JP4545211B2 - Moving object detection device - Google Patents

Description

  The present invention relates to a moving object detection apparatus and a moving object detection method for detecting a moving object by comparing moving images between frames, and in particular, moving that can detect a moving object stably regardless of noise of the moving image. The present invention relates to an object detection device and a moving object detection method.

  As a conventional moving object detection apparatus that detects moving objects by comparing moving images between frames, for example, there is one described in Patent Document 1. This conventional moving object detection device is intended to suppress erroneous detection of motion vectors in the background region without impairing the detection performance of the motion vectors in the moving object region, and has a different threshold between the moving object region and the background region. A value is used to switch whether or not to output a motion vector.

Specifically, first, the moving image is divided into a plurality of blocks for each frame, and the sum of absolute differences of pixel values indicating the degree of matching is calculated for each block between blocks. Next, the minimum value of the sum of absolute differences is obtained for each block to detect a motion vector. Next, for each block, a difference between a minimum value that is a sum of absolute differences and a sum of absolute differences that is a simple difference between frames of the same position blocks is taken, and this difference is compared with a threshold value. Here, if it is equal to or greater than the threshold value, a motion vector is output, and if it is less than the threshold value, a zero vector is output. Further, based on the result of the area determination performed based on the motion vector, as shown in FIG. 16, in the background area 1611, a predetermined first threshold th1 for the background area is used, and the moving object In the area 1612, a predetermined second threshold th2 for the moving object area is used.
JP-A-7-99657

  However, in such a conventional moving object detection device, the threshold value is not an appropriate value depending on the degree of noise of the moving image and the fluctuation of the noise, and the moving vector cannot be normally output. Therefore, there is a problem that a moving object cannot be detected stably.

  The present invention has been made to solve such a problem, and provides a moving object detection apparatus and a moving object detection method capable of stably detecting a moving object regardless of noise of a moving image. .

  The moving object detection device of the present invention includes a moving image capturing unit that captures a moving image for each frame, and divides the moving image into a plurality of blocks for each frame, and the degree of matching between each block between the frames and each block A degree-of-matching calculating unit that calculates a degree of matching, a moving-vector detecting unit that detects the moving vector by comparing the degree of matching for each block of the moving image, and the moving-vector of each moving-image block The first matching degree between the block corresponding to the start point and the block corresponding to the end point of the motion vector or the difference between the first matching degree and the second matching degree between the same-position blocks is compared with a threshold value. A moving vector detecting unit that switches whether to output the moving vector, and a moving object detecting unit that detects a moving object using the moving vector output via the moving vector switching unit. A noise measurement unit that measures noise of the moving image, a threshold value determination unit that determines the threshold according to the measurement result of the noise, and a motion related to the motion vector detected by the motion vector detection unit. A moving vector information recording unit that records vector information, wherein the noise measuring unit refers to the moving vector information, measures noise of the moving image based on a change state of the moving vector, and moves the moving image. The image is divided into a plurality of regions, the increase number of the motion vector is calculated for each region, and the threshold value determination unit determines the threshold value based on the plurality of increase numbers. ing.

  With this configuration, even if the noise level of the moving image increases, the output of an incorrect moving vector is suppressed. Therefore, a moving object can be detected stably regardless of the degree of noise or fluctuation of the moving image. can do.

  In addition, even if the noise level of the moving image rises, the noise level is easily and accurately measured by paying attention to the state of change of the moving vector by a simple process of calculating the increase number of the moving vector for each area. Therefore, the threshold value can be dynamically and accurately controlled without much load.

  In the moving object detection device of the present invention, the moving object detection unit detects a moving object by labeling a connected block in which a motion vector is detected in an adjacent block from among the blocks in which the motion vector is detected. It has a configuration.

  With this configuration, even if the noise level of the moving image increases, the erroneously output moving vector is ignored. Therefore, the moving object is more stable regardless of the noise level of the moving image and the fluctuation of the noise. Can be detected.

  The moving object detection device of the present invention has a configuration in which the moving object detection unit detects a moving object by attaching the same label to a block in which the direction of the motion vector is within a certain range.

  With this configuration, an erroneously output motion vector is ignored, and moving objects that move in different directions are distinguished from each other. Therefore, a moving object can be detected stably and in different directions. Even when moving objects are mixed, the moving object can be detected stably.

  In the moving object detection device of the present invention, the moving object detection unit attaches the same label to a block in which the direction of the motion vector is within a certain range and the size of the motion vector is within the certain range. It has a configuration to detect.

  With this configuration, an erroneously output motion vector is ignored, and moving objects that move in different directions and different speeds are distinguished from each other. Therefore, a moving object can be detected stably, Even when moving objects that move in different directions and at different speeds coexist, the moving objects can be detected stably.

  In the moving object detection device of the present invention, the degree-of-matching calculation unit calculates a sum of absolute differences of pixel values for each block between blocks, and the motion vector detection unit calculates each block of the moving image. A motion vector is detected by obtaining a minimum value of the sum of absolute differences, and the motion vector switching unit includes a first absolute difference between a block corresponding to a start point of the motion vector and a block corresponding to an end point of the motion vector. The difference between the value sum or the first difference absolute value sum and the second difference absolute value sum between the same position blocks is compared with the threshold value.

  With this configuration, even when the degree of matching between blocks between frames is detected as the sum of absolute differences of pixel values, moving objects can be detected stably regardless of the degree of noise in the moving image and the fluctuation in noise. can do.

  ADVANTAGE OF THE INVENTION According to this invention, the moving object detection apparatus and moving object display method which have the outstanding effect that a moving object can be detected stably irrespective of the noise of a moving image can be provided.

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

(First embodiment)
FIG. 1 is a block diagram showing a moving object detection apparatus 100 according to the first embodiment of this invention.

  In FIG. 1, a moving image capturing unit 101 captures a moving image output from an imaging device 10 such as a visible camera or a far infrared camera for each frame. The moving image recording unit 102 records a moving image for each frame. The difference absolute value sum calculation unit 103 (matching degree calculation unit) divides the moving image into a plurality of blocks for each frame, and calculates the difference absolute value sum of pixel values indicating the degree of matching for each block between blocks. Is calculated. The motion vector detection unit 104 detects a motion vector by obtaining the minimum value of the sum of absolute differences for each block of the motion image. The noise measuring unit 105 measures the noise of the moving image based on the information regarding the moving vector detected by the moving vector detecting unit 104. For example, as the information about the motion vector, the noise of the motion image is measured based on one of the detection position of the motion vector, the direction of the motion vector, the size of the motion vector, or a combination thereof. The threshold value storage unit 106 stores a plurality of threshold values. The threshold value determination unit 107 determines a threshold value for switching whether or not to output a motion vector according to a noise measurement result. For example, first, a predetermined reference value stored in the threshold value storage unit 106 is used as a threshold value. When noise increases, a threshold value switching candidate stored in the threshold value storage unit 106 is selected. The one corresponding to the noise state is selected and used. For each block of the moving image, the motion vector switching unit 108 calculates the sum of absolute differences between the block corresponding to the start point of the motion vector and the block corresponding to the end point of the motion vector (that is, the sum of absolute differences corresponding to the motion vector) The difference with the sum of absolute differences, which is a simple difference between frames of the same position blocks, is compared with a threshold value. The moving object detection unit 109 detects a moving object using the motion vector output via the motion vector switching unit 108. Specifically, a moving object is detected by labeling a connected block in which a motion vector is detected in an adjacent block from among blocks in which a motion vector is detected.

  The pixel value is a luminance value or other pixel level.

  The sum of absolute differences is obtained by comparing each pixel value in the target block with the current frame (hereinafter referred to as “current frame”) and the frame before the current frame (hereinafter referred to as “previous frame”). This is the sum of absolute values of differences from each pixel value. This sum of absolute differences is one of evaluation values (also referred to as “correlation values”) indicating the degree of image alignment between the target block and the comparison block (hereinafter referred to as “matching degree”). The degree of consistency is high.

  The motion vector indicates the amount of movement of the object. In the present embodiment, the comparison block is moved up, down, left, and right on the current frame, and the positional relationship between the comparison block and the target block that minimizes the sum of absolute differences from the target block of the previous frame is used as a motion vector. It is like that.

  Noise is generated in an imaging apparatus, a moving image transmission path, and the like, and is a change in pixel value that does not depend on a change in an imaging target.

  Next, the operation of the moving object detection device of the present embodiment will be described using the flowchart of FIG.

  First, a moving image is captured for each frame (S201). The captured moving image is recorded by the moving image recording unit 102 for each frame.

ここで、（ｂx，ｂy）はフレーム内の基準ブロックを原点とした各ブロックの位置であり、（ｍ，ｎ）は各ブロックの大きさであり、（ｘ，ｙ）は各ブロック内の各基準点を原点とした各画素の相対位置であり、（ｖx，ｖy）は各注目ブロックから各対比ブロックへのベクトルである。 Next, the moving image is divided into a plurality of blocks for each frame, and a sum of absolute differences of pixel values is calculated for each block of the previous frame taken out from the moving image recording unit 102 and each block of the captured current frame. (S202). Specifically, in the vicinity of each block of interest, the comparison block is moved on the current frame from −X to X in the horizontal direction and from −Y to Y in the vertical direction. A difference absolute value sum V (vx, vy) between each pixel value It-1 in each target block in the frame and each elementary value It in each comparison block in the current frame is calculated.

Here, (bx, by) is the position of each block with the reference block in the frame as the origin, (m, n) is the size of each block, and (x, y) is each block in each block. This is the relative position of each pixel with the reference point as the origin, and (vx, vy) is a vector from each block of interest to each comparison block.

次に、検出された動ベクトルに関する情報に基づいて動画像のノイズを測定する（Ｓ２０４）。ノイズの測定をどのように行うかについては、（１）動ベクトルの検出位置についての分布の状態に基づいて動画像のノイズを測定する、（２）動ベクトルの方向や大きさについての分布の状態に基づいて動画像のノイズを測定する、（３）動ベクトルの時間系列での変化の状態に基づいて動画像のノイズを測定する、などがある。これらのノイズ測定の具体的な例については、後述する第２の実施の形態から第４の実施の形態までにおいてそれぞれ詳細に説明する。 Next, for each block of the moving image, the minimum value of the sum of absolute differences is obtained and a moving vector is detected (S203). Specifically, as shown in Equation 2, for each block of interest, the minimum value Vmin of the sum of absolute differences with each contrast block is obtained, and a vector (vx, vy) to the contrast block that minimizes the sum of absolute differences is obtained. ) Is a motion vector.

Next, the noise of the moving image is measured based on the information about the detected moving vector (S204). Regarding how to measure the noise, (1) measure the noise of the moving image based on the distribution state of the detection position of the motion vector, and (2) the distribution of the direction and size of the motion vector. For example, the noise of the moving image is measured based on the state, and (3) the noise of the moving image is measured based on the state of the change in the time series of the moving vector. Specific examples of these noise measurements will be described in detail from the second embodiment to the fourth embodiment described later.

  Next, a threshold for switching whether or not to output a motion vector is determined according to the noise measurement result (S205). Specifically, a predetermined reference value stored in the threshold value storage unit 106 is used as an initial threshold value, and when the noise increases, the threshold value stored in the threshold value storage unit 106 is switched. The candidate corresponding to the noise state is selected and used.

  Next, for each block of the moving image, the difference V (0,0) between the difference absolute value sum Vmin corresponding to the motion vector and the difference absolute value sum V (0,0) corresponding to the inter-frame difference between the blocks at the same position. 0) −Vmin is compared with a threshold value to switch whether or not to output a motion vector (S206). FIG. 11 shows the original motion vector information 1110 obtained by calculating the minimum value of the difference absolute value sum V (vx, vy) between each block of interest in the previous frame and each contrast block in the current frame, and the threshold value. An example of moving vector information 1120 for detecting a moving object whose output is switched by comparison is shown. In FIG. 11, in the original vector information 1110, a plurality of motion vectors due to the movement of an object are detected in the moving object region 1112, and a plurality of motion vectors due to noise are detected in the background region 1111. In this embodiment, a threshold is determined according to the state of noise, and the difference between the sum of absolute differences corresponding to each motion vector and the sum of absolute differences between blocks at the same position is equal to or greater than the threshold. By switching to output a motion vector at a certain time and not to output a motion vector (output a zero vector) when it is less than the threshold value, the motion vector information 1120 for detecting the moving object is displayed in the moving object region 1122. A plurality of motion vectors due to the movement of the object are output, and a motion vector due to noise is hardly output in the background region 1121.

  Next, a moving object is detected from the output motion vector (S207). Specifically, a moving object is detected by labeling a connected block in which a motion vector is detected in an adjacent block from among blocks in which a motion vector is detected.

  The above steps are continued until the moving image is completed (S208).

  Note that each step shown in FIG. 2 may be executed by a computer.

  In addition, although an example has been described in which the step of measuring a noise (S204), the step of determining a threshold (S205), the step of switching a motion vector (S206), and the step of detecting a moving object (S207) are performed. A step of switching a vector (S206) and a step of detecting a moving object (S207) followed by a step of measuring noise (S204) and a step of determining a threshold (S205) are newly determined from the next frame. The motion vector may be switched with a threshold value.

  Further, in the step of switching motion vectors (S206), the difference V (0,0) −Vmin between the difference absolute value sum Vmin corresponding to the motion vector and the difference absolute value sum V (0,0) between the blocks at the same position. In the above example, the difference absolute value sum Vmin corresponding to the motion vector and the threshold value may be compared.

  Further, the example in which the threshold value is dynamically changed by extracting the threshold value candidate from the threshold value storage unit 106 has been described, but the difference between the threshold value reference value and the threshold value reference value is determined in advance. Alternatively, the threshold value may be changed dynamically. Moreover, the case where there are two types of threshold values is included.

  As described above, the moving object detection device of the present embodiment measures moving image noise and determines a threshold value for switching whether to output a moving vector according to the noise measurement result. Therefore, even if the moving image noise level increases, the output of the wrong moving vector will be suppressed. Therefore, it is possible to stably detect a moving object regardless of the degree of noise of the moving image and the fluctuation of the noise. Can do.

  In addition, since noise is measured based on information about a motion vector, noise that affects detection of the motion vector can be easily and accurately measured, and therefore the threshold value can be controlled dynamically and accurately. it can.

  In addition, since moving objects are detected by labeling connected blocks in which motion vectors are detected in adjacent blocks from among blocks in which motion vectors have been detected, the motion vectors of isolated blocks that are output incorrectly are ignored. Therefore, the moving object can be detected more stably.

(Second Embodiment)
FIG. 3 is a block diagram showing a moving object detection apparatus according to the second embodiment of the present invention.

  In FIG. 3, the same components as those of the moving object detection device of the first embodiment shown in FIG. 1 are denoted by the same reference numerals, and these components are described in the first embodiment. Therefore, detailed description is omitted.

  In the present embodiment, the position distribution measuring unit 305 measures the noise of the moving image based on the distribution state with respect to the detection position of the moving vector. Specifically, from the blocks in which the motion vector is detected, the isolated blocks in which the motion vector is not detected in the adjacent blocks are extracted, and the moving image is divided into a plurality of regions to calculate the number of isolated blocks for each region. To do. The threshold value determination unit 307 determines a threshold value based on the number of isolated blocks for each region.

  Next, the operation of the moving object detection device of the present embodiment will be described using the flowchart of FIG.

  In FIG. 4, from the step (S201) of capturing a moving image to the step (S203) of obtaining a minimum value of the sum of absolute differences for each block of the moving image and detecting a moving vector (S203) will be described in the first embodiment. It is the same process as each step.

  Next, an isolated block in which no motion vector is detected is extracted from adjacent blocks in which motion vectors are detected to neighboring blocks in the vicinity of 4 or 8, and a moving image is equally divided into a plurality of regions. The number of isolated blocks is calculated every time (S404). Next, a threshold for switching whether or not to output a motion vector is determined according to the number of isolated blocks for each region (S405).

  For example, based on the number of isolated blocks for each area, it is determined whether or not the isolated blocks are evenly distributed over the entire area. When the isolated blocks are evenly distributed, the threshold value is determined from the reference value from a predetermined reference value. Change to a larger threshold. FIG. 12 shows an example in which isolated blocks are evenly distributed over the entire area. In FIG. 12, the number of isolated blocks in each region is equal to or greater than a predetermined first number (for example, “3” or more), and the maximum number of isolated blocks in each region (“4” in FIG. 12). )) And the minimum number of isolated blocks in each region (“3” in FIG. 12) (“1” in FIG. 12) is a predetermined second number or less (eg, “4” or less). The number of isolated blocks in each region is equal to or greater than the first number, and the difference between the maximum number of isolated blocks and the minimum number of isolated blocks is equal to or less than the second number If so, change to a larger threshold.

  In FIG. 4, the step of switching whether or not to output a motion vector (S206), the step of detecting a moving object using the motion vector (S207), and the step of determining whether or not to end (S208) This is the same processing as each step described in the first embodiment.

  Each step shown in FIG. 4 may be executed by a computer.

  Further, when the noise is reduced and the number of isolated blocks is reduced, for example, when there is a region where no motion vector is detected, the threshold value may be returned to the reference value.

  As described above, the moving object detection device according to the present embodiment can easily and accurately reduce the noise level by paying attention to the distribution state of the detection position of the moving vector even if the noise level of the moving image increases. Therefore, the threshold value can be controlled dynamically and accurately. In addition, by calculating the number of isolated vectors for each region, the noise level can be easily and accurately measured, and therefore the threshold value can be controlled dynamically and accurately without imposing a heavy load. it can.

(Third embodiment)
FIG. 5 is a block diagram showing a moving object detection apparatus according to the third embodiment of the present invention.

  In FIG. 5, the same components as those of the moving object detection device of the first embodiment shown in FIG. 1 are denoted by the same reference numerals, and these components are described in the first embodiment. Therefore, detailed description is omitted.

  In the present embodiment, the frequency distribution measurement unit 505 generates information indicating the distribution of the direction of the motion vector and the size of the motion vector, and at least one of the direction of the motion vector and the distribution of the size of the motion vector. The uniformity of the distribution is determined. The threshold value determination unit 507 determines a threshold value based on the uniformity obtained by the frequency distribution measurement unit 505.

  Next, the operation of the moving object detection device of the present embodiment will be described using the flowchart of FIG.

  In FIG. 6, from the step (S201) of capturing a moving image to the step (S203) of obtaining a minimum value of the sum of absolute differences for each block of the moving image and detecting a moving vector (S203) will be described in the first embodiment. It is the same process as each step.

  Next, a histogram (frequency distribution information) for the direction of the motion vector is generated, the uniformity of the distribution is determined for the direction of the motion vector (S604), and a threshold value is determined based on this uniformity (S605). ).

  For example, a moving image is equally divided into a plurality of regions, and in all regions, a moving vector is detected in a block having a certain ratio or more with respect to the total number of blocks in each region, and the direction of the moving vector is determined by a histogram. When it is determined that each frequency is within a certain range, the threshold value is changed to a larger threshold value. FIG. 13 shows motion vector information 1310 that is a base when a motion vector due to noise is detected in the entire area when there is no moving object. In this way, when the motion vector due to noise has increased in the entire region compared to when there is no influence or little influence due to noise, the noise is large based on the histogram 1401 shown in FIG. The threshold value is changed. In the histogram 1401 of FIG. 14, each frequency in the direction of the motion vector is in the range from “5” to “7”, and it is determined that the frequency is almost uniformly distributed, and the threshold value is changed to a large threshold value. Is done.

  In FIG. 6, the step of switching whether or not to output a motion vector (S206), the step of detecting a moving object using the motion vector (S207), and the step of determining whether or not to end (S208) This is the same processing as each step described in the first embodiment.

  Each step shown in FIG. 6 may be executed by a computer.

  Further, when the noise becomes small and the direction and the magnitude vary, for example, when there is an area where no motion vector is detected, the threshold value may be returned to the reference value.

  As described above, the moving object detection apparatus according to the present embodiment can easily and accurately adjust the noise level by paying attention to the direction of the moving vector and the distribution of the magnitude even if the noise level of the moving image increases. Therefore, the threshold value can be controlled dynamically and accurately. In addition, the noise level can be easily and accurately measured by determining the uniformity of the distribution of the direction of the motion vector and the size of the motion vector, and therefore the threshold value can be set without much load. It can be controlled dynamically and accurately.

  Although the example in which the threshold value is determined based on the distribution state with respect to the direction of the motion vector has been described, the threshold value is determined based on the distribution state with respect to the direction of the motion vector and the size of the motion vector. You may do it. In addition, the noise of the moving image is measured based on the distribution state with respect to the magnitude of the moving vector, and the distribution state with respect to the direction of the moving vector is not measured.

(Fourth embodiment)
FIG. 7 is a block diagram showing a moving object detection apparatus according to the fourth embodiment of the present invention.

  In FIG. 7, the same components as those of the moving object detection device of the first embodiment shown in FIG. 1 are denoted by the same reference numerals, and these components are described in the first embodiment. Therefore, detailed description is omitted.

  In the present embodiment, the motion vector information recording unit 710 records motion vector information regarding the motion vector detected by the motion vector detection unit 104. The motion vector information includes at least a motion vector detection position. The motion vector information may include the direction and magnitude of the motion vector. The change measuring unit 705 measures the noise of the moving image based on the moving vector change state with reference to the moving vector information. Specifically, the moving image is divided into a plurality of regions, and the increase number of the moving vector is calculated for each region. The threshold value determination unit 707 determines a threshold value based on the number of motion vectors increased for each region.

  Next, the operation of the moving object detection device of the present embodiment will be described using the flowchart of FIG.

  In FIG. 8, from the step (S201) of capturing a moving image to the step (S203) of obtaining a minimum value of the sum of absolute differences for each block of the moving image and detecting a moving vector (S203) will be described in the first embodiment. It is the same process as each step.

  Next, the moving image is divided into a plurality of regions, and the increase number of the moving vector is calculated for each region (S804). Next, the threshold value is determined based on the number of motion vectors increased for each region (S805).

For example, when the number of motion vectors increases in all regions, the threshold value is changed from a predetermined reference value to a threshold value that is larger than this reference value.
Each step shown in FIG. 8 may be executed by a computer.

  Further, when the noise is reduced and the number of motion vectors is reduced, for example, when there is a region where no motion vectors are detected, the threshold value may be returned to the reference value.

  In FIG. 8, the step of switching whether or not to output a motion vector (S206), the step of detecting a moving object using the motion vector (S207), and the step of determining whether or not to end (S208) This is the same processing as each step described in the first embodiment.

  As described above, the moving object detection device according to the present embodiment can easily and accurately measure the noise level by paying attention to the change state of the motion vector even if the noise level of the moving image increases. Therefore, the threshold value can be controlled dynamically and accurately.

  In addition, the noise level can be easily and accurately measured by calculating the number of increase of the motion vector for each region. Therefore, the threshold value is dynamically and accurately controlled without much load. be able to.

(Fifth embodiment)
FIG. 9 is a block diagram showing a moving object detection apparatus according to the fifth embodiment of the present invention.

  In FIG. 9, the same components as those of the moving object detection device of the first embodiment shown in FIG. 1 are denoted by the same reference numerals, and these components are described in the first embodiment. Therefore, detailed description is omitted.

  In the present embodiment, the moving object detection unit 909 is a connected block in which a motion vector is detected in an adjacent block among blocks in which a motion vector is detected, and the direction of the motion vector is within a certain range. A moving object is detected by attaching the same label to a certain block.

  Next, the operation of the moving object detection device of the present embodiment will be described using the flowchart of FIG.

  In FIG. 10, the process from the step of capturing a moving image (S201) to the step of switching whether or not to output a moving vector (S206) is the same processing as each step described in the first embodiment.

  In the present embodiment, a connected block in which a motion vector is detected in an adjacent block in the vicinity of 4 or 8 out of blocks in which a motion vector is detected, and the direction of the motion vector is within a certain range. A moving object is detected by attaching the same label to a certain block (S1007). FIG. 15 shows an example of labeling. Based on the motion vector information 1510 indicating the detection position, direction, and size of the motion vector, each block is labeled, a first moving object region 1521 labeled “a”, and a label “b” The attached second moving object region 1522 is detected.

  The above steps are continued until the moving image is completed (S208).

  Each step shown in FIG. 10 may be executed by a computer.

  As described above, the moving object detection device according to the present embodiment ignores a motion vector that is output in error even if the noise level of the moving image increases, and therefore, the degree of noise of the moving image. It is possible to detect a moving object more stably regardless of fluctuations in noise and noise.

  In addition, since moving objects are detected by attaching the same label to blocks whose motion vector direction is within a certain range, moving objects moving in different directions are distinguished from each other, and moving objects moving in different directions are distinguished. Even when there is a mixture of moving objects, a moving object can be detected stably.

  In the fifth embodiment, an example has been described in which a moving object is detected by attaching the same label to blocks in which the direction of the motion vector is within a certain range. However, the present invention is not limited to this, and the motion vector is not limited thereto. In other words, the same label is attached to the blocks whose direction is within a certain range and the magnitude of the motion vector is within the certain range, and a moving object is detected.

  In this way, when a moving object is detected by attaching the same label to a block whose motion vector size is within a certain range, moving objects moving at different speeds are distinguished from each other, and thus different. Even when moving objects moving at a speed are mixed, the moving object can be detected stably.

  Further, in the first to fifth embodiments, the example in which the degree of matching between blocks is calculated as the sum of absolute differences of pixel values has been described, but the present invention is not limited to this. This includes calculating the degree of matching between blocks based on normalized cross-correlation values and other matching evaluation values.

The block diagram which shows the moving object detection apparatus of the 1st Embodiment of this invention The flowchart which shows the moving object detection process in the moving object detection apparatus of the 1st Embodiment of this invention. The block diagram which shows the moving object detection apparatus of the 2nd Embodiment of this invention The flowchart which shows the moving object detection process in the moving object detection apparatus of the 2nd Embodiment of this invention. The block diagram which shows the moving object detection apparatus of the 3rd Embodiment of this invention. The flowchart which shows the moving object detection process in the moving object detection apparatus of the 3rd Embodiment of this invention. The block diagram which shows the moving object detection apparatus of the 4th Embodiment of this invention. The flowchart which shows the moving object detection process in the moving object detection apparatus of the 4th Embodiment of this invention. The block diagram which shows the moving object detection apparatus of the 5th Embodiment of this invention The flowchart which shows the moving object detection process in the moving object detection apparatus of the 5th Embodiment of this invention. Explanatory drawing for demonstrating threshold determination in the moving object detection apparatus of the 1st Embodiment of this invention Explanatory drawing for demonstrating the state of distribution of the detection position of the motion vector in the moving object detection apparatus of the 2nd Embodiment of this invention. Explanatory drawing for demonstrating the state of the distribution of the direction of a motion vector in the moving object detection apparatus of the 3rd Embodiment of this invention. The figure which shows the frequency distribution information of the direction of a motion vector in the moving object detection apparatus of the 3rd Embodiment of this invention. Explanatory drawing for demonstrating the moving object detection in the moving object detection apparatus of the 1st Embodiment of this invention to the 5th Embodiment. Explanatory drawing for demonstrating the threshold value in the conventional moving object detection apparatus

Explanation of symbols

101 moving image capturing unit 102 moving image recording unit 103 difference absolute value sum calculating unit (matching degree calculating unit)
104 motion vector detection unit 105 noise measurement unit 106 threshold storage unit 107 threshold determination unit 108 motion vector switching unit 109 moving object detection unit

Claims (5)

  A moving image capturing unit that captures a moving image for each frame, and a degree of matching that divides the moving image into a plurality of blocks for each frame and calculates a degree of matching for each block between the blocks and each block A calculation unit; a motion vector detection unit that detects the motion vector by comparing the degree of matching for each block of the video; and a block corresponding to the start point of the motion vector and the motion vector for each block of the video Whether the motion vector is output by comparing the first matching degree with the block corresponding to the end point of the first or the difference between the first matching degree and the second matching degree between the same-position blocks with a threshold value A moving vector switching unit that switches between the moving vector detection unit, a moving object detection unit that detects a moving object using the moving vector output via the moving vector switching unit, and noise measurement of the moving image A noise measurement unit, a threshold value determination unit that determines the threshold value according to the measurement result of the noise, and motion vector information that records motion vector information related to the motion vector detected by the motion vector detection unit A recording unit, wherein the noise measurement unit refers to the moving vector information, measures noise of the moving image based on a change state of the moving vector, and divides the moving image into a plurality of regions. Then, the moving object detection device is characterized in that the increase number of the motion vector is calculated for each region, and the threshold value determination unit determines the threshold value based on the plurality of increase numbers. .   The moving object detection unit is configured to detect a moving object by attaching a label to a connected block in which a motion vector is detected in an adjacent block among blocks in which the motion vector is detected. Item 4. The moving object detection device according to Item 1.   The moving object detection device according to claim 2, wherein the moving object detection unit detects a moving object by attaching the same label to blocks in which the direction of the motion vector is within a certain range. .   The moving object detection unit detects a moving object by attaching the same label to blocks in which the direction of the motion vector is within a certain range and the size of the motion vector is within a certain range. The moving object detection device according to claim 2.   The matching degree calculation unit calculates a sum of absolute differences of pixel values for each block between blocks, and the motion vector detection unit calculates a minimum value of the sum of absolute differences for each block of the moving image. The motion vector switching unit obtains a motion vector, and the motion vector switching unit detects the first difference absolute value sum or the first difference absolute value between the block corresponding to the start point of the motion vector and the block corresponding to the end point of the motion vector. 5. The moving object detection according to claim 1, wherein a difference between the value sum and a second difference absolute value sum between the same position blocks is compared with the threshold value. apparatus.

Images