JP2005099953A - Image processor, object tracing system therewith, image processing method and image processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To more accurately perform a process for tracing an object such as a face from inside moving image data. <P>SOLUTION: The process for tracing the face from inside the moving image data comprising a static image data string is performed. When detecting the face by search of the whole range inside static image data in a search part, the tracing of the face is performed by a tracing part. The tracing part traces the face from inside the static image data of a tracing target on the basis of characteristics of the face already detected from inside one or more pieces of static image data in time after the static image data of the tracing object to trace the face in a tracing-back direction of a temporal axis of the static image data string. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus that performs processing for tracking an object from moving image data composed of still image data sequences that are data at different times, an object tracking system including the same, an image processing method, and image processing Regarding the program.

  In order to detect an object such as a face from image data by image processing, a complicated calculation is required, and there is a problem that it takes a long time for detection. However, moving images have the property that the correlation between frames is very high, so when performing object detection processing, the “size” of the detected object at a time prior to the detection target frame. By utilizing characteristics such as “position”, “angle”, “shape”, “brightness”, “color”, and their “trajectory”, the detection time can be shortened. In object detection processing in moving image data, a method of searching for an object in the forward direction of the time axis using the characteristics of the object detected in the frame at the previous time is widely used.

  In addition, as another background art, an example of a technique for detecting a human face from image data is disclosed in Non-Patent Document 1, and a speech section is estimated by detecting a face from a video image and detecting a mouth movement. A technique for improving the speech recognition result is disclosed in Non-Patent Document 2.

R. Stiefelhagen et al., “Simultaneous tracking of head pose in a panoramic view”, proc. International Conference on Pattern Recognition, Vol. 3, pp726-729, 2000 Kazumasa Murai, Kenichi Kumatani, Satoshi Nakamura, "A Robust End Point Detection by Speaker's Facial Motion", HSC2001 (International Workshop on Hands-Free Speech), 2001/4/9

  However, in the above-described method for searching for an object in the forward direction of the time axis, only the characteristics of the object at a time after the time when the object can be detected first can be obtained. Furthermore, the prediction accuracy of the future object is lowered only by the characteristics of the object at the time after the first detection time. As described above, in the method of searching for an object in the forward direction of the time axis, there is still room for improvement with respect to more accurate object tracking and more accurate prediction of future objects.

  An object of the present invention is to provide an image processing apparatus that can more accurately perform processing for tracking an object from moving image data, an object tracking system including the same, an image processing method, and an image processing program.

  An image processing apparatus according to the present invention is an image processing apparatus that performs processing for tracking an object from moving image data composed of still image data sequences each of which is data at different times. Search means for searching for, and when one or more objects can be detected by the search means, tracking means for tracking the one or more objects in other still image data, and the tracking means By tracking an object from the tracked still image data based on the characteristics of the object already detected from one or more still image data at a time later than the tracked still image data, the still image data The gist is to trace the object in the direction going back the time axis of the column.

  In the image processing apparatus according to the present invention, the one or more still image data at a time later than the tracking target still image data may include still image data at a time immediately after the tracking target still image data. . In the image processing apparatus according to the present invention, the characteristics of the object may include at least one of the position, size, angle, shape, brightness, color, and locus of the object.

  In the image processing apparatus according to the present invention, the tracking unit is based on characteristics including the position and size of an object already detected from one or more still image data at a time before the still image data to be tracked. By tracking an object from the still image data to be tracked, the object can be tracked in the direction along the time axis of the still image data string. In this way, it is possible to use both means for tracking an object in the direction going back the time axis of the still image data sequence and means for tracking the object in the forward direction along the time axis. It is also possible to track an object using a motion detection algorithm typified by an optical flow method using an image of the detected object and an image to be detected. In the image processing apparatus according to the present invention, the one or more still image data at a time prior to the tracking target still image data may include still image data at a time immediately before the tracking target still image data. .

  In the image processing apparatus according to the present invention, it is preferable that the object is a face. In the image processing apparatus according to the present invention, the image processing apparatus is preferably an apparatus that performs processing for tracking an object in real time from real-time moving image data. It is particularly effective to apply the present invention because complex calculation is required to detect a face image in real time.

  An object tracking system including an image processing device according to the present invention includes an image processing imaging device that images an object to obtain real-time moving image data including the object, and still image data already detected by the tracking unit. Based on the characteristics of the object, a characteristic estimation means for estimating the future characteristic of the object, a tracking device capable of tracking the object, and a control command value based on the future characteristic of the object estimated by the characteristic estimation means Then, by outputting to the tracking device, it is possible to provide control command calculation means for performing control so that the tracking device tracks the object. In the object tracking system according to the present invention, the tracking device is a tracking imaging device that can track and image an object, and the control command calculation means includes at least one of an imaging direction and a zoom of the tracking imaging device. It is preferable to calculate a control command value for controlling the.

  An image processing method according to the present invention is an image processing method for performing processing for tracking an object from moving image data composed of still image data sequences, each of which is data at different times. And a tracking step for tracking the one or more objects in other still image data when the search step detects one or more objects, the tracking step comprising: , By repeatedly tracking the object from the tracking target still image data based on the characteristics of the object already detected from the one or more still image data at a time later than the tracking target still image data, The gist is to trace an object in a direction going back the time axis of the still image data string.

  An image processing program according to the present invention is an image processing program for causing a computer to execute processing for tracking an object from moving image data composed of still image data sequences each of which is data at different times. The computer executes search processing for searching for an object from within, and tracking processing for tracking the one or more objects in other still image data when one or more objects can be detected by the search processing The tracking process tracks an object from the still image data to be tracked based on the characteristics of the object already detected from one or more still image data at a time later than the still image data to be tracked. Processing to track objects in the direction going back the time axis of the still image data string by processing The gist of the Ukoto.

  As described above, according to the present invention, it is possible to more accurately perform processing for tracking an object from moving image data by tracking the object in a direction going back the time axis of the still image data sequence.

  Hereinafter, embodiments of the present invention (hereinafter referred to as embodiments) will be described with reference to the drawings.

  FIG. 1 is a block diagram showing an outline of the configuration of an object tracking system including an image processing apparatus according to an embodiment of the present invention. The object tracking system according to the present embodiment is a system that tracks a human face using image processing, and its configuration is roughly divided into a camera 10, an image processing device 12, and a control device 14.

  A camera 10 as an image processing imaging apparatus captures an image of a human face that is an object of the present embodiment. A moving image signal captured by the camera 10 is input to the image processing device 12. Further, the camera 10 can change its imaging direction and zoom, and the imaging direction and zoom of the camera 10 can be controlled by the control device 14 as will be described later.

  The image processing device 12 performs processing for tracking a human face in real time from a moving image captured by the camera 10. The image processing apparatus 12 includes an image preprocessing unit 28, an image storage unit 29, a search unit 16, and a tracking unit 18.

  A moving image signal from the camera 10 is input to the image preprocessing unit 28. The moving image signal from the camera 10 is subjected to processing such as A / D conversion and quantization in the image pre-processing unit 28, and is converted into real-time moving image data composed of still image data sequences that are data at different times. Converted. Here, each of the still image data strings is temporarily stored in the image storage unit 29. Each of the still image data sequences stored in the image storage unit 29 is input to the search unit 16 or the tracking unit 18, and processing for tracking a human face in real time is performed. An example of a still image data string stored in the image storage unit 29 is shown in FIG.

  The search unit 16 searches for the human face in the latest still image data obtained by the image preprocessing unit 28 from still image data at predetermined frame intervals. The search unit 16 here searches for a human face using, for example, the entire range in still image data as a search range. Alternatively, by setting the face search range based on the difference value of each coordinate based on the background difference or the inter-frame difference, the face search range can be limited and the search time can be shortened.

  When the search unit 16 can detect one or more human faces, the tracking unit 18 continues to search for the human face from other still image data stored in the image storage unit 29. Track human face. The tracking unit 18 includes a forward search range setting unit 20, a forward search unit 22, a reverse search range setting unit 24, and a reverse search unit 26.

  The forward search range setting unit 20 determines the search range in the still image data to be tracked based on the characteristics of the human face already detected from one or more still image data at a time prior to the still image data to be tracked. Set. As a result, a search range in the still image data in the direction along the time axis, that is, in the time axis forward direction is set from the detected still image data in which the human face has already been detected. Here, the one or more still image data at a time before the tracking target still image data includes still image data at a time immediately before the tracking target still image data. The forward search unit 22 tracks the human face from the tracking target still image data by searching for the human face within the search range set by the forward search range setting unit 20. The face characteristics here include at least one of the position, size, angle, shape, brightness, color, and locus of the face.

  On the other hand, the reverse search range setting unit 24 searches the still image data to be tracked based on the characteristics of the human face already detected from one or more still image data at a time later than the still image data to be tracked. Set the range. As a result, a search range is set in the still image data in the direction going back in time with respect to the detected still image data in which the human face has already been detected, that is, in the direction opposite to the time axis. Here, the one or more still image data at a time after the tracking target still image data includes still image data at a time immediately after the tracking target still image data. Then, the reverse search unit 26 searches for a human face within the search range set by the reverse search range setting unit 24, thereby tracking the human face from the still image data to be tracked.

  As described above, in this embodiment, when the tracking unit 18 tracks a human face, as shown in FIG. 2, the still image in the direction along the time axis (time axis forward direction) is detected from the detected still image data. In addition to performing processing for tracking the faces 34-1 and 34-2 from the image data, the face 34-1 is detected from the still image data in the direction (time axis reverse direction) that goes back the time axis from the detected still image data. , 34-2 are also tracked.

  The control device 14 controls at least one of the imaging direction of the camera 10 and the zoom based on the tracking result of the facial characteristics obtained by the image processing device 12. The control device 14 includes a characteristic estimation unit 30 and a control command calculation unit 32.

  The characteristic estimation unit 30 estimates the future characteristics of the human face based on the characteristics of the human face in the still image data already detected by the search unit 16 and the tracking unit 18. The future characteristics of the face here include at least one of the future position and size of the face. Then, the control command calculation unit 32 calculates a control command value for controlling at least one of the imaging direction of the camera 10 and the zoom based on the future characteristics of the human face estimated by the characteristic estimation unit 30. 10 is output. Accordingly, at least one of the imaging direction and zoom of the camera 10 is controlled to track the human face. The camera 10 here also serves as a tracking imaging device for tracking and imaging a human face.

  Note that the image processing device 12 and the control device 14 can be realized by, for example, a computer in which a CPU, a ROM, a RAM, a hard disk, a communication I / F, and the like are connected by a bus. And the process by the search part 16 and the tracking part 18 mentioned later, and the process by the characteristic estimation part 30 and the control command calculating part 32 are realizable like software with the program run by a computer.

  Next, processing in the system of the present embodiment will be described using the flowcharts shown in FIGS. Here, FIG. 3 shows a flowchart for explaining processing by the search unit 16. 4 shows a flowchart for explaining the process of tracking the face in the direction going back the time axis by the tracking unit 18, and FIG. 5 explains the process of tracking the face in the direction along the time axis by the tracking unit 18. A flowchart is shown. FIG. 6 is a flowchart for explaining processing by the control device 14. In the following description, it is assumed that a plurality of types of faces are tracked from moving image data.

  In step (hereinafter referred to as S) 1 in the flowchart of FIG. 3, it is determined whether or not to search for a face by the search unit 16 from the latest still image data obtained by the image preprocessing unit 28. Here, the latest still image data is obtained every predetermined time by the image preprocessing unit 28. In the image processing apparatus 12 of the present embodiment, the search condition by the search unit 16 is set so that the search unit 16 searches for a face at predetermined frame intervals. If the determination result in S1 is NO, the determination in S1 is repeated until acquisition of a predetermined number of frames of still image data is completed after acquisition of the still image data that was searched last time. On the other hand, if the determination result in S1 is YES, the process proceeds to S2.

  In S2, the search unit 16 searches for a human face from still image data. Here, the detection of the human face from the image data can be detected by using, for example, a technique for detecting the skin color disclosed in Non-Patent Document 1. For the search range of the face, for example, the entire range in the still image data is searched. However, it takes a long time to search the entire range, which is typically longer than the frame interval period. Therefore, as described above, the search range can be narrowed down by using the background difference or the inter-frame difference. If the search of the face by the search part 16 is complete | finished, it will progress to S3.

  In S3, it is determined whether or not a new face is detected by the search unit 16. If the determination result in S3 is YES, that is, if a new face is detected, the process proceeds to S4. On the other hand, if the determination result in S3 is NO, that is, if a face has not been detected, or if a face that has already been detected by the search unit 16 and is being tracked by the tracking unit 18 is detected again, the process goes to S1. Returning, the processing after S1 is repeated.

  In S4, the tracking unit 18 starts processing for tracking the face newly detected in S3. Specifically, the processing shown in the flowcharts of FIGS. 4 and 5 is started. And it returns to S1 and the process below S1 is repeated. Here, even when a face is detected by the search unit 16 and a face is tracked by the tracking unit 18, the reason for repeating the face search by the search unit 16 is to track one face 34-1 as shown in FIG. This is because another type of face 34-2 may newly appear during processing. As described above, a plurality of types of faces can be detected by repeating the face search by the search unit 16 at predetermined frame intervals. Further, even if the tracking of the face by the tracking unit 18 fails once, the face can be re-detected by the searching unit 16, so that the tracking unit 18 can re-track the face.

  Next, the process shown in FIGS. 4 and 5 will be described. In the present embodiment, a case will be described in which the process of tracking a face in a direction retroactive to the time axis in FIG. 4 and the process of tracking a face in a direction along the time axis in FIG. 5 are performed in parallel. However, these processes do not necessarily have to be performed in parallel. For example, the process of tracking the face in the direction retroactive to the time axis in FIG. 4 is performed first and then in the direction along the time axis in FIG. It is also possible to perform processing for tracking a face. When a plurality of types of faces are detected by the search unit 16, the processes of FIGS. 4 and 5 are performed for each face.

  In S101 of the flowchart of FIG. 4, based on the characteristics of the face in the detected still image data, the search range in the still image data in the direction going back the time axis from the detected still image data is the reverse search range setting unit 24. Set by. Then, the process proceeds to S102. Hereinafter, a specific example of setting the search range will be described.

  As shown in FIG. 7, for example, when the detected still image data is only the frame f that has been detected by the search unit 16, the search range of the frame f-1 that goes back the time axis by one frame from the frame f It can be set by the center position of both eyes at f and the predicted amount of fluctuation (vertical direction, horizontal direction and depth) of the center position of both eyes in the frame interval period.

  Next, as shown in FIG. 8, for example, consider a case where the detected still image data is a frame f detected by the search unit 16 and a frame f−1 detected by the reverse search unit 26. In this case, for the search range of the frame f-2 that goes back the time axis by one frame from the frame f-1, the center of both eyes in the frame f-2 calculated by regression from the center positions of both eyes in the frames f and f-1 It can be set by the position and the predicted amount of fluctuation of the center position of both eyes in the frame interval period.

  Next, as shown in FIG. 9, for example, consider a case where the detected still image data is a frame f detected by the search unit 16 and frames f−1 and f-2 detected by the reverse search unit 26. In this case, for the search range of the frame f-3 that goes back the time axis by one frame from the frame f-2, the frame f-3 that is calculated by regression from the center position of both eyes in the frames f, f-1, and f-2. The center position of both eyes and the predicted amount of fluctuation of the center position of both eyes in the frame interval period can be set.

  Here, the position fluctuation prediction amount in the frame interval cycle can be set experimentally, for example. As an example, FIG. 10 shows the result of experimentally determining the probability distribution of the lateral displacement at the center position of both eyes in the frame interval period. As shown in FIG. 10, the probability distribution obtained from two frames has a smaller variation than the probability distribution obtained from one frame. The same experiment can be used to set the position fluctuation prediction amount in the frame interval period to be small according to the increase in the number of detected frames, so that the search range can be narrowed according to the increase in the number of detected frames. The search time can be shortened. In the above description, the center position of both eyes is used for setting the search range, but the search range can also be set using other positions of the face.

  In S102, the reverse search unit 26 searches for a face within the search range set by the reverse search range setting unit 24. As described above, the face can be detected by using the method disclosed in Non-Patent Document 1, for example. As for the face search by the inverse search unit 26 here, the search range is limited, and therefore the face search time can be greatly shortened. Typically, the search process can be completed in a time extremely shorter than the frame interval period. When the face search by the reverse search unit 26 ends, the process proceeds to S103.

  In S103, it is determined whether or not the face to be tracked can be detected by the reverse search unit 26. If the determination result in S103 is YES, the process returns to S101, and the processing in S101 and S102 is repeated for still image data that is backed by one frame on the time axis. On the other hand, when the determination result in S103 is NO, it is determined that the face to be tracked is out of the imaging range of the camera 10, and the tracking in the direction going back the time axis of the face is ended. In the determination of S103, when a face cannot be detected continuously for a plurality of frames, it may be determined that the face to be tracked is out of the imaging range of the camera 10.

  On the other hand, in S201 of the flowchart of FIG. 5, the search range in the still image data in the direction along the time axis is set to the forward search range setting based on the characteristics of the face in the detected still image data. Set in section 20. Then, the process proceeds to S202. The search range here can be set by the same method as when going back in time. Further, when the search range is set by the forward search range setting unit 20, since the face search is performed back in time by the face search by the reverse search unit 26 being performed in parallel, The detected face characteristics used for the setting include the face characteristics detected by the inverse search unit 26. For example, in the example of FIG. 9, the face characteristics in the frames f-2 and f-1 detected by the inverse search unit 26 are used for setting the search range in the frame f + 1. Thus, since the search range can be narrowed by setting the search range using the characteristics of the face detected by the reverse search unit 26, the search time by the forward search unit 22 can be further shortened. Can do.

  In S <b> 202, the face search is performed by the forward search unit 22 within the search range set by the forward search range setting unit 20. As described above, the face can be detected by using the method disclosed in Non-Patent Document 1, for example. As for the face search by the forward search unit 22 here, since the search range is limited, the face search time can be greatly shortened. Typically, the search process can be completed in a time extremely shorter than the frame interval period. When the face search by the forward search unit 22 ends, the process proceeds to S203.

  In S203, the forward search unit 22 determines whether or not the face to be tracked has been detected. If the determination result in S203 is YES, the process returns to S201, and the processes in S201 and S202 are repeated for still image data along one frame on the time axis. On the other hand, if the determination result in S203 is NO, it is determined that the face to be tracked is out of the imaging range of the camera 10, and the tracking of the face along the time axis is terminated. In the determination of S203, if a face cannot be detected continuously for a plurality of frames, it may be determined that the face to be tracked is out of the imaging range of the camera 10.

  Next, control of the camera 10 will be described. In S301 of the flowchart of FIG. 6, it is determined whether or not the camera 10 is to be controlled. If the face tracking result sufficient to estimate the future face characteristics is not yet obtained by the search unit 16 and the tracking part 18, the determination result in S301 is NO, and the tracking result sufficient to estimate the future face characteristics is obtained. Until S301, the determination in S301 is repeated. On the other hand, if the search unit 16 and the tracking unit 18 have obtained tracking results that can estimate the future characteristics of the face, the determination result in S301 is YES, and the process proceeds to S302.

  In S302, the characteristic estimation unit 30 estimates the future characteristics of the face after a predetermined time has elapsed. Here, for example, by using a regression equation obtained from each of the center positions of both eyes in a plurality of detected still image data, the future characteristics of the face after a predetermined time has elapsed can be estimated. At this time, the center position of both eyes in the detected still image data includes the center position of both eyes detected by the reverse search unit 26, and the future characteristics of the face are also utilized using the center position of both eyes detected by the reverse search unit 26. Therefore, it is possible to accurately estimate the future characteristics of the face. Then, the process proceeds to S303.

  In S <b> 303, a control command value for controlling the imaging direction of the camera 10 is calculated by the control command calculation unit 32. Here, the control command value is calculated so that the imaging range of the camera 10 includes the future face estimated by the characteristic estimation unit 30. Further, when the tracking results of a plurality of types of faces are obtained and the future characteristics of the plurality of types of faces are estimated, the control command value is set so that the imaging range of the camera 10 includes the plurality of types of future faces. Is calculated. For the calculation of the control command value, for example, a characteristic map relating to the control command value and the imaging direction of the camera 10 may be stored and calculated using the characteristic map. Then, the process proceeds to S304.

  In S304, the control command value calculated by the control command calculation unit 32 is output to the camera 10, and the imaging direction of the camera 10 is controlled. And it returns to S301 and the process after S301 is repeated. As for the control of the camera 10, a control command value for controlling the zoom may be calculated and output instead of the imaging direction control. Furthermore, a control command value for controlling both the imaging direction and the zoom may be calculated and output.

  As described above, according to the present embodiment, when performing the process of tracking a face from moving image data, it is possible to obtain a more characteristic of the face in the past by tracking the face in a direction going back in the time axis. it can. Furthermore, the face characteristics in the past can be reflected in the prediction of the future face characteristics. For example, in the case of trying to predict the start of speaking using the face tracking result in the moving image data as in Non-Patent Document 2, there is a time difference between the start of lip movement and the actual start of speaking. It is desirable that the movement of the lips in the past can be grasped from the moving image data. Therefore, by tracing the face in the direction going back on the time axis and grasping the movement of the lips in the past, it is possible to predict the beginning of the conversation earlier. As described above, in the present embodiment, the face can be tracked more accurately from the moving image data by tracking the face in the direction going back in time and obtaining the face tracking result in the past. Furthermore, the characteristics of the future face can be predicted more accurately.

  Then, the face tracking result in the direction going back the time axis obtained by the inverse search unit 26 can be used for tracking the face in the direction along the time axis. Specifically, the forward search range setting unit 20 sets the search range based on the face characteristics in the detected still image data including the face characteristics traced back along the time axis. More specifically, the face search time by the forward search unit 22 can be shortened.

  Further, the future characteristics of the face are estimated based on the characteristics of the face including the tracking result traced back in time, and the imaging direction and zoom of the camera 10 are set so that the imaging range of the camera 10 includes the estimated future face. By controlling at least one of the faces, the face can be accurately tracked and imaged by the camera 10.

  In the present embodiment, the image processing imaging device used for face detection image processing and the tracking imaging device that tracks and captures a face are shared by the camera 10, but the image processing imaging device and It is also possible to use separate cameras for the tracking imaging device. Furthermore, the tracking device of the present invention is not limited to a camera. For example, by controlling the directional characteristics of the directional microphone based on the future characteristics of the face, the directional characteristics of the directional microphone can be tracked to a human face.

  In the present embodiment, the case where the object to be tracked is a human face has been described, but the present invention can also be applied to the case of tracking other objects.

  Further, in the present embodiment, the case where the image processing apparatus performs processing for tracking an object in real time from real-time moving image data has been described. However, the image processing apparatus of the present invention is configured so that an object from moving image data stored in advance is stored. It can be applied even in the case of performing post-processing for tracking.

  As mentioned above, although embodiment of this invention was described, this invention is not limited to such embodiment at all, Of course, in the range which does not deviate from the summary of this invention, it can implement with a various form. It is.

It is a block diagram which shows the outline of a structure of an object tracking system provided with the image processing apparatus which concerns on embodiment of this invention. It is a figure explaining the tracking process of the face by the image processing apparatus which concerns on embodiment of this invention. It is a flowchart explaining the process performed by the search part. It is a flowchart explaining the process performed by the tracking part. It is a flowchart explaining the process performed by the tracking part. It is a flowchart explaining the process performed by a control apparatus. It is a figure explaining an example of the setting of the search range when tracking a face. It is a figure explaining an example of the setting of the search range when tracking a face. It is a figure explaining an example of the setting of the search range when tracking a face. It is a figure which shows the probability distribution of the horizontal displacement of the center position of both eyes in a frame space | interval period.

Explanation of symbols

  10 cameras, 12 image processing devices, 14 control devices, 16 search units, 18 tracking units, 20 forward search range setting units, 22 forward search units, 24 reverse search range setting units, 26 reverse search units, 28 image preprocessing units, 29 Image storage unit, 30 characteristic estimation unit, 32 control command calculation unit.

Claims (11)

An image processing apparatus that performs processing for tracking an object from moving image data composed of still image data sequences that are data at different times,
Search means for searching for an object from still image data;
Tracking means for tracking the one or more objects in other still image data when one or more objects can be detected by the search means;
With
The tracking means tracks an object from the still image data to be tracked based on the characteristics of the object already detected from one or more still image data at a time later than the still image data to be tracked. An image processing apparatus that tracks an object in a direction that goes back along the time axis of a still image data sequence. The image processing apparatus according to claim 1,
One or more still image data at a time later than the tracking target still image data includes still image data at a time immediately after the tracking target still image data. The image processing apparatus according to claim 1, wherein:
The image processing apparatus, wherein the characteristics of the object include at least one of a position, a size, an angle, a shape, brightness, a color, and a locus of the object. The image processing apparatus according to any one of claims 1 to 3,
The tracking means is configured to select from the still image data to be tracked based on the characteristics including the position and size of the object already detected from one or more still image data at a time before the still image data to be tracked. An image processing apparatus that tracks an object in a direction along a time axis of a still image data sequence by tracking the object. The image processing apparatus according to claim 4,
The image processing apparatus according to claim 1, wherein the one or more still image data at a time prior to the tracking target still image data includes still image data at a time immediately before the tracking target still image data. An image processing apparatus according to any one of claims 1 to 5,
An image processing apparatus, wherein the object is a face. The image processing apparatus according to any one of claims 1 to 6,
The image processing apparatus is an apparatus that performs processing for tracking an object in real time from real-time moving image data. An object tracking system comprising the image processing device according to claim 7,
In order to obtain real-time moving image data including an object, an image processing imaging device that images the object;
A characteristic estimation unit that estimates a future characteristic of the object based on the characteristic of the object in the still image data already detected by the tracking unit;
A tracking device capable of tracking objects;
Control command calculating means for controlling the tracking device to track the object by calculating a control command value based on the future characteristics of the object estimated by the characteristic estimating means and outputting the control command value to the tracking device; ,
An object tracking system comprising: The object tracking system according to claim 8, comprising:
The tracking device is a tracking imaging device capable of tracking and imaging an object,
The object tracking system, wherein the control command calculation means calculates a control command value for controlling at least one of an imaging direction and zoom of the tracking imaging device. An image processing method for performing processing for tracking an object from moving image data composed of still image data sequences each of which is data at different times,
A search step for searching for an object from still image data;
A tracking step of tracking the one or more objects in other still image data if the searching step detects one or more objects;
Including
The tracking step includes tracking an object from the still image data to be tracked based on the characteristics of the object already detected from one or more still image data at a time later than the still image data to be tracked. An image processing method characterized in that an object is traced in a direction going back on a time axis of a still image data sequence by repeating. An image processing program for causing a computer to execute processing for tracking an object from moving image data composed of still image data sequences each of which is data at different times,
Search processing for searching for objects from still image data,
A tracking process for tracking the one or more objects in other still image data when one or more objects can be detected by the search process;
To the computer,
The tracking process is a process of tracking an object from the still image data to be tracked based on the characteristics of the object already detected from one or more still image data at a time later than the still image data to be tracked. An image processing program that performs processing for tracking an object in a direction that goes back in the time axis of a still image data sequence.

Images