JP2009147607A - Person tracking device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a person tracking device capable of appropriately determining a skin color segment to be tracked from a plurality of skin color segments without needing many calculation resources even for real time tracking processing and of continuing to track a target moving object even when the skin color segment suddenly disappears. <P>SOLUTION: The person tracking device is provided with: a frame memory 12 for storing image frame data of an input image signal; a detection part 13 for making the image frame data a block to detect a moving object block; a creation part 14 for creating an adjacent moving object block as a segment; a detection part 15 for detecting a skin color area of the image frame data and weighting them; a search part 16 for detecting a moving object present within the image frame in past, for searching a moving object present nearby the currently moving object and weighting it; a determination part 17 for determining a moving object segment to be tracked; and a determination part 18 for determining position coordinates to be tracked within the moving object segment to be tracked. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention detects a moving object from an input image signal obtained from a surveillance camera, an image reproducing device, public broadcasting, or the like, detects a portion having a human skin color from the detected moving object, and The present invention relates to a person tracking device that can perform tracking of a person without interrupting the tracking even if a portion having a human skin color cannot be detected suddenly.

  In recent years, surveillance systems for crime prevention and disaster prevention have been introduced not only in facilities such as factories and shops but also in homes. Various new functions have been added to these monitoring systems. For example, there are a function for detecting a person from an image taken by a surveillance camera and a function for tracking a specific person thus detected.

Specifically, a moving object is detected from a moving image captured by a surveillance camera using a background subtraction method to detect a moving object, and the movement of the moving object is dynamically determined based on the moving object area. There is an object detection device that identifies a target moving object by template matching processing using a template created based on the analysis result (see Patent Document 1).
JP 2004-348303 A

  However, the template matching process described in Patent Document 1 requires a large amount of calculation for detecting a moving object, and requires many calculation resources to perform this calculation in real time. As a result, there is a problem that the cost is high and real-time detection processing of such a moving object is practically difficult.

  In addition, the template matching process described above is designed to detect a flesh-colored segment from the moving segments. However, even if the tracking target is the same person, there are a plurality of flesh-colored segments such as faces and both hands. However, since it is impossible to accurately determine which skin color segment is a detection target, there is a problem that tracking may fail.

  Furthermore, for example, when a person who is tracking the skin color segment of the face suddenly becomes behind and the face disappears, the skin color segment targeted for tracking will suddenly disappear, and in this case However, there was a problem that tracking might fail.

  Therefore, the present invention has been made in view of the above-described problems, and is for detecting a moving object to be tracked that appears in an input image signal obtained from a surveillance camera, an image reproducing device, or a public broadcast. Real-time tracking processing is possible because the amount of calculation is small, and even if the tracking target is the same person and there are multiple skin color segments such as the face and both hands, the skin color segment to be tracked is accurately identified from this An object of the present invention is to provide a person tracking device that can be determined and does not fail in tracking.

  In addition to the configuration of the person tracking apparatus, the present invention provides a person tracking apparatus that continues to track a target moving object without losing sight of the skin color segment even if the skin color segment to be tracked suddenly disappears. For the purpose.

  In order to solve the above-described problems, the present invention provides a person tracking device having the following configuration.

That is, as shown in FIG. 1, data relating to the movement of the skin color of an exposed portion (face, hand, foot, etc.) of a person extracted from an input image signal (for example, an image of the imaging area of the surveillance camera 101) Is a person tracking device 10 that detects and tracks the movement of the person,
Memory means (frame memory) 12 for sequentially storing frame data 1A and 1B obtained by converting the input image signal in units of image frames;
The previous and current blocks obtained by dividing the previous frame data 1A (FIG. 2) and the current frame data 1B (FIG. 3) stored in the memory means 12 into a plurality of blocks (for example, 300 blocks), respectively. Moving object detection means (moving object detection unit) that detects a difference between luminance data of blocks (blocks at the same position) exceeding a predetermined threshold value as a moving object block (FIG. 4) 13 and
Segment creation means (segment creation section) 14 for creating moving object segments 2 and 3 (FIG. 5) obtained by segmenting the moving object block output from the moving object detection means 13;
Based on the moving object segments 2 and 3 output from the segment creation means 14 and the current frame data 1B output from the memory means 12, the moving object includes a skin color block 4 having skin color data. Skin color segment detection means (skin color detection unit) 15 for detecting segment 3,
Based on the moving object segment 3 output from the flesh color segment detection means 15 and the moving object segment in the previous frame data that is most approximate to the moving object segment 3, the tracking object is continuously selected as the tracking target. A person tracking device comprising: tracking target determining means (tracking target determining unit) 17 for determining a moving object segment.

  The present invention enables real-time tracking processing because the amount of calculation for detecting a moving object to be tracked that appears in an input image signal obtained from a surveillance camera, an image playback device, or a public broadcast is reduced. In addition, even if the tracking target is the same person and there are multiple skin color segments such as the face and both hands, it is possible to accurately determine the skin color segment to be tracked from this, so tracking does not fail A person tracking device can be provided.

  In addition to the configuration of the person tracking apparatus, the present invention provides a person tracking apparatus that continues to track a target moving object without losing sight of the skin color segment even if the skin color segment to be tracked suddenly disappears. be able to.

  Hereinafter, embodiments of the person tracking device of the present invention will be described in the order of the following (1) to (5) with reference to FIGS.

(1) Configuration of person tracking device (2) Operation of person tracking device (3) First embodiment (4) Second embodiment (5) Other embodiments FIG. 1 shows the implementation of the person tracking device of the present invention. 2 is a diagram showing the previous image frame data, FIG. 3 is a diagram showing the current image frame data, FIG. 4 is a diagram showing a moving object block, and FIG. 5 is a diagram showing moving object segments. FIG. 6 is a flowchart showing the operation of the embodiment of the person tracking device of the present invention. In addition, the apparatus, method, etc. which are used in embodiment of this invention are examples, and of course, this invention is not limited to these.

  Hereinafter, the embodiment of the present invention will be described with respect to tracking a person appearing in an input image signal captured by a surveillance camera as an example, but the present invention is not limited to this, for example, Tracking a specific person appearing in a broadcast video after receiving a public broadcast, or tracking a specific person appearing in a playback video signal transmitted wirelessly from a remote video playback device It goes without saying that is also applicable.

(1) Configuration of Person Tracking Device As shown in FIG. 1, a person tracking device 10 according to an embodiment of the present invention includes a monitoring camera 101 that transmits an image captured in a monitoring area as an image signal, and a tracking process. Connected to a monitor device 102 that displays the image of the image.

  As shown in FIG. 1, the person tracking device 10 includes an image signal receiving unit 11, a frame memory 12, a moving object detection unit 13, a segment creation unit 14, a skin color detection unit 15, a neighborhood search unit 16, and a tracking target determination unit 17. A tracking position determination unit 18, a coordinate data transmission unit 19, a flesh color weight storage unit 20, a neighborhood weight storage unit 21, and a control unit 22. The control unit 22 performs operation control of each unit constituting the person tracking device 10.

  The person tracking device 10 having the above-described configuration is extracted from the captured image obtained by capturing the imaging area of the monitoring camera 101 and is exposed as a person (face, hand, foot, etc.) as shown in FIG. This is a person tracking device that detects and tracks data relating to the movement of the skin color as the movement of the person.

  The image signal receiving unit 11 described above receives the image signal transmitted from the monitoring camera 101, converts it into image frame data in units of image frames, and sequentially outputs them to the frame memory 12.

  The frame memory 12 stores the image frame data in units of image frames input from the image signal receiving unit 11 with a time stamp in the input order (time series order). The frame memory 12 sequentially stores image frame data, for example, first image frame data,..., Second image frame data, and third image frame data.

  In the following description, the second image frame data is “previous image frame data” 1A (FIG. 2), the third image frame data is “current image frame data” 1B (FIG. 3), and the first image The frame data is used as “past image frame data”.

  The previous image frame data 1A has data portions 1Aa to 1Ac as shown in FIG. 2, and the current image frame data 1B has data portions 1Ba to 1Bc as shown in FIG. Yes. The data parts 1Aa to 1Ac correspond to the data parts 1Ba to 1Bc, respectively. The movement of the data parts 1Ba and 1Bb with respect to the data parts 1Aa and 1Ab is indicated by arrows in FIG. There is no movement in the data portion 1Bc shown in FIG.

  The moving object detection unit 13 detects a moving part of the current image frame data 1B with respect to the previous image frame data 1A, and outputs this to the segment creation unit 14 as moving object block data. This moving part is detected by using a motion detection method that is an existing technique.

  That is, the moving object detection unit 13 reads the previous image frame data 1A and the current image frame data 1B stored in the frame memory 12, and sets the read image frame data 1A and 1B in advance. The block is divided into a certain number of blocks (for example, 300 blocks (vertical 15 blocks × horizontal 20 blocks) as shown in FIG. 4 described later).

  Each of the blocked image frame data 1A and 1B forms a block group in which a large number of blocks are aggregated. Then, the brightness of each block (previous block) constituting the block group of the previous image frame data 1A that has been blocked and each block (current block) constituting the block group of the current image frame data 1B that has been blocked. ) Find the difference in brightness from each brightness.

  As a result, when each luminance difference exceeds a predetermined threshold, the block is detected as moving object block data. The area of the moving object block data is an area obtained by combining the areas of the two blocks.

  The segment creation unit 14 receives the moving object block data from the moving object detection unit 13. The segment creation unit 14 detects all blocks adjacent to the respective outer edges of the input moving body block data, and merges the areas of all detected blocks into the original moving object block data area. Segment data is respectively generated and output to the frame memory 12 and the skin color detection unit 15. The two moving object segments 2 and 3 shown in FIG. 4 are generated in this way.

  Based on the image frame data 1B input from the frame memory 12 and the moving object segment data related to the moving object segments 2 and 3 input from the segment creation unit 14, the skin color detection unit 15 described above 3 is detected as a skin color block having “skin color” which is a human skin color as an index indicating a human.

  Here, “skin color” is generally defined as follows. That is, # F1BB93 in hexadecimal notation, (241, 187, 147) in RGB, (2, 23, 35, 0) in CMYK, (26 °, 39%, 95%) in HSV, and 5YR8 / 5 in Munsell value It is.

  On the other hand, the “skin color” handled by the person tracking device 10 of the present invention has a spread necessary for specifying as a human regardless of race, centering on the previously defined “skin color”. It also includes the “skin color” of the range. As used herein, “skin color” has this meaning.

  The skin color block described above corresponds to a block of a face or both hands exposed from clothes in the image frame data 1B.

  However, under the environment of the monitoring area where the surveillance camera 101 is installed, the monitor camera 101 is under various lighting (natural light, fluorescent light, incandescent light bulb, etc.), and the color balance adjustment of the surveillance camera 101 is performed according to the color temperature of the illumination light. It is necessary to make it in advance. Without such advance adjustment, even if the image frame data 1B obtained by imaging the skin color with the monitoring camera 101 is input to the skin color detection unit 15, the skin color detection unit 15 may not recognize this as the skin color. This is because, as a result, there is a possibility that human movement is erroneously detected.

  In this way, as shown in FIG. 5, the skin color detection unit 15 of the skin color block (diagonal halftone line portion) 4 constituting the moving object segment 3 that can detect the skin color of the two moving object segments 2 and 3. Perform weighting. That is, the skin color block 4 adds, for example, +10 to the block 5 which is not the skin color block. In this way, the total number of weights for all skin road blocks 4 is calculated, and the total points are output to the skin color weight storage unit 20 and the tracking target determination unit 17, respectively. The skin color weight storage unit 20 stores the total score.

  Then, the control unit 22 performs information processing by regarding the moving object segment 3 as a moving object segment composed of all skin color blocks 4.

  The neighborhood search unit 16 searches for the fact that moving object block data similar to the moving object block data detected by the moving object detection unit 13 has been detected so far within a time that is a predetermined time after the current time. .

  That is, the proximity search unit 16 sequentially reads out the image frame data stored in the frame memory 12 within a predetermined time from the current time set by the control unit 22, and is closest to the moving object block data currently grasped. Proximity moving object block data is detected. This search result is output to the tracking target determining unit 17.

  The tracking target determination unit 17 compares the neighborhood moving object block data input from the neighborhood search unit 16 with the total weighted scores for all the blocks that can be detected from the skin color detection unit 15. When it is detected that the latter weighted total score is most similar to the former moving object block data, the moving object segment is determined to be a tracking target. The tracking target determination unit 17 outputs the moving object segment data that is the tracking target to the tracking position determination unit 18.

  The tracking position determination unit 18 described above determines a position coordinate to be tracked based on the moving object segment 3 having the tracking target skin color block 4 determined by the tracking target determination unit 17. The tracking position determination unit 18 changes the setting so that the center-of-gravity coordinates of the moving object segment 3 including the skin color block 4 to be tracked is the moving object segment to be tracked. The barycentric coordinates indicate the center position in the direction in which the moving object will move.

  The tracking position determination unit 18 continues the moving object to be a tracking target based on the moving object segment 3 and the moving object segment in the previous frame data from the previous frame data 1A that is most approximate to the moving object segment 3. The position coordinate data relating to the barycentric coordinates from which the moving object segment to be tracked is determined will be output to the coordinate data transmission unit 19.

  The coordinate data transmission unit 19 compares the current position coordinate data with the position coordinate data to be transferred from now on, and uses the difference data as a drive signal as a PTZ (pan-tilt) mechanism or rotating pan head with the monitoring camera 101 attached. Is output to each driving means. By making the moving object to be tracked always fit in the center of the image in this way, the target person can be always photographed at the center position of the image frame.

(2) Operation of Person Tracking Device Next, the operation of the person tracking device (1) described above will be described in the order of operation along the flowchart shown in FIG. FIG. 7 is a diagram for explaining skin color block detection processing.

(A) First, image frame data obtained by imaging the monitoring area is received and stored in succession. ("Frame data reception" step S1).

(B) Next, the previous image frame data 1A and the current image frame data 1B stored in memory are read out and divided into a predetermined number of blocks (for example, 300 blocks) to be made into blocks. Then, the brightness of each block (previous block) constituting the block group of the previous image frame data 1A that has been blocked and each block (current block) constituting the block group of the current image frame data 1B that has been blocked. ) Find the difference in brightness from each brightness. As a result, when each luminance difference exceeds a predetermined threshold, the block is detected as a moving object block (moving object block data) ("detecting a moving object block" step S2).

(C) Next, when a moving object block is detected (YES), the process proceeds to the next step S4. If a moving object block is not detected (NO), the process returns to step S1 (“A moving object block has been detected?” Step S3).

(D) Next, when it is detected as a moving object block (YES), all the blocks adjacent to the outer edge of the moving object block are detected, and these are merged to obtain a moving object segment (moving object segment data) 2, 3 ("segment moving object block" step S4).

(E) Next, it is detected that all the blocks constituting the moving object segments 2 and 3 are skin-colored blocks (“whether or not all moving object blocks are skin-colored areas) Step S5).

  That is, the skin color detection unit 15 creates an average pixel of the block for each of the moving object segments 2 and 3, and converts the average pixel to YUV format data if the average pixel is not YUV format data. The YUV format is a format in which color is expressed by three pieces of information: luminance signal (Y), difference between luminance signal and blue component (U), and difference between luminance signal and red component (V). A larger amount of data is assigned to luminance information by utilizing a property that is more sensitive to changes in brightness than changes in brightness.

  The flesh color block 4 (FIG. 5) is expressed by a rectangular area composed of the difference between the luminance signal and the blue component (U1 · U2) and the difference between the luminance signal and the red component (V1 · V2) in the U-V coordinate shown in FIG. Is done. The values of U1, U2, V1, and V2 are respectively set to optimum values according to the shooting situation (color balance) of the monitoring camera 101.

(F) Next, the skin color detection unit 15 determines that the average pixel (U, V) of the blocks constituting the moving object segment 3 is the following (Equation 1),
U1 ≦ U ≦ U2 and V1 ≦ V ≦ V2 (Formula 1)
("Since at least one moving object block that is a skin color area has been detected?" Step S6).

As a result of this determination, when the conditional expression (Expression 1) is satisfied, the moving object block is detected as the skin color block 4. For example, in FIG. 5, the face, arms, and legs of a running woman are detected as the skin color block 4.

(G) Next, the moving object segment 3 having the flesh color block 4 is weighted. The “weight” is determined by the installation environment and the target object of the monitoring camera 101, and is set so that skin color is emphasized in order to track a person under a situation where many moving objects other than the person move, for example. To do. As an example of weighting, when the number of blocks of a certain moving object segment is 10, of which the skin color block 4 is 3 and the other moving object block 5 is 7, the weight of the skin color block 4 is doubled and (3 × 2) Give a score of + 7 = 13.

  That is, the weight of the skin color block 4 is proportional to the width of the skin color distribution. Note that the moving object segment 2 that does not include the flesh color block 4 is not weighted. For example, when the number of blocks of the moving object segment is 10, the score is kept at 10. The score of the moving object segment 2 and the score of the weighted moving object segment 3 possessed by the skin color block 4 are stored in the skin color weight storage unit 20 (“weight moving object block which is a skin color area” step S7).

(H) The neighborhood searching unit 16 detects a moving object that has existed in the frame in the past. Specifically, it is determined whether or not the moving object segment 3 has been detected in the image frame data captured by the monitoring camera 101 during a predetermined time. This determination is performed by searching for a frame with the moving object segment 3 stored in the frame memory 12 (“Are moving objects detected during a preset time?” Step S8).

(I) When the moving object segment 3 has been detected (YES), the neighborhood search unit 16 selects a frame closest to the current time from the past frames with moving object segments, and the moving object in this frame The position of segment 3 is specified. Further, in the latest received frame, the moving object segment closest to the specified position is searched, the searched moving object segment is set as a neighboring segment, and weighting is performed.

  The weighted moving object segment 3 score is stored in the neighborhood weight storage unit 21 (“weight the moving object closest to the latest moving object detected in the past” step S9).

  Here, the weighting relationship regarding the skin color detection unit 15 and the neighborhood search unit 16 will be described. The person tracking device 10 can reliably recognize and track the moving object segment to be tracked by freely combining the weights of the moving object segment 3 holding the skin color block 4 and the weights of the neighboring segments. To do.

  For example, when tracking a moving object segment that is a person, if only the moving object segment 3 that holds the flesh color block 4 is tracked, the person performs an action such as looking back, and the flesh color block is instantaneously displayed. The moving object segment 3 that holds 4 may disappear from the image frame. In such a case, a neighboring segment of the person immediately before the moving object segment 3 disappears is used, and this neighboring segment is tracked as a moving object segment to be tracked until the moving object segment 3 having the skin color block 4 appears again. To do.

  As another example of weight setting, if the weight is set so that the neighboring segment always has a score higher than any other moving object segment, once the moving object segment that is the target person is detected, it will be The moving object segment is preferentially tracked. Also, if you want to track only a person even if it is unspecified, set the weight of the moving object segment with the skin color block higher than the weight of the neighboring segment, and even if it is not necessarily the same person as the person detected last time, the skin color A moving object including, that is, a person is tracked with priority.

(J) The tracking target determination unit 17 has the highest score among the score of the moving object segment 3 possessed by the skin color block stored in the skin color weight storage unit 20 and the score of the neighboring segment stored in the neighborhood weight storage unit 21. The moving object segment is determined as the tracking target (“select the segment with the highest score among all the segments” step S10).

(K) The tracking position determination unit 18 determines whether or not the moving object segment 3 selected by the tracking target determination unit 17 is a moving object segment having the skin color block 4 (“skin color in the segment selected in S10” Is there a moving object block that is an area? "Step S11).

(L) Next, when it is determined that the moving object segment has the flesh-color block 4 (YES), the center of gravity of only the area composed of the flesh-color block or a plurality of flesh-color blocks is calculated, and the coordinates of the center of gravity are tracked. Determine (“Determine the coordinates to be tracked in consideration of the deviation of the skin color area in the segment selected in S10” Step S12).

(M) On the other hand, when it is not determined that the moving object segment 3 has the flesh color block 4 (NO), the coordinates of the center of gravity of the moving object segment 3 are simply determined as the tracking position (“the segment selected in S10” The barycentric coordinates are obtained and set as the tracking target ”step S13).

(N) Next, the coordinate data transmission unit 19 outputs the coordinates of the center of gravity as a tracking target to an external device such as a PTZ mechanism or a rotating pan head (outputs the coordinates of the determined center of gravity of the segment to the outside). ) Step S14).

  Also for the image frames received thereafter, the person tracking device 10 repeatedly executes the above steps S1 to S14 and continues the tracking process of the target person.

(3) First Example With respect to an example in which a moving object segment having a skin color block to be tracked is determined from among moving object segments having a plurality of skin color blocks, two image frames shown in FIGS. 2 and 3 are used. explain.

  FIG. 2 shows a past image frame, and FIG. 3 shows a current image frame. Comparing the frame data 1A and 1B in FIGS. 2 and 3, the person and the car are moving in different directions. FIG. 4 shows the moving object segment 2 detected when the car moves and the moving object segment 3 detected when the person moves.

  The moving object segment 3 of the person is a segment in which the moving object block formed at the place where the person existed before the movement and the moving object block where the person exists after the movement are combined (see step S4 in FIG. 6). ).

  In the image frame of FIG. 5 in which the flesh color block 46 is extracted from the moving object segment 3 shown in FIG. 4, no flesh color block is detected in the moving object segment 2, but the flesh color block 4 is detected in the moving object segment 3. Yes.

  In this case, the moving object segment 3 including the flesh color block 4 is weighted to increase the score (see step S7 in FIG. 6). The tracking target determining unit 17 determines the moving object segment 3 that holds the flesh color block 4 as a tracking target. Since the tracking target moving object segment 3 and the flesh color segment 4 are included, the tracking position determination unit 18 calculates the center of gravity of the area composed of the flesh color block 4 (see steps S11 and S12 in FIG. 6).

  Thereafter, pan / tilt control is performed and the center of gravity of the image is set so as to be at the center of the photographed image on which the surveillance camera 101 is installed, so that the position where the person to be tracked is located as in the frame data 1C shown in FIG. It can be reset so that 1Cb becomes the center of the screen to be imaged.

(4) Second Example An example in which the target moving object segment 3 is tracked even if the skin color segment disappears from the image frame to be captured will be described using the two frame data 1D and 1E shown in FIGS. To do. 8 shows past frame data 1D, and FIG. 9 shows current frame data 1E. Comparing the frame data 1D and 1E, the person J and the person K are moving in different directions. The arrows in FIG. 9 indicate the moving direction. In FIG. 8, the person K faces the front, but in FIG. 9, the person K looks back and there is no skin color block.

  In FIG. 10, the moving object segment 6 detected when the person J moves, the moving object block created at the place where the person K existed and the moving object block created at the place after the movement are shown. Is displayed (see step S4 in FIG. 6).

  In the moving object segment 7, a flesh color block (not shown) is detected, but not in the moving object segment 6. In this case, the weight of the skin color block is added to the moving object segment 7 (see step S19 in FIG. 6). Since the moving object segment 7 is the segment closest to the centroid 8 of the person detected last time, the moving object segment 7 is weighted as a neighboring segment (see steps S8 to S9 in FIG. 6).

  Here, if the target moving object is set to be tracked even if the skin color segment disappears from the captured frame data 1D, the score is always the highest for the neighboring segments even if the skin color block 6 is not held. The weight is set as follows.

  In this case, the moving object segment 7 of the person K is selected as a tracking target (see step S10 in FIG. 6). Since the moving object segment 6 has no skin color block, the center of gravity of the moving object segment 6 itself is calculated as it is (see steps S11 and S13 in FIG. 6). Further, by setting the center of gravity to be at the center of the captured image by pan / tilt control, the position where the person K that is the tracking target is located at the center of the image as in the frame data 1F of FIG. it can.

(5) Other Embodiments Since any person can be tracked, only the person is to be tracked. In the image frame of FIG. 10, the weight of the skin color block is weighted by the weight of the neighboring segment. In this case, the weight of the flesh color block detected by the flesh color detection unit 15 is prioritized over the neighboring segment, whereby the moving object segment 9 including the person J different from the previous tracking target becomes the tracking target. Thereafter, the center of gravity of the moving object segment 9 is calculated, and the position of the person J is captured as shown in the frame data 1G of FIG. 13 by setting the center of gravity to be the center of the captured image by pan / tilt control. The center of the video.

  As described above, in the person tracking device 10 according to the embodiment of the present invention, it is possible to quickly determine and track a person-like area by weighting the skin color block area of the moving object. Furthermore, once a moving object segment that looks like a person is detected, the moving object segment detected in the vicinity of the moving object segment is weighted so that it can be set to continue tracking the person even if other person-like areas are detected. it can.

The block diagram which shows the structure of embodiment of the person tracking device of this invention The figure which shows the last picture frame data Diagram showing the current image frame data Diagram showing moving object block Diagram showing moving object segment The flowchart which shows operation | movement of embodiment of the person tracking device of this invention. Coordinate diagram showing skin color area The figure which shows the last picture frame data Diagram showing the current image frame data Diagram showing moving object segment Diagram showing image frame data Diagram showing image frame data Diagram showing image frame data

Explanation of symbols

1A, 1B Frame data 2, 3 Block 4, 5 Moving object segment 6 Skin color block 10 Human tracking device 11 Image signal receiving unit 12 Frame memory (memory means)
13 Moving object detection unit (moving object detection means)
14 Segment creation section (segment creation means)
15 skin color detection unit (skin color segment detection means)
16 Neighborhood search part 17 Tracking object determination part (Tracking object determination means)
18 Tracking position determination unit 19 Coordinate data transmission unit 20 Skin color weight storage unit 21 Neighborhood weight storage unit 101 Monitoring camera 102 Monitor

Claims (1)

A person tracking device that extracts data from an input image signal and detects and tracks data relating to the movement of the skin color of an exposed portion of the person as the movement of the person,
Memory means for sequentially storing frame data obtained by converting the input image signal in units of image frames;
The difference in luminance data between the previous and current blocks obtained by dividing the previous frame data and the current frame data stored in the memory means into a plurality of blocks, respectively, exceeds a predetermined threshold value. When detected, a moving object detection means for detecting the block as a moving object block;
Segment creating means for creating a moving object segment obtained by segmenting the moving object block output from the moving object detecting means;
Skin color for detecting a moving object segment including a skin color block having skin color data based on the moving object segment output from the segment creating means and the current frame data output from the memory means Segment detection means;
Based on the moving object segment output from the flesh-colored segment detecting means and the moving object segment in the previous frame data that is most approximate to the moving object segment, the moving object to be subsequently tracked A person tracking device comprising tracking target determining means for determining a segment.

Images