JP5590915B2 - Door open / close detection device and door open / close detection method - Google Patents

Description

  The present invention relates to a door opening / closing detection apparatus and a door opening / closing detection method for detecting presence / absence of door opening / closing by performing video processing using a moving image signal as input and outputting information on presence / absence of door opening / closing.

  In recent years, from the viewpoint of crime prevention, surveillance cameras have been widely used in anticipation of the effect of deterring crime in building management. In particular, surveillance cameras are often attached to elevators and automatic doors. A recording device is connected to such a surveillance camera, and it is common to record surveillance video. However, since there is a limit to the recording capacity required for recording, only significant surveillance video is carefully selected and recorded. Thus, there is a demand for effective use of recording capacity and longer recording time.

  As a monitoring device using an elevator monitoring camera, for example, there is one described in Patent Document 1, and this document shows a passenger detection method and a door open / close detection method using a camera image. Thus, by automatically detecting passengers and door opening / closing, it is possible to automatically determine the significance as a monitoring video to be recorded.

JP 2000-238968 A

  In the conventional door opening / closing detection method as described in Patent Document 1, when a motion sensor or the like is installed in an elevator door to detect the door opening / closing, it is necessary to repair the control panel having different specifications for each elevator model. In order to solve the problem of extra costs in various situations such as design, production, arrangement, inventory management, etc., door opening / closing detection using image processing is performed so that control panel repair is not required ing. In the image, in a very small fixed region near the upper part of the elevator door in the image, the opening / closing of the door is detected using the background subtraction method on the premise that no person passes. However, in general, an elevator monitoring camera is often installed obliquely downward in the back corner of an elevator car, and the vicinity of the upper part of the elevator itself is often not captured depending on the installation position of the monitoring camera. It was. Even if the image is captured, it is necessary to manually set the fixed area after installing the camera, and a tall person may cover the fixed area, which may cause false detection. It was.

  The present invention has been made to solve the above-described problems, and a door opening / closing detection device and door opening / closing detection capable of reliably detecting door opening / closing without being affected by the position of the door in the image. The purpose is to obtain a method.

The door opening / closing detection device according to the present invention is a door opening / closing detection device that detects whether doors are opened or closed by inputting each picture including a door image, and receives a plurality of pictures as inputs, and locally A motion vector calculation unit that calculates a simple motion and outputs it as a motion vector distribution, and receives a motion vector distribution as an input and recognizes a motion vector similar to the neighborhood as a motion vector group. The motion vector group centroid position estimation unit for estimating the centroid position, the motion vector distribution and the centroid position estimation result of the motion vector group are input, and the two motion vector groups exist in opposite directions and in pairs. when widely distributed motion vector to the line direction, which comprises an act there as determining the detection result determining unit of the casement door opening and closing the opening portion slides on both sides A.

The door opening / closing detection apparatus according to the present invention receives the motion vector distribution and the center of gravity position estimation result of the motion vector group as input, the two motion vector groups exist in opposite directions and in pairs, and the motion vector in the normal direction thereof When the door is widely distributed, it is determined that there is a double-open door opening / closing operation where the opening slides on both sides . Therefore, the door opening / closing is reliably detected without being affected by the position of the door in the image. be able to.

It is explanatory drawing which shows typically the elevator car to which the door opening / closing detection apparatus which concerns on Embodiment 1-5 of this invention is applied. It is explanatory drawing which shows the structure of the input / output of the door opening / closing detection apparatus which concerns on Embodiment 1 of this invention. It is an internal block diagram of the door opening / closing detection apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the example of an image | video inside the elevator car to which the door opening / closing detection apparatus which concerns on Embodiment 1-5 of this invention is applied. It is explanatory drawing which shows the example of motion vector distribution inside the elevator car to which the door opening / closing detection apparatus which concerns on Embodiment 1-4 of this invention is applied. It is explanatory drawing which shows the example which groups the motion vector distribution inside the elevator car with which the door opening / closing detection apparatus which concerns on Embodiment 1-4 of this invention is applied. It is a flowchart which shows operation | movement of the door opening / closing detection apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the evaluation value of the vector horizontal component in the vector evaluation part of the door opening / closing detection apparatus which concerns on Embodiment 1-4 of this invention. It is explanatory drawing which shows the evaluation value of the vector vertical component in the vector evaluation part of the door opening / closing detection apparatus which concerns on Embodiment 1-5 of this invention. It is explanatory drawing which shows the evaluation value in the recursive distance evaluation part of the door opening / closing detection apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the relationship between the block position in a neighborhood vector similarity evaluation part of the door opening / closing detection apparatus which concerns on Embodiment 1 of this invention, and a coefficient. It is explanatory drawing which shows the evaluation value in the distance evaluation part of the door opening / closing detection apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the structure of the input / output of the door opening / closing detection apparatus which concerns on Embodiment 2 of this invention. It is an internal block diagram of the door opening / closing detection apparatus which concerns on Embodiment 2 of this invention. It is explanatory drawing which shows the structural example of the coded picture which concerns on Embodiment 2 of this invention. It is explanatory drawing which shows the structure of the input / output of the door opening / closing detection apparatus which concerns on Embodiment 3, 4 of this invention. It is an internal block diagram of the door opening / closing detection apparatus which concerns on Embodiment 3 of this invention. It is an internal block diagram of the video recording data processing part in the door opening / closing detection apparatus which concerns on Embodiment 4 of this invention. It is explanatory drawing which shows the example of motion vector distribution inside the elevator car with which the door opening / closing detection apparatus which concerns on Embodiment 5 of this invention is applied. It is explanatory drawing which shows the example which groups the motion vector distribution inside the elevator car with which the door opening / closing detection apparatus which concerns on Embodiment 5 of this invention is applied. It is explanatory drawing which shows the evaluation value of the motion vector horizontal component in the vector evaluation part of the door opening / closing detection apparatus which concerns on Embodiment 5 of this invention.

Embodiment 1 FIG.
Embodiment 1 of the present invention will be described below. In the first embodiment, an elevator as shown in FIG. 1 will be described as an example of door opening / closing detection. FIG. 1 is an explanatory view schematically showing an elevator car 101, an elevator door 102, an elevator passenger 103, and a surveillance camera 104 inside the elevator car. The monitoring camera 104 is installed at the upper corner of the back side in the elevator car 101 so as to photograph the direction of the door 102. Further, a door opening / closing detection device 201 is connected in the form as shown in FIG. 2, and a moving image signal 204 that is a monitoring image inside the elevator car 101 is output to the door opening / closing detection device 201. .

  The door opening / closing detection apparatus 201 receives the moving image signal 204 as input, performs door opening / closing detection determination, and outputs a door opening / closing detection result signal / moving image signal / motion vector distribution set 205 to the recording apparatus 202. The recording apparatus 202 receives the door opening / closing detection result signal / moving image signal / motion vector distribution set 205 as an input, and uses international standard video encoding methods such as Motion-JPEG and MPEG based on the door opening / closing detection result signal. Compression encoding is performed and the recorded data 203 is stored.

  FIG. 3 shows an internal block of the door opening / closing detection apparatus 201 according to the first embodiment. The door opening / closing detection apparatus 201 includes a moving image signal processing unit 301, a motion vector group barycentric position estimation unit 302, a detection result determination unit 303, a moving image signal storage memory 304, and a motion vector distribution storage memory 305. First, the principle of door opening / closing detection in the door opening / closing detection apparatus 201 will be described.

  FIG. 4 is an example of an image obtained by photographing the inside of the elevator car 101 with the monitoring camera 104 of FIG. The door 102 is in a closed state, and a passenger 103 is on it. It is assumed that the door 102 is a double door. When the passenger 103 arrives at the target floor, the door 102 opens and the passenger 103 tries to get out of the elevator. Therefore, by performing video processing for calculating a local motion vector in the video, as shown in FIG. Simple motion vectors are detected. The left door motion vector 501, the right door motion vector 502, and the passenger motion vector 503 are detected as shown in FIG. 5 when viewed from the passenger 103.

  At this time, considering grouping of similar motion vectors, grouping is performed as shown in FIG. If the horizontal right direction on the drawing is positive and the horizontal left direction is negative, a negative vector group 601 that groups the left door motion vector 501, a plus vector group 602 that groups the right door motion vector 502, and passenger motion There are three types of vector groups, a vector group 603 obtained by grouping the vectors 503. At this time, since the door 102 is a double door, the motion vectors included in the negative vector group 601 grouping the left door motion vector 501 and the plus vector group 602 grouping the right door motion vector 502 are , They are opposite to each other. In addition, the motion vector distribution has a distribution shape that is long in the normal direction of the direction of the motion vector. That is, the principle of door opening / closing detection is to analyze the distribution of motion vectors and determine that the door opening / closing operation is performed when such a pair of motion vectors is detected.

  FIG. 7 is a flowchart showing the operation of the door opening / closing detection apparatus 201 according to the first embodiment. First, video processing is performed to calculate a local motion vector in the video, and a motion vector distribution is calculated (step ST701). Next, as a preparation for grouping motion vectors, the center-of-gravity position of each motion vector group is estimated based on the motion vector distribution calculated in step ST701 (step ST702). Based on the position of the center of gravity estimated in step ST702, an evaluation value is calculated as to whether or not the motion vector group has a reverse pair and a long distribution in the normal direction as shown in the vector groups 601 and 602 of FIG. (Step ST703). Furthermore, door open / closed detection is determined based on the evaluation value calculated in step ST703 (step ST704). In these ST701 to ST704, door opening / closing detection is performed. Here, step ST701 in the flowchart of FIG. 7 corresponds to the operation of the moving image signal processing unit 301 of FIG. 3, and similarly, step ST702 is detected by the motion vector group barycenter position estimating unit 302, and steps ST703 and ST704 are detected. This corresponds to the operation of the result determination unit 303. Hereinafter, each block of FIG. 3 will be described based on the flow of the flowchart shown in FIG.

The moving image signal processing unit 301 receives the moving image signal 321 and outputs the motion vector distribution 323. A motion vector calculation unit 317 is provided as an internal block. The moving image signal 321 is two moving image signals existing in the moving image signal storage memory 304, and a motion vector between the two images is calculated. The motion vector calculation method divides an image into rectangular blocks and calculates a motion vector in each block. As a method of calculating a motion vector in each block, there is a method of calculating (v _x , v _y ) that minimizes f in (Equation 1) below using a general block matching method, for example.

  The above (Formula 1) is an example in the case of a block size of 16 × 16 pixels. y represents a luminance value, and t and t + 1 represent time. The motion vector for each block calculated in this way is stored in the motion vector distribution storage memory 305 as a motion vector distribution.

The motion vector group center-of-gravity position estimation unit 302 receives the motion vector distribution 324 stored in the motion vector distribution storage memory 305 and the previous door opening / closing determination result 332 output from the detection result determination unit 303 as input, and the motion vector group The estimated center-of-gravity position 330 is output. The motion vector group center-of-gravity position estimation unit 302 includes four calculation blocks: a vector evaluation unit 311, a recursive distance evaluation unit 312, a neighborhood vector similarity evaluation unit 313, and an estimated center-of-gravity position calculation unit 314. The outputs of the three blocks of the vector evaluation unit 311, the recursive distance evaluation unit 312, and the neighborhood vector similarity evaluation unit 313 are all evaluation values, and based on the respective evaluation values, the estimated centroid position calculation unit 314 performs the centroid position. Is estimated. As the centroid position, two kinds of centroid positions of the minus vector group 601 and the plus vector group 602 in FIG. 6 are calculated. With respect to the evaluation values calculated by the evaluation units 311, 312, and 313, negative / plus evaluation values are calculated for each block, and each block belongs to which group of minus / plus or both. The evaluation value _{fminus_estimation1} / _{fplus_estimation1} for evaluating _{whether there} is _any is calculated.

Based on the evaluation value _{fminus_estimation1} (m, n) / _{fplus_estimation1} (m, n) of the block position (m, n), the estimated center-of-gravity position calculation unit 314 is expressed by (Expression 2) and (Expression 3) below. As described above, the two types of horizontal centroid positions _{fx_minus} / _{fx_plus} of the minus vector group 601 and the plus vector group 602 are estimated and output as estimated centroid positions 330. In the following equation, M is the number of horizontal blocks in the image, and N is the number of vertical blocks in the image.

  Since the first embodiment has been described with an example of an elevator, the center-of-gravity position in the horizontal direction is calculated. In general, elevator surveillance cameras are arranged so that the area of the person's back height plus alpha is best seen from the bottom of the door, and the relationship between the camera distortion and elevator car shape and the orientation of the surveillance camera In the case where the vector groups 601 and 602 shown in FIG. 6 are found, the direction of the line where the elevator doors are in contact with each other is inevitably reflected in the vertical direction even though the reflection may be slightly oblique. In addition, it is sufficient to estimate the position of the center of gravity in the horizontal direction of each group.

Regarding the calculation of the evaluation value _{fminus_estimation1} (m, n) / _{fplus_estimation1} (m, n) of the block position (m, n), the vector evaluation unit 311, the recursive distance evaluation unit 312 and the neighborhood vector similarity evaluation unit 313 Calculation is performed by taking the product of evaluation values 327, 328, and 329, which are output values. Hereinafter, the evaluation units 311, 312, and 313 will be described.
The vector evaluation unit 311 receives the motion vector distribution 324 as input and evaluates the direction and magnitude of the motion vector at the block position (m, n). If the horizontal direction of the motion vector is negative to the left, the possibility of belonging to the negative vector group is high. Conversely, if the horizontal direction of the motion vector is positive to the right, the possibility of belonging to the positive vector group is high. Therefore, the evaluation value is calculated as, for example, (Equation 4) and (Equation 5) below. (Expression 4) and (Expression 5) are shown in FIG. (Equation 4) (Equation 5) and FIG. 8 are examples in which the motion vector (v _x , v _y ) calculation range in the motion vector calculation unit 317 is ± 16.

Also, the vertical component of the motion vector is reflected in a substantially vertical direction with respect to the direction of the line where the elevator doors 102 are in contact with each other. Therefore, it can be evaluated that the vertical component of the motion vector is nearly zero, the higher the possibility that the motion vector belongs to the vector groups 601 and 602. Therefore, the evaluation value for the vertical component of the motion vector is calculated as, for example, (Equation 6). . (Equation 6) is shown in FIG. FIG. 9 is also an example in the case where the motion vector (v _x , v _y ) calculation range in the motion vector calculation unit 317 is ± 16.

  The evaluation value 327 that is the output of the vector evaluation unit 311 is calculated as the product of minus / plus e1 and e2, respectively.

Recursive distance evaluating unit 312, the horizontal block position of center of gravity position _{f x_minus} / _{f x_plus} when an input motion vector distribution 324, the single previous door opening and closing detection is determined to Yes (m, n) The value inversely proportional to the distance is used as the evaluation value. For example in the form in _{f minus_e3 (m, n)} as shown in FIG. 10 _{/ f} evaluated as _{plus_e3} (m, n), is output as evaluation value 328. In order to save the above-described center-of-gravity position _{fx_minus} / _{fx_plus} when it is determined that the previous door opening / closing detection is _present , the estimated center-of-gravity position 330 and the door opening / closing determination result 332 are input. .

The neighborhood vector similarity evaluation unit 313 receives the motion vector distribution 324 as input, and evaluates the overall similarity of motion vectors in the vicinity of the block position (m, n). A coefficient as shown in FIG. 11 is used for weighted addition based on the block position (m, n) and the motion vector similarity between the blocks of the target block. The motion vector similarity between blocks is set to a larger value as they are more similar. Although there is a negative value in FIG. 11, it is designed so that a small evaluation value is obtained if they are similar. Since the elevator door opening / closing vector groups 601 and 602 are vertically long and the one that bulges horizontally may not be the elevator door opening / closing, if there is a motion vector similar to the horizontal position, something other than the doors such as passengers The door opening / closing evaluation value is designed to be a small evaluation value. A value obtained by weighting and adding the motion vector similarity between the blocks with the coefficient in FIG. 11 is output as an evaluation value 329.
The above is the description of the motion vector group center-of-gravity position estimation unit 302.

Next, the detection result determination unit 303 will be described.
The detection result determination unit 303 includes the motion vector distribution 325 stored in the motion vector distribution storage memory 305, the estimated centroid position 330 output from the motion vector group centroid position estimation unit 302, the vector evaluation value 327, and the neighborhood vector similarity. The degree evaluation value 329 is input, and the door open / close determination result 332 is output. The detection result determination unit 303 includes two calculation blocks, a distance evaluation unit 315 and a door open / close determination unit 316. Each calculation block will be described below.

The distance evaluation unit 315 receives the motion vector distribution 325 and the estimated centroid position 330 as an input, and _{uses a} distance evaluation value 331 as a value that is inversely proportional to the horizontal distance of the block position (m, n) from the estimated centroid position _{fx_minus} / _{fx_plus.} Output as. For example in the form in _{f minus_e3n (m, n)} as shown in FIG. _{12 / f plus_e3n (m, n} ) and evaluated as to output as the respective distances evaluation value 331. As shown in FIG. 12, only a block in the immediate vicinity of the centroid position estimated by the motion vector group centroid position estimation unit 302 is evaluated highly. This is an evaluation method based on the fact that the vector groups 601 and 602 shown in FIG. 6 are thin and vertically long.

The door opening / closing determination unit 316 receives the vector evaluation value 327, the distance evaluation value 331, and the neighborhood vector similarity evaluation value 329 as inputs, and outputs the door opening / closing determination result 332 as an output. The various evaluation values to be input are negative / plus evaluation values at the block position (m, n). In each block, the product of three kinds of evaluation values for each minus / plus is calculated, and the evaluation value _{fminus_estimation2} (m, n) / _{fplus_estimation2} (m, n) of the block position (m, n) is calculated. Performs addition of negative / positive for each per total block _{f minus_estimation2 (m, n) /} f plus_estimation2 (m, n), the negative / positive respectively for the total value, that threshold evaluation with a fixed value A, If both plus / minus exceeds the threshold, the door is opened / closed, otherwise the door opening / closing is not detected, and the door opening / closing determination result 332 is output.

The presence / absence of door opening / closing detection is determined for each image. On the other hand, in each block (m, n), a negative / plus evaluation value _{fminus_estimation2} (m, n) / _{fplus_estimation2} (m, n) is calculated. By performing the threshold evaluation is also a fixed value B with respect to the value of _{f Minus_estimation2} in each block _{(m, n) / f plus_estimation2} (m, n), the block belongs to the negative groups, positive groups It can be classified into three types: belonging or not belonging to either.

  At this time, not only the door opening / closing operation but also the door opening / closing operation for the case where blocks with a large motion vector are gathered in blocks that do not belong to either minus or plus even though it is determined that there is a door opening / closing operation. It is determined that other animals (passengers, etc.) are also moving.

  Also, if it is determined that there is a door opening / closing operation and all the blocks that do not belong to either are small motion vectors, it is considered that there is no other moving object, and there was only a door opening / closing operation in an elevator without moving passengers Judge that.

  When there is no door opening / closing operation, it is judged that other moving objects (passengers, etc.) other than the door opening / closing are moving when blocks with large motion vectors are gathered regardless of the negative / positive evaluation value. .

  By configuring the door opening / closing detection apparatus 201 shown in the first embodiment as described above, it detects if the vector groups 601 and 602 exist while ignoring the motion vectors such as the vector group 603 shown in FIG. It is possible to detect the presence / absence of a door opening / closing operation. Accordingly, it is possible to detect the door opening / closing operation even if the vicinity of the upper part of the elevator itself, which has been a problem, does not appear depending on the installation position of the monitoring camera. In other words, the surveillance camera 104 is installed so that a portion of the elevator door 102 through which a person passes is always photographed, and therefore, most of the door 102 is generally photographed. In the first embodiment, since the door opening / closing detection can be performed if most of the door 102 is reflected, the door opening / closing detection can be performed regardless of the position of the door 102 in the image. Manual setting is also unnecessary.

  Further, in the first embodiment, when it is detected that there is a door opening / closing operation, it is possible to check the presence or absence of a moving passenger by examining the motion vector distribution in a block that does not belong to either minus / plus. Yes, it can also be used to determine the significance of whether or not the video should be recorded. Therefore, the door opening / closing determination result 332 as an output of the door opening / closing detection apparatus 201 not only for the presence / absence of the door opening / closing operation but also for the presence / absence of other moving objects. In addition, it is possible to exclude the video of only the door opening / closing operation in the elevator without moving passengers from the recording target, and as a result, there is an advantage that the recording time can be extended.

As described above, according to the door opening / closing detection apparatus of the first embodiment, a door opening / closing detection apparatus that detects whether doors are opened or closed by inputting each picture including a door image, and inputs a plurality of pictures. A motion vector calculation unit 317 that calculates a local motion between pictures and outputs the motion vector distribution; and a motion vector distribution as an input, and when there are motion vectors similar to the vicinity, these are expressed as a motion vector group. The motion vector group centroid position estimation unit 302 that estimates the centroid position of the motion vector group and the motion vector distribution and the centroid position estimation result of the motion vector group are input, and the two motion vector groups are in opposite directions. in and with existing in pairs, when the motion vector in the normal direction thereof is widely distributed, the operation there and determine the casement door opening and closing the opening portion slides on both sides Since comprises a detection result determining unit 303, without being affected by the position of the door in the video, it is possible to detect reliably the door opening and closing.

Further, according to the door opening / closing detection method of the first embodiment, there is provided a door opening / closing detection method using a door opening / closing detection device that receives each picture including a door image as input and detects the presence / absence of door opening / closing. The motion vector calculation step that calculates the local motion between each picture and outputs it as a motion vector distribution, and the motion vector distribution as an input when there is a motion vector similar to the neighborhood. As a motion vector group, the motion vector group centroid position estimation step for estimating the centroid position of the motion vector group and the motion vector distribution and the centroid position estimation result of the motion vector group are input. opposite to and with existing in pairs, when the motion vector in the normal direction thereof is widely distributed, casement door opening and closing the opening portion slides on both sides Since it comprises a detection result determining step of determining that there operation, without being affected by the position of the door in the video, it is possible to realize a door opening and closing detecting apparatus capable of detecting a reliable door opening and closing.

  Further, according to the door opening / closing detection apparatus of the first embodiment, the detection result determination unit 303 determines that the animal is not an animal when there is a motion vector group other than the motion vector group that determines that the door opening / closing operation has been performed. Since it was determined that the body existed, it is possible to confirm the presence or absence of moving objects such as moving passengers.

Embodiment 2. FIG.
The second embodiment of the present invention will be described below. Also in the second embodiment, an elevator as shown in FIG. 1 will be described as an example of door opening / closing detection.

  The monitoring camera 104 according to the second embodiment is connected to a recording device 1301 in a form as shown in FIG. 13 and outputs a moving image signal 204 that is a monitoring image inside the elevator car to the recording device 1301.

  The recording device 1301 receives the moving image signal 204 and outputs a motion vector distribution 1304 used in the process of compression encoding using an international standard video encoding method such as MPEG to the door opening / closing detection device 1302. At this time, motion vectors of international standard video coding systems such as MPEG are generally vectors that are directed to the past starting from the future, and there are cases in which the direction is the opposite of the actual direction of motion. In such a case, it is known that each motion vector in the motion vector distribution 1304 is a motion vector having a direction opposite to the motion vector for MPEG, and a motion vector having the same direction as the actual motion is used. It will be treated as a motion vector distribution. The door opening / closing detection device 1302 receives the motion vector distribution 1304, detects the presence / absence of a door opening / closing operation and the presence / absence of another animal body based on the motion vector distribution, and detects only the presence / absence of the door opening / closing operation. In addition, the presence / absence of other moving objects is included in the door open / close determination result 332 and output.

  The recording device 1301 determines whether or not to store the recording data 1303 based on the door opening / closing determination result 332.

  FIG. 14 shows an internal block diagram of door opening / closing detection device 1302 according to the second embodiment. Compared with the door open / close detection device 201 according to the first embodiment, the moving image signal processing unit 301 and the moving image signal storage memory 304 are not provided, and the other blocks are the same in configuration and operation. The illustration and description of the inside of the group gravity center position estimation unit 302 and the detection result determination unit 303 are omitted.

  In general, when compression coding is performed by an international standard video coding system such as MPEG, resynchronization is possible by periodically inserting I pictures in the form shown in FIG. This lump is generally called GOP (Group Of Picture). Since door open / close detection is performed for each image, for example, as shown in FIG. 15, when a picture in the middle of a GOP is determined to be recorded, it cannot be played back even if only that P picture is saved, so the GOP to which the P picture belongs The picture can be played back by saving.

  By configuring the door opening / closing detection device 1302 shown in the second embodiment as described above, the door opening / closing detection can be performed in an existing surveillance video recording system including a monitoring camera and a recording device, as compared with the first embodiment. There is an advantage that the device 1302 can be easily incorporated. The reason is that, in the case of the first embodiment, since the door opening / closing detection device 201 is incorporated between the monitoring camera 104 and the recording device 202, two interface developments are required, whereas in the second embodiment, the recording device is provided at the subsequent stage. This is because the door opening / closing detection device 1302 only needs to be arranged, and interface development is completed in one place. Further, the door opening / closing detection device 1302 of the second embodiment can simplify the device configuration as compared with the door opening / closing detection device 201 of the first embodiment, which is also a great advantage. This is because the motion vector required when the recording apparatus 1301 performs compression encoding using an international standard video encoding system such as MPEG can be reused and supplied to the door open / close detection apparatus 1302.

As described above, according to the door open / close detection device of the second embodiment, each picture including the door video and the motion vector distribution of the entire picture of the motion vector indicating the local motion in the picture are input. A door opening / closing detection device 1302 for detecting presence / absence of door opening / closing, which receives a motion vector distribution as an input, recognizes a motion vector similar to the vicinity, and recognizes these as a motion vector group, and the center of gravity of the motion vector group The motion vector group centroid position estimation unit 302 for estimating the position, the motion vector distribution and the centroid position estimation result of the motion vector group are input, and the two motion vector groups exist in opposite directions and in pairs. If the motion vectors are widely distributed in the linear direction, it comprises an operation there and determines detection result determining unit 303 of the casement door opening and closing the opening portion slides on both sides Since, in addition to the effect of the first embodiment, easy to incorporate a door opening and closing detecting device 1302 to the existing monitoring video recording system, also, it is possible to simplify the structure than the first embodiment.

Embodiment 3 FIG.
The third embodiment of the present invention will be described below. In the third embodiment, as an example of door opening / closing detection, in the elevator as shown in FIG. 1, the surveillance video inside the elevator car is compressed and encoded by an international standard video encoding system such as MPEG. The recorded data will be described as an example.

  The apparatus configuration of the third embodiment is shown in FIG. Recording data 1602 is input / output to / from the door opening / closing detection device 1601. That is, the door opening / closing detection device 1601 receives the recording data 1603 and outputs multiplexed data 1604 obtained by adding metadata of the door opening / closing operation detection signal and the other moving object detection signal to the recorded data in units of pictures.

  FIG. 17 shows an internal block diagram of door opening / closing detection device 1601 in the third embodiment. In comparison with the door open / close detection device 1302 of the second embodiment, a recorded data processing unit 1701, a detection result metadata generation multiplexing unit 1702, and a recorded data storage memory 1703 are added. Since the other blocks are the same in configuration and operation as those in the first and second embodiments, description thereof will be omitted.

  The recording data 1603 input to the door opening / closing detection device 1601 is stored in the recording data storage memory 1703. The recording data stored in the recording data storage memory 1703 is output as recording data 1721 input to the recording data processing unit 1701 and recording data 1723 input to the detection result metadata generation multiplexing unit 1702, respectively.

  The recording data processing unit 1701 receives the recording data 1721 and outputs the motion vector distribution 1722. A motion vector extraction unit 1711 is provided as an internal block, and a motion vector is extracted from the recorded data in this block. At this time, motion vectors of international standard video coding systems such as MPEG are generally vectors that go to the past starting from the future, and are therefore opposite to the actual direction of motion. In such a case, for each motion vector in the motion vector distribution 1722, a motion vector having the opposite direction to the extracted motion vector for MPEG is used, and the motion vector has the same direction as the actual motion. A motion vector distribution 1722 is output as the motion vector distribution of the motion vectors.

  The detection result metadata generation multiplexing unit 1702 receives the recording data 1723 and the door opening / closing determination result 332 including the presence / absence of a door opening / closing operation and the presence / absence of another moving object, and inputs each picture in the recording data 1723 and the presence / absence of a door opening / closing operation. Then, the generation of metadata for associating the presence or absence of other animals and the multiplexing of the recorded data and the metadata are performed, and the multiplexed data 1604 of the recorded data and the metadata is output.

  By configuring the door opening / closing detection device 1601 shown in the third embodiment as described above, unlike the first and second embodiments, the door opening / closing detection and other animals are detected for the recorded data. Body presence / absence detection can be performed. There is an advantage that a significant scene search can be easily performed by taking correspondence between the door opening / closing detection result and the detection result of the presence or absence of other animals and the recorded data.

As described above, according to the door opening / closing detection apparatus of the third embodiment, each picture of the digital video signal including the door video is divided into predetermined unit areas, and the bit stream is compressed and encoded in that unit. Is a door open / close detection device 1601 that detects whether door is open or closed, and receives a bit stream, extracts a motion vector used for motion compensation prediction from the bit stream, and outputs it as a motion vector distribution 1701, a motion vector group centroid position estimation unit 302 that receives a motion vector distribution as input and recognizes motion vectors similar to those in the vicinity as a motion vector group and estimates the centroid position of the motion vector group; The motion vector distribution and the motion vector group centroid position estimation result are input, and the two motion vector groups are in opposite directions. One well as present in pairs, when the motion vector in the normal direction thereof is widely distributed, and operations there and determines the detection result of the casement door opening and closing judging unit 303 with an opening to slide on both sides, the bit stream if the decision result And a detection result metadata generation multiplexing unit 1702 that outputs a bit stream obtained by multiplexing the metadata in which the pictures included in the bit stream are associated with the determination results. In addition to the above effect, there is an effect that door opening / closing detection and other animal body presence / absence detection can be performed on already recorded video data.

Embodiment 4 FIG.
The fourth embodiment of the present invention will be described below. In the fourth embodiment, as an example of door opening / closing detection, in the elevator as shown in FIG. 1, the monitoring video inside the elevator car is compressed and encoded by an international standard video encoding method such as MPEG. The recorded data will be described as an example.

  The configuration of the door opening / closing detection device of the fourth embodiment on the drawing is the same as that of the third embodiment shown in FIG. The difference between the fourth embodiment and the third embodiment is the configuration of the recorded data processing unit 1701 in the internal block of the door opening / closing detection device 1601 shown in FIG. FIG. 18 shows an internal block diagram of the recorded data processing unit 1801 of the fourth embodiment. Since other than this is the same as the door opening / closing detection device 1601 of the third embodiment, the recording data processing unit 1801 will be described below.

  The recorded data processing unit 1801 receives the recorded data 1721 as an input and outputs a motion vector distribution 1722. This input / output is the same as in the third embodiment. The recorded data processing unit 1801 includes a picture type detection unit 1811, a motion vector extraction unit 1711, and a motion vector calculation unit 1812 as internal blocks. Compared with the recorded data processing unit 1701 in the third embodiment, the recorded data type A detection unit 1811 and a motion vector calculation unit 1812 are added.

  The picture type detection unit 1811 detects the picture type included in the recorded data, and in the case of an I picture, since a motion vector is not included, the motion vector calculation unit 1812 calculates a motion vector. On the other hand, when the picture type is P picture or B picture, the motion vector is included in the motion vector extraction unit 1711 because the motion vector is included. At this time, motion vectors of international standard video coding systems such as MPEG are generally vectors that go to the past starting from the future, and are therefore opposite to the actual direction of motion. In such a case, for each motion vector in the motion vector distribution 1722, a motion vector having the opposite direction to the extracted motion vector for MPEG is used, and the motion vector has the same direction as the actual motion. A motion vector distribution 1722 is output as the motion vector distribution of the motion vectors.

  By configuring the door open / close detection device shown in the fourth embodiment as described above, it is possible to detect whether the door is open / closed or to detect other animals with respect to an I picture which is recorded data having no motion vector as compared with the third embodiment. There is an advantage that body presence / absence detection can be performed.

  As described above, according to the door open / close detection device of the fourth embodiment, the recorded data processing unit 1801 extracts a picture type from a bit stream, and if a picture type performs motion compensation prediction, a motion vector is extracted from the bit stream. In the case of a picture type that is extracted and does not perform motion compensation prediction, the local motion between each picture is calculated and output as a motion vector distribution. In addition to the effects of the third embodiment, the I picture In addition, there is an effect that it is possible to detect the presence / absence of door opening / closing and the presence / absence of other animal bodies.

Embodiment 5 FIG.
The fifth embodiment of the present invention will be described below. As an example of door opening / closing detection, an elevator as shown in FIG. 1 will be described as an example. Unlike Embodiments 1 to 4, the door opening / closing detection will be described in the case of a single door instead of a double door. I do.

  In the case of the double door 102 in the first to fourth embodiments, the left door motion vector 501 is grouped in the form shown in FIGS. 5 and 6 where the horizontal rightward direction is positive and the horizontal leftward direction is negative. When the minus vector group 601 and the plus vector group 602 obtained by grouping the motion vector 502 of the right door are used, the presence / absence of the door opening / closing operation is determined by detecting the vector groups 601 and 602.

  FIG. 19 shows the state of motion vector distribution in the elevator car in the case of single opening in the fifth embodiment. The single door 1902 is an example in which the door is opened from the left to the right as viewed from the passenger 103, and the motion vector 1901 of the left door is “doubled” compared to the motion vector 502 of the right door. The point of “the same direction” in the “length” is different from the case of the double door 102.

  Also in the fifth embodiment, when the distribution state of motion vectors is analyzed and two motion vector groups such as vector groups 2001 and 602 are detected as a set as shown in FIG. The determination is the principle of door opening / closing detection in the fifth embodiment.

  At this time, the difference from the first to fourth embodiments is that the motion vector of the left door is “double the length” and “same direction” compared to the motion vector of the right door. Therefore, the internal operation of the vector evaluation unit 311 in FIG. 3 and each expression in the first embodiment and the subscript “minus” in the expressions in FIGS. 8, 9, 10 and 12 are “plus1” and “plus” is “plus2”. By changing to, it is possible to detect whether the door is open or closed. This will be described below.

  FIG. 8 shows an example of the evaluation value of the horizontal component of the motion vector in the vector evaluation unit 311. This is based on the fact that if the horizontal component of the motion vector is negative, the possibility of belonging to the negative vector group is high, and if the horizontal component of the motion vector is positive, the possibility of belonging to the positive vector group is high. Therefore, the vector evaluation unit 311 according to the fifth embodiment is configured as shown in FIG. 21, for example, and the evaluation value is obtained when the left door is “double the length” of the right door and “same direction”. It is configured to be higher. Also, based on such a vector evaluation value 327, the detection result determination unit 303 has two motion vector groups in the same direction and one motion vector group exists twice as large as the other motion vector group. And when it distributes widely in the normal line direction, it determines with the operation | movement of the door opening and closing of a single swing having been performed.

The door opening / closing detection apparatus shown in the fifth embodiment is configured as described above, and in the first to fourth embodiments, the double-open door is described as an example. There is an advantage that the presence or absence of other animal bodies can be detected.
In the fifth embodiment, the case where there are two doors 102 has been described. However, the present invention can be similarly applied even when there are three or more doors 102. That is, the detection result determination unit 303 has R (R is an integer of 2 or more) motion vector groups in the same direction, and the largest motion vector group exists in the size of R times the smallest motion vector group. And when it distributes widely in the normal line direction, it can apply similarly by determining with the operation | movement of the door opening and closing of single doors having been performed.

As described above, according to the door opening / closing detection apparatus of the fifth embodiment, the detection result determination unit 303 has the largest motion vector group in which R (R is an integer of 2 or more) motion vector groups are in the same direction. Is present in the size of R times the smallest motion vector group and is widely distributed in the normal direction, it is determined that the door opening / closing operation in which the opening slides to one side has been performed. Therefore, it is possible to cope with the detection of single door opening / closing, and the same effect as in the first embodiment can be realized.

  Moreover, although this Embodiment 1-5 demonstrated the operation | movement of the door opening / closing detection apparatus and the door opening / closing detection method using the elevator as an example of door opening / closing detection, it moves like the door of an elevator (for example, automatic Needless to say, similar effects can be obtained for doors and the like.

  Moreover, although Embodiment 5 demonstrated the case where it applied to Embodiment 1, you may combine with Embodiments other than Embodiment 1, such as combining with any of Embodiment 2-4.

  101 car, 102 units, 103 passengers, 104 surveillance cameras, 201, 1302, 1601 door opening / closing detection device, 204 moving image signal, 205 door opening / closing detection result signal / moving image signal / motion vector distribution set, 301 moving image signal processing 302, motion vector group centroid position estimation unit, 303 detection result determination unit, 317, 1812 motion vector calculation unit, 311 vector evaluation unit, 312 recursive distance evaluation unit, 313 neighborhood vector similarity evaluation unit, 314 estimated centroid position calculation , 315 Distance evaluation unit, 316 Door open / close determination unit, 501 Left door motion vector, 502 Right door motion vector, 503 Passenger motion vector, 601 Minus vector group, 602 Plus vector group, 603 vector group, 1304 Motion vector distribution, 16 3 Recording Data, 1701, 1801 Recording Data Processing Unit, 1702 Detection Result Metadata Generation Multiplexing Unit, 1711 Motion Vector Extraction Unit, 1811 Picture Type Detection Unit, 1901 Left Door Motion Vector, 1902 Single Door, 2001 Vector Group .

Claims (7)

A door opening / closing detection device that detects whether a door is opened or closed by inputting each picture including a door image,
A motion vector calculation unit that receives a plurality of pictures, calculates a local motion between the pictures, and outputs a motion vector distribution;
A motion vector group centroid position estimator that takes the motion vector distribution as an input and recognizes them as a motion vector group when there are motion vectors similar to the vicinity; and estimates a centroid position of the motion vector group;
When the motion vector distribution and the centroid position estimation result of the motion vector group are input, the two motion vector groups exist in opposite directions and in pairs, and the motion vector is widely distributed in the normal direction A door opening / closing detection apparatus comprising: a detection result determination unit that determines that there is a double door opening / closing operation in which the opening slides on both sides . A door opening / closing detection device that detects whether a door is opened or closed by inputting each picture including a door video and a motion vector distribution of the entire picture of a motion vector indicating local motion in the picture,
A motion vector group centroid position estimator that takes the motion vector distribution as an input and recognizes them as a motion vector group when there are motion vectors similar to the vicinity; and estimates a centroid position of the motion vector group;
When the motion vector distribution and the centroid position estimation result of the motion vector group are input, the two motion vector groups exist in opposite directions and in pairs, and the motion vector is widely distributed in the normal direction A door opening / closing detection apparatus comprising: a detection result determination unit that determines that there is a double door opening / closing operation in which the opening slides on both sides . A door opening / closing detection device for detecting whether doors are opened or closed by dividing each picture of a digital video signal including a door video into predetermined unit areas and inputting a bit stream compressed and encoded in that unit. ,
A recording data processing unit that takes the bitstream as an input, extracts a motion vector used for motion compensation prediction from the bitstream, and outputs it as a motion vector distribution;
A motion vector group centroid position estimator that takes the motion vector distribution as an input and recognizes them as a motion vector group when there are motion vectors similar to the vicinity; and estimates a centroid position of the motion vector group;
When the motion vector distribution and the centroid position estimation result of the motion vector group are input, the two motion vector groups exist in opposite directions and in pairs, and the motion vector is widely distributed in the normal direction A detection result determination unit that determines that there is a double door opening / closing operation in which the opening slides on both sides ;
A detection result metadata generation multiplexing unit that receives the bit stream and the determination result as input, and outputs a bit stream obtained by multiplexing metadata included in the bit stream and associated with the determination result; A door open / close detection device comprising:   The recorded data processing unit extracts a picture type from the bit stream, extracts a motion vector from the bit stream if the picture type performs motion compensation prediction, and localizes between each picture if the picture type does not perform motion compensation prediction. 4. The door opening / closing detection apparatus according to claim 3, wherein the movement is calculated and output as a motion vector distribution. The detection result determination unit includes R (R is an integer of 2 or more) motion vector groups in the same direction, the largest motion vector group exists in the size of R times the smallest motion vector group, and 5. The device according to claim 1, wherein when the door is widely distributed in the normal direction, it is determined that the door opening / closing operation in which the opening slides to one side is performed. Door open / close detection device.   The detection result determination unit determines that the moving object exists when there is a motion vector group other than the motion vector group that is a basis for determining that the door opening / closing operation has been performed. The door opening / closing detection apparatus of any one of Claim 5. A door opening / closing detection method using a door opening / closing detection device that detects whether a door is opened or closed by inputting each picture including a door image.
A motion vector calculating step that receives a plurality of pictures, calculates a local motion between the pictures, and outputs a motion vector distribution;
A motion vector group centroid position estimating step for estimating the centroid position of the motion vector group and recognizing them as a motion vector group when there is a motion vector similar to the neighborhood when the motion vector distribution is input;
When the motion vector distribution and the centroid position estimation result of the motion vector group are input, the two motion vector groups exist in opposite directions and in pairs, and the motion vector is widely distributed in the normal direction A door opening / closing detection method comprising: a detection result determination step for determining that there is a double door opening / closing operation in which the opening slides on both sides .

Images