JP4268644B2 - Mobile object detection system, mobile object detection apparatus, mobile object detection method, and mobile object detection program - Google Patents

Description

  The present invention relates to a moving body detection system, a moving body detection device, a moving body detection method, and a moving body detection program for detecting movement information such as the number of passersby, vehicles, and the like, movement direction, and movement time. .

  2. Description of the Related Art Conventionally, various techniques for detecting movement information such as the number of moving objects such as passers-by and roads, vehicles, and the like, movement directions, and movement times have been devised.

  For example, in Patent Documents 1 and 2, two images (hereinafter, collectively referred to as stereo images) captured using time-synchronization using two image capturing devices (such as a video camera) are used according to a moving image and a still image. In other words, a technique for calculating a three-dimensional coordinate of a moving object such as a vehicle and identifying the moving object based on the three-dimensional coordinate is disclosed. ing.

  In particular, in the moving body detection technique disclosed in Patent Document 1, a rigid body (in this case, a moving body) is specified (determining whether or not it is a rigid body) based on the degree of variation in Euclidean distance. In the moving body detection technique disclosed in the above, when the moving body is a person, the person is specified by performing clustering processing from the height of each person.

Recently, a technique for specifying the coordinates (shape) of a three-dimensional object in the image based on the stereo image (hereinafter referred to as a stereo matching method) has also been devised (see, for example, Patent Document 3). ).
JP-A-11-66319 Japanese Patent Laid-Open No. 10-49718 Japanese Patent Laid-Open No. 3-167678

However, the conventional technique has the following problems.
That is, in the techniques disclosed in Patent Documents 1, 2, and 3, only the position of the moving body (space area occupied by the moving body) in space (three-dimensional coordinate system) is specified, and the specified movement is performed. It is difficult to acquire movement information over time related to the body.

  An object of the present invention is to accurately specify a position of a moving body in a space and to accurately acquire movement information with the passage of time of the position-specified moving body.

In order to solve the above-mentioned problem, the invention described in claim 1
Based on the left and right imaging devices arranged on the stereo baseline and the time-sequential multiple frames of the stereo still images taken by the two imaging devices in time synchronization, the images are copied to the stereo still images. In a mobile body detection system comprising a mobile body detection device for detecting a mobile body
The mobile object detection device is:
Among the plurality of frames of stereo still images, for each of the left and right still images, an inter-frame difference image representing a difference between two frames of still images and one of the two frames of still images and the background Creating a background difference image representing a difference between the images, and then combining the inter-frame difference image and the background difference image to identify a moving body region in the composite image;
Three-dimensional measuring means for assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference images;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. A different label for each color is assigned to the body region, the three-dimensional coordinates are assigned to the labeled mobile body regions, and the labeled mobile body regions are separated based on the three-dimensional coordinates. Body separation means;
Mobile object tracking that determines whether or not the mobile objects corresponding to each time are the same according to the degree of overlap of the mobile object regions corresponding to two different times, and tracks the mobile object based on the determination result Means,
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects Moving body measuring means for measuring the direction,
It is provided with.

Moreover, in order to solve the said subject, invention of Claim 2 is the following.
Among the time-sequential multiple frames of stereo still images that are captured in time synchronization by the left and right imaging devices arranged on the stereo baseline, for each left and right still image, between two frames of still images An inter-frame difference image representing a difference and a background difference image representing a difference between any one of the two frames of still images and the background image are created, and then the inter-frame difference image and the background difference image Moving body region specifying means for specifying the moving body region in the composite image by combining
Three-dimensional measuring means for assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference images;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. A different label for each color is assigned to the body region, the three-dimensional coordinates are assigned to the labeled mobile body regions, and the labeled mobile body regions are separated based on the three-dimensional coordinates. Body separation means;
Mobile object tracking that determines whether or not the mobile objects corresponding to each time are the same according to the degree of overlap of the mobile object regions corresponding to two different times, and tracks the mobile object based on the determination result Means,
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects Moving body measuring means for measuring the direction,
It is provided with.

Moreover, in order to solve the said subject, invention of Claim 3 is the following.
Among the time-sequential multiple frames of stereo still images that are captured in time synchronization by the left and right imaging devices arranged on the stereo baseline, for each left and right still image, between two frames of still images An inter-frame difference image representing a difference and a background difference image representing a difference between any one of the two frames of still images and the background image are created, and then the inter-frame difference image and the background difference image Identifying a moving body region in the composite image by combining
Assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference images;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. Assigning a different label for each color to the body region, assigning the three-dimensional coordinates to each labeled mobile body region, and separating each labeled mobile body region based on the three-dimensional coordinates When,
Determining whether or not the mobile objects corresponding to the respective times are the same according to the degree of overlap of the mobile object regions corresponding to two different times, and tracking the mobile object based on the determination results;
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects Measuring the direction, and
It is characterized by including.

Moreover, in order to solve the said subject, invention of Claim 4 is the following.
In the computer that performs image analysis,
Among the time-sequential multiple frames of stereo still images that are captured in time synchronization by the left and right imaging devices arranged on the stereo baseline, for each left and right still image, between two frames of still images An inter-frame difference image representing a difference and a background difference image representing a difference between any one of the two frames of still images and the background image are created, and then the inter-frame difference image and the background difference image A function for identifying a moving body region in the composite image by combining
A function of assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference image;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. A function of assigning a different label for each color to a body region, assigning the three-dimensional coordinates to each labeled mobile body region, and separating each labeled mobile body region based on the three-dimensional coordinates When,
According to the degree of overlap of the moving body region corresponding to two different times, it is determined whether or not the moving body corresponding to each time is the same, and the function of tracking the moving body based on the determination result;
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects A function to measure the direction,
Is realized.

  According to the present invention, not only the position of a moving body such as a passerby (a space area occupied by the moving body) can be accurately identified in the space, but also the movement information associated with the passage of time related to the identified moving body can be accurately acquired. Therefore, advanced mobile object detection can be realized.

  Hereinafter, preferred first and second embodiments of the present invention will be described in order with reference to the accompanying drawings.

(First embodiment)
A mobile object detection system 100 according to a first embodiment to which the present invention is applied will be described in detail with reference to the drawings.
The mobile object detection system 100 is based on a plurality of time-series stereo images taken in time synchronization by two imaging devices arranged on a stereo baseline, and the number of mobile objects such as passers-by passing on a road or the like. And movement information such as movement direction and movement time. In the first embodiment, as an example, a case will be described in which movement information of passers-by passing through the predetermined station staircase 1 shown in FIGS. 1A and 1B is acquired. The present invention is not limited to this, and the present invention can also be applied to vehicles passing on a road, vehicles entering / exiting a parking lot, and the like. Hereinafter, the mobile means a passerby.

First, the configuration of the moving object detection system 100 will be described.
As shown in FIG. 2, the moving body detection system 100 includes a stereo base line device 10, a moving body detection device 20, a database 30, and the like.

  The stereo baseline device 10 includes photographing devices 10a and 10b such as video cameras. In addition, the stereo base line device 10 is installed on the station staircase 1 in the arrangement as shown in FIGS. 1A and 1B, and a state in which the moving body passes through the station staircase 1 is shown in the imaging devices 10 a and 10 b. Shoot with time synchronization. The arrangement of the stereo base line device 10 is not limited to that shown in FIGS. 1A and 1B, and can be freely set.

  The moving body detection device 20 includes a moving body area specifying means 21, a congestion level specifying means 22, a moving body specifying means 23, and a movement information acquiring means 24. The moving body area specifying means 21, the congestion level specifying means 22, and the moving body specifying means are provided. The moving body detection process shown in the flowchart of FIG. 3 is performed using the means 23 and the movement information acquisition means 24.

  The mobile body detection device 20 executes a mobile body detection program stored in advance in a built-in memory (not shown), thereby moving the mobile body area specifying means 21, the congestion level specifying means 22, the mobile body specifying means 23, and the movement information acquiring means 24. Each is controlled and the moving body detection process shown in the flowchart of FIG. 3 is performed.

  Note that the mobile object region specifying means 21, the congestion level specifying means 22, the mobile object specifying means 23, and the movement information acquiring means 24 may represent functions of the mobile object detection program, or may be hardware. Good.

  The database 30 stores past moving object detection results (moving information such as the number of moving objects passing through the station staircase 1, moving direction, moving time, etc.) and various types of information used in the moving object detection process (for example, 3D coordinate system information set as a photogrammetric orientation element, a background image of the station staircase 1, a moving body template, etc.).

  Here, the three-dimensional coordinate system information is stored in a three-dimensional space including the station staircase 1 when the stereo base line device 10 is installed at a position as shown in FIGS. 1A and 1B, for example. This is information representing a coordinate system set for the user. When the installation position of the stereo base line device 10 is changed, the three-dimensional coordinate system is reset according to the change.

  The background image of the station staircase 1 is an image of only the station staircase 1 where no moving object is shown. For example, a plurality of background images taken in each time zone such as morning, noon, evening, and night are prepared.

  For example, as shown in FIG. 4, a plurality of types of human shapes corresponding to the congestion level specified in the mobile body detection process are prepared as the mobile body template.

FIG. 4 shows four types of mobile templates corresponding to four levels of congestion levels 1 to 4. The moving body template A1 is used when the congestion level is 1 (for example, the average interval between the moving body and the moving body is within 0 to 0.1 m), and the moving body template A2 is the congestion level 2 (for example, the moving body and the moving body Mobile unit template A3 has a congestion level 3 (for example, the average interval between the mobile unit and the mobile unit is within 0.5 to 1 m). ) And the moving body template A4 is used when the congestion level is 4 (for example, the average interval between the moving body and the moving body is 1 m or more).
The congestion level and the criterion for determining the congestion level are not limited to those described above, and can be freely set.

Next, the moving body detection process performed by the moving body detection device 20 will be described with reference to FIG.
First, each process of step S1 and S2 demonstrated below is performed by the mobile body area | region identification means 21. FIG.

  That is, a stereo moving image photographed by the stereo base line device 10 is extracted as a still image every predetermined time (for example, 1 second or the like, a time during which the movement of the moving body can be tracked), and a plurality of stereo still images are extracted. Created in time series.

  Then, based on one of the created stereo still images and the length of a predetermined object (for example, the width of the station stair 1 etc.) captured in the stereo still image, the station stair 1 A three-dimensional coordinate system as a photogrammetric orientation element is set for a three-dimensional space including (step S1).

After step S1, for each of the created stereo still images, the space area of each moving object (three-dimensional coordinates for specifying the three-dimensional shape of the moving object) captured in the stereo still image is Based on the set three-dimensional coordinate system, it is calculated by the stereo matching method (step S2).
Further, after step 2, the congestion level specifying means 22 performs steps S3 and S4 described below.

  That is, after step S2, a background image that matches the time zone in which the image was taken is selected, and a difference area with respect to the background image (indicating an area occupied by the moving object) for each of the created stereo still images. .) Is extracted (step S3).

  After step S3, the area ratio for the entire image of the extracted difference area is calculated, and the congestion level in the station staircase 1 (for example, four levels of congestion as shown in FIG. 4) is calculated according to the calculated area ratio. .) Is specified (step S4).

  Furthermore, after step S4, each process of steps S5 and S6 described below is performed by the moving body specifying means 23.

  That is, after step S4, a mobile template is selected according to the specified congestion level (step S5).

  After step S5, the mobile object template selected in step S5 is assigned to the space area of each passerby calculated in step S2, and the mobile object is specified (step S6).

  Here, a state in which the mobile object template is assigned to the space area of the mobile object is shown in FIG. The image B1 shows a state where the moving body template B12 is moved along the moving path B11 in order to detect the space area B13 of the moving body, and the image B2 detects the space area B13, and the moving body template A state in which B12 is assigned to the space area B13 is shown.

  Each process of step S1-S5 demonstrated above is performed with respect to each of the several stereo still image produced in the said time series.

  Further, after step S6, the movement information acquisition means 24 tracks the movement of each moving object based on the plurality of time-series stereo still images, and for the same moving object, the space of the moving object is tracked. The moving body template assigned to the region is assigned the same color and the movement of the moving body is recorded, and the movement information of each moving body (based on the moving body template) is recorded based on the color attached to the moving body template. The moving direction, the moving time, etc.) are calculated (step S7).

  Here, as a method for tracking the movement of the moving object, for example, there is a method using an optical flow. This method is a method for estimating the motion of a moving object by calculating an optical flow (movement vector) at each point of the moving object. Specifically, two methods called a matching method and a gradient method are generally used. ing.

FIG. 6 shows a state in which the movement of the moving body is tracked using the optical flow. Images C1 and C2 are still images corresponding to each of the left (Left) / right (Right) of the stereo still images at time L, and images C3 and C4 have elapsed by ΔT from time L. Of the stereo still images after the above, the still images corresponding to the left / right. The smaller ΔT, the more processing time is required, but the tracking of the moving object becomes more accurate.

  In the images C3 and C4, only two optical flows are shown for simplification of illustration.

  Further, as a method for obtaining movement information of each moving body based on the color attached to the moving body template, for example, there is a method using a Munsell color system for color identification.

  As described above, the moving body detection system 100 detects a road or the like based on a plurality of time-series stereo still images that are captured in time synchronization by the two imaging devices 10a and 10b arranged on the stereo baseline. The spatial area (three-dimensional coordinates) occupied by each moving mobile object is specified, the difference area with respect to the background image of the stereo still image is extracted, the area ratio of the difference area with respect to the entire image area is calculated, and the area ratio is calculated. Based on this, the congestion level of the mobile body is specified, and the mobile body templates A1 to A4 corresponding to the congestion level are assigned to the space area of the specified mobile body to specify the mobile body.

  Furthermore, the movement of a moving object is tracked based on a plurality of time-series stereo still images, and the movement object template assigned to the space area of the same moving object is assigned the same color to record the movement of each moving object. The movement information (movement direction, movement time, etc.) of the moving body is acquired based on the color.

  Therefore, not only can the mobile object position (the space area occupied by the mobile object) such as the mobile object be accurately identified in the three-dimensional space, but also the movement information associated with the passage of time related to the identified mobile object can be accurately acquired. Advanced mobile object detection can be realized.

  The description in the first embodiment shows an example of the moving body detection system and the moving body detection device according to the present invention, and the present invention is not limited to this. The detailed configuration and detailed operation of the moving object detection system 100 and the moving object detection device 20 in the first embodiment can be changed as appropriate without departing from the spirit of the present invention.

  For example, the moving body region specifying unit 21 extracts a stereo moving image captured by the stereo base line device 10 as a still image every predetermined time, and creates a plurality of stereo still images in time series. Alternatively, the photographing apparatuses 10a and 10b can photograph a plurality of time-series stereo still images by continuous shooting. In this case, the imaging devices 10a and 10b may be either a video camera or a digital camera as long as continuous shooting is possible.

  Moreover, the stereo image image | photographed with imaging | photography apparatus 10a, 10b may be made to input into the mobile body detection apparatus 20 via recording media, such as a videotape and DVD (Digital Versatile Disk), or a stereo base line apparatus. 10 and the mobile object detection device 20 may be provided in advance, and the stereo base line device 10 may input to the mobile object detection device 20 via the data line.

(Second Embodiment)
Next, a mobile object detection system 100a according to a second embodiment to which the present invention is applied will be described in detail with reference to the drawings.
Similar to the moving body detection system 100, the moving body detection system 100 a travels on a road or the like based on a plurality of time-series stereo images captured in time synchronization by two imaging devices arranged on the stereo baseline. This is for acquiring movement information such as the number of moving bodies such as passersby, the moving direction, and the moving time. In the second embodiment, as in the first embodiment, the movement information of passers-by passing through the predetermined station staircase 1 shown in FIGS. 1A and 1B is acquired. However, the present invention is not limited to this, and the present invention can be applied to a vehicle passing on a road, a vehicle entering and leaving a parking lot, and the like. Hereinafter, the mobile means a passerby.

First, the configuration of the moving object detection system 100a will be described.
As shown in FIG. 7, the moving object detection system 100 a includes a stereo base line device 10, a database 30, and a moving object detection device 40. Here, the stereo base line device 10 and the database 30 are the same as those described in the first embodiment, and the same reference numerals are given for the sake of brevity and description thereof is omitted.

  The moving object detection device 40 includes a moving object region specifying means 41, a three-dimensional measuring means 42, a moving object separating means 43, a moving object tracking means 44, and a moving object measuring means 45, and using these means, the flowchart of FIG. The moving body detection process shown in FIG.

  The mobile object detection device 40 executes a mobile object detection program 41, a three-dimensional measurement means 42, a mobile object separation means 43, a mobile object tracking means 44, by executing a mobile object detection program stored in advance in a built-in memory (not shown). Each of the moving body measuring means 45 is controlled to perform a moving body detection process shown in FIG.

  The moving body region specifying means 41, the three-dimensional measuring means 42, the moving body separating means 43, the moving body tracking means 44, and the moving body measuring means 45 may represent each function of the moving body detection program. It may be based on hardware.

Next, a moving body detection process performed by the moving body detection device 40 will be described with reference to FIG.
First, the state of the station staircase 1 is photographed as a stereo image (moving image) by the stereo base line device 10 (step S1a), and the station staircase 1 without a moving body is a left and right background image (still image, left background). An image, which is referred to as a right background image) is taken by the stereo baseline device 10 (step S1b).

  Next, each process (steps S10 to S14) of the moving body region specifying process D1 will be described. The moving body area specifying process D1 is executed by the moving body area specifying means 41.

  Of the stereo images (moving images), still images (still images at times t1 and t2) for two left and right frames separated from each other by a predetermined time (for example, a short time such as 0.1 second) are acquired. . The acquired still images for two frames on each of the left and right sides are referred to as a t1 left image, a t2 left image (hereinafter referred to as a left still image), a t1 right image, and a t2 right image (hereinafter referred to as a right still image).

  Next, the t1 left image, the t2 left image, the t1 right image, the t2 right image, and the left and right background images are subjected to HSV (Hue Saturation Value; hue, saturation, brightness) processing, and the hue from the RGB color system. Conversion into the HSV color system based on saturation and lightness (Step S10).

  After step S10, inter-frame difference images (left inter-frame difference image and right inter-frame difference image) are respectively applied to the t1 left image and the t2 left image and to the t1 right image and the t2 right image. Create (step S11). Here, the inter-frame difference is a difference (change region) between images at different times, and corresponds to an edge portion of the moving object region.

  The background difference based on the saturation and the hue for the t1 left image (or t2 left image) and the left background image, and for the t1 right image (or t2 right image) and the right background image, respectively. Each image is created (step S12). The background difference corresponds to the inside of the moving object area.

  Here, since the area extracted as the difference is different between the hue and the saturation, the two are combined, and the combined image is used as the final background difference image (left background difference image, right background difference image). .

  After steps S11 and S12, each of the left frame difference image and the right frame difference image is subjected to black and white binarization processing and expansion / contraction processing (step S13).

  By the above expansion and contraction processing, each background difference area (inside the moving body area) of the left background difference image and right background difference image, and each frame difference area (in the moving body area) of the left frame difference image and right frame difference image. Edge portions) can be adapted to each other well.

  After step S13, the left frame difference image and the left background difference image are combined (referred to as a left combined image), a moving body region for the left image is specified, and the right frame difference image and the right background difference image are combined. Then, the moving body region for the right image is specified (referred to as the right composite image) (step S14).

  Next, each process (steps S11, S21 to S23) of the three-dimensional measurement process D2 will be described. The three-dimensional measurement process D2 is executed by the three-dimensional measurement unit 42.

  The left and right inter-frame difference images created in step S11 are divided into n sections up to sections K1 to Kn (n is a natural number) as shown in FIG. 9A. The luminance value for each pixel in the horizontal direction of the image (the direction indicated by symbol R in the figure) is integrated (step S21).

  An example of the luminance value integration result is shown in FIG. Here, the integration results for category 1 are shown. The horizontal axis represents the position in the horizontal direction of the image using pixel coordinates, and the vertical axis represents the integrated luminance value. Here, the pixel coordinates are coordinates for uniquely specifying a position in the image using pixel numbers assigned to all the pixels constituting the image.

Next, measurement points for assigning three-dimensional coordinates to the image are specified.
Of the luminance value integration results calculated in step S21, locations corresponding to the peak value (for example, locations indicated by reference signs P1, P2, etc. in the figure, a plurality of locations) are extracted for each of the categories K1 to Kn. The position (measurement point) in the difference image between the left and right frames taking the peak value is specified by pixel coordinates (step S22).

  Then, template matching is performed in order to make the measurement point (position where the peak value is obtained) exactly match between the left and right frame difference images. For example, referring to FIG. 10, a position (right peak point) in the right interframe difference image that exactly corresponds to a predetermined position (referred to as a left peak point) that takes a peak value in the left interframe difference image. Will be described with reference to template matching.

First, a predetermined image vicinity centered on the left peak point V1 (referred to as a left template T1) is extracted, and then wider than the left template V1 centered on the right peak point V1 corresponding to the left peak point (for example, A predetermined image area (referred to as right template T2) is extracted. Then, in the left template T2, a location having the highest correlation (that is, the correlation of pixel values) with the right template T1 is specified, and further, based on this result, the position corresponding to the left peak point in the right interframe difference image is (This location may be different from the original right peak point.)

  After step S22, the three-dimensional coordinate value is assigned to the specified measurement point using the orientation measurement result S1c acquired in advance based on stereo measurement (step S23).

  Here, the orientation measurement refers to obtaining measurement values such as the inclinations of the imaging devices 10a and 10b, and using the measurement values, the same points (pixel coordinates) on the images respectively captured by the imaging devices 10a and 10b. ) Is specified, a parameter is calculated so that three-dimensional coordinates representing the real space can be obtained using the measured value.

  Further, the three-dimensional coordinates are transformed using the measurement values obtained by the orientation measurement (the inclinations of the imaging devices 10a and 10b, etc.), and the direction of any one of the coordinate axes (for example, the z-axis) is in the vertical direction. It is preferable to match. If such a coordinate system is used, it becomes easy to perform the moving body separation process D3 described later.

  Next, each process (steps S30 to S32) of the moving object separation process D3 will be described. The moving body separating process D3 is executed by the moving body separating means 43.

  After step S14 of the above-described moving body region specifying process D1, each moving body region specified in step S14 is labeled (numbered) based on the color.

  First, the t1 left image (or t2 left image) and the t1 right image (or t2 right image) are subjected to color reduction processing and smoothing processing under the RGB color system. By the color reduction process and the smoothing process, the number of regions labeled based on the color is limited, and unnecessary labeling can be avoided.

  Based on the t1 left image (or t2 left image) and t1 right image (or t2 right image) after the color reduction processing and the smoothing processing, and each moving body region specified in step S14, Colors are assigned and labeled (step S30).

  Thereafter, the three-dimensional coordinates specified in step S23 are assigned to each moving object region labeled in step S30, and each moving object is separated by cluster analysis (ward method) based on the three-dimensional coordinates (steps). S32). At this time, if the distance between the labels is equal to or less than a predetermined threshold (moving body separation coefficient), the labels are grouped and assigned again to a single label. Here, the moving body separation coefficient is set in advance or is set each time by the operator (step S31).

  Next, the moving body tracking process D4 (step S40) will be described. The moving body tracking process D4 is executed by the moving body tracking means 44.

  Here, each moving body acquired in step S32 is tracked every time using a lap rate (described below). Hereinafter, the tracking method using the wrap rate will be described with reference to FIG.

  As shown in FIG. 11, at time t, three mobile bodies (labeled L11, L12, and L13, respectively) are separated and identified by the above method (ie, cluster analysis), and at time t + Δt, It is assumed that five moving bodies (labeled L21, L22, L23, L24, and L25, respectively) are separated and specified by this method.

  Here, the moving body region of each moving body at time t overlaps the moving body region of each moving body at time t + Δt. As a result of the superposition, when the ratio of overlapping areas (lap ratio) is 50% or more, it is determined that the moving bodies at the time t and the time t + Δt are the same, and the movement at the time t + Δt. Relabeled as a body.

  For example, the moving body L11 at time t overlaps with the moving bodies L21 and L22 at time t + Δt, but since both have a lap ratio of 50% or more, the moving body L11 and the moving bodies L21 and L22 are the same moving body. It is determined that

  Further, although the moving body L12 at time t overlaps with the moving body L23 at time t + Δt, since the lap rate is 50% or more, it is determined that the moving body L23 and the moving body L12 are the same moving body. .

  Further, the moving body L13 at time t overlaps with the moving bodies L24 and L25 at time t + Δt, but the moving body L13 and the moving body L25 are not the same because the wrap rate of the moving body L25 is less than 50%. Since the lap rate is 50% or more for the moving body L24, it is determined that the moving body L13 and the moving body L24 are the same.

  Furthermore, since the tracking process of the mobile body is sequentially performed over a relatively long time, the mobile body can be accurately measured. Hereinafter, this moving body measuring method will be described with reference to FIG.

  In FIG. 12, three mobile bodies (labeled L101, L102, and L103, respectively) are identified at time t10, and three mobile bodies (L111, L112, and L113, respectively) are identified at time t11. At time t12, three mobile bodies (labeled at L121, L122, and L123, respectively) are specified, and at time t13, four mobile bodies (each at L131) are specified. , L132, L133, and L134) are identified, and at time t14, four mobile bodies (labeled L141, L142, L143, and L144, respectively) are identified.

  At time t10, the two mobile bodies L101 and L102 are determined to be different mobile bodies. However, at time t11, the mobile bodies L101 and L102 are both moved in a single movement by the mobile body tracking process based on the lap rate. It is determined that it is a part of the body L111. Thereafter, at times t12, t13, and t14, the mobile body is sequentially labeled as L121, L131, and L141.

  In addition, since the moving bodies L103 and L112 determined to be the same moving body before time t11 have disappeared after time t12, they are not added to the measurement of the number of moving bodies.

  In addition, at time t12, the mobile body L122 determined to be a single mobile body is determined to be two mobile bodies after time t13. One is a moving body labeled L132 and L142, and the other is a moving body labeled L133 and L143.

  In addition, at time t11, the moving body L113 determined to be a single moving body is determined to be a single moving body (moving bodies labeled at L123, L134, and L144) after time t12. The

  Next, the moving body measurement process D5 (step S50) will be described. The moving body measuring process D5 is executed by the moving body measuring means 45.

  Here, in step S40, it is determined in which direction each moving body tracked for each time is moving, and the number of moving bodies in each moving direction such as up / down is measured.

  That is, as shown to Fig.13 (a), a background image (namely, image showing the station staircase 1) is divided | segmented into three measurement areas M1-M3, and each measurement area M1- of the mobile body area | region of each mobile body. The proportion of M3 is calculated every hour (t1 to t10). An example of the calculation result is shown in FIGS. 13 (b) and 13 (c).

  In FIG. 13B, when the moving body moves from the measurement area M1 to the measurement area M3 (downward direction), the ratio of the moving body to each of the measurement areas M1 to M3 is displayed, and FIG. When the mobile body moves from the measurement area M3 to the measurement area M1 (upward direction), the ratio of the mobile body to the measurement areas M1 to M3 is displayed.

  Further, since it is necessary to track the number of moving bodies over a certain amount of time, the number of moving bodies is measured for each moving direction such as up / down in the measurement area M2 located in the approximate center of the image representing the station staircase 1.

  As described above, the moving object detection system 100a creates an inter-frame difference image and a background difference image based on two left and right images, and specifies a moving object region from the difference images (moving object). Area specifying process D1), further, assigning three-dimensional coordinates to the inter-frame difference image (three-dimensional measurement process D2), labeling the moving body area for each color, and labeling each moving body area 3D coordinates are assigned to each, and each moving body is separated and specified by cluster analysis based on the 3D coordinates (moving body separation processing D3), and further, the moving body is tracked by time using the lap rate (moving) Body tracking processing D4), and further, based on the tracking result, the moving direction and the number of moving bodies are measured (moving body measuring processing D5).

  Therefore, not only can the mobile object position (the space area occupied by the mobile object) such as the mobile object be accurately identified in the three-dimensional space, but also the movement information associated with the passage of time related to the identified mobile object can be accurately acquired. Advanced mobile object detection can be realized.

  The description in the second embodiment shows an example of the moving object detection system and the moving object detection device according to the present invention, and the present invention is not limited to this. The detailed configuration and detailed operation of the moving object detection system 100a and the moving object detection device 40 in the second embodiment can be changed as appropriate without departing from the spirit of the present invention.

  For example, stereo images captured by the imaging devices 10a and 10b may be input to the moving object detection device 40 via a recording medium such as a video tape or a DVD (Digital Versatile Disk), or a stereo baseline device. 10 and the mobile object detection device 40 may be provided in advance, and the stereo base line device 10 may input to the mobile object detection device 40 via the data line.

The example of installation of the stereo baseline apparatus which the mobile body detection system to which the present invention is applied is shown, (a) is a view of the installation location of the stereo baseline apparatus from above, and (b) is shown in (a). It is the figure which looked at the installation place of a stereo base line apparatus from the side. It is a block diagram which shows the structure of the mobile body detection system to which this invention is applied. It is a flowchart explaining the mobile body detection process performed by the mobile body detection system to which this invention is applied. It is a figure which shows the mobile body template used in the case of the mobile body detection process shown to the flowchart of FIG. It is a figure which shows a mode that the mobile body template shown in FIG. 4 is allocated to the space area | region of a mobile body. It is a figure which shows a mode that the motion of a moving body is tracked using an optical flow. It is a block diagram which shows the structure of the other mobile body detection system to which this invention is applied. It is a flowchart explaining the mobile body detection process performed by the mobile body detection system shown in FIG. (A) It is a figure which shows the division | segmentation area | region used as the integration object of the luminance value in a difference image between frames, (b) It is a figure which shows the integrated value of the luminance value in one division with respect to each pixel coordinate of an image horizontal direction. . It is a figure for demonstrating template matching. It is a figure for demonstrating the mobile body tracking method using a lap | wrap rate. It is a figure which shows a mode that a mobile body is tracked. (A) is a figure which shows a mode that three measurement areas are provided in the image which shows a station stairs, (b) is a mobile body in each measurement area, when a mobile body moves to a down direction. It is a figure which shows the ratio which overlaps for every time, (c) is a figure which shows the ratio for which a mobile body overlaps each measurement area for every time when a mobile body moves to an up direction.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Station stairway 10 Stereo base line apparatus 10a Imaging apparatus 10b Imaging apparatus 20 Moving body detection apparatus 21 Moving body area identification means 22 Congestion level identification means 23 Moving body identification means 24 Movement information acquisition means 30 Database 40 Moving body detection apparatus 41 Moving body area Identifying means 42 Three-dimensional measuring means 43 Moving object separating means 44 Moving object tracking means 45 Moving object measuring means 100, 100a Moving object detection system

Claims (4)

Based on the left and right imaging devices arranged on the stereo baseline and the time-sequential multiple frames of the stereo still images taken by the two imaging devices in time synchronization, the images are copied to the stereo still images. In a mobile body detection system comprising a mobile body detection device for detecting a mobile body
The mobile object detection device is:
Among the plurality of frames of stereo still images, for each of the left and right still images, an inter-frame difference image representing a difference between two frames of still images and one of the two frames of still images and the background Creating a background difference image representing a difference between the images, and then combining the inter-frame difference image and the background difference image to identify a moving body region in the composite image;
Three-dimensional measuring means for assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference images;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. A different label for each color is assigned to the body region, the three-dimensional coordinates are assigned to the labeled mobile body regions, and the labeled mobile body regions are separated based on the three-dimensional coordinates. Body separation means;
Mobile object tracking that determines whether or not the mobile objects corresponding to each time are the same according to the degree of overlap of the mobile object regions corresponding to two different times, and tracks the mobile object based on the determination result Means,
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects Moving body measuring means for measuring the direction,
A moving body detection system comprising: Among the time-sequential multiple frames of stereo still images that are captured in time synchronization by the left and right imaging devices arranged on the stereo baseline, for each left and right still image, between two frames of still images An inter-frame difference image representing a difference and a background difference image representing a difference between any one of the two frames of still images and the background image are created, and then the inter-frame difference image and the background difference image Moving body region specifying means for specifying the moving body region in the composite image by combining
Three-dimensional measuring means for assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference images;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. A different label for each color is assigned to the body region, the three-dimensional coordinates are assigned to the labeled mobile body regions, and the labeled mobile body regions are separated based on the three-dimensional coordinates. Body separation means;
Mobile object tracking that determines whether or not the mobile objects corresponding to each time are the same according to the degree of overlap of the mobile object regions corresponding to two different times, and tracks the mobile object based on the determination result Means,
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects Moving body measuring means for measuring the direction,
A moving body detection apparatus comprising: Among the time-sequential multiple frames of stereo still images that are captured in time synchronization by the left and right imaging devices arranged on the stereo baseline, for each left and right still image, between two frames of still images An inter-frame difference image representing a difference and a background difference image representing a difference between any one of the two frames of still images and the background image are created, and then the inter-frame difference image and the background difference image Identifying a moving body region in the composite image by combining
Assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference images;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. Assigning a different label for each color to the body region, assigning the three-dimensional coordinates to each labeled mobile body region, and separating each labeled mobile body region based on the three-dimensional coordinates When,
Determining whether or not the mobile objects corresponding to the respective times are the same according to the degree of overlap of the mobile object regions corresponding to two different times, and tracking the mobile object based on the determination results;
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects Measuring the direction, and
A moving body detection method comprising: In the computer that performs image analysis,
Among the time-sequential multiple frames of stereo still images that are captured in time synchronization by the left and right imaging devices arranged on the stereo baseline, for each left and right still image, between two frames of still images An inter-frame difference image representing a difference and a background difference image representing a difference between any one of the two frames of still images and the background image are created, and then the inter-frame difference image and the background difference image A function for identifying a moving body region in the composite image by combining
A function of assigning left and right common three-dimensional coordinates to each difference area in the created left and right inter-frame difference image;
For each of the left and right still images, the color of any one of the two frames of still images is reduced, the still image after the color reduction is overlapped with the specified moving body region, and each movement after the overlapping is performed. A function of assigning a different label for each color to a body region, assigning the three-dimensional coordinates to each labeled mobile body region, and separating each labeled mobile body region based on the three-dimensional coordinates When,
According to the degree of overlap of the moving body region corresponding to two different times, it is determined whether or not the moving body corresponding to each time is the same, and the function of tracking the moving body based on the determination result;
Dividing the still image into a plurality of measurement areas in advance, determining which measurement area of the plurality of measurement areas the measurement object has passed in which order, and the number and movement of the movement objects A function to measure the direction,
A moving body detection program for realizing.

Images