JP4624396B2 - Situation judging device, situation judging method, situation judging program, abnormality judging device, abnormality judging method and abnormality judging program - Google Patents

Description

  The present invention relates to a situation determination apparatus, a situation determination method, a situation determination program, and an abnormality determination that analyze a captured image to detect a degree of congestion or movement of a person in a public space where a plurality of persons move, such as a station / airport. The present invention relates to an apparatus, an abnormality determination method, and an abnormality determination program.

  In recent years, with the increasing demand for safety and security, surveillance cameras are being installed in public spaces such as stations and airports and important facilities. In the past, surveillance cameras were constantly monitoring surveillance cameras, but from the viewpoint of increasing the number of surveillance cameras and preventing oversight due to fatigue of the surveillance staff, efforts to reduce labor and increase the efficiency of surveillance by recognizing images Has been done.

  Patent Documents 1 and 2 have been proposed as techniques for counting people at a monitoring location. Patent Document 1 extracts a foreground based on a background difference and counts the number of people crossing a monitoring area orthogonal to a person's passage. By preparing a plurality of monitoring areas and using the dispersion of the count values between the monitoring areas, a robust count against disturbance is realized.

  Patent Document 2 uses a camera in which the optical axis is installed vertically downward above a passage, extracts a motion vector at a boundary line provided on an image for counting, and integrates a vertical component of the motion vector with respect to the boundary line. Thus, the number of people passing through the passage is counted.

  On the other hand, Patent Document 3 has been proposed as a technique for extracting a feature corresponding to the number of people from an image and calculating the degree of congestion instead of counting each person. Patent Document 3 describes the number of changes within a certain time for each pixel or each local region on the premise that “if the number of passersby is large, the number of changes between images taken at different times increases”. The degree of congestion is obtained based on the number of changes.

Japanese Patent Laid-Open No. 2002-074371 Japanese Patent Laid-Open No. 2005-135339 JP 2004-102380 A

  However, the above-described prior art has the following problems. That is, in Patent Document 1, since the background difference is used, it is difficult to apply to a place where the illumination change is large, and it is difficult to count each person at the time of congestion. Similarly, in Patent Document 2, it is difficult to count each person at the time of congestion. Since Patent Document 3 is based on the premise that a person is moving, it is difficult to calculate the degree of congestion in a situation where a moving person and a stationary person are mixed.

  The present invention has been made in view of such circumstances, and a situation determination apparatus, a situation determination method, a situation determination program, an abnormality determination apparatus, an abnormality that can easily determine the situation of a monitoring place and the degree of congestion An object is to provide a determination method and an abnormality determination program.

To achieve the above object, the apparatus for judging the condition of the present invention is an apparatus for judging the condition determining the movement status and / or degree of congestion of people by analyzing the moving image captured, the moving image a plurality of divided into local regions, calculates a local image change ratio calculating means for calculating a histogram representing a frequency distribution of the local image change ratio calculated in the local image change ratio calculating means for local image change ratio of each of the local region And a situation determination unit that compares the calculated histogram with a reference histogram obtained in advance to determine a person's movement situation and / or a person's congestion level.

  According to this configuration, the local change rate in the captured image is obtained for a plurality of local regions, the histogram for the plurality of regions of the local change rate is obtained, and this histogram is analyzed. Spatial characteristics (for example, motion bias, etc.) with a corresponding rate of change can be detected, and it is possible to determine the general situation of the monitoring place and the degree of congestion.

Further, the apparatus for judging the condition of the present invention, local image change ratio calculating means for each of the local regions, obtains the representative motion vector obtained from the captured images at a first time interval of the moving image, the The size of the representative motion vector is compared with a predetermined threshold value to determine the presence or absence of movement of the local region, and the number of second time intervals longer than the first time interval divided by the first time interval Then, the local image change rate is calculated as a ratio to the number of the first time intervals determined as having the motion .

  According to this configuration, the local change rate in the captured image is obtained for a plurality of local regions, the histogram for the plurality of regions of the local change rate is obtained, and this histogram is analyzed. Spatial characteristics (for example, motion bias, etc.) with a corresponding rate of change can be detected, and it is possible to determine the general situation of the monitoring place and the degree of congestion.

In the above configuration, the situation determination unit includes a reference histogram storage unit that holds a previously obtained reference histogram, and a histogram comparison unit that compares the calculated histogram with the reference histogram .

Further, in the above configuration, the movement situation is determined as a movement of a large number of people, a movement of a small number of people, or a movement path bias, and the number of areas determined to have movement in the local area is large. When the local image change rate is medium or large, it is determined that the movement is a large number of people, and the number of regions determined to be moving is large in the local region, and the local image is determined in the calculated histogram. When the rate of change is medium or large, it is determined that the movement is a small number of people, and the number of regions determined to be moving is small in the local region, and the local image change rate is medium or low in the calculated histogram. When the large frequency is large, it is determined that the movement path is biased .

  According to this configuration, since it is possible to detect that the moving route is biased with a simple process, the presence of the queue and the congestion level in a place where there is a train queue such as a station, for example. Can be estimated. Of course, the situation and the degree of congestion can also be estimated in a situation where there is no bias in the travel route. In addition, when the deviation of the movement route is detected in a passage that normally does not generate a matrix, it is possible to estimate that free movement of a person is hindered by some kind of obstacle.

  The abnormality determination device of the present invention is an abnormality determination device that determines an abnormal situation by analyzing a moving image or a plurality of still images captured by an imaging unit installed on a platform of a station. Train arrival information is obtained using a device, train arrival detection means for detecting arrival of a train to the platform, situation determination result by the situation determination device and train arrival information by the train arrival detection means And an abnormality determination unit that determines that an abnormality occurs when it is determined that a person's movement route is biased as a result of the situation determination after a predetermined time later.

  According to this configuration, it is possible to determine an abnormal congestion state that is different from normal using the situation type or the degree of congestion obtained by the situation determination device.

  Moreover, the said structure WHEREIN: When the said abnormality determination means determined with abnormality, the notification means which notifies to the contact address defined beforehand was further provided.

  According to this configuration, it is possible to provide auxiliary information to the monitor and to promptly notify a predetermined contact address.

Also, status determination method of the present invention is a situation determining method of determining a movement status and / or degree of congestion of people by analyzing the moving image captured divides the moving image into a plurality of local regions , calculates the local image change ratio calculating step of calculating a local image change ratio of each of the local region, a histogram representing a frequency distribution of the local image change ratio of the local each area calculated by the local image change ratio calculating step And a situation determination step of comparing the calculated histogram with a reference histogram obtained in advance to determine a person's movement situation and / or a person's congestion level.

  According to this method, a local change rate in a captured image is obtained for a plurality of local regions, a histogram for the plurality of regions of the local change rate is obtained, and this histogram is analyzed. Spatial characteristics (for example, motion bias, etc.) with a corresponding rate of change can be detected, and it is possible to determine the general situation of the monitoring place and the degree of congestion.

Further, in the situation determination method, the movement situation is determined as a movement of a large number of people, a movement of a small number of people, or a movement path bias, and the number of areas determined to have movement in the local area is large, and the calculation is performed. When the local image change rate in the histogram is large or large, it is determined that there is a large number of people moving, and the number of regions determined to have movement in the local region is large. When the local image change rate is medium or large, when the frequency is small, it is determined that the movement is a large number of people, the number of regions determined to be moving is small in the local region, and the local image change rate is determined in the calculated histogram. When the medium or large frequency is large, it is determined that the movement is large .

  According to this method, it is possible to detect that the moving route is biased by a simple process. For example, in a place where there is a train queue such as a station, the presence of the queue and its congestion level Can be estimated. Of course, the situation and the degree of congestion can also be estimated in a situation where there is no bias in the travel route. In addition, when the deviation of the movement route is detected in a passage that normally does not generate a matrix, it is possible to estimate that free movement of a person is hindered by some kind of obstacle.

  The abnormality determination method of the present invention is an abnormality determination method for determining an abnormal situation by analyzing a moving image or a plurality of still images captured by an imaging unit installed on a platform of a station. Using the situation determination step for executing the method, the train arrival detection step for detecting the arrival of the train to the platform, the situation determination result by the situation determination device and the train arrival information by the train arrival detection means, And a failure determination step of determining an abnormality when it is determined that the movement route of the person is biased as a result of the situation determination after a lapse of a predetermined time after the train arrival information is obtained.

  According to this method, it is possible to determine an unusually crowded state different from normal using the situation type or the degree of congestion obtained by the situation judging method.

Moreover, the situation determining program of the present invention, a computer, a situation determining program for determining the movement status and / or degree of congestion of people by analyzing the moving image captured, the moving image a plurality of divided into local regions, the frequency distribution of the local and the local image change ratio calculating step of calculating a local image change ratio for each region, the local image change ratio of the local each area calculated by the local image change ratio calculating step And a situation determination step of comparing the calculated histogram with a previously obtained reference histogram to determine a person's movement situation and / or a person's congestion level.

  According to this program, the local change rate in a captured image is obtained for a plurality of local regions, a histogram for the plurality of regions of the local change rate is obtained, and this histogram is analyzed. Spatial characteristics (for example, motion bias, etc.) with a corresponding rate of change can be detected, and it is possible to determine the general situation of the monitoring place and the degree of congestion.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments for carrying out the invention will be described in detail with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a schematic configuration of a situation determination apparatus according to Embodiment 1 of the present invention. In FIG. 1, the situation determination apparatus according to the present embodiment includes an image input unit 100, an image storage unit 110, a local image change detection unit 120, a local image change information storage unit 130, and a local image change rate calculation unit 140. A local image change rate accumulation unit 150, a local image change rate histogram calculation unit 160, and a situation determination unit 170.

  FIG. 2 is a diagram illustrating an example in which the situation determination apparatus of FIG. 1 is installed on a platform of a railway station. The railway station includes a platform PH (hereinafter abbreviated as a platform) PH, a side wall WL of the platform, a staircase ST that is an entrance to the platform, and a track RL. The camera CM is installed so that the longitudinal direction of the home PH is the optical axis direction and is connected to the situation determination device SD in order to capture an image of a person existing in the home PH. FIG. 3 shows an example of a captured image of the camera CM. The angle of view, installation position, and optical axis direction of the camera CM are determined so that the captured image includes the home PH, the entrance / exit portion of the staircase ST, and the side wall WL. The situation determination device SD corresponds to the situation determination device in FIG.

  Next, the operation of the situation determination apparatus of the present embodiment will be described using the flowchart of FIG. First, the image input step S100 is executed by the image input unit 100. In step S100, one frame of an image captured by the camera CM is captured in a format that allows digital image processing, and then stored in the image storage unit 110. When the camera CM is an analog camera, the image is AD-converted, subjected to compression processing such as encoding as necessary, and then stored in the image storage unit 110. When the camera CM is a digital camera, an image is input through a digital line and stored in the image storage unit 110. In the present embodiment, it is assumed that a 10 fps digital moving image is input and a frame image at the current time is sequentially stored. When the camera CM is an analog camera, the image storage unit 110 stores an analog image such as a VTR, and AD conversion may be performed immediately before output to the local image change detection unit 120.

Next, the local image change detection step S <b> 110 is executed by the local image change detection unit 120. Here, as shown in FIG. 5, 0.1 (seconds) equal to the imaging period is set as the first time interval TS ₁ , and the image at the current time t _k among the frame images accumulated in the image accumulation unit 110. And two frame images of the image t _{k-1 one} time before TS are extracted, and a change in the local region is detected. As a local region dividing method, it is determined in advance according to the installation of the camera CM as shown in FIG. Here, the total number of local regions is N _R. Further, here, in order to correct the influence of perspective projection, the size of the local region is set large at a point close to the camera (lower screen) and small at a point far from the camera (upper screen). A case where a motion vector is used as a change detection method will be described below.

(1) Calculation of motion vector for each pixel from two frame images For calculation of a motion vector, a gradient method such as the Lucus-Kanade method in Non-Patent Document 1 or a block matching method can be used. Here, it is desirable to use only a highly reliable motion vector in the subsequent processing by the Good Features to Track method in Non-Patent Document 2 or an evaluation value (SAD, SSD, etc.) at the time of matching. FIG. 7 shows an example of a highly reliable motion vector among the motion vectors obtained from two frame images for the entire screen.

Non-Patent Document 1: BDLucas and T. Kanade. “An iterative image registration technique with an application to stereo vision”. IJCAI, 1981.
Non-Patent Document 2: Jianbo Shi, Carlo Tomasi, "Good Features to Track", IEEE Conference on Computer Vision and Pattern Recognition, pp.593-600, 1994 (CVPR'94)

(2) Integration of motion vectors for each local area Hereinafter, a subscript k is used for motion vectors extracted from two frame images of an image at the current time t _k and an image t _{k-1 one} time before TS. , I is the subscript i for the motion vector included in the i th (1 ≦ i ≦ N _R ) local region, and j is the n th _{k of the} NV _{k, i} motion vectors included in the i th local region. The motion vector (u _{k, i, j} , v _{k, i, j} ) is expressed by using the subscript j for the motion vector of 1 ≦ j ≦ NV _{k, i} ).

Then, by calculating the average value using (Equation 1) and (Equation 2), a representative motion vector (mu _{k, i} , mv _{k, i} ) representing the local region i is calculated. The length of the motion vector on the image is affected by the perspective projection, and is calculated to be smaller than the actual human speed at the top of the screen. Therefore, as shown in (Equation 3) and (Equation 4), the motion vector after correcting the influence may be averaged to calculate the representative motion vector. Here, w _kij is the weighting coefficient for correcting the size of the j-th motion vectors in the i-th local region at time t _k. The start point of the motion vector is set so as to increase as it is at the top of the screen.

(3) Determine the presence or absence of motion by performing threshold processing on the representative motion vector. The size of the representative motion vector (mu _{k, i} , mv _{k, i} ) representing the local region i is determined in advance. The presence or absence of motion is determined using the determined threshold value. If it is equal to or greater than the threshold, it is determined that there is motion, and if it is less than the threshold, it is determined that there is no motion. At this point, local change information M _{k, i} expressing the presence / absence of a change (motion) for each local region i in binary is obtained. If there is a motion, M _{k, i} = 1, otherwise M _{k, i} = 0. FIG. 8 shows the local region local change information represented by an image. In FIG. 8, the local region where the change exists is shaded for expression. The local change information M _k is expressed as binary vector information obtained by combining the number of local areas with the presence or absence of change (binary information) for each local area as shown in (Formula 5), as shown in FIG. Accumulated in the accumulation unit 130. In FIG. 5, frame image I _k and local image change information M _k are accumulated at time t _k , and frame image I _k−1 and local image change information M _k−1 are stored at time t _k−1 . Indicates that it has been accumulated.

Returning to FIG. 4, next, the local image change rate calculation step S <b> 120 is executed by the local image change rate calculation unit 140. Here, set as the interval TS ₂ second time 10 (seconds). For each local region, the rate of change (movement) occurring during the past TS ₂ (seconds) is obtained. Change because the information is being sought at intervals TS ₁ the first time, determine the rate of change with TS 2 _/ TS ₁ pieces of change information. When TS ₁ = 0.1 (seconds) and TS ₂ = 10 (seconds), 100 pieces of change information are used. 9, in the i-th local region, the local change information _{M k} at time _{t k,} a _i, among values of the local change information _{M k-TS2 / TS1 + 1} , i at time _{t k-TS2 / TS1 + 1} is 1 Is counted (that is, there is a change), and the value is defined as C _{k, i} . Then, the local change rate RT _{k, i} is calculated by dividing by the total number TS ₂ / TS ₁ . The possible value of the local change rate RT _{k, i} is [0,1]. The local change rate RT _k is stored in the local image rate storage unit 150 as shown in FIG. 10 as vector information obtained by combining the change rates for each local region by the number of local regions as shown in (Expression 6). In FIG. 10, the local image change rate RT _k is accumulated at time t _k and the local image change rate RT _k−1 is accumulated at time t _k−1 .

Next, the local image change rate histogram calculation step S 130 is executed by the local image change rate histogram calculation unit 160. Here calculates the _{N R} local change ratio _{RT k, i} histograms of the current time _{t k.} Since the local change rate RT _{k, i} can take a value [0, 1], if the width of the histogram class is BW, the class number is 1 / BW. Here, the class width BW = 0.1 and the number of classes is 10. An example of the calculated local image change rate histogram is shown in FIG. The horizontal axis represents the local image change rate, and the vertical axis represents the frequency (number of regions). The total frequency (integrated value of the histogram) is the total number of regions N _R.

  Here, let us consider the movement situation of people on the platform of the station shown in FIG. Consider the following three patterns of movement.

(1) Free movement of a large number of people For example, there is a situation (FIG. 12) in which many passengers get off after the arrival of a train and move toward the staircase ST. Since the movement is a large number of people, people move along many movement routes such as AR1 to AR3.

(2) Free movement of a small number of people For example, there is a situation (FIG. 13) in which a passenger who gets off the train moves through the stairs ST and then a passenger on the next train moves through an empty platform. To do. Although the movement is a small number of people, the movement paths of people vary because the home is vacant. As a result, as in the case of movement by a large number of people, people move along many movement routes such as AR1 to AR3.

(3) Uneven travel route For example, in a situation where there is a queue waiting for a train, a new passenger who has moved to the platform through the stairs ST passes through a vacant place on the platform and is on the near side of the home screen. There is a situation (FIG. 14) of moving toward. Since there is a person WP waiting for the train, the moving path of the moving person MP is limited to the vicinity of the arrow AR1.

  FIG. 15 shows a summary of the trend of the number of regions where changes occur (having a change rate of a certain value or more) and the magnitude of the local image change rate for these three movement status patterns. Further, typical examples of local image change rate histograms as shown in FIG. 11 relating to these three movement patterns are shown in FIG. 16 for (1) movement of a large number of people, (2) movement of a small number of people, and (3) deviation of a movement path, respectively. It shows in FIG. 17, FIG.

  First, regarding the situation where a large number of people move, a supplementary explanation will be given using the image of FIG. 19 in which local region division is performed in FIG. When a large number of people are moving, changes (movements) occur in many areas in the image. Since the density of people is high, when attention is paid to a certain local region, moving people pass one after another, and the local image change rate increases in many local regions. Therefore, the tendency is as shown in the first column of FIG. Similarly, in the local image change rate histogram of FIG. 16, the number of local regions where the local image change rate is medium to large increases.

  A supplementary explanation of the situation in which a small number of people move freely will be given using the image of FIG. Even when a small number of people are moving as in the case of a large number of people moving, since people pass through various routes in a vacant home, changes (movements) occur in many areas in the image. However, since the density of people is small, when attention is paid to a certain local region, the frequency of movement of a person passing is reduced, and the local image change rate is reduced. Therefore, the tendency is as shown in the second column of FIG. Similarly, in the local image change rate histogram of FIG. 17, the number of local regions where the local image change rate is medium increases, but the number of local regions where the local image change rate is large decreases.

  With respect to the situation where the movement route is biased, supplementary explanation will be given using the image of FIG. 21 in which local region division is performed in FIG. When there is a queue for waiting for a train, the movement path of the person is limited to the vicinity of the arrow AR1. Since the person waiting for the train does not move except sometimes moving on the spot, the area where the change (movement) occurs is almost limited to the area where the moving person MP exists, and the number of areas where the change occurs is small. If the flow rate of people who flow into the home PH through the stairs ST is about the same as when moving with a small number of people (Fig. 12), the number of people passing through the local area on the movement route is small because the movement route is limited. It becomes larger than when moving. Therefore, the tendency is as shown in the third column of FIG. Similarly, in the local image change rate histogram of FIG. 18, the number of local regions where the local image change rate is medium and large is smaller than in the case of movement of a large number of people in FIG. The difference is that there are at least a number of local regions having a large local image change rate.

  As described above, the local image change rate histogram is calculated by the local image change rate histogram calculation step S130 (local image change rate histogram calculation unit 160).

FIG. 22 shows local image change rate histograms extracted from three scenes (multi-person movement / small-group movement / movement path bias) in an actual moving image. Each, even though corresponding to a histogram of the local rate of change _{RT k'(certain} time _{t k'and} at a certain time _{t k',} as described in FIG. 10, the local rate of change of one unit time is past TS2 / TS1 amino Calculated from local change information). FIG. 22 shows the results when the total number of local regions N _R = 162 (pieces), TS ₁ = 0.1 (seconds), TS ₂ = 10 (seconds), and the width of the histogram class is BW = 0.1. It is. It can be seen from FIG. 22 that the tendency is the same as that shown in FIGS.

  Next, the situation determination step S140 is executed by the situation determination unit 170. It has already been explained that the shapes of the local image change rate histograms are as shown in FIGS. 16 to 18 in the three situations of large number of people movement / small number of people movement / movement route bias. In the present embodiment, the local image change rate histogram in these three situations is obtained in advance and stored as a reference histogram, and compared with the local image change rate histogram in the situation to be determined. Determine if there is.

  FIG. 23 shows the internal configuration of the situation determination unit 170 in the present embodiment. The situation determination unit 170 includes a reference histogram storage unit 200 and a histogram comparison unit 210. The reference histogram storage unit 200 has at least one local image change rate histogram determined in advance as a reference for each situation, any of the situation (multi-person movement / small-person movement / movement path bias). ) And stored. The histogram comparison unit 210 compares the local image change rate histogram calculated by the local image change rate histogram calculation unit 160 and the reference histogram stored in the reference histogram storage unit 200 with the situation to be determined. Then, it is determined to which reference histogram the local image change rate histogram in the situation to be judged is most similar, and the situation associated with the most similar histogram is output as the result of situation judgment. As a method of calculating the similarity between histograms, the histogram intersection of (Equation 7), the Bhattaccharyya coefficient of (Equation 8), the normalized correlation of (Equation 9), and the like can be used.

  Returning to FIG. 4, finally, step S <b> 150 is executed by a control unit (not shown in FIG. 1). If a processing end instruction is input by the operator of the apparatus through the input unit (not shown in FIG. 1), the process returns to the image input step S100, and the frame image at the next time is processed. If no process end instruction has been input, the process ends.

In the present embodiment, in the local image change detection step S110 executed by the local image change detection unit 120, two frames having the time interval TS ₁ among the frame images stored in the image storage unit 110. In order to detect a change in a local region between images, a motion vector is used, but another method may be used. However, since the movement direction and speed can be known by using the motion vector, more detailed situation determination and congestion degree determination are possible. Inter-frame differences can be used to simply detect the presence or absence of changes. As shown in FIG. 24, when the inter-frame difference between the two frames captured at time t _k−1 and time t _k is calculated, an image D _k (x, y) is obtained, but whether or not there is a change for each local region may be determined depending on whether or not the average value d _k, i of the luminance difference in the local region i is equal to or greater than a predetermined threshold value. In FIG. 24, S _i is a set of pixels constituting the region i, and NP _i is the number of elements.

As shown in FIG. 25, after a difference image D _k (x, y) in which the value of each pixel is composed of a luminance difference is obtained, binarization indicating presence / absence of change is performed, and binarization difference The image BD _k (x, y) may be obtained, and whether or not there is a change for each local region may be determined based on the number of pixels bd1 _{k, i} having a change in the local region. When the areas of the local regions are different from each other, whether or not there is a change for each local region may be determined by bd2 _{k, i obtained} by dividing (normalizing) the number of pixels with change by the area of the local region.

Further, since the luminance value-based inter-frame difference is affected by illumination changes, the difference between the images obtained by performing edge extraction and binarization on the input image as shown in FIG. 26 (XOR processing for each pixel) ) To obtain an edge inter-frame difference image ED _k (x, y), and the presence / absence of a change in each local region may be determined based on the number of pixels ed1 _{k, i} having a change in the local region. When the area of each local region is different, whether or not there is a change for each local region may be determined by ed2 _{k, i obtained} by dividing (normalizing) the number of pixels with change by the area of the local region.

  As described above, according to the situation determination apparatus of the present embodiment, the time change rate of the luminance value in the local region of the captured image and the histogram of the time change rate for the plurality of local regions are analyzed. It is possible to determine the movement status of In particular, since it is possible to detect that the movement route is biased by a simple process, it is possible to estimate the presence of the queue at a place where a train queue such as a station exists. In addition, when the deviation of the movement route is detected in a passage that normally does not generate a matrix, it is possible to estimate that free movement of a person is hindered by some kind of obstacle.

(Embodiment 2)
The configuration of the situation determination apparatus according to the second embodiment of the present invention is represented in FIG. 1 similarly to the situation determination apparatus according to the first embodiment described above. Therefore, the description of FIG. 1 is omitted. However, while the internal configuration of the situation determination unit 170 is represented in FIG. 23 in the first embodiment, it is represented in FIG. 27 in the present embodiment. That is, the situation determination unit 170 includes a feature extraction unit 300, an identification reference storage unit 310, and an identification unit 320. The feature extraction unit 300 extracts feature amounts for situation determination or congestion degree determination from the histogram calculated by the local image change rate histogram calculation unit 160. In the identification reference storage unit 310, the relationship between the feature amount extracted by the feature extraction unit 300 and the situation type or the congestion degree index for situation determination or congestion degree determination is obtained and stored in advance. The identification unit 320 identifies the situation type or the degree of congestion using the feature amount extracted by the feature extraction unit 300 and the identification standard stored in the identification standard storage unit 310.

  The flowchart of the situation determination method according to the present embodiment is represented in FIG. 4 as in the first embodiment. Here, the processing up to the local image change rate histogram calculation step S130 and the processing in step 150 is the same as in the first embodiment, and thus the description thereof is omitted.

  Hereinafter, the operation in the situation determination step S140 will be described with reference to the block diagram of FIG. First, the feature extraction unit 300 extracts a feature amount for situation determination or congestion determination from the histogram calculated by the local image change rate histogram calculation unit 160. Here, a two-dimensional feature amount is extracted from the local image change rate histogram. The feature amount extraction method is shown below.

(1) In the local image change rate histogram, the number of regions having a medium change rate (threshold value TH ₁ ) or more is counted, and this number of regions is set as RN ₁ .
(2) In the local image change rate histogram, the number of regions having a large change rate (threshold value TH ₂ or more; TH ₂ > TH ₁ ) is counted, and this number of regions is defined as RN ₂ .
(3) When RN ₁ ≠ 0, (f ₁ , f ₂ ) = (RN ₁ / N _R , RN ₂ / RN ₁ ), and when RN ₁ = 0, (f ₁ , f ₂ ) = (RN ₁ / N _R , 0) is a feature amount.

Assuming TH ₁ = 0.4 and TH ₂ = 0.7, the values of RN ₁ and RN ₂ in the three situations (multi-person movement / small-person movement / movement route bias) in FIG. The values of f ₁ and f ₂ are as shown in the table of FIG.

In the identification reference storage unit 310, the relationship between the feature amount extracted by the feature extraction unit 300 and the situation type or the congestion degree index for situation determination or congestion degree determination is obtained and stored in advance. FIG. 30 shows a distribution of two-dimensional feature values (f ₁ , f ₂ ) extracted in advance from three scenes of moving images. Each point in the distribution is extracted from a local change rate histogram at one time. That is, feature extraction is performed from each of the three scenes at a plurality of times. As information stored in the identification reference storage unit 310, a distribution as shown in FIG. 30 is used.

  The identification unit 320 identifies the situation type or the degree of congestion using the feature amount extracted by the feature extraction unit 300 and the identification standard stored in the identification standard storage unit 310.

A method for identifying the situation type will be described below. Assume that the information of FIG. 30 is stored in the identification reference storage unit 310. It is assumed that each histogram calculated by the local image change rate histogram calculation unit 160 is one in FIG. 22, and the situation type is unknown. At this time, the feature quantity extracted by the feature extraction unit 300 is a set of values in any row in the table shown in FIG. The feature type extracted by the feature extraction unit 300 searches for the closest feature point among the feature amounts stored in the identification reference storage unit 310 (nearest neighbor method), and the situation type associated with the feature point Is output, it is possible to determine the situation of the scene corresponding to the feature amount extracted by the feature extraction unit 300. FIG. 31 shows the relationship between the values shown in the table of FIG. 29 and the distribution of the two-dimensional feature values (f ₁ , f ₂ ) shown in FIG. It can be seen that the situation judgment can be correctly performed by using the nearest neighbor method.

  Although the nearest neighbor method has been described above, any other method may be used as long as it is a method for identifying unknown samples using a large number of training samples that are supervised. For example, Support Vector Machine (SVM), discriminant analysis (linear discriminant analysis, secondary discriminant analysis), neural network, or the like can be used. Since each method is explained in various documents, the explanation is omitted. For example, see Non-Patent Document 3 for SVM and Non-Patent Document 4 for discriminant analysis.

Non-Patent Document 3: "Introduction to Support Vector Machine", Nero Christianiani, John Shawe-Taylor, Takeshi Ohkita, Kyoritsu Shuppan, 2005/03
Non-Patent Document 4: "Multivariate Analysis Method Modern Human Statistics 2", Haruo Yanai et al., Asakura Shoten, 1979/01

Next, a method for determining the degree of congestion will be described. First, the change tendency of the indicators RN ₁ and RN ₂ and the feature quantities f ₁ and f ₂ when the congestion progresses is divided into a case of free movement (the movement route is not limited) and a case where the movement route is biased. Give an explanation.

<In the case of free movement (the movement route is not limited)>
Region having a local image change rate of medium or higher as the state shifts from [1] to [2] in FIG. 32 (in the figure, the numerical values are surrounded by circles, and the other numerical values 3 to 6 are the same). The number RN ₁ increases, but the number of regions RN ₂ having a large local image change rate does not increase so much. This is because when the number of people increases in a situation where the movement route is not restricted, the person passes various routes (they rarely pass the same route as the person). Therefore, the increase in the feature amount f ₁ is mainly performed, and the position of the two-dimensional feature amount space (f ₁ , f ₂ ) changes from [1] to [2] in FIG.

[3] in FIG. 32 is a state in which a person's movement route is spread almost throughout the station platform. In the transition from the state [2] to the state [3] in FIG. 32, as in the transition from [1] to [2], the feature amount f ₁ is mainly increased, and the two-dimensional feature amount space ( In f ₁ , f ₂ ), the position changes from [2] to [3] in FIG.

In the state of [4] in FIG. 32, since the density of people increases, the area having a large local image change rate increases. The increase in the number of regions RN ₂ having a large local image change rate is larger than the increase in the number of regions RN ₁ having a medium or higher local image change rate, and the two-dimensional feature space (f ₁ , f ₂ ). In FIG. 33, the position changes from [3] to [4] in FIG. 33 (the gradient of movement of the points on which the feature values are plotted increases).

  32, when the situation changes in the order of [4] → [3] → [2] → [1], the movement of the feature quantity in the feature quantity space of FIG. 33 is [4] → [3] → [ 2] → [1] The reverse path is followed.

<When the travel route is biased>
The movement of the plot in the feature amount space of FIG. 33 when changing from the state [1] to the state [2] in FIG. 32 is the same as the case of free movement. From the state of [2], people begin to form a queue waiting for trains. In [5] of FIG. 32 where the procession has started to be formed, the movement path of the person is limited by the procession. Once the matrix is formed and the movement path is restricted, the increase in the number of regions RN ₁ having a medium or higher local image change rate reaches a peak. The index RN ₁ starts to decrease when the queue is extended and the movement path is further restricted. For the other indicators RN _2, the movement path is limited, increase the frequency of human passage per fixed Datosuruto unit area flow of people, RN ₂ is increased. In the two-dimensional feature space (f ₁ , f ₂ ) in FIG. 33, the position changes as [2] → [5].

In [6] of FIG. 32, the number of people in the queue further increases, and the movement route of the person is further restricted. Therefore, the number of regions RN ₁ having a medium or higher local image change rate is decreased, and the number of regions RN ₂ having a large local image change rate is increased. In the two-dimensional feature space (f ₁ , f ₂ ) in FIG. 33, the position changes as [5] → [6].

  Next, a method for calculating the degree of congestion will be described. Here, an example using the subspace method will be described. Refer to Chapter 14 of Non-Patent Document 5 for details on the subspace method. Only the outline will be described here. Processes (1) to (2) are performed in advance.

(1) A plurality of feature amounts are extracted for each situation such as movement of a large number of people / movement of a small number of people / movement route deviation, and a distribution as shown in FIG. 30 is obtained.

(2) Principal component analysis is performed on the two-dimensional distribution of each situation, and a straight line of the first principal component (principal axis) is defined as a one-dimensional subspace. FIG. 34 shows the partial spaces of the three situations.

  A method for determining a situation to which a feature amount extracted from a scene for which the degree of congestion is desired belongs and a method for calculating the degree of congestion will be described in (3).

(3) The subspace to which the feature quantity whose state is unknown belongs is determined. When the input feature vector f is projected onto the subspace as shown in FIG. 35, the subspace having the maximum projected component || Pf || is selected as the identification result. Up to this point, this corresponds to another embodiment for identifying the situation type described above. As shown in FIG. 36, by previously associating a position on the partial space (here, a straight line) with an index representing the degree of congestion, the end point position of a vector obtained by projecting the feature vector f into the partial space (FIG. 35). (The position of the arrow POS), the congestion degree index can be obtained. As the congestion degree index, the closest congestion level (integer) in FIG. 36 may be selected, or interpolation may be performed based on the end point position to calculate the congestion degree level in decimal.

  Non-Patent Document 5: “Computer Vision Technology Review and Future Prospects”, Takashi Matsuyama, Yoshinori Kuno, Satoshi Imiya, New Technology Communications, 1998/06

  In the present embodiment, the degree of congestion is calculated using a position projected onto the partial space, but other methods such as multiple regression analysis may be used instead of principal component analysis. Since the details of the multiple regression analysis are described in Non-Patent Document 4, description thereof is omitted.

  In the present embodiment, three situations (multi-person movement / small-person movement / movement path bias) are defined as the situation types, but “normal movement” and “movement-path bias” are a combination of multi-person movement and small-person movement. 2 situations may be identified, or more situations than the 3 situations may be classified. For example, the situation may be classified into four situations as shown in FIG. 37 so that the approximation error when performing the approximation in the partial space is reduced.

  As described above, according to the situation determination apparatus of the present embodiment, the time change rate of the luminance value in the local region of the captured image and the histogram of the time change rate for the plurality of local regions are analyzed. In addition, it is possible to determine the movement status of the vehicle and to calculate the congestion level. In particular, since it is possible to detect that the travel route is biased with simple processing, it is possible to estimate the presence of the queue and the level of congestion in a place where there is a train queue such as a station. is there. Of course, the situation and the degree of congestion can also be estimated in a situation where there is no bias in the movement route. In addition, when the deviation of the movement route is detected in a passage that normally does not generate a matrix, it is possible to estimate that free movement of a person is hindered by some kind of obstacle.

(Embodiment 3)
FIG. 38 is a block diagram showing a schematic configuration of the abnormality determination device according to Embodiment 3 of the present invention. In FIG. 38, the abnormality determination device of the present embodiment is configured to include a situation determination device 500, a train arrival detection unit 510, an abnormality determination unit 520, and a notification unit 530. Similar to the first embodiment described above, this abnormality determination device is installed on the platform of the railway station in FIG.

  The situation determination device 500 is represented by the block diagram of FIG. 1, and transmits the determined situation type or congestion level to the abnormality determination unit 520. Details such as the operation of the situation determination apparatus 500 are as described in the first embodiment or the second embodiment. It is assumed that the situation type to be determined includes at least a situation where a person's movement route is biased.

  The train arrival detection unit 510 is a means for detecting arrival of a train on the platform. Any method may be used as long as it can detect the arrival of a train, such as by image recognition, by a sensor other than an image, or by using train operation data. When image recognition is used, appearance information such as the color and shape of the train is registered in advance, and the arrival of the train may be determined by matching with the registered template. The motion vector described in Embodiment 1 can also be used. If a plurality of motion vectors having similar directions and intensities appear in a predetermined area (near the track) in the image, it may be determined that the train has arrived because the movement of the rigid body has been detected. As a sensor other than the image, for example, a load sensor can be provided under the track, and the arrival of the train can be determined based on the load value. In addition, by using a laser emitter and a light receiver, it is possible to detect the reflected light from the train of laser light, or to detect that the light from the light emitter is blocked by the train, and determine the arrival of the train. . As an example of using train operation data, train arrival can be known by receiving communication from a train or an operation command center.

  The abnormality determination unit 520 determines an abnormality from both the result of the situation determination or the congestion degree determination from the situation determination device 500 and the train arrival information from the train arrival detection unit 510. Details of the processing will be described later. When the abnormality determination unit 520 determines that an abnormality has occurred, the reporting unit 530 reports to a predetermined contact address.

  Hereinafter, the operation of the abnormality determination device of the present embodiment will be described using the flowchart of FIG. The situation determination apparatus 500 executes the image input step S100 to the situation determination step S140. The configuration of the situation determination apparatus 500 is shown in FIG. 1, and the details of the means for executing each process from step S100 to step S140 and the operation of each process have been described in the first embodiment or the second embodiment. Since it is as it is, description is abbreviate | omitted. Here, a normal time-series change in the congestion degree index and the situation determination result output by the situation determination apparatus 500 will be described with reference to FIG. FIG. 40 is a schematic diagram showing, in a time series, the degree of congestion and the situation determination result output by the situation determination apparatus described in the second embodiment.

<Congestion index>
The horizontal axis represents the time, the train has arrived at time t ₃ and t _6. It is the home for immediately after the time t ₁ in which the front of the train arrived in the degree of congestion there is no person who is 0. Until the arrival of the train at time t _3, the crowds gradually increase because the people to get on the train flow into the platform. At time t ₃ when the train arrives, people waiting for the train enter the train and passengers from the train Go down to the home and head towards the staircase ST in FIG. Passengers the time t the degree of congestion in ₄ has moved to the all stairs ST back to zero again. After that, a repeat of from time _{t 1} to _{t 4.}

<Situation judgment result>
From time t ₁ to t _{2, since} there is no train waiting queue at the platform, it is determined as “situation 1: small number of people moving” as the situation determination result. Began to occur from time t ₂ train queue in the home, status determination as a result: it is determined that the "situation 2 moving path bias". When the train arrives at time t ₃ , passengers head for the staircase ST in FIG. 2 at a time, so that the situation determination result is “Situation 3: Multi-person movement”.

  Next, the train arrival detection step S200 is executed by the train arrival detection unit 510. The processing here is as described when the operation of the train arrival detection unit 510 is described.

  Next, step S210 is executed by a control unit (not shown in FIG. 38). If train arrival information exists at the current time (in the case of YES), the process proceeds to an abnormality determination step S220, and if train arrival information does not exist (in the case of NO), the process proceeds to step S150.

Next, the abnormality determination step S220 is executed by the abnormality determination unit 520. The processing in this step is performed as follows. Train congestion from the arrival is measured in advance time (e.g. from time t ₃ of t ₄ interval) to return to 0 again, previously determined threshold DT based on that value. When the time DT elapses from the time when the train arrival detection unit 510 detects the arrival of the train, when the congestion degree index does not fall below a predetermined threshold or the situation determination result does not become “situation 1: small number of people moving” Determined as abnormal.

  As an example of an abnormal situation, when all of the waiting customers who were at the platform cannot get on the train because the train arrived is full, some people remain at the platform, or congestion occurs at the stairs ST, and the time DT is Even if it passes, the person who got off the train may not be able to move to the stairs ST. It should be noted that since the time for clearing the congestion is affected by the number of people in the train queue and the number of people getting off the train, it is desirable to set appropriate values for the time zone and the day of the week as the threshold DT.

  Next, step S230 is executed by a control unit (not shown in FIG. 38). If it is determined that there is an abnormality in the abnormality determination step (in the case of YES), the process proceeds to notification step S240; otherwise (in the case of NO), the process proceeds to step S150. The reporting step S240 is executed by the reporting unit 530. The processing here is as described when the operation of the reporting unit 530 is described. Finally, step S150 is executed by a control unit not shown in FIG. If a processing end instruction is input by the operator of the apparatus through the input unit (not shown in FIG. 38) (YES), the process returns to the image input step S100 to process the frame image at the next time. If no process end instruction is input (NO), the process ends.

  As described above, according to the abnormality determination device of the present embodiment, it is possible to determine an abnormal congestion state different from normal using the situation type or the degree of congestion obtained by the situation determination device.

  The situation determination device, the situation determination method, and the situation determination program according to the present invention determine a situation type related to the movement of a person, or determine the degree of congestion of a person, and a public in which a plurality of people such as stations and airports move It is possible to provide an apparatus, a method, and a program capable of saving labor and increasing efficiency in image monitoring in a space. Further, the abnormality determination device, the abnormality determination method, and the abnormality determination program according to the present invention determine an abnormal congestion situation related to the movement of a person and perform image monitoring in a public space where a plurality of people move such as a station / airport. It is possible to provide an apparatus, a method, and a program that can detect an abnormality early in order to save labor and improve efficiency and prevent an accident associated with abnormal congestion.

The block diagram which shows schematic structure of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows a mode that the situation determination apparatus which concerns on Embodiment 1 of this invention was installed in the station of a railway The figure which shows the picked-up image with the camera CM which concerns on Embodiment 1 of this invention. The flowchart for demonstrating the condition determination method implemented with the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the relationship between the frame image accumulate | stored in the image storage part of the condition determination apparatus which concerns on Embodiment 1 of this invention, the local change information accumulate | stored in a local image change information storage part, and time. The figure which shows a mode that the local region division was performed with respect to the captured image with the camera CM which concerns on Embodiment 1 of this invention. The figure which shows the example of the motion vector extracted in the local image change detection part of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the element of the local change information detected in the local image change detection part of the condition determination apparatus which concerns on Embodiment 1 of this invention, and is accumulate | stored in a local image change information storage part. The figure which shows the element of the local change rate calculated in the local image change rate calculation part of the condition determination apparatus which concerns on Embodiment 1 of this invention, and accumulate | stores in a local image change rate storage part. Frame images stored in the image storage unit of the situation determination apparatus according to Embodiment 1 of the present invention, local change information stored in the local image change information storage unit, and local images stored in the local image change rate storage unit Diagram showing the relationship between rate of change and time The figure which shows the example of the local image change rate histogram calculated in the local image change rate histogram calculation part of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the example of the image which imaged the movement condition of many people in operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the example of the image which imaged the movement condition of a small number of persons in the operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the example of the image which imaged the condition where a person moves in the place where the queue which waits for a train exists in operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. In the operation of the situation determination apparatus according to the first embodiment of the present invention, a table showing the tendency of the number of regions where the change occurs and the local image change rate with respect to three situations. The figure which shows the characteristic of the local image change rate histogram at the time of multi-person movement in operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the characteristic of the local image change rate histogram at the time of small group movement in operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the characteristic of the local image change rate histogram in case operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention has a bias | inclination in a human movement path | route. The figure which shows a mode that the local region division | segmentation was performed to the image which imaged the movement condition of many people in operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows a mode that local region division was performed to the image which imaged the movement condition of a small number of persons in the operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows a mode that local region division was performed to the image which imaged the condition where the bias | inclination exists in a person's movement path | route in operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the example of the local image change rate histogram calculated | required from the actual moving image in operation | movement of the condition determination apparatus which concerns on Embodiment 1 of this invention. The block diagram which shows the internal structure of the condition determination part of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the other implementation example of the process in the local image change detection part of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the other implementation example of the process in the local image change detection part of the condition determination apparatus which concerns on Embodiment 1 of this invention. The figure which shows the other implementation example of the process in the local image change detection part of the condition determination apparatus which concerns on Embodiment 1 of this invention. The block diagram which shows the internal structure of the condition determination part of the condition determination apparatus which concerns on Embodiment 2 of this invention. The figure which shows the value of the parameter | index calculated | required in the feature amount extraction of the feature extraction part of the condition determination apparatus which concerns on Embodiment 2 of this invention. The figure which shows the value of the feature-value extracted by the feature extraction part of the condition determination apparatus which concerns on Embodiment 2 of this invention. The figure which shows distribution of the two-dimensional feature-value extracted from the moving image of 3 scenes by the feature extraction part of the condition determination apparatus which concerns on Embodiment 2 of this invention. The figure which shows the relationship between each value of the feature-value extracted by the feature extraction part of the condition determination apparatus which concerns on Embodiment 2 of this invention, and distribution of a two-dimensional feature-value. The figure which is used for description of the congestion degree determination process in the situation determination apparatus according to the second embodiment of the present invention, and shows an image obtained by capturing six scenes at the station platform The figure which is used for description of the congestion degree determination process in the situation determination apparatus according to Embodiment 2 of the present invention, and shows the distribution in the two-dimensional feature amount space of feature amounts corresponding to six scenes. The figure which used for explanation of the congestion degree judgment processing in the situation judgment device concerning Embodiment 2 of the present invention, and displayed each partial space calculated for the distribution of the feature quantity in three situations simultaneously with the distribution. The figure for demonstrating the subspace method used for description of the congestion degree determination process in the condition determination apparatus which concerns on Embodiment 2 of this invention. The figure which shows a mode that it was used for description of the congestion degree determination process in the condition determination apparatus which concerns on Embodiment 2 of this invention, and matched the parameter | index showing a congestion degree with the position on a partial space. The figure which used for explanation of the congestion degree judgment processing in the situation judgment device concerning Embodiment 2 of the present invention, and displayed each partial space calculated simultaneously with the distribution of the feature quantity in four situations. The block diagram which shows schematic structure of the abnormality determination apparatus which concerns on Embodiment 3 of this invention. The flowchart for demonstrating the abnormality determination process performed with the abnormality determination apparatus which concerns on Embodiment 3 of this invention. The graph which shows the time transition in the normal state of the congestion degree and condition classification which the condition determination apparatus of the abnormality determination apparatus which concerns on Embodiment 3 of this invention outputs

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Image input part 110 Image storage part 120 Local image change detection part 130 Local image change information storage part 140 Local image change rate calculation part 150 Local image change rate storage part 160 Local image change rate histogram calculation part 170 Situation determination part 200 Reference histogram Storage unit 210 Histogram comparison unit 300 Feature extraction unit 310 Identification reference storage unit 320 Identification unit 500 Situation determination device 510 Train arrival detection unit 520 Abnormality determination unit 530 Notification unit PH Station platform WL Platform side wall ST Platform entrance / exit of people Stairs for RL Track CM Camera SD Situation Judgment Device AR1 to AR3 Arrows representing the movement path of a person WP Person waiting for a train MP Person moving

Claims (11)

A situation determination device that determines a movement situation and / or a congestion degree of a person by analyzing a captured moving image ,
Wherein dividing the moving image into a plurality of local regions, and local image change ratio calculating means for calculating the local image change ratio of each of the local region,
A histogram representing the frequency distribution of the local image change rate for each of the local regions calculated by the local image change rate calculating means is calculated, and the calculated histogram and the reference histogram obtained in advance are compared to determine the movement situation of the person And / or situation determination means for determining the degree of congestion of people,
Situation judgment device with The local image change ratio calculating means for each local region, obtains a representative motion vector obtained from the captured images at a first time interval of the moving image,
Comparing the magnitude of the representative motion vector with a predetermined threshold to determine the presence or absence of motion in the local region;
The local image change rate as a ratio between the number of times obtained by dividing the second time interval longer than the first time interval by the first time interval and the number of the first time intervals determined as having motion. The situation determination apparatus according to claim 1, which calculates The situation determination means includes a reference histogram storage means for holding a reference histogram obtained in advance ,
Situation determining apparatus according to claim 1 or claim 2 Ru is composed of a histogram comparing means for comparing the reference histogram and histograms the calculated. As the movement situation, it is determined as a situation of movement of many people, movement of a small number of people or movement route bias,
When the number of areas determined to have movement in the local area is large, and the local image change rate in the calculated histogram is medium or large, it is determined as multi-person movement,
When the number of areas determined to have motion in the local area is large, and the local image change rate in the calculated histogram is determined to be a small number of people when the medium or large frequency is small,
3. The movement path bias is determined when the number of areas determined to have motion in the local area is small and the local image change rate in the calculated histogram is large at a medium or large frequency . Situation judgment device. An abnormality determination device that determines an abnormal situation by analyzing a moving image or a plurality of still images captured by an imaging unit installed on a platform of a station,
A situation determination device according to claim 4 ;
Train arrival detection means for detecting the arrival of the train to the platform;
Determined Ri said status determination result as human movement path polarized by using the train arrival information after a predetermined time has passed after the train arrives information is obtained by the state determination result to the train arrival detecting means according to the status determining device An abnormality determining means for determining an abnormality when being performed;
An abnormality determination device comprising: The abnormality determination device according to claim 5 , further comprising a notification unit configured to notify a predetermined contact address when the abnormality determination unit determines that an abnormality has occurred. A situation determination method for determining a movement situation and / or a degree of congestion of a person by analyzing a captured moving image ,
The moving image is divided into a plurality of local regions, and local image change ratio calculating step of calculating a local image change ratio of each of the local region,
A histogram representing the frequency distribution of the local image change rate for each local area calculated in the local image change rate calculating step is calculated, and the calculated histogram and a reference histogram obtained in advance are compared to determine the movement situation of the person And / or a situation determination step for determining the degree of congestion of the person;
Situation judgment method with The local image change rate calculation step for each local region obtains a representative motion vector obtained from images captured at a first time interval of the moving image,
Comparing the magnitude of the representative motion vector with a predetermined threshold to determine the presence or absence of motion in the local region;
The local image change rate as a ratio between the number of times obtained by dividing the second time interval longer than the first time interval by the first time interval and the number of the first time intervals determined as having motion. The situation determination method according to claim 7, which calculates As the movement situation, it is determined as a situation of movement of many people, movement of a small number of people or movement route bias,
When the number of areas determined to have movement in the local area is large, and the local image change rate in the calculated histogram is medium or large, it is determined as multi-person movement,
When the number of areas determined to have motion in the local area is large, and the local image change rate in the calculated histogram is determined to be a small number of people when the medium or large frequency is small,
When the number of areas determined to have motion in the local area is small and the local image change rate in the calculated histogram is medium or large, the path is determined to be biased.
The situation determination method according to claim 7. An abnormality determination method for determining an abnormal situation by analyzing a moving image or a plurality of still images captured by an imaging unit installed on a platform of a station,
A situation determination step for executing the situation determination method according to claim 9;
A train arrival detection step for detecting the arrival of the train to the platform;
Determining a moving route deviation of human as the status determination result after a predetermined period of time after the obtained the train arrival information by using a train arrival information by the train arrival detecting means and status determination result by the situation determining unit An abnormality determination step for determining an abnormality,
An abnormality determination method comprising: A situation determination program for causing a computer to determine a movement situation and / or a degree of congestion of a person by analyzing a captured moving image ,
The moving image is divided into a plurality of local regions, and local image change ratio calculating step of calculating a local image change ratio of each of the local region,
A histogram representing the frequency distribution of the local image change rate for each local area calculated in the local image change rate calculating step is calculated, and the calculated histogram and a reference histogram obtained in advance are compared to determine the movement situation of the person And / or a situation determination step for determining the degree of congestion of the person;
Situation determination program with

Images