定部、を更に有し、
前記尤度マップ生成部は、
前記部位領域以外の領域については、前記部位領域に対応する前記部位が位置することの尤もらしさを低くした前記学習尤度マップを生成する、
請求項7記載の姿勢状態推定装置。 - 前記尤度マップ生成部は、
前記画像データに含まれる平行線に基づいて、前記尤度マップを生成し、
前記姿勢状態推定部は、
前記平行線から取得される前記推定尤度マップの主要なエッジ方向と、前記学習尤度マップの主要なエッジ方向とを用いて、一致度を算出する、
請求項1記載の姿勢状態推定装置。 - 前記画像データの画像における被写体の面を凹凸で区分したマップである凹凸マップを生成する凹凸マップ推定部、を更に有し、
前記姿勢状態推定部は、
更に、前記姿勢状態に予め対応付けられた前記凹凸マップである学習凹凸マップと、前記画像データに基づいて生成された前記凹凸マップである推定尤度マップとの一致度に基づいて、前記物体の姿勢状態を推定する、
請求項1記載の姿勢状態推定装置。 - 関節により接続された複数の部位を有する物体を撮影した画像データに基づいて前記物体の姿勢状態の推定を行う姿勢状態推定方法であって、
前記画像データから、少なくとも2つ以上の前記部位について、各部位が位置することの尤もらしさの分布を示す尤度マップを生成するステップと、
前記姿勢状態に予め対応付けられた前記尤度マップである学習尤度マップと、前記画像データに基づいて生成された前記尤度マップである推定尤度マップとの一致度を判定するステップと、
前記一致度が高いとき、当該学習尤度マップと対応付けられた前記姿勢状態を、前記物体の姿勢状態として推定するステップと、を有する、
姿勢状態推定方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201180054681.4A CN103221977B (zh) | 2010-12-09 | 2011-11-22 | 姿势状态估计装置及姿势状态估计方法 |
JP2012547682A JP5837508B2 (ja) | 2010-12-09 | 2011-11-22 | 姿勢状態推定装置および姿勢状態推定方法 |
US13/991,021 US9262674B2 (en) | 2010-12-09 | 2011-11-22 | Orientation state estimation device and orientation state estimation method |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2010274673 | 2010-12-09 | ||
JP2010-274673 | 2010-12-09 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012077287A1 true WO2012077287A1 (ja) | 2012-06-14 |
Family
ID=46206802
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2011/006499 WO2012077287A1 (ja) | 2010-12-09 | 2011-11-22 | 姿勢状態推定装置および姿勢状態推定方法 |
Country Status (4)
Country | Link |
---|---|
US (1) | US9262674B2 (ja) |
JP (1) | JP5837508B2 (ja) |
CN (1) | CN103221977B (ja) |
WO (1) | WO2012077287A1 (ja) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104487999A (zh) * | 2012-07-26 | 2015-04-01 | 松下电器产业株式会社 | 姿势估计装置、姿势估计方法以及姿势估计程序 |
US20190318500A1 (en) * | 2018-04-17 | 2019-10-17 | Fujitsu Limited | Method for estimating body orientation |
US10796186B2 (en) | 2017-07-03 | 2020-10-06 | Fujitsu Limited | Part recognition method, information processing apparatus, and imaging control system |
JP2020201558A (ja) * | 2019-06-06 | 2020-12-17 | Kddi株式会社 | 姿勢推定装置、方法およびプログラム |
US10902056B2 (en) | 2013-11-06 | 2021-01-26 | Samsung Electronics Co., Ltd. | Method and apparatus for processing image |
JP7419993B2 (ja) | 2020-07-02 | 2024-01-23 | コニカミノルタ株式会社 | 信頼度推定プログラム、信頼度推定方法、および信頼度推定装置 |
Families Citing this family (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013008935A1 (ja) * | 2011-07-13 | 2013-01-17 | 日東電工株式会社 | オンデマンド型電力制御システム、オンデマンド型電力制御システムプログラム及びそのプログラムを記録したコンピュータ読み取り可能な記録媒体 |
JP5818773B2 (ja) * | 2012-11-22 | 2015-11-18 | キヤノン株式会社 | 画像処理装置、画像処理方法及びプログラム |
US10026002B2 (en) * | 2013-10-01 | 2018-07-17 | Nec Corporation | Object detection apparatus, method for detecting object, and learning apparatus |
WO2015064144A1 (ja) * | 2013-10-30 | 2015-05-07 | オリンパスイメージング株式会社 | 撮像装置、撮像方法およびプログラム |
CA2958888C (en) * | 2014-08-28 | 2023-02-28 | Retailmenot, Inc. | Reducing the search space for recognition of objects in an image based on wireless signals |
US9552070B2 (en) * | 2014-09-23 | 2017-01-24 | Microsoft Technology Licensing, Llc | Tracking hand/body pose |
US10121260B2 (en) | 2014-10-30 | 2018-11-06 | Panasonic Intellectual Property Management Co., Ltd. | Orientation estimation method and orientation estimation device |
CN105069413B (zh) * | 2015-07-27 | 2018-04-06 | 电子科技大学 | 一种基于深度卷积神经网络的人体姿势识别方法 |
WO2017203986A1 (ja) * | 2016-05-25 | 2017-11-30 | ソニー株式会社 | 情報処理装置、情報処理方法、プログラム、および見守りシステム |
JP6824838B2 (ja) * | 2017-07-07 | 2021-02-03 | 株式会社日立製作所 | 作業データ管理システム及び作業データ管理方法 |
JP7093075B2 (ja) * | 2018-04-09 | 2022-06-29 | 東芝エネルギーシステムズ株式会社 | 医用画像処理装置、医用画像処理方法、およびプログラム |
JP2019200560A (ja) * | 2018-05-16 | 2019-11-21 | パナソニックIpマネジメント株式会社 | 作業分析装置および作業分析方法 |
JP6973298B2 (ja) * | 2018-05-31 | 2021-11-24 | トヨタ自動車株式会社 | 物体監視装置 |
CN108921047B (zh) * | 2018-06-12 | 2021-11-26 | 江西理工大学 | 一种基于跨层融合的多模型投票均值动作识别方法 |
JP7010778B2 (ja) * | 2018-06-29 | 2022-01-26 | 国立大学法人東海国立大学機構 | 観測位置推定装置、その推定方法、及びプログラム |
JP2020129018A (ja) * | 2019-02-07 | 2020-08-27 | 株式会社日立製作所 | 動作評価システムおよび方法 |
CN111679248B (zh) * | 2020-05-15 | 2023-04-21 | 黑龙江工程学院 | 一种基于海底水平l型阵列的目标方位和距离联合稀疏重构定位方法 |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08214289A (ja) * | 1994-12-06 | 1996-08-20 | Olympus Optical Co Ltd | 時系列画像解析装置及びその解析方法 |
JP2005242759A (ja) * | 2004-02-27 | 2005-09-08 | National Institute Of Information & Communication Technology | 行動・意図推定システム、行動・意図推定方法、行動・意図推定プログラムおよびそのプログラムを記録したコンピュータ読み取り可能な記録媒体 |
JP2005351814A (ja) * | 2004-06-11 | 2005-12-22 | Konica Minolta Holdings Inc | 検出装置および検出方法 |
JP2007004732A (ja) * | 2005-06-27 | 2007-01-11 | Matsushita Electric Ind Co Ltd | 画像生成装置及び画像生成方法 |
JP2007310707A (ja) * | 2006-05-19 | 2007-11-29 | Toshiba Corp | 姿勢推定装置及びその方法 |
JP2009288917A (ja) * | 2008-05-28 | 2009-12-10 | Sony Corp | 情報処理装置、情報処理方法、およびプログラム |
JP2010514064A (ja) * | 2006-12-21 | 2010-04-30 | 本田技研工業株式会社 | ラベル付けを用いた人体姿勢の推定および追跡 |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4492036B2 (ja) | 2003-04-28 | 2010-06-30 | ソニー株式会社 | 画像認識装置及び方法、並びにロボット装置 |
JP3775683B2 (ja) * | 2003-08-21 | 2006-05-17 | 松下電器産業株式会社 | 人物検出装置および人物検出方法 |
US7542858B2 (en) * | 2005-06-03 | 2009-06-02 | Lsi Corporation | Simulated battery logic testing device |
JP4710426B2 (ja) * | 2005-06-14 | 2011-06-29 | 富士ゼロックス株式会社 | 画像処理装置、画像処理方法及び画像処理プログラム |
JP4148281B2 (ja) | 2006-06-19 | 2008-09-10 | ソニー株式会社 | モーションキャプチャ装置及びモーションキャプチャ方法、並びにモーションキャプチャプログラム |
JP5215740B2 (ja) * | 2008-06-09 | 2013-06-19 | 株式会社日立製作所 | 移動ロボットシステム |
KR101626065B1 (ko) * | 2009-10-13 | 2016-05-31 | 삼성전자주식회사 | 마커리스 모션 캡쳐링 장치 및 방법 |
-
2011
- 2011-11-22 JP JP2012547682A patent/JP5837508B2/ja not_active Expired - Fee Related
- 2011-11-22 US US13/991,021 patent/US9262674B2/en not_active Expired - Fee Related
- 2011-11-22 WO PCT/JP2011/006499 patent/WO2012077287A1/ja active Application Filing
- 2011-11-22 CN CN201180054681.4A patent/CN103221977B/zh active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08214289A (ja) * | 1994-12-06 | 1996-08-20 | Olympus Optical Co Ltd | 時系列画像解析装置及びその解析方法 |
JP2005242759A (ja) * | 2004-02-27 | 2005-09-08 | National Institute Of Information & Communication Technology | 行動・意図推定システム、行動・意図推定方法、行動・意図推定プログラムおよびそのプログラムを記録したコンピュータ読み取り可能な記録媒体 |
JP2005351814A (ja) * | 2004-06-11 | 2005-12-22 | Konica Minolta Holdings Inc | 検出装置および検出方法 |
JP2007004732A (ja) * | 2005-06-27 | 2007-01-11 | Matsushita Electric Ind Co Ltd | 画像生成装置及び画像生成方法 |
JP2007310707A (ja) * | 2006-05-19 | 2007-11-29 | Toshiba Corp | 姿勢推定装置及びその方法 |
JP2010514064A (ja) * | 2006-12-21 | 2010-04-30 | 本田技研工業株式会社 | ラベル付けを用いた人体姿勢の推定および追跡 |
JP2009288917A (ja) * | 2008-05-28 | 2009-12-10 | Sony Corp | 情報処理装置、情報処理方法、およびプログラム |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104487999A (zh) * | 2012-07-26 | 2015-04-01 | 松下电器产业株式会社 | 姿势估计装置、姿势估计方法以及姿势估计程序 |
US10902056B2 (en) | 2013-11-06 | 2021-01-26 | Samsung Electronics Co., Ltd. | Method and apparatus for processing image |
US10796186B2 (en) | 2017-07-03 | 2020-10-06 | Fujitsu Limited | Part recognition method, information processing apparatus, and imaging control system |
US20190318500A1 (en) * | 2018-04-17 | 2019-10-17 | Fujitsu Limited | Method for estimating body orientation |
US10909718B2 (en) * | 2018-04-17 | 2021-02-02 | Fujitsu Limited | Method for estimating body orientation |
JP2020201558A (ja) * | 2019-06-06 | 2020-12-17 | Kddi株式会社 | 姿勢推定装置、方法およびプログラム |
JP7054392B2 (ja) | 2019-06-06 | 2022-04-13 | Kddi株式会社 | 姿勢推定装置、方法およびプログラム |
JP7419993B2 (ja) | 2020-07-02 | 2024-01-23 | コニカミノルタ株式会社 | 信頼度推定プログラム、信頼度推定方法、および信頼度推定装置 |
Also Published As
Publication number | Publication date |
---|---|
JP5837508B2 (ja) | 2015-12-24 |
US20130301882A1 (en) | 2013-11-14 |
JPWO2012077287A1 (ja) | 2014-05-19 |
US9262674B2 (en) | 2016-02-16 |
CN103221977B (zh) | 2015-10-14 |
CN103221977A (zh) | 2013-07-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2012077287A1 (ja) | 姿勢状態推定装置および姿勢状態推定方法 | |
JP5873442B2 (ja) | 物体検出装置および物体検出方法 | |
JP5715833B2 (ja) | 姿勢状態推定装置および姿勢状態推定方法 | |
Xia et al. | Human detection using depth information by kinect | |
US9480417B2 (en) | Posture estimation device, posture estimation system, and posture estimation method | |
JP5820366B2 (ja) | 姿勢推定装置及び姿勢推定方法 | |
Pala et al. | Multimodal person reidentification using RGB-D cameras | |
JP6125188B2 (ja) | 映像処理方法及び装置 | |
US9330307B2 (en) | Learning based estimation of hand and finger pose | |
US8824781B2 (en) | Learning-based pose estimation from depth maps | |
JP7094702B2 (ja) | 画像処理装置及びその方法、プログラム | |
US9117138B2 (en) | Method and apparatus for object positioning by using depth images | |
JP2012123667A (ja) | 姿勢推定装置および姿勢推定方法 | |
JP2015184054A (ja) | 同定装置、方法及びプログラム | |
CN108694348B (zh) | 一种基于自然特征的跟踪注册方法及装置 | |
Hachiuma et al. | Recognition and pose estimation of primitive shapes from depth images for spatial augmented reality | |
Schulz et al. | Pedestrian recognition from a moving catadioptric camera | |
Iwashita et al. | Person Identification using Shadow Analysis. | |
JP6085160B2 (ja) | 物体認識システム | |
CN115829829A (zh) | 图像处理方法、电子设备、存储介质及程序产品 | |
Prous et al. | Laser Range Data Registration using Spin Images. | |
Ghassemian | Track,” IEEE International Conference on Computer Vision & Pattern Recognition, pp. 593–600, 1994.[37] JR Shewchuk,“Delaunay Refinement Mesh Generation,” Ph. D. Thesis | |
Soundararajan et al. | Mesh Generation,” Ph. D. Thesis, Carnegie Mellon University, pp. 593–600, 1997. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11847499 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2012547682 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13991021 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 11847499 Country of ref document: EP Kind code of ref document: A1 |