JP4866159B2 - Moving body detection device - Google Patents

Description

The present invention relates to a moving body detection device used in, for example, a monitoring system for a driving state of a car on a road, a railroad track, a monitoring system at a railroad crossing, and the like, particularly from a background image of video information input by a TV camera or the like. The present invention relates to a moving body detection apparatus suitable for extracting a moving body.
In the present invention, the moving body detection apparatus according to the present invention can be applied to a system such as a parking lot monitoring system in addition to the system described above. An example will be described.

  2. Description of the Related Art A monitoring system for the state of travel of an automobile on a road is provided with a moving body detection device that extracts a moving body from video information input by a TV camera or the like.

For example, in Patent Document 1, as a method for detecting a moving object, a method for examining changes in the entire image area that have been proposed in the past requires a large-scale calculation, and the background and individual vehicle areas are automatically detected. For the reason that it is difficult to perform separation / detection in an objective and stable manner, the pixel data of the detection slit portion along the vehicle traveling direction set in advance in the image area is detected over time, Vehicles that perform line segment detection on the changed part, determine which part of the image area is the background or which part is the vehicle, and the amount of movement on the image, and superimpose them as separate images for each individual vehicle A technique is disclosed in which the extraction can be performed separately for each.
Japanese Patent Laid-Open No. 5-112191

However, in the detection method of the moving body of Patent Document 1, a change of the pixel data of the detection slit portion along the vehicle traveling direction set in advance in the image region is detected over time, and a line segment is detected for the changed portion. This is a method to determine which part of the image area is the background or which part is the vehicle, so the detection slit must be a slit along the vehicle traveling direction, and the moving direction of the moving body is not decided in advance Could become undetectable.
An object of the present invention is to provide a moving body detection apparatus that can accurately detect a moving body even if the moving direction is indefinite.

  A feature of the present invention is that a pixel readout unit that reads out pixel information corresponding to the detection slit from a plurality of detection slits set in parallel with a frame of pixel information stored in the pixel information storage unit, and a pixel readout unit Storage unit that stores pixel information read out from the pixel unit in time series, a background color calculation unit that calculates a background color from the pixel information stored in the storage unit, and a background color and pixel information calculated from the background color calculation unit A moving body pixel detection unit that detects pixel corresponding to the detection slit other than the background color by comparing pixel information from the storage unit, and a plurality of pixels corresponding to the detection slit obtained from the moving body pixel detection unit A block forming unit that divides into blocks, and a moving object detection unit that detects the presence or absence of a moving object based on the existence form of blocks adjacent to the block obtained by the blocking unit. It is characterized in that is.

  According to the present invention described above, since the pixels of the moving object are blocked and the moving object is estimated from the presence form of the block, it is possible to detect the moving object whose moving direction is indefinite.

Next, typical embodiments of the present invention will be described in detail with reference to the drawings.
The block diagram shown in FIG. 1 shows the functions of the moving object detection apparatus according to the present invention. In practice, these functions are executed by a software program of a computer.

  A video input unit 101 including a TV camera or the like captures a scene of a vehicle running state to be monitored, converts it into a video signal, and sends the video signal to the moving object detection apparatus 100. In the moving object detection apparatus 100, the video signal is framed as 25 to 30 still images per second, and the still image is digitized and stored in a memory in the moving object detection apparatus 100 for each frame.

  The moving object detection device 100 reads the memory contents, determines whether the pixel on the frame image is the background or the moving object, cuts out the moving object, according to the processing program stored at another address in the memory, Estimate the direction and speed of movement. The extracted image of the moving object and other processing results associated therewith are transmitted to the result output means 108 including a display or the like.

The result output means 108 outputs the background image processed by the moving object detection apparatus 100, the image of the moving object, the movement direction, and the speed on the screen.
The moving body detection apparatus 100 includes an image memory 102, a storage memory 103, a background color determination means 104, a moving body pixel detection means 105, a blocking means 106, a moving body detection means 107, and the like. Here, the moving body detection apparatus 100 has a function of storing video information obtained by converting into digital image data for each frame.

  Further, the storage memory 103 has a function of storing pixel information read by the detection slit set on one frame of the image memory 102 in time series. Note that storing in time series means that the video signal is set to 25 to 30 still images per second, and the still images are stored and managed at predetermined time intervals.

  Here, as shown in FIG. 3, the detection slits set on the frame are set with a plurality of detection slits 303 (here, 13 detection slits) on the frame 301, and the storage memory 103 stores the detection slits on the frame. Thus, pixel information of 13 detection slits 303 is stored. As described above, the pixel information of the 13 detection slits 303 is stored in the storage memory 103 in time series, but other than the car image 302 is stored as the background.

  The background color determining unit 104 has a function of determining a background color from the pixel data of the storage memory 103, and this uses a known background color determining method in which pixels having no temporal change in density are processed as a background color. This can be achieved.

  The moving object pixel detecting unit 105 has a function of detecting pixels of the moving object by comparing the pixel data from the image memory 102 and the pixel data from the background color determining unit 104, and this is also a difference image from the background. This can be realized by using a known pixel detection method such as taking the logical sum of

  The blocking unit 106 has a function of managing a plurality of pixels detected as a moving body as pixels in block units. As shown in FIG. 4, this concept of blocking defines a block in which a plurality of pixels are detected among the pixels detected by the detection slit 303. The number of pixels constituting the block differs depending on the pixel extracted by the detection slit 303, and one or more blocks exist on one slit. In this way, the pixels detected as the moving body are grouped into a plurality of pixels by the blocking unit 106.

  The moving object detection means 107 has a function of detecting the moving object by estimating the existence of the moving object based on the existence form of the blocked pixels. That is, as shown in FIG. 5, it is determined whether or not there is a pixel representing a moving body that is continuous in eight directions of upper, lower, left, right, upper right, lower right, upper left, lower left in the detection slit pixel. Is detected.

The mobile object detection apparatus 100 as described above is actually executed by a computer program, and its flowchart will be described with reference to FIG.
First, the moving image input from the image input means 101 is digitized as a still image in step 1 (hereinafter, step will be described as S) and read and stored in the image memory 102 in time series.

  Next, in the image stored in the image memory 102 in S2, pixel information at a preset position is read by a plurality of detection slits 303 as shown in FIG. The pixel information of the slit is stored in the storage memory 103. The storage memory 103 stores time-series pixel information as described above.

  When the pixel information of the plurality of detection slits 303 is stored in S3, the process proceeds to S4, and the background color for each pixel from the slits SL1 to SL13 of the detection slit 303 is determined. This background color is set as a background color for pixels having no density change from time-series pixel information. In this case, determination of the background color is started from SL1 of the detection slit 303.

  When the background color of SL1 of the detection slit 303 is determined in S4, the process proceeds to S5, and the presence of the moving object pixel is detected on a pixel-by-pixel basis. This can be realized by using a method of comparing the pixel data of the two and calculating the logical sum of the difference image with the background.

  When the background color of SL1 of the detection slit 303 is determined in S4 and S5 and the detection of the moving body pixel is completed, the process proceeds to S6 and the background color is determined in S4 and S5 for all the detection slits 303 from SL2 to SL13. Perform pixel detection.

When the pixel processing is executed for all the detection slits 303, the process proceeds to S7 to block the detection slits. Here, this blocking will be described with reference to FIG.
As shown in FIG. 4, the detection slit 303 has a plurality of pixels arranged continuously, and the block is defined as a group of the plurality of continuous pixels. The number of pixels constituting the block differs depending on the pixel extracted by the detection slit 303, and one or more blocks exist on one slit. In this way, the pixels detected as the moving body are grouped into a plurality of pixels by the blocking unit 106.

When the blocking of all the detection slits 303 is completed in S7, the process proceeds to S8 to detect the moving body based on the blocked pixels. Here, the detection method of this moving body is demonstrated based on FIG.
FIG. 5 is a diagram showing the association of adjacent blocks, and whether there is a moving body adjacent to the block 501 detected as a moving body in the eight directions around the top, bottom, left, right, top right, bottom right, top left, bottom left. The mobile object is detected by.

  The case shown in (a) shows the case where the block 502 is adjacent to the block 501 in the vertical direction, and the case shown in (b) is the case where the block 502 is adjacent to the block 501 in the horizontal direction. The case shown in (c) shows the case where the block 502 is adjacent to the block 501 in an oblique direction. In these cases, the block 501 and the block 502 are detected as one moving body.

On the other hand, the example shown in (d) shows the case where the block 503 does not exist adjacent to the block 501, and since the block 503 detected as a moving object is not adjacent to the block 501, the block 501 and the block 501 Reference numeral 503 is detected as a separate moving object.
In this way, a mobile unit in block units is detected as one mobile unit when adjacent blocks are the same mobile unit.

If a moving body is detected in S8, the process proceeds to S9, where the detected information on the moving body is combined with an input image in the image memory 102, and the result output means 108 superimposes and outputs each image.
In this way, since the pixels of the moving object are blocked and the moving object is estimated from the presence form of the block, it is possible to detect the moving object whose moving direction is indefinite.

Next, a method for detecting the moving direction of the moving body will be described.
FIG. 6 is a flowchart of the method of detecting the moving direction of the moving object from the moving object detection result described in the above embodiment. The detection result of the moving object is fetched in S10, and the process proceeds to S11. The information is stored in the storage memory 103, and the process further proceeds to S12 to detect the moving direction. As a moving direction detection method executed in S12, the moving direction is estimated from the table 701 formed in the storage memory 103 shown in FIG.

  In FIG. 7, the moving body information is stored in a table 701 in time series with the position of the slit with respect to the upper left position of the screen, the position of the block formed in this slit, the number of slits and the number of blocks indicating the size of the moving body.

  In the detection of the moving direction in S12, the time-series information value change of each information value read from the table 701 is checked, and the value becomes smaller as time (t) elapses as shown in (a), or When the value increases as time (t) elapses as in (b), it is estimated that the moving body is moving up and down, left and right, or approaching and moving away from the screen.

  In addition, as shown in (c), the information value changes in a vibration state as time (t) elapses, but the change is not large. If the information value changes within a certain range, for example, a tree branch swayed by the wind Presumed to be a moving body. Therefore, in this case, processing is performed assuming that the mobile body is not a detection target.

  FIG. 8 is a diagram for explaining an example of estimating the moving direction of the moving body executed in S12. The moving direction is determined by the state change of the moving body 801 before moving and the moving body 802 after moving from the above-described information values in time series. Is estimated.

  In the case of the left movement shown in (a), the moving body information stored in the table 701 is in a state where the value of the upper left slit position decreases and the values of the other items do not change.

  In the case of the upward movement shown in (b), the moving body information stored in the table 701 is in a state where the value of the upper left block position decreases and the values of the other items remain unchanged.

In the case of the approach shown in (c), the moving body information stored in the table 701 is in a state where the values of the upper left slit position and the upper left block position are decreased and the values of the slit number and the block number are increased.
As described above, the moving direction of the moving body is estimated from the time-series change of each information value in the table 701, and the moving direction is detected.

  Next, an example of using a plurality of input videos will be briefly described. In FIG. 9, an input video 901 is a video in which four videos from video 1 to video 4 are connected to a 2 × 2 grid, and the moving body detection device blocks the four videos of the input video 901 into blocks. By grouping blocks and performing a moving body detection process for each group, it is possible to realize detection of a moving body that is not affected by the connection even when the images are connected.

FIG. 10 illustrates the grouping of blocks described above. A plurality of blocks 1001 are grouped into a group 1002. At this time, the blocks 1001 are grouped according to the connected boundaries in the input video 901, and the moving object detection process can be performed for each group.
As described above, according to the present embodiment, it is possible to detect a moving body whose moving direction is indefinite, and further to detect the movement of the moving body in the three-dimensional direction.

It is a block diagram of the moving body detection apparatus which is one Example of this invention. It is an operation | movement flowchart figure of a moving body detection apparatus. It is the figure which showed the setting state of the detection slit in a flame | frame. It is a figure explaining the example which divides | segments a slit into a block. It is a figure explaining the example which detects a mobile body from the aggregate state of a block. It is an operation | movement flowchart figure of the moving direction detection method of a moving body. It is a figure explaining the detection method of the moving direction of a moving body. It is a figure explaining the example of a movement of a mobile body. It is a figure explaining the input image | video which connected the several image | video. It is a figure explaining grouping of a block.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 100 ... Moving body detection apparatus, 101 ... Image | video input means, 102 ... Image memory, 103 ... Storage memory, 104 ... Background color determination means, 105 ... Moving body pixel detection means, 106 ... Blocking means, 107 ... Moving body detection means 107, 108 ... result output means, 301 ... frame, 302 ... moving body, 303 ... slit, 501, 502, 503 ... block.

Claims (2)

Video input means for inputting video, pixel information storage means for storing video information input from the video input means as pixel information, mobile object detection means for detecting a mobile object from the input video, and detection results In the moving body detection apparatus having the detection result output means for outputting,
The moving body detecting means includes
A pixel readout unit that reads out pixel information corresponding to the detection slit from a plurality of detection slits set in parallel to a frame of pixel information stored in the pixel information storage unit;
An accumulation unit for storing pixel information read from the pixel reading unit in a time series in units of pixels;
A background color determination unit for determining a background color from the pixel information stored in the storage unit;
A moving object pixel detector for detecting a moving object pixel by comparing pixel information from the pixel information storage means as the background color calculated from the background color determination section,
Among the moving body pixels detected by the moving body pixel detection unit, a blocking unit that manages a plurality of moving body pixels continuous in one detection slit as a block;
Among the blocks managed by the blocking unit, another block existing in the eight directions around the upper, lower, right, upper right / lower right, upper left / lower left with respect to the block by the adjacent detection slit is the same moving body. And a moving body detecting unit that detects the moving body based on the determination.   2. The moving body detecting means according to claim 1, further comprising a moving direction detecting unit that detects a movement of the moving body in a three-dimensional direction of up and down, left and right, or approaching and moving away from changes in information of the detection slit and the block. A moving object detection apparatus characterized by the above.

Images