JP2007219603A - Person tracking device, person tracking method and person tracking program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a person tracking device for highly precisely tracking a person even when a plurality of persons exist or approach each other in a pickup image. <P>SOLUTION: This person tracking device is provided with: a moving distance distribution information storage means 16 for storing moving distance distribution information showing the distribution of face size information changing according to the movement of a person and face moving distance information; a face size information acquisition means 14 for acquiring the face size information of the person; a face moving distance information acquisition means 15 for acquiring the face moving distance information of the person; and an identical person estimating means 17 for estimating whether the person is the identical person or a different person. The moving distance distribution information includes an identical person estimation area and a different person estimation area. The identical person estimation means 17 estimates whether the person is the identical person or a different person by judging which of the identical person estimation area and the different person estimation area, the location in the moving distance distribution information to be determined by the face size information and the face moving distance information exists. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a person tracking device, a person tracking method, and a person tracking program that detect a person's face from a captured image and track the detected person.

A conventional person tracking device extracts a person area from a photographed image, and based on a proportion of the human body (for example, the size ratio of each part of the human body such as the head, upper body, lower body, etc.) Dividing the image into blocks, one of the divided blocks (for example, the block in which the head is reflected) is used as a template, and each successive captured image is sequentially matched with the template, so that a person can be identified from the captured images. The movement is detected and the movement is tracked (see, for example, Patent Document 1).
JP 2004-94518 A

  However, in the conventional person tracking device, in the person tracking, when there are a plurality of persons in the captured image and they are approaching or intersecting, a tracking failure such as connecting others to each other occurs. There was a problem.

  The present invention has been made to solve the conventional problems, and performs high-precision person tracking even when a plurality of persons are present in a captured image or when a plurality of persons are close to each other. It is an object to provide a person tracking device, a person tracking method, and a person tracking program.

  The person tracking device of the present invention detects a person's face from a captured image, and in the device for tracking the detected person, a movement distance indicating a distribution of face size information and face movement distance information that changes as the person moves Moving distance distribution information storing means storing distribution information, face size information acquiring means for acquiring face size information of the person, face moving distance information acquiring means for acquiring face movement distance information of the person, The same person estimation means for estimating whether the person is the same person or another person, and the movement distance distribution information is the same person estimation area for estimating the person as the same person, and another person Another person estimation area to be estimated, and the same person estimation means has a position in the movement distance distribution information determined from the face size information and the face movement distance information in the same person estimation area Luke, or by either present in the by Person estimation region, whether the person is the same person, is configured to estimate whether the other person.

  With this configuration, even if there are a plurality of persons on the screen and persons are close to each other, highly accurate person tracking can be performed.

  Further, the person tracking device of the present invention comprises tracking history recording means for recording a series of face images estimated as the same person by the same person estimating means as a tracking history.

  With this configuration, a person tracking history can be created and recorded.

  The person tracking device according to the present invention further includes a tracking history correcting unit that corrects the tracking history by classifying the tracking history into the same person and different persons when the tracking history includes face images of a plurality of persons. It is configured.

  With this configuration, the created person tracking history can be corrected and recorded.

  Further, the person tracking device of the present invention is configured to include face orientation information acquisition means for acquiring face orientation information from the detected face.

  With this configuration, it is possible to perform highly accurate person tracking by performing person tracking from face orientation information.

  Further, the person tracking device of the present invention is configured to include imaging means angle information acquisition means for acquiring installation angle information or angle of view information of the imaging means.

  With this configuration, the installation angle information or the angle-of-view information of the imaging unit can be reflected in the movement distance distribution information, and highly accurate person tracking can be performed.

  In the person tracking device of the present invention, the movement distance distribution information storage means includes a plurality of movement distance distribution information and a movement distance distribution information selection means for selecting the plurality of movement distance distribution information. ing.

  With this configuration, it is possible to select optimal travel distance distribution information and perform highly accurate person tracking.

  Further, in the person tracking device of the present invention, two adjacent face images of the detected continuous face images are compared with each other, and whether the face is the same person when the change in the face size is large or the change in the face direction is large. It is configured to include a tracking history confirmation means for confirming the user with the user.

  With this configuration, the user can easily check and correct the tracking result on the screen, and a highly accurate person tracking result can be created.

  Further, the person tracking method of the present invention is a person tracking method for detecting a person's face from a captured image and tracking the detected person, and distribution of face size information and face moving distance information that change as the person moves. Storing the movement distance distribution information indicating, the step of acquiring the face size information of the person, the step of acquiring the face movement distance information of the person, and whether the person is the same person or a different person The movement distance distribution information includes the same person estimation area for estimating the person as the same person, and another person estimation area for estimating the person as a different person, and the face size information Depending on whether the position in the movement distance distribution information determined from the face movement distance information exists in the same person estimation area or the different person estimation area Whether it is the same person, to estimate if it was another person.

  With this configuration, even if there are a plurality of persons on the screen and persons are close to each other, highly accurate person tracking can be performed.

  The person tracking program of the present invention causes a computer to execute the person tracking method.

  With this configuration, even if there are a plurality of persons on the screen and persons are close to each other, highly accurate person tracking can be performed.

The present invention can track a person with high accuracy even when there are a plurality of persons in the captured image or when they are approaching. Further, it is possible to provide a person tracking device, a person tracking method, and a person tracking program that allow a user to check and correct a tracking result and create a highly accurate person tracking result.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing the configuration of the person tracking apparatus according to Embodiment 1 of the present invention.

  In FIG. 1, the person tracking apparatus according to the first embodiment of the present invention includes an image pickup means 11 such as a camera for picking up an image of a monitoring place, etc., and a video signal output from the image pickup means 11 converted into digital image data etc. A video input unit 12 that is input to the processing device, a face detection unit 13 that receives the image data output from the video input unit 12 and detects a face in the captured image, and a face detected by the face detection unit 13 Face size information acquisition means 14 for acquiring face size information from the face, face movement distance information acquisition means 15 for acquiring face movement distance information from the face detected by the face detection means 13, and face size information and face movement distance information. Information of the movement distance distribution information storage means 16 that stores the movement distance distribution information 21 (see FIG. 8) showing the distribution of the above, the face size information acquisition means 14 and the face movement distance information acquisition means 15. Thus, the same person estimation means 17 for estimating whether the faces detected by the face detection means 13 are the same person based on the estimation of whether or not they are the same face, and the same person estimation means 17 estimates the same person. A tracking history recording means 18 for recording a series of images as a tracking history, and when a series of face images includes a plurality of face images from the recorded tracking history, A tracking history correction unit 19 that corrects the tracking history by classifying into groups, a tracking history confirmation unit 20 that prompts the user to check according to a comparison result between adjacent images of two detected consecutive face images, And a display unit 22 for displaying the tracking history recorded by the tracking history recording unit 18.

  Next, each of the above means will be described in more detail.

  The imaging means 11 is configured by a device such as a surveillance camera that outputs video in real time, and outputs a video signal. The imaging means 11 may be a device such as a hard disk recorder that outputs recorded video.

  The video input means 12 is a device that connects the image pickup means 11 and the processing device, such as data movement, conversion of a video signal from the image pickup apparatus 11 into a digital image 51 (RGB color image) as shown in FIG. I do.

  The face detection means 13 detects the face area 52 of the person 53 from the digital image 51 as shown in FIG.

  As a method for detecting the face of the person 53, whether or not the face is a face by learning in advance with respect to a template matching method using a face template such as a front face or a side face, or several types of gray rectangle filters for extracting facial features. There is an Ada Boost method that determines a discrimination function in an identification space for estimating whether it is a non-face, and combines this discriminator in several stages to perform final face estimation. It is not limited.

  The face size information acquisition means 14 acquires the face width W and height H, which are face size information, of the face area 52 in the display area of the display means 22 (see FIG. 7). In other words, the face size information acquisition unit 14 obtains the number of pixels in the horizontal direction and the vertical direction in the digital image 51 of the face region 52 detected by the face detection unit 13, and determines the face width W and height H respectively. Size information.

  As the face size information, average values of the width W and the height H in a plurality of continuous digital images 51 may be used. The total number of pixels may be used as face size information.

  The face movement distance information acquisition means 15 obtains the centroid position of each face area 52 from the face area 52 currently detected by the face detection means 13 and the face area 52 detected a short time ago, and the face position information is obtained from each centroid position. Get travel distance information. Here, the movement distance of the face is the pixel coordinate indicating the center of gravity of the currently detected face area.

  And the pixel coordinates indicating the barycentric position of the face area before the minute time

  Then,

And it is required.

  Here, the minute time may be a time for one frame between the frame of the current captured image and a frame one frame before, or may be a time for several frames.

  The movement distance distribution information storage unit 16 stores the face size information obtained by the face size information acquisition unit 14 and the face movement distance information obtained by the face movement distance information acquisition unit 15. Further, the movement distance distribution information storage unit 16 stores movement distance distribution information 12 indicating the distribution of the face size determined by the distance between the imaging unit 11 and the face position of the person and the movement distance of the face. In the captured image, the face size of the person who is close to the image pickup means 11 is large, while the face size of the person who is far from the image pickup means 11 is displayed small. When the person moves at the same speed, the moving distance of the face near the imaging means 11 is long, while the moving distance of the face far from the imaging means 11 is displayed short. Thus, in the captured image, there is a relationship that the movement distance is long when the face size of the person is large, and the movement distance is short when the face size of the person is small.

FIG. 8 is a diagram showing the relationship between the face size information and the movement distance information before and after the minute time, and is an example of the movement distance distribution information 21. In FIG. 8, for example, in the case of a person A whose face size is SA, it is possible to move the range from the distance 0 to the distance L1, and in the case of the person B whose face size is SB, the distance from the distance 0 This indicates that it is possible to move the range up to L2. Therefore, the person A does not move to a distance L3 longer than the distance L1, for example. When the distance is L3 as the movement distance information, it can be determined that it is not the movement of the person A but the movement of the person B other than the person A.

  Therefore, in the movement distance distribution information 21 of FIG. 8, if the position obtained from the face size information and the face movement distance information is in the same person estimation area 61 below the same person estimation curve 63, it can be estimated as the same person. If it is in the different person estimation area 62 above the same person estimation curve 63, it can be estimated as a different person.

  The relationship between the face size information and the movement distance information in FIG. 8 includes the movement conditions (speed, direction, etc.) of the person, the installation conditions of the video input means 11 (installation position and orientation (horizontal angle, vertical angle), etc.), etc. The face size information and the movement distance information (same person estimation curve 63) must be set according to conditions.

  For example, when the shooting direction of the video input unit 11 and the moving direction of the person are parallel, the change in the face size information is large and the change in the face moving distance information is small. On the other hand, when the shooting direction of the video input means 11 and the moving direction of the person are at right angles, the change in the face size information is small and the change in the face moving distance information is large.

  The same person estimation means 17 estimates whether the currently detected face image and the face image detected a minute before are the same person's face. In this estimation, the position determined from the acquired face size information and face movement distance information is the same person estimation area 61 or another person estimation area 62 of the movement distance distribution information 21 stored in the movement distance distribution information storage means 16 in advance. Of these, it is performed depending on which region exists.

  For example, in FIG. 8, when the obtained face size information is SA and the face movement distance information is L4 (<L1), it is in the same person estimation area 61 and is estimated to be the same person. When the face size information is SA and the face movement distance information is L3 (> L1), the face size information is in the different person estimation area 62, so it is estimated that the person is a different person.

  The tracking history recording means 18 records a series of face images estimated as the same person or different persons and various tracking information corresponding to each face image as tracking history information, and displays the recorded contents on the display means 22. To do.

  FIG. 9 is an explanatory diagram when the recorded content of the tracking history information is displayed on the display means 22.

  The face images 101a, 101b, 101c, 101d, and 101e are a group of face images when the same person is estimated by the same person estimating means 17, and are displayed in the right direction on the screen in order from the oldest detected time. In the example shown in FIG. 9, a face image, detected time, and face size information are displayed on the screen as tracking information corresponding to each face image. In FIG. 9, for example, the face image 101a detected at time t0 has face size information of 30 pix, and the face image 101b detected at time t1 indicates that face size information is 38 pix. .

  In addition to this, the tracking information includes the center position of the detected face on the screen, the moving direction after the person appears, the moving speed of the person, the collation of the detected face image with the registered face image. You may display the face recognition result etc. when performing.

  As described above, a series of face images estimated by the same person estimating means 17 as the same person can be recorded in the tracking history recording means 18 as tracking history information.

  Next, a method for correcting the tracking history information recorded in the tracking history recording means 18 will be described.

  The tracking history correction means 19 corrects the tracking history information by classifying the tracking history information recorded in the tracking history recording means 18 into a group of the same person and other groups.

  10 and 11 are diagrams for explaining the tracking history information correction method.

  In FIG. 10, face images 111 a, 111 b, 111 c, 111 d, and 111 e indicate face image groups that are estimated to be the same person by the same person estimation means 17.

    In FIG. 10, the face size information from time t0 to time t3 increases from 30 pix to 58 pix approximately in proportion to the change in time, but the face size information from time t3 to time t4 increases from 58 pix to 80 pix. It has changed greatly.

    The reason for this large change is that when the same person moves faster than the moving speed from time t3 to time t4, the face size information becomes large, or the person in the face image at time t3 and time t4 is different. There is a case.

    If the position obtained from the face size information and the face movement distance information is in the same person estimation area 61 shown in FIG. 8, the same person estimation means 17 may estimate another person as the same person.

    Generally, when the change in tracking information is large within adjacent minute times, the possibility of being the same person is reduced.

  Therefore, the user can determine whether the person is the same person or a different person by checking the change of the tracking information for the face images adjacent in the face image group estimated to be the same person. It is possible to perform tracking with high accuracy.

  The tracking history confirmation unit 20 displays a message on the display unit 22 that recommends confirmation to the user when there is a large change in tracking information for adjacent face images in the face image group of the same person.

  The user may determine whether or not they are the same person based on the displayed tracking history information, or may determine based on a message from the tracking history confirmation unit 20. By displaying the message, the user can easily check whether the tracked person is the same person, and the burden can be reduced. The message may be displayed when the change in the tracking information within a very short time is greater than or equal to a certain threshold value or less. The means for recommending the user to confirm is not limited to the message, but may be a frame display, an alarm sound, or the like.

  In the example illustrated in FIG. 10, the tracking history confirmation unit 20 has the face size of the face image 111d at time t3 of 58 pix in the face images 111a, 111b, 111c, 111d, and 111e estimated by the same person estimation unit 17 as the same person. Since the face size of the face image 111e at time t4 after a minute time is 80 pix, and the change is large compared to the change in the face size before that, a confirmation frame 112 is displayed on the screen of the display means 22, and the user Recommended to check.

  When the user confirms and determines that the face image 111d and the face image 111e are not the same person, as shown in FIG. 11, the user selects a face image 111e that does not belong to the same person (the face image 111e on the screen). The selection frame 113 is displayed.

  FIG. 11 shows an example in which the user determines that the face images 111a, 111b, 111c, and 111d are the person A and the face image 111e is the person B.

  As described above, it is possible for the user to select the face image of the same person and the face image of another person for the face images selected by the same person estimating means 17 as the group of the same person.

  Next, a tracking history correction method will be described.

  In FIG. 13, the tracking history information 131 is a time series display of face images estimated by the same person estimation means 17 as the same person. Here, reference numeral 135 denotes a face image of the person A, and 136 denotes a face image of the person B. A selection frame 134 indicated by a thick frame indicates a face image that is designated on the screen by the tracking history correcting unit 19 as determined by the user as a different person.

  The face images designated on the screen are classified into two groups of the same person group and another person group (face image indicated by the selection frame 134) from the tracking history information 131.

  As a result, the tracking history information 131 including the person A and the person B includes two face images of the correction tracking history information 132 including only the person A and the correction tracking history information 133 including only the person B not including the person A. Classified into series.

  As described above, the contents of the tracking history information 131 estimated to be the same person are reviewed, classified into groups of the same person and other persons, the tracking history is corrected, and a highly accurate tracking history is stored in the tracking history recording means 18. It can be stored, and the performance of the person search using this database can be improved.

  In the tracking history correction means 19, the tracking history information recorded by the tracking history recording means 18 may be confirmed while the user confirms the contents of the display means 22, or a template matching method using a template or the like. The image recognition means may be mounted on the tracking history correction means 19.

  Next, the operation of the person tracking apparatus according to the first embodiment will be described with reference to the flowchart shown in FIG.

  The video input unit 11 captures a monitoring place or the like, and the video input unit 12 converts the video signal from the video input unit 11 into a digital image 51 or the like (step S1).

  The face detection means 13 detects the face area 52 of the person 53 from the digital image 51 (step S2).

  The face size information acquisition unit 14 acquires face size information from the face area detected by the face detection unit 13 (step S3), and the face movement distance information acquisition unit 15 acquires the face size information acquired by the face size information acquisition unit 14. The face movement distance information is acquired from (step S4).

  The acquired face size information and face movement distance information are stored in the movement distance distribution information storage means 16.

The same person estimation means 17 includes the face size information, face movement distance information stored in the movement distance distribution information storage means 16, and the imaging means and the face position of the person stored in advance in the movement distance distribution information storage means 16. The movement distance distribution information 21 showing the distribution of the face size information determined by the distance and the face movement distance information is referred to (step S5), and it is estimated whether the persons in the captured image are the same person or different persons. (Step S6).

  The tracking history recording means 18 records a series of face images estimated as the same person by the same person estimating means 17 and various tracking information corresponding to each face image as tracking history information (step S7), and the tracking history is recorded. The screen is displayed on the display means 22 (step S9).

  The tracking history recording means 18 records a series of face images estimated as different persons by the same person estimating means 17 and various tracking information corresponding to each face image as tracking history information (step S8). Is displayed on the display means 22 as required (step S9).

  The tracking history confirmation unit 20 confirms whether there is an image with a large change in tracking information for adjacent face images in the series of tracking history information estimated by the same person estimation unit 17 as the same person (step S10). If there is an image whose change is large (step S10, Yes), a frame is displayed on the corresponding image (step S11).

  If there is no image with a large change in tracking information (step S10, No), the process proceeds to step S12.

  Whether or not the change in the tracking information is large may be determined using a threshold value determined as appropriate.

  When there is a face image selected by the user based on the frame display of the tracking history confirmation unit 20, the tracking history correction unit 19 corrects the tracking history (separating the face image selected by the user from the tracking history) (step S12: Yes). Step S13) is performed, the group that does not belong to the same person is stored in the tracking history recording unit 18 (Step S14), the face image that the user did not select is corrected as the group of the same person, and the tracking history is recorded. 18 (step S14).

  When there is no face image selected by the user (No at Step S12), the tracking history correcting unit 19 stores the tracking history in the tracking history recording unit 18 (Step S14).

  As described above, in the first embodiment of the present invention, whether the detected person is the same person or a different person, or the position determined from the acquired face size information and face movement distance information is the movement distance distribution information. It can be estimated according to which of the same person estimation area 61 and the different person estimation area 62 of the movement distance distribution information 21 stored in the storage means 16 exists, and a series of estimated persons. Can be recorded as a tracking history, and the recorded tracking history can be corrected.

(Embodiment 2)
The second embodiment of the present invention is different from the first embodiment in that a face orientation information acquisition unit 23 and a movement distance distribution selection unit 24 are added, and the movement distance distribution information storage unit 16 stores a plurality of movement distance distribution information 21. The storage is different.

  FIG. 3 shows the configuration of the person tracking apparatus according to Embodiment 2 of the present invention.

  In FIG. 3, the same components as those of the person tracking device of the first embodiment are denoted by the same reference numerals, and the added components will be described below.

  The face orientation information acquisition unit 23 acquires the face orientation information on the screen of the display unit 22 of the face area 52 detected by the face detection unit 13.

  As shown in FIG. 14A, in the method of detecting the face direction of the person 53, two small ellipses are arranged in the detected face information by the template matching method used in the face detection method or the like. As shown in FIG. 14B, the angle θ formed by the line connecting the centers of gravity of the areas determined to be eyes and the horizontal line is acquired as face orientation information. The acquired face orientation information is stored in the movement distance distribution information storage means 16.

  Note that the above face orientation information acquisition method is an example, and the present invention is not limited to this.

  The movement distance distribution selection means 24 selects the movement distance distribution information 21 according to the face direction information when the face direction information is stored in the movement distance distribution information storage means 16, and based on the selected movement distance distribution information 21, The same person estimation means 17 estimates whether or not the detected face images are the same face.

  In general, when the face orientation information is around 0 ° (the face appears on the front face in the display means 22), the distance traveled on the display means 22 within a short time is small, and the face orientation information is large ( When the face of the display means 22 is facing sideways), the distance traveled within a minute time is increased.

  FIG. 15 is a diagram for explaining the principle that the moving distance in the display means 22 differs depending on the face orientation information.

  FIG. 15A is a plan view showing a state in which a person whose face direction information is near 0 ° and FIG. 15B is a person whose face direction information is large move along the distance L1, respectively. c) is the captured image of FIG. 15A, and FIG. 15D is the captured image of FIG. The imaging means 11 is set so that the height from the horizon is substantially the same as the face height and the optical axis is horizontal.

  As shown in FIG. 15 (c), even if the person whose face orientation information is around 0 ° moves the distance L1, the face moving distance on the display means 22 is almost zero. On the other hand, as shown in FIG. 15D, in the case of a person with large face orientation information, when the distance L1 is moved, the distance L2, which is the horizontal component, is acquired as face movement distance information on the display means 22.

  Therefore, the face moving distance is larger when the face direction information is large than when the face direction information is around 0 °.

  FIG. 16 is a diagram showing the relationship between the face size information and the face movement distance information when the face direction information is large, selected by the movement distance distribution selecting unit 24, and when the face direction information is around 0 °. Correction is performed to add to the face movement distance information an increase in face movement distance information due to the occurrence of a face orientation.

  The same person estimation means 17 moves from the face size information and the center position of the currently detected face area and the center position detected a short time ago by the same method as the method performed in the first embodiment of the present invention. Get distance information. If the position obtained from the acquired face size information and face movement distance information is in the same person estimation area 61 below the same person estimation curve 63, the position is estimated as the same person, and the position above the same person estimation curve 63 is determined. If it is in the person estimation area 62, it is estimated as a different person.

The movement distance distribution information storage means 16 preferably has a plurality of movement distance distribution information 21 corresponding to the face orientation information.

  Further, when the face direction is near 0 °, a threshold is set, and within the threshold, the movement distance distribution information 21 of the face direction is adopted, and when the threshold is exceeded, the movement distance distribution information 21 according to the face direction information is adopted. And the threshold value may be set as appropriate.

  FIG. 12 shows an example in which the same person is estimated using face orientation information.

  The same person estimation means 17 estimates the five images of the face images 121a, 121b, 121c, 121d, and 121e as the same person, but the tracking history confirmation means 20 indicates that the face orientation of the face image 121d at time t3 is the same. At 0 °, the face orientation of the face image 121e at time t4 after a minute time is 15 °, and the change is large compared to the change in the face orientation before that, so the selection frame 122 is displayed on the screen of the display means 22. Display and encourage users to confirm.

  When the user confirms and determines that the face image 121d and the face image 121e are not the same person, as shown in FIG. 12, the user selects a face image 121e that does not belong to the same person (the face image 121e on the screen). And a thick selection frame 123 is displayed.

  FIG. 12 illustrates an example in which the user estimates the face images 121a, 121b, 121c, and 121d as the person A and the face image 121e as the person B.

  Since the operations of the tracking history recording unit 18 and the tracking history correcting unit 19 are the same as those of the person tracking apparatus according to the first embodiment of the present invention, the description thereof is omitted.

  Next, the operation of the person tracking apparatus according to the second embodiment of the present invention will be described based on the flowchart shown in FIG.

  Since the operation of the second embodiment of the present invention is basically the same as the operation of the first embodiment, different points will be described.

  The face direction information acquisition unit 23 acquires the face direction information on the screen of the display unit 22 of the face area 52 detected by the face detection unit 13 (step S21). The acquired face orientation information is stored in the movement distance distribution information storage means 16 together with the face size information and the face movement distance information.

  The movement distance distribution information storage means 16 selects the movement distance distribution information 21 stored in advance according to the presence / absence of face direction information and the content (angle etc.) of the face direction information (step S22).

  Since step S6 and subsequent operations are the same as those in the first embodiment of the present invention, description thereof is omitted.

  As described above, in the second embodiment of the present invention, the position determined from the acquired face size information, face movement distance information, and face direction information, whether the detected person is the same person or a different person, It can be estimated depending on which region of the same person estimation area 61 and the different person estimation area 62 of the movement distance distribution information 21 stored in the movement distance distribution information storage means 16 is present. The estimated series of face images can be recorded as a tracking history, and the recorded tracking history can be corrected.

(Embodiment 3)
The third embodiment of the present invention is different from the first embodiment in that an imaging means angle information acquisition means 25 is added.

  In FIG. 5, the same components as those of the person tracking apparatus according to the first embodiment of the present invention are denoted by the same reference numerals, and the additional components will be described below.

  5, the person tracking apparatus according to the third embodiment of the present invention includes an imaging unit angle information acquisition unit 25 that acquires imaging unit angle information such as an installation angle of the imaging unit 11.

  Generally, the imaging means 11 in the person tracking device is attached at a high place so that the person can be looked down from above. Accordingly, the image pickup means 11 is attached with an angle (hereinafter referred to as an image pickup means installation angle) where the optical axis of the image pickup means 11 is in a horizontal plane.

  When the imaging means installation angle is small, the face movement distance of a person who moves on the display means 22 within a very short time is small, while when the imaging means installation angle is large, the face movement distance is large.

  17 and 18 are diagrams for explaining the relationship between the imaging means installation angle and the face movement distance.

  FIG. 17 shows a horizontal direction of the optical axes of the image pickup means 11a and 11b for a person 53 moving toward the image pickup means 11a having a small image pickup means installation angle 54a and an image pickup means 11b having a large image pickup means installation angle 54b. It is the figure shown from. Here, the person 53 is imaged by the two imaging units 11a and 11b.

  FIG. 18A shows an image picked up by the image pickup means 11a, and FIG. 18B shows an image picked up by the image pickup means 11b.

  In the case of imaging with the imaging unit 11a, the distance L3 toward the imaging unit 11a hardly appears in the image of FIG. 18A, and lateral movement appears in the image as the distance L4.

  In the case of imaging with the imaging unit 11b, the distance L3 appears as the distance L5 in the image of FIG. 18B, and the lateral movement appears as the distance L4 as in the imaging device 11a.

  Thus, when the imaging means installation angle 54a is small, the face movement distance 55 of the person who moves on the display means 22 within a short time is small, while when the imaging means installation angle 54b is large, the face movement distance 55 is growing.

  In addition, the imaging unit 11 has a unique angle of view, and a range that can be captured is determined by the angle of view. Even when a person or the like moves in the same time, if the imaging unit 11 has a wide angle of view, the display device When the face moving distance on 22 is short and the angle of view is small, the face moving distance becomes large.

  The imaging means angle information acquisition means 25 acquires information on the imaging means installation angles 54a and 54b of the imaging means 11 and imaging means angle information such as field angle information.

  The movement distance distribution selection means 24 selects the movement distance distribution information 21 reflecting the imaging means angle information from the movement distance distribution information 21 stored in the movement distance distribution information storage means 16.

  FIG. 19 is a diagram showing the relationship between the face size information and the face movement distance information when the imaging means angle information selected by the movement distance distribution selection means 24 is reflected. FIG. 19 shows an example in which the face movement distance information increases with the same face size information by reflecting the imaging means angle information. However, depending on the imaging means angle information, the face with the same face size information is shown. The travel distance information may be reduced.

  The method for estimating whether or not they are the same person from the detected face image is the same as the method described in the first embodiment of the present invention, and the description thereof is omitted here.

  Next, the operation of the person tracking apparatus according to the third embodiment of the present invention will be described based on the flowchart shown in FIG.

  Since the operation of the third embodiment of the present invention is basically the same as the operation of the first embodiment, different points will be described.

  The imaging means angle information acquisition means 25 acquires imaging means angle information such as the imaging means installation angle and the angle of view information of the imaging means 11 (step S31), and the acquired imaging means angle information includes face size information and face movement. It is stored in the movement distance distribution information storage means 16 together with the distance information.

  The movement distance distribution information storage means 16 selects the movement distance distribution information 21 stored in advance according to the content of the imaging means angle information (step S22).

  Since step S5 and subsequent operations are the same as those in the first embodiment of the present invention, the description thereof is omitted.

  As described above, the third embodiment of the present invention determines whether the detected person is the same person or a different person, or a position determined from the acquired face size information, face movement distance information, and angle information of the imaging means. Can be estimated according to which of the same person estimation area 61 and the different person estimation area 62 of the movement distance distribution information 21 stored in the movement distance distribution information storage means 16 is present. A series of face images estimated as persons can be recorded as a tracking history, and the recorded tracking history can be corrected.

  The person tracking device, the person tracking method, and the person tracking program according to the present invention can perform tracking of a person with high accuracy even when there are a plurality of persons in the captured image or when they are close to each other. In addition, the user can easily check and correct the tracking results, and can create highly accurate person tracking results. This database is used to monitor human behavior and search for people. It is useful as a person tracking device for constructing

1 is a block diagram showing a configuration of a person tracking apparatus according to a first embodiment of the present invention. The flowchart which shows the operation | movement procedure of the person tracking device of Embodiment 1 of this invention. The block diagram which shows the structure of the person tracking device of Embodiment 2 of this invention. The flowchart which shows the operation | movement procedure of the person tracking device of Embodiment 2 of this invention. The block diagram which shows the structure of the person tracking device of Embodiment 3 of this invention. The flowchart which shows the operation | movement procedure of the person tracking device of Embodiment 3 of this invention. The figure explaining detection of face size information The figure which shows the relationship between face size information and face movement distance information Diagram explaining list display of tracking history Figure explaining the display of tracking history confirmation contents Diagram explaining how to modify tracking history Diagram explaining how to modify tracking history Diagram explaining how to modify the tracking history database The figure explaining the acquisition method of face angle A figure explaining the moving distance on the screen depending on the face orientation The figure which shows the relationship between face size information and face movement distance information when there is a face orientation The figure explaining the installation angle of an imaging means The figure explaining the movement distance on the screen by the installation angle of an imaging means The figure which shows the relationship between the face size information reflecting the imaging means angle information, and face movement distance information

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Image pickup means 12 Image | video input means 13 Face detection means 14 Face size information acquisition means 15 Face movement distance information acquisition means 16 Movement distance distribution information storage means 17 Same person estimation means 18 Tracking history recording means 19 Tracking history correction means 20 Tracking history confirmation Means 21 Movement distance distribution information 22 Display means 23 Face orientation information acquisition means 24 Movement distance distribution selection means 25 Imaging means angle information acquisition means

Claims (9)

In a person tracking device that detects a person's face from a captured image and tracks the detected person,
A moving distance distribution information storing means storing moving distance distribution information indicating a distribution of face size information and face moving distance information that change with the movement of a person;
Face size information acquisition means for acquiring face size information of the person;
Face movement distance information acquisition means for acquiring the face movement distance information of the person;
The same person estimation means for estimating whether the person is the same person or another person,
The movement distance distribution information has the same person estimation area for estimating that the person is the same person, and another person estimation area for estimating that the person is a different person,
Whether the same person estimation means has a position in the movement distance distribution information determined from the face size information and the face movement distance information in the same person estimation area or in another person estimation area The person tracking apparatus characterized by estimating whether the person is the same person or a different person. 2. The person tracking apparatus according to claim 1, further comprising tracking history recording means for recording a series of face images estimated by the same person estimating means as the same person as a tracking history. The tracking history correcting means for correcting the tracking history by classifying the tracking history into the same person and another person when the tracking history includes face images of a plurality of persons. 2. The person tracking device according to 2. 4. The person tracking apparatus according to claim 1, further comprising face direction information acquisition means for acquiring face direction information from the detected face. The person tracking apparatus according to claim 1, further comprising an imaging unit angle information acquisition unit that acquires installation angle information or angle of view information of the imaging unit. 6. The travel distance distribution information storage means comprises a plurality of travel distance distribution information and a travel distance distribution information selection means for selecting the plurality of travel distance distribution information. The person tracking device according to Crab. Tracking history confirmation means for comparing two adjacent face images of detected consecutive face images and confirming to the user whether the face is the same person when the face size change is large or the face orientation change is large. The person tracking device according to claim 1, further comprising a person tracking device. In a person tracking method for detecting a person's face from a captured image and tracking the detected person,
Storing movement distance distribution information indicating a distribution of face size information and face movement distance information that changes with the movement of a person;
Obtaining face size information of the person;
Obtaining the face movement distance information of the person;
Estimating whether the person is the same person or another person,
The movement distance distribution information has the same person estimation area for estimating that the person is the same person, and another person estimation area for estimating that the person is a different person,
The person is the same person depending on whether the position in the movement distance distribution information determined from the face size information and the face movement distance information is in the same person estimation area or in the different person estimation area Or a person tracking method characterized by estimating whether the person is a different person. A person tracking program for causing a computer to execute the person tracking method according to claim 8.

Images