IMAGE SELECTION DEVICE, IMAGE SELECTION METHOD, AND PROGRAM
The image processing device improves image classification and search accuracy by employing skeletal structure estimation and normalization techniques to handle changes in orientation and body shape, enabling flexible recognition and search of postures.
Patent Information
- Application Number
- JP2022556719
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2020-10-13
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2040-10-13
AI Technical Summary
Existing image classification and selection technologies struggle with accuracy when classifying or selecting images based on posture information, particularly when the desired state of a person is unspecified or partially hidden, making it difficult to flexibly search and classify desired postures.
An image processing device that utilizes skeletal structure estimation to detect two-dimensional skeletal structures from images, calculates feature amounts, and performs classification and search processes based on the similarity of these features, incorporating methods to normalize and robustly handle changes in orientation and body shape.
Enhances the accuracy of image classification and search by enabling flexible recognition of postures, allowing for classification and search of desired postures even when unspecified or partially hidden, using normalized feature amounts that are robust to changes in orientation and body shape.
Smart Images

Figure 0007589744000003 
Figure 0007589744000004 
Figure 0007589744000005
Abstract
Description
[Technical field]
[0001] The present invention relates to an image selection device, an image selection method, and a program. [Background technology]
[0002] In recent years, in surveillance systems and the like, techniques for detecting and searching for states such as posture and behavior of a person from images of a surveillance camera are used. For example, Patent Documents 1 and 2 are known as related techniques. Patent Document 1 discloses a technique for searching for a similar posture of a person based on key joints of the person's head, limbs, etc. contained in a depth image. Patent Document 2 discloses a technique for searching for a similar image using posture information such as tilt added to an image, which is not related to the posture of a person. In addition, Non-Patent Document 1 is known as a technique related to human skeleton estimation.
[0003] Patent Document 3 describes detecting skeletal information of a person from an image and analyzing the person's movements using the skeletal information. Patent Document 4 describes searching for images using posture information of a person as a search query. In Patent Document 4, the posture information is calculated by dividing the feature points and features. Point to point Patent Document 5 describes that when selecting at least one image from a plurality of images, the evaluation results of a plurality of evaluation items are used, and that the evaluation items used include image brightness, the number of faces included in the image, the positions of the faces, the orientation of the faces, the size of the faces, the facial expressions, the importance of the event at the time of image capture, and the number of similar images. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Special Publication No. 2014-522035 [Patent Document 2] JP 2006-260405 A [Patent Document 3] JP 2019-200241 A [Patent Document 4] JP 2019-091138 A [Patent Document 5] JP 2007-080014 A [Non-patent literature]
[0005] [Non-Patent Document 1] Zhe Cao, Tomas Simon, Shih-En Wei, Yaser Sheikh, "Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields", The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, P. 7291-7299 Summary of the Invention [Problem to be solved by the invention]
[0006] Pose information may be used to classify a plurality of images or to select an image. The present inventor has considered how to improve the accuracy of such classification or selection. One example of the objective of the present invention is to improve the accuracy of image classification or selection. [Means for solving the problem]
[0007] According to the present invention, there is provided an information generating means for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection means for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; An image selection device is provided, comprising:
[0008] According to the present invention, a computer an information generation process for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection process for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; An image selection method is provided that:
[0009] According to the present invention, a computer includes: an information generating function for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection function that classifies the plurality of target images using the posture information and the other information, or selects at least one target image from the plurality of target images; A program will be provided to help students develop these skills. Effect of the Invention
[0010] According to the present invention, it is possible to improve accuracy when classifying or selecting images. [Brief description of the drawings]
[0011] The above objects, as well as other objects, features and advantages, will become more apparent from the following preferred embodiments and the accompanying drawings.
[0012] [Figure 1] 1 is a configuration diagram showing an overview of an image processing device according to an embodiment; [Diagram 2] 1 is a configuration diagram showing a configuration of an image processing device according to a first embodiment. [Diagram 3] 4 is a flowchart showing an image processing method according to the first embodiment. [Figure 4] 4 is a flowchart showing a classification method according to the first embodiment. [Diagram 5] 4 is a flowchart showing a search method according to the first embodiment. [Figure 6]FIG. 4 is a diagram showing an example of detection of a skeletal structure according to the first embodiment. [Figure 7] FIG. 2 is a diagram showing a human body model according to the first embodiment. [Figure 8] FIG. 4 is a diagram showing an example of detection of a skeletal structure according to the first embodiment. [Figure 9] FIG. 4 is a diagram showing an example of detection of a skeletal structure according to the first embodiment. [Figure 10] FIG. 4 is a diagram showing an example of detection of a skeletal structure according to the first embodiment. [Figure 11] 4 is a graph showing a specific example of the classification method according to the first embodiment. [Figure 12] FIG. 4 is a diagram showing a display example of a classification result according to the first embodiment. [Figure 13] FIG. 2 is a diagram for explaining a search method according to the first embodiment. [Figure 14] FIG. 2 is a diagram for explaining a search method according to the first embodiment. [Figure 15] FIG. 2 is a diagram for explaining a search method according to the first embodiment. [Figure 16] FIG. 2 is a diagram for explaining a search method according to the first embodiment. [Figure 17] FIG. 4 is a diagram showing a display example of search results according to the first embodiment. [Figure 18] FIG. 11 is a configuration diagram showing a configuration of an image processing device according to a second embodiment. [Figure 19] 10 is a flowchart showing an image processing method according to the second embodiment. [Figure 20] 13 is a flowchart showing a specific example 1 of a height pixel number calculation method according to the second embodiment. [Figure 21] 13 is a flowchart showing a specific example 2 of the height pixel number calculation method according to the second embodiment. [Figure 22] 13 is a flowchart showing a specific example 3 of a height pixel number calculation method according to the second embodiment. [Diagram 23] 11 is a flowchart showing a normalization method according to the second embodiment. [Figure 24] FIG. 11 is a diagram showing a human body model according to a second embodiment. [Diagram 25] FIG. 11 is a diagram showing an example of detection of a skeletal structure according to the second embodiment. [Figure 26] FIG. 11 is a diagram showing an example of detection of a skeletal structure according to the second embodiment. [Figure 27] FIG. 11 is a diagram showing an example of detection of a skeletal structure according to the second embodiment. [Figure 28] FIG. 11 is a diagram showing a human body model according to a second embodiment. [Figure 29] FIG. 11 is a diagram showing an example of detection of a skeletal structure according to the second embodiment. [Diagram 30] 13 is a histogram for explaining a height pixel number calculation method according to the second embodiment. [Diagram 31] FIG. 11 is a diagram showing an example of detection of a skeletal structure according to the second embodiment. [Diagram 32] FIG. 11 is a diagram showing a three-dimensional human body model according to the second embodiment. [Diagram 33] FIG. 11 is a diagram for explaining a height pixel number calculation method according to the second embodiment. [Diagram 34] FIG. 11 is a diagram for explaining a height pixel number calculation method according to the second embodiment. [Diagram 35] FIG. 11 is a diagram for explaining a height pixel number calculation method according to the second embodiment. [Diagram 36] FIG. 11 is a diagram for explaining a normalization method according to the second embodiment. [Figure 37] FIG. 11 is a diagram for explaining a normalization method according to the second embodiment. [Figure 38] FIG. 11 is a diagram for explaining a normalization method according to the second embodiment. [Figure 39] FIG. 2 is a diagram illustrating an example of a hardware configuration of an image processing apparatus. [Diagram 40] FIG. 13 is a diagram illustrating a first example of a functional configuration of a search section related to a search method 6. [Diagram 41] 13 is a diagram showing an example of a screen that an image selection unit causes to be displayed on a user's terminal or a display unit; FIG. [Diagram 42] 41 is a flowchart showing an example of a process performed by a search unit shown in FIG. 40. [Diagram 43] FIG. 41 is a diagram showing a modified example of FIG. 40. [Diagram 44] 44 is a flowchart showing an example of the operation of a search unit shown in FIG. 43. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0013] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In all the drawings, the same components are given the same reference numerals and the description will be omitted as appropriate.
[0014] (Considerations leading to the embodiment) In recent years, image recognition technology that utilizes machine learning such as deep learning has been applied to various systems. For example, it is being applied to surveillance systems that monitor using images from surveillance cameras. By utilizing machine learning in surveillance systems, it is becoming possible to grasp to some extent the posture, behavior, and other conditions of people from images.
[0015] However, such related technologies may not always be able to grasp the state of a person desired by a user on demand. For example, the state of a person that a user wants to search and grasp may be specified in advance, or may not be specifically specified, such as an unknown state. In such cases, the user may not be able to specify in detail the state of the person that he or she wants to search for. Furthermore, a search or the like may not be possible if part of the person's body is hidden. In the related technologies, the state of a person can only be searched for based on specific search criteria, making it difficult to flexibly search and classify the state of a desired person.
[0016] Therefore, the inventors have considered a method of using a skeleton estimation technique such as that described in Non-Patent Document 1 in order to recognize the state of a person desired by a user from an image on demand. In related skeleton estimation techniques such as OpenPose disclosed in Non-Patent Document 1, a person's skeleton is estimated by learning various patterns of image data with correct answers. In the following embodiment, such a skeleton estimation technique is utilized to enable flexible recognition of a person's state.
[0017] Note that the skeletal structure estimated by a skeletal estimation technique such as OpenPose is composed of "key points" which are characteristic points such as joints, and "bones (bone links)" which indicate the links between the key points. For this reason, in the following embodiments, the skeletal structure will be described using the terms "key points" and "bones", but unless otherwise specified, "key points" correspond to the "joints" of a person, and "bones" correspond to the "bones" of a person.
[0018] (Outline of the embodiment) FIG. 1 shows an overview of an image processing device 10 according to an embodiment. As shown in FIG. 1, the image processing device 10 includes a skeleton detection unit 11, a feature calculation unit 12, and a recognition unit 13. The skeleton detection unit 11 detects two-dimensional skeleton structures of a plurality of people based on a two-dimensional image acquired from a camera or the like. The feature calculation unit 12 calculates feature amounts of the plurality of two-dimensional skeleton structures detected by the skeleton detection unit 11. The recognition unit 13 performs recognition processing of the states of the plurality of people based on the similarity of the plurality of feature amounts calculated by the feature calculation unit 12. The recognition processing is classification processing and search processing (selection processing) of the states of the people. Therefore, the image processing device 10 also functions as an image selection device.
[0019] In this manner, in the embodiment, a two-dimensional skeletal structure of a person is detected from a two-dimensional image, and the state of the person is classified and detected based on the feature amount calculated from the two-dimensional skeletal structure. cable By performing the above recognition processing, the state of a desired person can be flexibly recognized.
[0020] (Embodiment 1) Hereinafter, embodiment 1 will be described with reference to the drawings. FIG. 2 shows the configuration of an image processing device 100 according to this embodiment. The image processing device 100 constitutes an image processing system 1 together with a camera 200 and a database (DB) 110. The image processing system 1 including the image processing device 100 is a system that classifies and searches for states such as posture and behavior of a person based on the skeletal structure of the person estimated from an image. The image processing device 100 also functions as an image selection device.
[0021] The camera 200 is an imaging unit such as a surveillance camera that generates a two-dimensional image. The camera 200 is installed at a predetermined location and captures an image of a person or the like in an imaging area from the installation location. The camera 200 is directly connected to the image processing device 100 so as to be able to output the captured image (video) to the image processing device 100, or is connected via a network or the like. The camera 200 may be provided inside the image processing device 100.
[0022] The database 110 is a database that stores information (data) necessary for processing by the image processing device 100, processing results, and the like. The database 110 stores images acquired by the image acquisition unit 101, detection results by the skeletal structure detection unit 102, data for machine learning, features calculated by the feature calculation unit 103, classification results by the classification unit 104, search results by the search unit 105, and the like. The database 110 is directly connected to the image processing device 100 as necessary so as to be able to input and output data, or is connected via a network or the like. The database 110 may be provided inside the image processing device 100 as a non-volatile memory such as a flash memory, a hard disk device, or the like.
[0023] As shown in FIG. 2, the image processing device 100 includes an image acquisition unit 101, a skeletal structure detection unit 102, a feature amount calculation unit 103, a classification unit 104, a search unit 105, an input unit 106, and a display unit 107. The configuration of each unit (block) is an example, and may be configured with other units as long as the method (operation) described later is possible. The image processing device 100 is realized by a computer device such as a personal computer or a server that executes a program, but may be realized by one device or by multiple devices on a network. For example, the input unit 106, the display unit 107, etc. may be external devices. The image processing device 100 may include both the classification unit 104 and the search unit 105, or only one of them. Both or one of the classification unit 104 and the search unit 105 is a recognition unit that performs recognition processing of a person's state.
[0024] The image acquiring unit 101 acquires a two-dimensional image including a person captured by the camera 200. The image acquiring unit 101 acquires, for example, an image including a person (a video including a plurality of images) captured by the camera 200 during a predetermined monitoring period. Note that the image may not be acquired from the camera 200, but may be acquired from a database 110 or the like that includes a person that has been prepared in advance.
[0025] The skeletal structure detection unit 102 detects the two-dimensional skeletal structure of a person in an image based on the acquired two-dimensional image. The skeletal structure detection unit 102 detects the skeletal structure of all people recognized in the acquired image. The skeletal structure detection unit 102 detects the skeletal structure of a person based on the features of the recognized person's joints and the like using a skeletal estimation technology using machine learning. The skeletal structure detection unit 102 uses a skeletal estimation technology such as OpenPose in Non-Patent Document 1, for example.
[0026] The feature amount calculation unit 103 calculates the feature amount of the detected two-dimensional skeletal structure, and stores the calculated feature amount in the database 110 in association with the image to be processed. The feature amount of the skeletal structure indicates the feature amount of a person, and is an element for classifying and searching the state of a person based on the person's skeleton. Usually, this feature amount includes a plurality of parameters (for example, classification elements described later). The feature amount may be the feature amount of the entire skeletal structure, may be the feature amount of a part of the skeletal structure, or may include a plurality of feature amounts like each part of the skeletal structure. The method for calculating the feature amount may be any method such as machine learning or normalization, and a minimum value or a maximum value may be obtained as normalization. As an example, the feature amount is a feature amount obtained by machine learning the skeletal structure, a size of the skeletal structure from the head to the foot on the image, etc. The size of the skeletal structure is the vertical height or area of a skeletal region including the skeletal structure on the image. The vertical direction (height direction or vertical direction) is the vertical direction (Y-axis direction) in the image, for example, a direction perpendicular to the ground (reference surface). Moreover, the left-right direction (horizontal direction) is the left-right direction in the image (X-axis direction), for example, a direction parallel to the ground.
[0027] In order to perform classification and search as desired by the user, it is preferable to use features that are robust to classification and search processing. For example, if the user desires classification and search that are not dependent on the orientation or body type of a person, features that are robust to the orientation or body type of a person may be used. By learning the skeletons of people facing in various directions in the same posture or the skeletons of people with various body types in the same posture, or by extracting features only in the up-down direction of the skeleton, features that are not dependent on the orientation or body type of a person can be obtained.
[0028] The classification unit 104 classifies (clusters) the multiple skeletal structures stored in the database 110 based on the similarity of the feature amounts of the skeletal structures. It can also be said that the classification unit 104 classifies the states of multiple people based on the feature amounts of the skeletal structures as a recognition process of the state of the person. The similarity is the distance between the feature amounts of the skeletal structures. The classification unit 104 may classify based on the similarity of the feature amounts of the entire skeletal structures, may classify based on the similarity of the feature amounts of a part of the skeletal structures, or may classify based on the similarity of the feature amounts of a first part (e.g., both hands) and a second part (e.g., both feet) of the skeletal structures. Note that the posture of the person may be classified based on the feature amounts of the skeletal structures of the person in each image, or the behavior of the person may be classified based on the change in the feature amounts of the skeletal structures of the person in multiple images that are consecutive in time series. That is, the classification unit 104 can classify the state of the person, including the posture and behavior of the person, based on the feature amounts of the skeletal structures. For example, the classification unit 104 classifies multiple skeletal structures in multiple images captured during a predetermined monitoring period. The classification unit 104 obtains the similarity between the feature amounts to be classified, and classifies the skeletal structures having high similarity into the same cluster (a group of similar postures). As with the search, the classification conditions may be specified by the user. The classification unit 104 stores the classification results of the skeletal structures in the database 110, and displays them on the display unit 107.
[0029] The search unit 105 searches for a skeletal structure having a high similarity to the feature amount of the search query (query state) from among a plurality of skeletal structures stored in the database 110. It can also be said that the search unit 105 searches for a state of a person that meets the search criteria (query state) from among a plurality of states of a person based on the feature amount of the skeletal structure as a recognition process of the state of the person. As in classification, the similarity is the distance between the feature amounts of the skeletal structures. The search unit 105 may search based on the similarity of the feature amount of the entire skeletal structure, or may search based on the similarity of the feature amount of a part of the skeletal structure, or may search based on the similarity of the feature amount of a first part (e.g., both hands) and a second part (e.g., both feet) of the skeletal structure. Note that the posture of a person may be searched based on the feature amount of the skeletal structure of the person in each image, or the behavior of a person may be searched based on the change in the feature amount of the skeletal structure of the person in a plurality of images that are consecutive in time series. That is, the search unit 105 can search for a state of a person including the posture and behavior of the person based on the feature amount of the skeletal structure. For example, the search unit 105 searches for the feature amounts of a plurality of skeletal structures in a plurality of images captured during a predetermined monitoring period as the search target, similar to the classification target. In addition, a skeletal structure (posture) specified by the user from among the classification results displayed by the classification unit 104 is used as a search query (search key). The search query may be selected from a plurality of unclassified skeletal structures, not limited to the classification results, or the user may input a skeletal structure to be used as a search query. The search unit 105 searches for feature quantities that are highly similar to the feature quantities of the skeletal structure of the search query from among the feature quantities to be searched. The search unit 105 stores the search results of the feature quantities in the database 110 and displays them on the display unit 107.
[0030] The input unit 106 is an input interface that acquires information input by a user who operates the image processing device 100. For example, the user is a monitor who monitors a person in a suspicious state from an image of a surveillance camera. The input unit 106 is, for example, a GUI (Graphical User Interface), and information according to a user's operation is input from an input device such as a keyboard, a mouse, or a touch panel. For example, the input unit 106 accepts a skeletal structure of a specified person from among the skeletal structures (postures) classified by the classification unit 104 as a search query.
[0031] The display unit 107 is a display unit that displays the results of the operation (processing) of the image processing device 100, and is, for example, a display device such as a liquid crystal display or an organic EL (Electro Luminescence) display. The display unit 107 displays the classification results of the classification unit 104 and the search results of the search unit 105 on a GUI according to the similarity, etc.
[0032] 39 is a diagram showing an example of the hardware configuration of the image processing device 100. The image processing device 100 includes a bus 1010, a processor 1020, a memory 1030, a storage device 1040, an input / output interface 1050, and a network interface 1060.
[0033] The bus 1010 is a data transmission path for transmitting and receiving data among the processor 1020, memory 1030, storage device 1040, input / output interface 1050, and network interface 1060. However, the method of connecting the processor 1020 and the like to each other is not limited to bus connection.
[0034] The processor 1020 is implemented by a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or the like.
[0035] The memory 1030 is a main storage device realized by a RAM (Random Access Memory) or the like.
[0036] The storage device 1040 is an auxiliary storage device realized by a hard disk drive (HDD), a solid state drive (SSD), a memory card, a read only memory (ROM), or the like. The storage device 1040 stores program modules that realize each function of the image processing device 100 (for example, the image acquisition unit 101, the skeletal structure detection unit 102, the feature amount calculation unit 103, the classification unit 104, the search unit 105, and the input unit 106). The processor 1020 loads each of these program modules onto the memory 1030 and executes them, thereby realizing each function corresponding to the program module. The storage device 1040 may also function as a database 110.
[0037] The input / output interface 1050 is an interface for connecting the image processing device 100 to various input / output devices. When the database 110 is located outside the image processing device 100, the image processing device 100 may be connected to the database 110 via the input / output interface 1050.
[0038] The network interface 1060 is an interface for connecting the image processing device 100 to a network. This network is, for example, a LAN (Local Area Network) or a WAN (Wide Area Network). The network interface 1060 may connect to the network wirelessly or by wire. The image processing device 100 may communicate with the camera 200 via the network interface 1060. When the database 110 is located outside the image processing device 100, the image processing device 100 may connect to the database 110 via the network interface 1060.
[0039] Figures 3 to 5 show the operation of image processing device 100 according to this embodiment. Figure 3 shows the flow from image acquisition to search processing in image processing device 100, Figure 4 shows the flow of classification processing (S104) in Figure 3, and Figure 5 shows the flow of search processing (S105) in Figure 3.
[0040] 3, the image processing device 100 acquires an image from the camera 200 (S101). The image acquisition unit 101 acquires images of people in order to classify and search based on their skeletal structures, and stores the acquired images in the database 110. The image acquisition unit 101 acquires, for example, a plurality of images captured during a predetermined monitoring period, and performs the following processes on all people included in the plurality of images.
[0041] Next, the image processing device 100 detects the skeletal structure of the person based on the acquired image of the person (S102). Fig. 6 shows an example of detecting the skeletal structure. As shown in Fig. 6, an image acquired from a surveillance camera or the like includes multiple people, and the skeletal structure is detected for each person included in the image.
[0042] Fig. 7 shows the skeletal structure of the human body model 300 detected at this time, and Fig. 8 to Fig. 10 show examples of detected skeletal structures. The skeletal structure detection unit 102 detects the skeletal structure of the human body model (two-dimensional skeletal model) 300 as shown in Fig. 7 from a two-dimensional image using a skeletal estimation technique such as OpenPose. The human body model 300 is a two-dimensional model made up of key points such as a person's joints and bones connecting each key point.
[0043] The skeletal structure detection unit 102, for example, extracts feature points that can be key points from an image, and detects each key point of a person by referring to information obtained by machine learning the image of the key points. In the example of Fig. 7, the following key points of a person are detected: head A1, neck A2, right shoulder A31, left shoulder A32, right elbow A41, left elbow A42, right hand A51, left hand A52, right hip A61, left hip A62, right knee A71, left knee A72, right foot A81, and left foot A82. Furthermore, the bones of the person connected by these key points are detected as bones, including bone B1 connecting the head A1 and neck A2, bones B21 and B22 connecting the neck A2 to the right shoulder A31 and left shoulder A32 respectively, bones B31 and B32 connecting the right shoulder A31 and left shoulder A32 to the right elbow A41 and left elbow A42 respectively, bones B41 and B42 connecting the right elbow A41 and left elbow A42 to the right hand A51 and left hand A52 respectively, bones B51 and B52 connecting the neck A2 to the right hip A61 and left hip A62 respectively, bones B61 and B62 connecting the right hip A61 and left hip A62 to the right knee A71 and left knee A72 respectively, and bones B71 and B72 connecting the right knee A71 and left knee A72 to the right foot A81 and left foot A82 respectively. The skeletal structure detection unit 102 stores the detected skeletal structure of the person in the database 110.
[0044] Fig. 8 shows an example of detecting a person standing upright. In Fig. 8, an image of a person standing upright is captured from the front, and bones B1, B51 and B52, B61 and B62, and B71 and B72 are detected without overlapping when viewed from the front, and bones B61 and B71 of the right foot are slightly more bent than bones B62 and B72 of the left foot.
[0045] Fig. 9 shows an example of detecting a person in a crouching state. In Fig. 9, the person in a crouching state is imaged from the right side, and bones B1, B51 and B52, B61 and B62, and B71 and B72 are detected as seen from the right side, and bones B61 and B71 of the right foot and bones B62 and B72 of the left foot are greatly bent and overlap each other.
[0046] Fig. 10 shows an example of detecting a person who is lying down. In Fig. 10, the person lying down is imaged from the diagonal front left, and bones B1, B51 and B52, B61 and B62, and B71 and B72 seen from the diagonal front left are detected, and bones B61 and B71 of the right foot and bones B62 and B72 of the left foot are bent and overlapped.
[0047] Next, as shown in FIG. 3, the image processing device 100 calculates the feature amount of the detected skeletal structure (S103). For example, when the height or area of the skeletal region is used as the feature amount, the feature amount calculation unit 103 extracts a region including the skeletal structure and calculates the height (number of pixels) and area (pixel area) of the region. The height and area of the skeletal region are calculated from the coordinates of the ends of the extracted skeletal region and the coordinates of the key points of the ends. The feature amount calculation unit 103 stores the calculated feature amount of the skeletal structure in the database 110. The feature amount of the skeletal structure is also used as posture information indicating the posture of a person, together with the above-mentioned key points and bones.
[0048] In the example of Figure 8, a skeletal region including all bones is extracted from the skeletal structure of a person standing upright. In this case, the top end of the skeletal region is key point A1 on the head, the bottom end of the skeletal region is key point A82 on the left foot, the left end of the skeletal region is key point A41 on the right elbow, and the right end of the skeletal region is key point A52 on the left hand. Therefore, the height of the skeletal region is calculated from the difference in the Y coordinates of key point A1 and key point A82. In addition, the width of the skeletal region is calculated from the difference in the X coordinates of key point A41 and key point A52, and the area is calculated from the height and width of the skeletal region.
[0049] In the example of Figure 9, a skeletal region including all bones is extracted from the skeletal structure of a crouching person. In this case, the top end of the skeletal region is key point A1 on the head, the bottom end of the skeletal region is key point A81 on the right foot, the left end of the skeletal region is key point A61 on the right hip, and the right end of the skeletal region is key point A51 on the right hand. Therefore, the height of the skeletal region is calculated from the difference in the Y coordinates of key points A1 and A81. The width of the skeletal region is calculated from the difference in the X coordinates of key points A61 and A51, and the area is calculated from the height and width of the skeletal region.
[0050] In the example of Figure 10, a skeletal region including all bones is extracted from the skeletal structure of a person lying down in the left-right direction of the image. In this case, the top end of the skeletal region is the left shoulder key point A32, the bottom end of the skeletal region is the left hand key point A52, the left end of the skeletal region is the right hand key point A51, and the right end of the skeletal region is the left foot key point A82. Therefore, the height of the skeletal region is calculated from the difference in the Y coordinates of key points A32 and A52. The width of the skeletal region is calculated from the difference in the X coordinates of key points A51 and A82, and the area is calculated from the height and width of the skeletal region.
[0051] Next, as shown in Fig. 3, the image processing device 100 performs a classification process (S104). In the classification process, as shown in Fig. 4, the classification unit 104 calculates the similarity of the calculated feature amounts of the skeletal structures (S111), SimilarityBased on this, the skeletal structures are classified (S112). The classification unit 104 obtains the similarity of the feature amounts between all the skeletal structures stored in the database 110 to be classified, and classifies (clusters) the skeletal structures (postures) with the highest similarity into the same cluster. Furthermore, the classification is repeated by obtaining the similarity between the classified clusters until a predetermined number of clusters are obtained. FIG. 11 shows an image of the classification result of the feature amounts of the skeletal structures. FIG. 11 shows an image of the cluster analysis by two-dimensional classification elements, and the two classification elements are, for example, the height of the skeletal region and the area of the skeletal region. In FIG. 11, the feature amounts of a plurality of skeletal structures are classified into three clusters C1 to C3 as a result of classification. The clusters C1 to C3 correspond to each posture, such as a standing posture, a sitting posture, and a lying posture, and the skeletal structures (people) are classified into each similar posture.
[0052] In this embodiment, various classification methods can be used by classifying based on the feature amount of the person's skeletal structure. The classification method may be set in advance, or may be set arbitrarily by the user. In addition, classification may be performed by the same method as the search method described later. In other words, classification may be performed by classification conditions similar to the search conditions. For example, the classification unit 104 performs classification by the following classification methods. Any of the classification methods may be used, or arbitrarily selected classification methods may be combined.
[0053] (Classification method 1) Classification based on multiple hierarchies The classification is performed by hierarchically combining classification based on the skeletal structure of the entire body, classification based on the skeletal structure of the upper and lower body, classification based on the skeletal structure of the arms and legs, etc. In other words, classification may be performed based on the feature amounts of the first and second parts of the skeletal structure, and further classification may be performed by weighting the feature amounts of the first and second parts.
[0054] (Classification method 2) Classification using multiple images in chronological order Classification is based on the feature values of the skeletal structure in multiple consecutive images in a time series. For example, feature values may be accumulated in the time series direction and classification may be based on the cumulative value. Furthermore, classification may be based on the change (amount of change) in the feature values of the skeletal structure in multiple consecutive images.
[0055] (Classification method 3) Classification ignoring the left and right sides of the skeletal structure Classify opposite skeletal structures on the right and left sides of a person as the same skeletal structure.
[0056] Furthermore, the classification unit 104 displays the classification results of the skeletal structure (S113). The classification unit 104 acquires the necessary skeletal structure and person images from the database 110, and displays the skeletal structure and person for each similar posture (cluster) on the display unit 107 as the classification results. FIG. 12 shows a display example in which postures are classified into three. For example, as shown in FIG. 12, posture areas WA1 to WA3 for each posture are displayed in the display window W1, and the skeletal structure and person (image) of the posture corresponding to each posture area WA1 to WA3 are displayed. The posture area WA1 is, for example, a display area for a standing posture, and displays the skeletal structure and person similar to the standing posture classified into cluster C1. The posture area WA2 is, for example, a display area for a sitting posture, and displays the skeletal structure and person similar to the sitting posture classified into cluster C2. The posture area WA3 is, for example, a display area for a lying posture, and displays the skeletal structure and person similar to the sitting posture classified into cluster C3. 3 The system displays skeletal structures and people in positions similar to those in a sleeping position, classified as .
[0057] Subsequently, as shown in FIG. 3, the image processing device 100 performs a search process (S105). In the search process, as shown in FIG. 5, the search unit 105 accepts an input of a search condition (S121) and searches for a skeletal structure based on the search condition (S122). The search unit 105 accepts an input of a search query, which is a search condition, from the input unit 106 in response to a user's operation. When inputting a search query from a classification result, for example, in the display example of FIG. 12, the user specifies (selects) a skeletal structure of a posture to be searched from among posture areas WA1 to WA3 displayed in the display window W1. Then, the search unit 105 uses the skeletal structure specified by the user as a search query to search for a skeletal structure having a high similarity in feature amount from among all skeletal structures stored in the database 110 as a search target. The search unit 105 calculates a similarity between the feature amount of the skeletal structure of the search query and the feature amount of the skeletal structure of the search target, and extracts a skeletal structure having a calculated similarity amount higher than a predetermined threshold value. The feature amount of the skeletal structure of the search query may be a feature amount calculated in advance, or a feature amount calculated at the time of the search. Note that the search query may be input by moving each part of the skeletal structure according to the user's operation, or the posture demonstrated by the user in front of the camera may be used as the search query.
[0058] In this embodiment, similarly to the classification method, a variety of search methods can be used by searching based on the feature amount of a person's skeletal structure. The search method may be preset, or may be arbitrarily set by the user. For example, the search unit 105 performs a search using the following search methods. Any of the search methods may be used, or arbitrarily selected search methods may be combined. A search may be performed by combining a plurality of search methods (search conditions) using a logical expression (e.g., AND (logical product), OR (logical sum), NOT (negation)). For example, a search may be performed using the search conditions "(posture with right hand raised) AND (posture with left leg raised)".
[0059] (Search method 1) Search using height features only By performing a search using only the height feature of a person, the effect of lateral changes in the person can be suppressed, and robustness against changes in the person's orientation and body type is improved. For example, as in skeletal structures 501 to 503 in Fig. 13, even if the person's orientation or body type is different, the height feature does not change significantly. For this reason, the skeletal structures 501 to 503 can be determined to have the same posture during search (classification).
[0060] (Search Method 2) Partial Search When a part of a person's body is hidden in an image, a search is performed using only information on recognizable parts. For example, as in the case of skeletal structures 511 and 512 in FIG. 14, even if the key point of the left foot cannot be detected because the left foot is hidden, a search can be performed using the feature amounts of other detected key points. For this reason, in the case of skeletal structures 511 and 512, it is possible to determine that the postures are the same at the time of search (classification). In other words, classification and search can be performed using the feature amounts of some key points, not all key points. In the example of skeletal structures 521 and 522 in FIG. 15, although the directions of both feet are different, it is possible to determine that the postures are the same by using the feature amounts of the key points (A1, A2, A31, A32, A41, A42, A51, A52) of the upper body as a search query. In addition, a search can be performed by weighting the part (feature point) to be searched, or a threshold value for similarity judgment can be changed. When a part of the body is hidden, a search can be performed by ignoring the hidden part, or a search can be performed by taking the hidden part into account. By including hidden parts in the search, it is possible to search for postures in which the same body part is hidden.
[0061] (Search method 3) Search ignoring the left and right sides of the skeletal structure Skeletal structures with opposite skeletal structures on the right and left sides of a person are searched for as the same skeletal structure. For example, a posture in which the right hand is raised and a posture in which the left hand is raised can be searched for (classified) as the same posture, as in skeletal structures 531 and 532 in Fig. 16. In the example of Fig. 16, skeletal structures 531 and 532 have different positions for key point A51 on the right hand, key point A41 on the right elbow, key point A52 on the left hand, and key point A42 on the left elbow, but the positions of the other key points are the same. Of the key points A51 on the right hand and A41 on the right elbow of skeletal structure 531, and the key points A52 on the left hand and A42 on the left elbow of skeletal structure 532, if the key points of one of the skeletal structures are flipped left to right, they will be in the same position as the key points of the other skeletal structure. Furthermore, of the key points A52 on the left hand and A42 on the left elbow of skeletal structure 531, and the key points A51 on the right hand and A41 on the right elbow of skeletal structure 532, if the key points of one of the skeletal structures are flipped left to right, they will be in the same position as the key points of the other skeletal structure, so they are determined to be the same posture.
[0062] (Search method 4) Search by vertical and horizontal features After performing a search using only the features of the person's vertical direction (Y-axis direction), the obtained results are further searched using the features of the person's horizontal direction (X-axis direction).
[0063] (Search method 5) Searching for multiple images in chronological order Searches are performed based on the feature values of the skeletal structure in multiple consecutive images in a time series. For example, the feature values may be accumulated in the time series direction, and searches may be performed based on the cumulative value. Furthermore, searches may be performed based on the change (amount of change) in the feature values of the skeletal structure in multiple consecutive images.
[0064] Furthermore, the search unit 105 displays the search results of the skeletal structure (S123). The search unit 105 acquires the necessary skeletal structure and person images from the database 110, and displays the skeletal structure and person obtained as the search results on the display unit 107. For example, when a plurality of search queries (search conditions) are specified, the search results are displayed for each search query. FIG. 17 shows a display example when a search is performed using three search queries (postures). For example, as shown in FIG. 17, in the display window W2, the skeletal structures and people of the specified search queries Q10, Q20, and Q30 are displayed at the left end, and the skeletal structures and people of the search results Q11, Q21, and Q31 of each search query are displayed side by side to the right of the search queries Q10, Q20, and Q30.
[0065] The order in which the search results are displayed next to the search query may be the order in which the corresponding skeletal structures were found, or the order in which the similarity is highest. When a partial search is performed with weights applied to the parts (feature points), the results may be displayed in order of similarity calculated using the weights. The results may also be displayed in order of similarity calculated only from the parts (feature points) selected by the user. In addition, images (frames) before and after the search result image (frame) may be cut out for a certain period of time and displayed.
[0066] (Search Method 6) In this search method, the search unit 105 uses information indicating a person's posture (hereinafter referred to as posture information) as a search query. An example of this posture information is the skeletal structure described above. The search query is generated, for example, by processing a query image. Then, the search unit 105 selects at least one image (hereinafter referred to as a target image) including a person similar to the posture indicated by the search query from a plurality of target images. At this time, the search unit 105 uses information related to the person that is different from the posture information (hereinafter referred to as other information) together with the posture information. The target image may be a still image or a video having a plurality of frame images. In addition to the function of selecting the target image, the search unit 105 also has a function of classifying a plurality of target images into a plurality of image groups that are similar to each other.
[0067] 40 is a diagram showing a first example of the functional configuration of the search unit 105 related to this search method. In this diagram, the search unit 105 has a function of classifying a plurality of target images into a plurality of image groups, and includes an information generation unit 610 and an image selection unit 620.
[0068] The information generating unit 610 generates posture information of a person included in each of a plurality of target images and other information related to the person from the respective target images. An example of posture information is the feature amount of a skeletal structure, as described above. An example of the feature amount of a skeletal structure is a plurality of key points and bones, and may further include the height and area of a skeletal region. An example of a method for calculating the feature amount of a skeletal structure is as described above. A part of the processing performed by the information generating unit 610 is similar to that of, for example, the skeletal structure detecting unit 102 and the feature amount calculating unit 103.
[0069] The image selection unit 620 classifies the target images into a plurality of image groups using the posture information and other information generated by the information generation unit 610. For example, the image selection unit 620 calculates the similarity between the target images using the posture information and other information, and classifies the target images whose similarity satisfies a criterion into the same image group.
[0070] Here, the above-mentioned other information will be described. The other information is information about a person with respect to the posture information, and is information different from the posture information. The other information includes, for example, at least one of color information of the person and / or the person's wearables, color information of a target area including both the person and the surroundings of the person, the person's face, the person's gender, the person's age group, the person's body shape, and the person's position in the target image.
[0071] First, the "color information of at least one of a person and the person's clothing" will be described. This color information includes at least one of the colors of the person's face and skin, and the colors of clothing, hats, and other clothing. For example, the information generating unit 610 generates the color information by cutting out an area corresponding to a person from the image to be processed and processing the area. The information generating unit 610 may use posture information of the person when cutting out the area corresponding to the person. As an example, the information generating unit 610 estimates the area corresponding to the person from the posture information. This color information is represented by the proportion of each color in the area corresponding to at least one of a person and the person's clothing. Here, the color is represented by a combination of the brightness of each of RGB, for example.
[0072] The color information may be calculated for each part of the person (e.g., face, arms, torso, and legs), for example, for each key point, or may be calculated at once for the entire area corresponding to at least one of the person and the person's clothing. In the former case, the image selection unit 620 calculates the similarity of the surrounding area for each part, for example, for each key point, and calculates the color similarity between the two images by calculating (e.g., adding) the similarity for each part. In this case, weighting may be set for multiple parts. The image selection unit 620 uses this weighting when calculating the similarity for each part. For example, when the calculation is addition, the weighting coefficient for the part that should be considered important is larger than the weighting coefficient for the other parts. Note that this weighting is set, for example, according to a user input.
[0073] Next, the "color information of the target area" will be described. This color information includes color information of the person and the person's clothing, as well as color information of the person's surroundings (for example, the background and the ground (floor if indoors)). The reason for using the color information of the target area is that when searching or classifying images, the person's environment can be an indicator along with the person's posture. This color information is also shown as the proportion of each color in the entire target area. Here, the colors are shown, for example, as a combination of the respective luminances of RGB. Note that the target area may be a part of the image (for example, a rectangular area), or it may be the entire image.
[0074] Also, the "person's face, person's gender, person's age group, and person's body shape" are determined by, for example, image processing.
[0075] The "position of the person in the target image" is also determined by image processing. The position of the person in the image may also be an index when searching or classifying images, and may be used as the other information described above.
[0076] In the example shown in the figure, a plurality of target images that become a population when the image selection unit 620 classifies images are stored in the image storage unit 630. The target images stored in the image storage unit 630 are repeatedly updated. This update includes both adding target images and deleting target images, but generally, the number of target images stored in the image storage unit 630 increases as time passes. In the example shown in the figure, the image storage unit 630 is a part of the search unit 105, i.e., the image processing device 10. However, the image storage unit 630 may be located outside the image processing device 10. The image storage unit 630 may be a part of the database 110 described above, or may be provided separately from the database 110.
[0077] FIG. 41 is a diagram showing an example of a screen that the image selection unit 620 displays on the user's terminal 700 or display unit 107. The screen shown in this figure is a screen for the user to input the weights of posture information and other information used when classifying images. The screen shown in this figure includes a field 710 for inputting a weighting coefficient α1 for posture information and a field 720 for inputting a weighting coefficient α2 for other information. Here, when one weighting coefficient is input, the other weighting coefficient may be automatically calculated and displayed. Then, the image selection unit 620 determines the similarity between the two images as, for example, "similarity of posture information×α1+similarity of other information×α2".
[0078] Fig. 42 is a flowchart showing an example of the process performed by search unit 105 shown in Fig. 40. First, information generating unit 610 acquires a plurality of target images from image storage unit 630 (step S300). At this time, information generating unit 610 may acquire all target images stored in image storage unit 630, or may acquire only some of the target images.
[0079] Next, the information generation unit 610 processes each of the target images to generate pose information (step S310) and other information (step S320).The image selection unit 620 then uses the pose information and other information to calculate the similarity between the target images acquired in step S300, and classifies the target images into image groups using the similarity (step S330).
[0080] Then, the image selection unit 620 outputs information indicating the classification result, for example, to be displayed on the screen of the terminal 700 or on the display unit 107 (step S340).
[0081] Fig. 43 is a diagram showing a modified example of Fig. 40. The search unit 105 shown in this figure acquires a query image including a human posture, and selects an image similar to the query image (hereinafter, referred to as a target image) from target images.
[0082] In the example shown in the figure, the search unit 105 includes a query acquisition unit 640 in addition to an information generation unit 610, an image selection unit 620, and an image storage unit 630. The query acquisition unit 640 acquires a query image. This query image may be selected from among the target images stored in the image storage unit 630, or may be newly input by the user.
[0083] Fig. 44 is a flowchart showing an example of the operation of the search unit 105 shown in Fig. 43. First, the query acquisition unit 640 acquires a query image (step S400). The information generation unit 610 acquires a plurality of target images from the image storage unit 630 (step S410). Then, the information generation unit 610 generates posture information for each of the query image and the plurality of target images (step S420) and also generates other information (step S430).
[0084] Then, the image selection unit 620 selects at least one target image from the multiple target images (step S440). As an example, the image selection unit 620 calculates a similarity to the query image for each of the multiple target images based on the posture information. The image selection unit 620 also calculates a similarity to the query image for each of the multiple target images based on other information. The image selection unit 620 then uses these two similarities to select a target image. For example, as described with reference to FIG. 41, the image selection unit 620 「 The integrated similarity is calculated using "similarity of posture information x α1 + similarity of other information x α2", and the subject image whose integrated similarity satisfies the criteria is selected as the target image.
[0085] After that, the image selection unit 620 outputs information indicating the selection result to be displayed on, for example, the screen of the terminal 700 or the display unit 107 (step S450).
[0086] As described above, in this embodiment, it is possible to detect a person's skeletal structure from a two-dimensional image, and to perform classification and search based on the feature amount of the detected skeletal structure. This allows classification into similar poses with high similarity, and also allows searching for similar poses with high similarity to a search query (search key). By classifying and displaying similar poses from an image, it is possible to grasp the pose of a person in an image without the user specifying a pose, etc. Since the user can specify the pose of a search query from the classification results, it is possible to search for a desired pose even if the user does not know in advance the pose he or she wants to search in detail. For example, classification and search can be performed using the whole or part of a person's skeletal structure as a condition, which enables flexible classification and search.
[0087] According to the search method 6, when classifying target images into a plurality of image groups or selecting images similar to a query image from the target images, other information related to the person is used in addition to the posture information, which improves the accuracy of classifying or selecting images.
[0088] (Embodiment 2) Hereinafter, embodiment 2 will be described with reference to the drawings. In this embodiment, a specific example of feature amount calculation in embodiment 1 will be described. In this embodiment, feature amounts are obtained by normalizing using the person's height. The rest is the same as embodiment 1.
[0089] Fig. 18 shows the configuration of an image processing device 100 according to this embodiment. As shown in Fig. 18, the image processing device 100 further includes a height calculation unit 108 in addition to the configuration of embodiment 1. Note that the feature calculation unit 103 and the height calculation unit 108 may be integrated into one processing unit.
[0090] The height calculation unit (height estimation unit) 108 calculates (estimates) the height (called height pixel count) of a person in a two-dimensional image when standing upright, based on the two-dimensional skeletal structure detected by the skeletal structure detection unit 102. The height pixel count can also be said to be the height of a person in a two-dimensional image (the length of the person's entire body in two-dimensional image space). The height calculation unit 108 finds the height pixel count (number of pixels) from the length of each bone of the detected skeletal structure (the length in two-dimensional image space).
[0091] In the following examples, concrete examples 1 to 3 are used as a method for calculating the height pixel number. Any of the concrete examples 1 to 3 may be used, or a combination of a plurality of arbitrarily selected methods may be used. In concrete example 1, the height pixel number is calculated by adding up the lengths of the bones from the head to the feet among the bones of the skeletal structure. If the skeletal structure detection unit 102 (skeletal estimation technology) does not output the top of the head and the feet, it can be corrected by multiplying a constant as necessary. In concrete example 2, the height pixel number is calculated using a human body model that indicates the relationship between the length of each bone and the length of the whole body (height in a two-dimensional image space). In concrete example 3, the height pixel number is calculated by fitting (applying) a three-dimensional human body model to the two-dimensional skeletal structure.
[0092] The feature amount calculation unit 103 in this embodiment is a normalization unit that normalizes the skeletal structure (skeletal information) of a person based on the calculated height pixel number of the person. The feature amount calculation unit 103 stores the feature amount (normalization value) of the normalized skeletal structure in the database 110. The feature amount calculation unit 103 normalizes the height of each key point (feature point) included in the skeletal structure on the image by the height pixel number. In this embodiment, for example, the height direction is the up-down direction (Y axis direction) in the two-dimensional coordinate (XY coordinate) space of the image. In this case, the height of the key point can be obtained from the Y coordinate value (pixel number) of the key point. Alternatively, the height direction may be the direction of the vertical projection axis (vertical projection direction) obtained by projecting the direction of the vertical axis perpendicular to the ground (reference plane) in the three-dimensional coordinate space of the real world onto the two-dimensional coordinate space. In this case, the height of the keypoint can be obtained from a value (number of pixels) along a vertical projection axis obtained by projecting an axis perpendicular to the ground in the real world onto a two-dimensional coordinate space based on the camera parameters. The camera parameters are image capture parameters, such as the attitude, position, image capture angle, and focal length of the camera 200. The camera 200 captures an image of an object whose length and position are known in advance, and the camera parameters can be obtained from the image. Distortion occurs at both ends of the captured image, and the vertical direction in the real world and the up-down direction of the image may not match. In response to this, the degree to which the vertical direction in the real world is tilted in the image can be found by using the parameters of the camera that captured the image. Therefore, the value of the keypoint along the vertical projection axis projected onto the image based on the camera parameters is normalized by height, and the keypoint can be characterized in consideration of the deviation between the real world and the image. The left-right direction (horizontal direction) refers to the left-right direction (X-axis direction) in the two-dimensional coordinate (XY coordinate) space of the image, or the direction parallel to the ground in the three-dimensional coordinate space of the real world projected onto the two-dimensional coordinate space.
[0093] Fig. 19 to Fig. 23 show the operation of the image processing device 100 according to this embodiment. Fig. 19 shows the flow from image acquisition to search processing in the image processing device 100, Fig. 20 to Fig. 22 show the flow of specific examples 1 to 3 of the height pixel number calculation processing (S201) in Fig. 19, and Fig. 23 shows the flow of the normalization processing (S202) in Fig. 19.
[0094] 19, in this embodiment, height pixel number calculation processing (S201) and normalization processing (S202) are performed as the feature amount calculation processing (S103) in embodiment 1. The rest is the same as in embodiment 1.
[0095] Following image acquisition (S101) and skeletal structure detection (S102), the image processing device 100 performs height pixel number calculation processing based on the detected skeletal structure (S201). In this example, as shown in Fig. 24, the height of the skeletal structure of a person standing upright in the image is taken as height pixel number (h), and the height of each key point of the skeletal structure in the state of the person in the image is taken as key point height (yi). Specific examples 1 to 3 of height pixel number calculation processing will be described below.
[0096] <Specific Example 1> In specific example 1, the height pixel number is calculated using the bone lengths from the head to the feet. In specific example 1, as shown in Fig. 20, the height calculation unit 108 obtains the length of each bone (S211) and sums up the obtained lengths of each bone (S212).
[0097] The height calculation unit 108 obtains the length of the bones on the two-dimensional image from the head to the feet of the person, and calculates the height pixel number. That is, from the image in which the skeletal structure is detected, the length (number of pixels) of each of the bones in FIG. 24, bone B1 (length L1), bone B51 (length L21), bone B61 (length L31), and bone B71 (length L41), or bone B1 (length L1), bone B52 (length L22), bone B62 (length L32), and bone B72 (length L42), is obtained. The length of each bone can be obtained from the coordinates of each key point in the two-dimensional image. The total of these, L1+L21+L31+L41, or L1+L22+L32+L42, is multiplied by a correction constant to calculate the height pixel number (h). When both values can be calculated, for example, the longer value is set as the height pixel number. That is, each bone will be the longest in the image when it is photographed from the front, and will appear shorter when tilted in the depth direction relative to the camera. Therefore, it is more likely that a longer bone was photographed from the front, and is closer to the true value. For this reason, it is preferable to select the longer value.
[0098] In the example of Fig. 25, bone B1, bones B51 and B52, bones B61 and B62, and bones B71 and B72 are detected without overlapping. The totals of these bones, L1+L21+L31+L41 and L1+L22+L32+L42, are calculated, and the height pixel number is determined by multiplying the detected bone length, for example, L1+L22+L32+L42 on the left leg side, which is longer, by a correction constant.
[0099] 26, bone B1, bone B51 and bone B52, bone B61 and bone B62, and bone B71 and bone B72 are detected, and bones B61 and B71 of the right foot overlap with bones B62 and B72 of the left foot. The totals of these bones, L1+L21+L31+L41 and L1+L22+L32+L42, are calculated, and the height pixel number is determined by multiplying the detected bone length L1+L21+L31+L41 on the right foot side, which is longer, by a correction constant.
[0100] 27, bone B1, bone B51 and bone B52, bone B61 and bone B62, and bone B71 and bone B72 are detected, and bones B61 and B71 of the right foot overlap with bones B62 and B72 of the left foot. The totals of these bones, L1+L21+L31+L41 and L1+L22+L32+L42, are calculated, and the height pixel number is determined by multiplying the detected bone length L1+L22+L32+L42 on the left foot side, which is longer, by a correction constant.
[0101] In Example 1, height can be calculated by adding up the lengths of the bones from head to toe, so the height pixel count can be calculated in a simple manner. In addition, because it is sufficient to detect at least the skeleton from head to toe using machine learning-based skeleton estimation technology, the height pixel count can be estimated with high accuracy even when the entire person is not necessarily captured in the image, such as when the person is crouching.
[0102] <Specific Example 2> In specific example 2, the height pixel number is calculated using a two-dimensional skeleton model that indicates the relationship between the length of the bones included in the two-dimensional skeleton structure and the length of the entire body of a person in two-dimensional image space.
[0103] FIG. 28 shows a human body model (two-dimensional skeletal model) 301 used in the concrete example 2, which shows the relationship between the length of each bone in the two-dimensional image space and the length of the whole body in the two-dimensional image space. As shown in FIG. 28, the relationship between the length of each bone of an average person and the length of the whole body (the ratio of the length of each bone to the length of the whole body) is associated with each bone of the human body model 301. For example, the length of the head bone B1 is the length of the whole body×0.2 (20%), the length of the right hand bone B41 is the length of the whole body×0.15 (15%), and the length of the right foot bone B71 is the length of the whole body×0.25 (25%). By storing such information of the human body model 301 in the database 110, the average whole body length can be obtained from the length of each bone. In addition to the human body model of the average person, a human body model may be prepared for each person's attributes such as age, sex, and nationality. This makes it possible to obtain the whole body length (height) appropriately according to the person's attributes.
[0104] In specific example 2, as shown in Fig. 21, height calculation unit 108 acquires the length of each bone (S221). Height calculation unit 108 acquires the length of all bones (length in two-dimensional image space) in the detected skeletal structure. Fig. 29 shows an example in which a person in a crouching state is imaged from diagonally behind to the right and a skeletal structure is detected. In this example, the person's face and left side are not captured, so the head bone and the left arm and left hand bones cannot be detected. Therefore, the length of each of the detected bones B21, B22, B31, B41, B51, B52, B61, B62, B71, and B72 is acquired.
[0105] Next, the height calculation unit 108 calculates the height pixel number from the length of each bone based on the human body model as shown in Fig. 21 (S222). The height calculation unit 108 refers to a human body model 301 showing the relationship between each bone and the length of the whole body as shown in Fig. 28, and calculates the height pixel number from the length of each bone. For example, since the length of the right hand bone B41 is the whole body length × 0.15, the height pixel number based on the bone B41 is calculated by dividing the length of the bone B41 by 0.15. Also, since the length of the right foot bone B71 is the whole body length × 0.25, the height pixel number based on the bone B71 is calculated by dividing the length of the bone B71 by 0.25.
[0106] The human body model to be referenced at this time is, for example, an average human body model, but a human body model may be selected according to the attributes of the person, such as age, sex, and nationality. For example, when a person's face is captured in a captured image, the attributes of the person are identified based on the face, and a human body model corresponding to the identified attributes is referenced. The attributes of the person can be recognized from the facial features of the image by referring to information obtained by machine learning of the face for each attribute. Also, when the attributes of the person cannot be identified from the image, an average human body model may be used.
[0107] Furthermore, the height pixel count calculated from the bone lengths may be corrected by camera parameters. For example, when a person is photographed looking down from a high position with a camera, the horizontal length of the shoulder width bones in the two-dimensional skeletal structure is not affected by the camera's inclination angle, but the vertical length of the neck-waist bones decreases as the camera's inclination angle increases. As a result, the height pixel count calculated from the horizontal length of the shoulder width bones tends to be larger than the actual height pixel count. Therefore, by utilizing camera parameters, the angle at which the camera is looking down on the person can be known, and this inclination angle information can be used to correct the two-dimensional skeletal structure to one as if photographed from the front. This allows the height pixel count to be calculated more accurately.
[0108] Next, the height calculation unit 108 calculates the optimum value of the height pixel number as shown in FIG. 21 (S223). The height calculation unit 108 calculates the optimum value of the height pixel number from the height pixel number calculated for each bone. For example, a histogram of the height pixel number calculated for each bone as shown in FIG. 30 is generated, and the largest height pixel number among them is selected. In other words, a height pixel number longer than the others is selected from a plurality of height pixel numbers calculated based on a plurality of bones. For example, the top 30% are considered to be valid values, and in FIG. 30, the height pixel numbers of bones B71, B61, and B51 are selected. The average of the selected height pixel numbers may be calculated as the optimum value, or the largest height pixel number may be calculated as the optimum value. In order to calculate the height from the length of the bones in the two-dimensional image, the bones are viewed from the front. Imaging If this is not possible, i.e. if the bone is photographed tilted in the depth direction as viewed from the camera, the length of the bone will be shorter than when photographed from the front. In this case, a value with a larger height pixel count is more likely to have been photographed from the front than a value with a smaller height pixel count, and is therefore a more likely value. Therefore, the larger value is set as the optimal value.
[0109] In Example 2, the height pixel number is calculated based on the detected bones of the skeletal structure using a human body model showing the relationship between the bones in the two-dimensional image space and the length of the whole body, so that even if all the bones from the head to the feet cannot be obtained, the height pixel number can be calculated from some of the bones. In particular, the height pixel number can be estimated with high accuracy by adopting the larger value among the values calculated from multiple bones.
[0110] <Specific Example 3> In specific example 3, a two-dimensional skeletal structure is fitted to a three-dimensional human body model (three-dimensional skeletal model), and the skeletal vector of the entire body is obtained using the height pixel count of the fitted three-dimensional human body model.
[0111] In specific example 3, as shown in Fig. 22, height calculation unit 108 first calculates camera parameters based on images captured by camera 200 (S231). Height calculation unit 108 extracts an object whose length is known in advance from a plurality of images captured by camera 200, and obtains camera parameters from the size (number of pixels) of the extracted object. Note that the camera parameters may be obtained in advance, and the obtained camera parameters may be acquired as necessary.
[0112] Next, the height calculation unit 108 adjusts the position and height of the three-dimensional human body model (S232). The height calculation unit 108 prepares a three-dimensional human body model for calculating the number of height pixels for the detected two-dimensional skeletal structure, and places it in the same two-dimensional image based on the camera parameters. Specifically, the "relative positional relationship between the camera and the person in the real world" is specified from the camera parameters and the two-dimensional skeletal structure. For example, assuming that the camera position is at coordinates (0,0,0), the coordinates (x,y,z) of the position where the person is standing (or sitting) are specified. Then, the two-dimensional skeletal structure and the three-dimensional human body model are superimposed by assuming an image when the three-dimensional human body model is placed at the same position (x,y,z) as the specified person and captured.
[0113] FIG. 31 shows an example in which a person who is crouching is imaged from the left diagonal front and a two-dimensional skeletal structure 401 is detected. The two-dimensional skeletal structure 401 has two-dimensional coordinate information. It is preferable that all bones are detected, but some bones may not be detected. For this two-dimensional skeletal structure 401, a three-dimensional human body model 402 as shown in FIG. 32 is prepared. The three-dimensional human body model (three-dimensional skeletal model) 402 has three-dimensional coordinate information and is a skeletal model having the same shape as the two-dimensional skeletal structure 401. Then, as shown in FIG. 33, the prepared three-dimensional human body model 402 is arranged and superimposed on the detected two-dimensional skeletal structure 401. In addition to the superimposition, the height of the three-dimensional human body model 402 is adjusted to match the two-dimensional skeletal structure 401.
[0114] The three-dimensional human body model 402 prepared at this time may be a model in a state close to the posture of the two-dimensional skeletal structure 401, as shown in Fig. 33, or may be a model in an upright state. For example, a three-dimensional human body model 402 in an estimated posture may be generated using a technology that uses machine learning to estimate a posture in three-dimensional space from a two-dimensional image. By learning information on the joints in the two-dimensional image and the joints in three-dimensional space, it is possible to estimate a three-dimensional posture from a two-dimensional image.
[0115] Next, the height calculation unit 108 fits the three-dimensional human body model to the two-dimensional skeletal structure as shown in FIG. 22 (S233). In a state where the three-dimensional human body model 402 is superimposed on the two-dimensional skeletal structure 401 as shown in FIG. 34, the height calculation unit 108 deforms the three-dimensional human body model 402 so that the postures of the three-dimensional human body model 402 and the two-dimensional skeletal structure 401 match. That is, the height, body orientation, and joint angles of the three-dimensional human body model 402 are adjusted and optimized so that there is no difference with the two-dimensional skeletal structure 401. For example, the joints of the three-dimensional human body model 402 are rotated within the range of human movement, and the entire three-dimensional human body model 402 is rotated and the overall size is adjusted. The fitting of the three-dimensional human body model to the two-dimensional skeletal structure is performed in a two-dimensional space (two-dimensional coordinates). In other words, a 3D human body model is mapped onto a 2D space, and the 3D human body model is optimized to a 2D skeletal structure, taking into consideration how the deformed 3D human body model changes in the 2D space (image).
[0116] Next, height calculation unit 108 calculates the height pixel number of the fitted three-dimensional human body model as shown in FIG. 22 (S234). When the difference between three-dimensional human body model 402 and two-dimensional skeletal structure 401 disappears and the postures match as shown in FIG. 35, height calculation unit 108 calculates the height pixel number of three-dimensional human body model 402 in that state. With optimized three-dimensional human body model 402 in an upright state, the length of the entire body in two-dimensional space is calculated based on the camera parameters. For example, the height pixel number is calculated based on the bone length (number of pixels) from the head to the feet when three-dimensional human body model 402 is upright. As in specific example 1, the bone lengths from the head to the feet of three-dimensional human body model 402 may be summed up.
[0117] In specific example 3, a 3D human body model is fitted to a 2D skeletal structure based on the camera parameters, and height pixel count is calculated based on the 3D human body model. This makes it possible to estimate height pixel count with high accuracy even when not all bones are viewed from the front, i.e., even when all bones are viewed at an angle, resulting in a large error.
[0118] <Normalization Processing> As shown in FIG. 19, the image processing device 100 performs normalization processing (S202) following the height pixel number calculation processing. In the normalization processing, as shown in FIG. 23, the feature amount calculation unit 103 calculates the key point height (number of pixels) of all key points included in the detected skeletal structure. The key point height is the length (number of pixels) in the height direction from the lowest end of the skeletal structure (for example, a key point of one of the feet) to the key point. Here, as an example, the key point height is obtained from the Y coordinate of the key point in the image. As described above, the key point height may be obtained from the length in the direction along the vertical projection axis based on the camera parameters. For example, in the example of FIG. 24, the height (yi) of the neck key point A2 is a value obtained by subtracting the Y coordinate of the right foot key point A81 or the left foot key point A82 from the Y coordinate of the key point A2.
[0119] Next, the feature amount calculation unit 103 specifies a reference point for normalization (S242). The reference point is a point that serves as a reference for expressing the relative height of the key point. The reference point may be set in advance, or may be selectable by the user. The reference point is preferably the center of the skeletal structure or higher than the center (upward in the vertical direction of the image), and for example, the coordinates of the key point of the neck are set as the reference point. Note that the reference point is not limited to the neck, and the coordinates of the head or other key points may be set as the reference point. The reference point is not limited to the key point, and any coordinates (for example, the center coordinates of the skeletal structure, etc.) may be set as the reference point.
[0120] Next, the feature calculation unit 103 normalizes the keypoint height (yi) by the height pixel count (S243). The feature calculation unit 103 normalizes each keypoint using the keypoint height, reference point, and height pixel count of each keypoint. Specifically, the feature calculation unit 103 normalizes the relative height of the keypoint with respect to the reference point by the height pixel count. Here, as an example focusing on only the height direction, only the Y coordinate is extracted, and normalization is performed with the reference point being the neck keypoint. Specifically, the Y coordinate of the reference point (neck keypoint) is set as (yc), and the feature (normalized value) is calculated using the following formula (1). Note that when a vertical projection axis based on the camera parameters is used, (yi) and (yc) are converted into values in the direction along the vertical projection axis.
number
[0121] For example, if there are 18 keypoints, the coordinates of the 18 keypoints (x0, y0), (x1, y1), ... (x17, y17) are converted into 18-dimensional features using the above formula (1) as follows:
number
[0122] FIG. 36 shows an example of the feature amount of each key point calculated by the feature amount calculation unit 103. In this example, the neck key point A2 is set as the reference point, so the feature amount of the key point A2 is 0.0, and the feature amounts of the right shoulder key point A31 and the left shoulder key point A32, which are at the same height as the neck, are also 0.0. The feature amount of the head key point A1, which is higher than the neck, is -0.2. The feature amount of the right hand key point A51 and the left hand key point A52, which are lower than the neck, is 0.4, and the feature amount of the right foot key point A81 and the left foot key point A82 are 0.9. If the person raises his / her left hand from this state, the left hand becomes higher than the reference point as shown in FIG. 37, so the feature amount of the left hand key point A52 is -0.4. On the other hand, since normalization is performed using only the coordinate of the Y axis, the feature amount does not change even if the width of the skeletal structure changes as shown in FIG. 38 compared to FIG. 36. That is, the feature amount (normalized value) in this embodiment indicates the feature in the height direction (Y direction) of the skeletal structure (key point), and is not affected by changes in the lateral direction (X direction) of the skeletal structure.
[0123] As described above, in this embodiment, a person's skeletal structure is detected from a two-dimensional image, and each key point of the skeletal structure is normalized using the height pixel count (height when standing upright in two-dimensional image space) calculated from the detected skeletal structure. By using this normalized feature amount, robustness when classification, search, etc. are performed can be improved. That is, the feature amount of this embodiment is not affected by lateral changes of the person as described above, and is therefore highly robust against changes in the person's orientation and body shape.
[0124] Furthermore, in this embodiment, since this can be realized by detecting the skeletal structure of a person using a skeletal estimation technology such as OpenPose, there is no need to prepare learning data for learning the posture, etc. of a person. Also, by normalizing the key points of the skeletal structure and storing them in a database, classification and search of posture, etc. of a person is possible, so classification and search can be performed even for unknown postures. Also, by normalizing the key points of the skeletal structure, clear and easy-to-understand features can be obtained, so that users can be more convinced of the processing results, unlike black-box algorithms such as machine learning.
[0125] Although the embodiments of the present invention have been described above with reference to the drawings, these are merely examples of the present invention, and various configurations other than those described above can also be adopted.
[0126] In addition, in the flowcharts used in the above description, multiple steps (processes) are described in order, but the order of execution of the steps performed in each embodiment is not limited to the order described. In each embodiment, the order of the steps shown in the figures can be changed to the extent that the content is not affected. In addition, the above-mentioned embodiments can be combined to the extent that the content is not contradictory.
[0127] A part or all of the above-described embodiments can be described as, but are not limited to, the following supplementary notes. 1. An information generating means for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection means for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; An image selection device comprising: 2. In the image selection device according to the above item 1, The method further includes: The information generating means further generates posture information and the other information of the person included in the query image; The image selection means further uses the posture information and the other information of the query image and each of the plurality of target images to select the at least one target image. 3. In the image selection device according to the above 1 or 2, The other information includes color information of at least one of the person and the person's clothing. 4. In the image selection device according to claim 3, A weight is set for each of the plurality of body parts of the person, The image selection means uses the weighting to classify the plurality of target images or to select at least one target image from the plurality of target images. 5. In the image selection device according to any one of 1 to 4 above, The other information includes color information of a region of interest that includes both the person and the surroundings of the person. 6. In the image selection device according to any one of 1 to 5 above, The other information includes at least one of the person's face, gender, age group, and body type. 7. In the image selection device according to any one of claims 1 to 6, The other information is a position of the person within the target image. 8. In the image selection device according to any one of 1 to 6 above, The image selection means includes: displaying an input screen on a terminal for allowing a user to input weightings of the posture information and the other information; An image selection device that classifies the plurality of target images or selects the at least one target image using the weighting input on the input screen. 9. The computer an information generation process for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection process for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; Image selection method. 10. In the image selection method according to claim 9, The computer includes: A query acquisition process is further performed to acquire a query image including a person; The information generating process further includes generating posture information and the other information of the person included in the query image; The image selection method further comprises the step of selecting the at least one target image using the posture information and the other information of the query image and each of the plurality of target images in the image selection process. 11. In the image selection method according to claim 9 or 10, An image selection method, wherein the other information includes color information of at least one of the person and the person's clothing. 12. In the image selection method according to claim 11, A weight is set for each of the plurality of body parts of the person, The image selection method, wherein the computer classifies the plurality of target images or selects at least one target image from the plurality of target images using the weighting in the image selection process. 13. In the image selection method according to any one of claims 9 to 12, The method of image selection, wherein the other information includes color information of a region of interest that includes both the person and the surroundings of the person. 14. In the image selection method according to any one of claims 9 to 13, The other information includes at least one of the person's face, gender, age group, and body shape. 15. In the image selection method according to any one of claims 9 to 14, The method of claim 1, wherein the other information is a position of the person within the target image. 16. In the image selection method according to any one of claims 9 to 14, In the image selection process, the computer displaying an input screen on a terminal for allowing a user to input weightings of the posture information and the other information; The image selection method further comprises: classifying the plurality of target images using the weightings input on the input screen; or selecting the at least one target image. 17. On the computer, an information generating function for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection function that classifies the plurality of target images using the posture information and the other information, or selects at least one target image from the plurality of target images; A program that allows you to have 18. In the program according to 17 above, The computer is further provided with a query acquisition function for acquiring a query image including a person; The information generation function further generates posture information and the other information of the person included in the query image; The image selection function further includes a program for selecting the at least one target image using the posture information and the other information of the query image and each of the plurality of target images. 19. In the program according to 17 or 18 above, The other information includes color information of at least one of the person and the person's clothing. 20. In the program according to claim 19, A weight is set for each of the plurality of body parts of the person, A program in which the image selection function uses the weighting to classify the plurality of target images or select at least one target image from the plurality of target images. 21. In the program according to any one of the above 17 to 20, The other information includes color information of a target area that includes both the person and the area surrounding the person. 22. In the program according to any one of 17 to 21 above, The other information includes at least one of the person's face, gender, age group, and body shape. 23. In the program according to any one of the above 17 to 22, The other information is the position of the person within the target image. 24. In the program according to any one of the above 17 to 22, The image selection function includes: displaying an input screen on a terminal for allowing a user to input weightings of the posture information and the other information; A program for classifying the plurality of target images or selecting the at least one target image using the weightings input on the input screen. [Explanation of symbols]
[0128] 1. Image Processing System 10 Image processing device (image selection device) 11 Skeleton detection section 12 Feature Calculation Unit 13 Recognition part 100 Image processing device (image selection device) 101 Image acquisition unit 102 Skeletal structure detection unit 103 Feature Calculation Unit 104 Classification Department 105 Search Department 106 Input section 107 Display section 108 Height Calculator 110 Database 200 Cameras 300, 301 Human body model 401 2D Skeleton Structure 402 3D Human Body Model 610 Information generation section 620 Image Selection 630 Image storage unit 640 Query Acquisition Unit 700 terminals
Claims
1. an information generating means for generating, from each of a plurality of target images, posture information of a person included in the target image and other information related to the person; an image selection means for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; Equipped with the other information includes color information of at least one of the person and an item worn by the person; A weight is set for each of the plurality of body parts of the person, The image selection means classifies the plurality of target images using the weighting, or selects at least one target image from the plurality of target images. Image selection device.
2. An information generating means for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection means for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; Equipped with The image selection means includes: displaying an input screen on a terminal for allowing a user to input weightings of the posture information and the other information; An image selection device that classifies the plurality of target images or selects the at least one target image using the weighting input on the input screen.
3. 3. The image selection device according to claim 1, The method further includes: The information generating means further generates posture information and the other information of the person included in the query image; The image selection means further uses the posture information and the other information of the query image and each of the plurality of target images to select the at least one target image.
4. The image selection device according to any one of claims 1 to 3, The other information includes color information of a region of interest that includes both the person and the surroundings of the person.
5. The image selection device according to any one of claims 1 to 4, The other information includes at least one of the person's face, gender, age group, and body shape.
6. The image selection device according to any one of claims 1 to 5, The other information is a position of the person within the target image.
7. The computer an information generation process for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection process for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; Do the following: the other information includes color information of at least one of the person and an item worn by the person; A weight is set for each of the plurality of body parts of the person, In the image selection process, the computer classifies the plurality of target images using the weighting, or selects at least one target image from the plurality of target images. Image selection method.
8. A computer comprising: an information generation process for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection process for classifying the plurality of target images or selecting at least one target image from the plurality of target images using the posture information and the other information; Do the following: In the image selection process, the computer displaying an input screen on a terminal for allowing a user to input weightings of the posture information and the other information; The image selection method further comprises: classifying the plurality of target images using the weightings input on the input screen; or selecting the at least one target image.
9. On the computer, an information generating function for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection function that classifies the plurality of target images using the posture information and the other information, or selects at least one target image from the plurality of target images; Let them have the other information includes color information of at least one of the person and an item worn by the person; A weight is set for each of the plurality of body parts of the person, The image selection function uses the weighting to classify the plurality of target images or to select at least one target image from the plurality of target images. program.
10. A computer comprising: an information generating function for generating posture information and other information related to a person included in each of a plurality of target images from the target images; an image selection function that classifies the plurality of target images using the posture information and the other information, or selects at least one target image from the plurality of target images; Let them have The image selection function includes: displaying an input screen on a terminal for allowing a user to input weightings of the posture information and the other information; A program for classifying the plurality of target images or selecting the at least one target image using the weightings input on the input screen.
Citation Information
Patent Citations
Image information updating system, image inputting device, image processing device, image updating device, image information updating method, image information updating program, and recording medium recording the program
JP2006260405A
Image evaluation device and method, and program
JP2007080014A
Apparatus and method for retrieving object poses
JP2014522035A
Robbery prevention system and software
JP2016057908A
Obtaining metrics for position using frames classified by associative memory
JP2016058078A
