JP2010211763A - Image monitoring device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image monitoring device capable of correctly detecting the presence or absence of a disaster in a monitoring space where a person is present. <P>SOLUTION: An image sensor 2 of the image monitoring device 1 includes: an imaging section 20 for sequentially imaging the monitoring space; a storing section 21 for storing the image imaged by the imaging section 20; an earthquake detecting means 220 for detecting an earthquake occurrence; a line segment extracting means 222 for extracting a line segment element from the image imaged by the imaging section 20; and a disaster determining means 223 for calculating variation amount of a number of line segment elements between the image imaged before the earthquake is detected and the image imaged after the earthquake is no longer detected, and determining a disaster occurrence when the variation amount exceeds a predetermined reference amount. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an image monitoring apparatus that detects an occurrence of a disaster by processing an image obtained by imaging a monitoring space.

  In recent years, since large-scale disasters such as large earthquakes have occurred one after another, it is recognized that it is important to quickly grasp the disaster situation immediately after a disaster occurs. It can be said that grasping the situation of disaster immediately after the occurrence of a disaster is one of the important elements in order to take an accurate initial action in business continuity management (BCM).

  In the monitoring device described in Patent Document 1, the difference between the images before and after the occurrence of the earthquake is determined, and it is determined that there is a disaster if the area of the difference is a predetermined value or more.

JP 2006-285644 A

  However, in the prior art, if a resident or employee exists in the monitoring space, the difference is extracted even in the area of the occupant. Furthermore, if there are a plurality of people in the room, the area of the difference increases by the number of people.

  For this reason, the conventional technology cannot correctly detect the presence or absence of a disaster in a monitoring space where people exist.

  The present invention has been made in view of the above problems, and an object thereof is to provide an image monitoring apparatus capable of correctly detecting the presence or absence of a disaster even in a monitoring space where a person exists.

In order to solve such a problem, the present invention includes an imaging unit that sequentially images a monitoring space, a storage unit that stores an image captured by the imaging unit, an earthquake detection unit that detects occurrence of an earthquake, and an imaging unit. Line segment extraction means for extracting line segment elements from the captured image, and change in the number of line segment elements between the image captured before the earthquake is detected and the image captured after the earthquake is no longer detected There is provided an image monitoring apparatus comprising: a damage determination unit that calculates an amount and determines the occurrence of a damage when the amount of change exceeds a preset reference amount.
According to such a configuration, the presence / absence of damage is determined based on the change in the number of line segment elements that are exclusively extracted from artifacts and difficult to extract from natural objects. Can be determined.

Further, in a preferred aspect of the present invention, the line segment extraction unit further measures the inclination of each line segment element, and the damage determination unit calculates the amount of change for the line segment element whose inclination is in a substantially vertical direction in the monitoring space. To do.
According to such a configuration, since the amount of change is calculated for a substantially vertical line segment element that changes significantly between a relatively orderly monitoring space before the disaster and a messy monitoring space after the disaster, higher accuracy is obtained. Can be used to determine if there is a disaster. Moreover, since a relatively large number of line segment elements are distributed in the monitoring space before the disaster, highly reliable determination is possible.

Further, in a preferred aspect of the present invention, the line segment extraction means further measures the length of each line segment element, and the damage determination means changes the line segment element whose length is equal to or greater than a preset threshold value. Calculate the amount.
According to such a configuration, since the amount of change is calculated by excluding short line segment elements that are also extracted from natural objects, it is possible to determine the damage with higher accuracy even in a monitoring space where a person exists.

Further, in a preferred aspect of the present invention, the line segment extraction means further measures the length of each line segment element, and the damage determination means totals the number of line segment elements for each predetermined length section, The change amount is calculated by accumulating the section change amount for each length section with a larger weight as the length section is longer.
According to such a configuration, since the amount of change is calculated by increasing the sensitivity with respect to a long line segment element that is difficult to be moved by the occupant, it is possible to determine the damage with higher accuracy even in a monitoring space where a person exists.

In a preferred aspect of the present invention, the damage determination means calculates the amount of change for a line segment element whose length is equal to or greater than a preset threshold length.
According to such a configuration, since the amount of change is calculated by excluding short line segment elements that are also extracted from natural objects, it is possible to determine the damage with higher accuracy even in a monitoring space where people exist.

Further, in a preferred aspect of the present invention, the damage determination means sets the length section so that the number of line segment elements extracted from the pre-detection image in each of the length sections is equal to or greater than a preset lower limit number. Set.
According to such a configuration, the reliability of the section change amount is ensured, and the change amount that is the sum of the section change amounts can be calculated with high reliability. Therefore, it is possible to perform highly reliable damage determination.

  In a preferred aspect of the present invention, the damage determination means suppresses variations in determination due to differences in monitoring space by normalizing the change amount with the number of line segment elements extracted from the pre-detection image. To do.

In a preferred aspect of the present invention, the image monitoring apparatus further includes a distortion correction unit that performs correction to cancel distortion caused by lens characteristics of the imaging unit with respect to the image captured by the imaging unit, and a line segment extraction unit Extracts a line segment element from the image corrected by the distortion correcting means.
According to such a configuration, the influence of lens distortion is eliminated and the accuracy of extracting a true line segment element is improved, so that the occurrence of a disaster can be determined more accurately.

  According to the image monitoring apparatus of the present invention, it is possible to accurately detect a disaster even in a monitoring space where a person exists.

It is the figure which showed the structure of the image monitoring apparatus 1. It is the figure which showed the structure of the image sensor 2. It is the figure which showed the timing chart of a damage determination process It is the figure which showed the flowchart of an image monitoring process It is the figure which showed the flowchart of the damage determination process concerning 1st embodiment. It is the figure which showed the mode of the damage determination process concerning 1st embodiment. It is the figure which showed the flowchart of the damage determination process concerning 2nd embodiment. It is the figure which showed the mode of the damage determination process concerning 2nd embodiment. It is the figure which showed the flowchart of the damage determination process concerning 3rd embodiment. It is the figure which showed the mode of the damage determination process concerning 3rd embodiment.

  As an example of a preferred embodiment of the present invention, an image monitoring device that determines the presence or absence of a disaster in a building and determines the safety of the occupants will be described.

<First embodiment>
[Configuration of image monitoring device]
With reference to FIG. 1, the structure of the image monitoring apparatus 1 concerning 1st embodiment of this invention is demonstrated.

  The image monitoring device 1 includes an image sensor 2 connected to a receiving device 5 via a controller 3 and a communication network 4.

  The image sensor 2 is installed in a room which is a monitoring space, and processes an image obtained by imaging the monitoring space to determine the presence or absence of a disaster and the safety of a room occupant. The determination result is output to the controller 3. The image sensor 2 and the controller 3 are connected by a local transmission network such as a sensor LAN 7. The sensor LAN 7 may be connected to a security sensor 6-1 or a disaster prevention sensor 6-2.

  The controller 3 is installed in the building 8 that accommodates the monitoring space, and sends the monitoring information input from the image sensor 2 to the communication network 4 by specifying the receiving device 5 set in advance as a notification destination as the destination.

  The communication network 4 is a wide area network such as the Internet or a telephone line, and transmits monitoring information to a receiving device 5 that is a notification destination.

  The receiving device 5 is a mobile phone possessed by a family member of the room, a supervisor, or the like, a PC used at a work place, a center device of a security center, or the like. The receiving device 5 includes a receiving unit (not shown) that receives monitoring information from the communication network 4, and a display unit (not shown) that displays the received monitoring information and transmits it to a safety confirmation applicant.

  The configuration of the image sensor 2 is shown with reference to FIG.

  The image sensor 2 includes an imaging unit 20, a storage unit 21, and a communication unit 23 connected to a control unit 22.

  The imaging unit 20 is a so-called surveillance camera. The imaging unit 20 images the monitoring space at predetermined time intervals, and sequentially outputs the captured images to the control unit 22. Hereinafter, the image is referred to as a monitoring image, and the unit of time that is recorded at the predetermined time interval is referred to as time. The imaging unit 20 is preferably installed above the monitoring space so as to include a lot of the floor surface in its visual field. By doing so, it is possible to obtain more image changes due to the articles falling or dropping from the monitoring image. In the present embodiment, the imaging unit 20 is installed on the corner ceiling of the room with its optical axis directed near the center of the floor of the room. In another embodiment, the imaging unit 20 includes a fisheye lens and is installed in the center of the ceiling of the room with its optical axis directed vertically downward.

  The storage unit 21 is a memory device such as a ROM (Read Only Memory) or a RAM (Random Access Memory). The storage unit 21 stores various programs and various data, and inputs and outputs them with the control unit 22. Various data includes a past image 211 and a distortion correction coefficient 212.

  The past image 211 is a past monitoring image such as a monitoring image one hour before (hereinafter also referred to as an immediately preceding image) or a monitoring image before an earthquake is detected (hereinafter also referred to as an undetected image).

  The distortion correction coefficient 212 is a coefficient for performing correction to cancel distortion generated in the monitoring image due to the lens characteristics of the imaging unit 20. A distortion correction coefficient 212 corresponding to the lens characteristics of the imaging unit 20 is set in advance.

  The control unit 22 is an arithmetic device such as a DSP (Digital Signal Processor) or MCU (Micro Control Unit). The control unit 22 reads each program describing the operations of the earthquake detection unit 220, the distortion correction unit 221, the line segment extraction unit 222, the disaster determination unit 223, the safety determination unit 224, and the like from the storage unit 21, and executes each program. Function as.

  The earthquake detection means 220 detects the occurrence of an earthquake in the monitoring space and outputs a detection result. Specifically, the earthquake detection unit 220 compares the monitoring images continuously captured, that is, the monitoring image captured by the imaging unit 20 at the current time (hereinafter also referred to as the current image) and the past image 211 are stored. If a change over almost the entire image is detected by comparison with the immediately preceding image, it is determined that an earthquake has occurred, and if no change is detected, it is determined that no earthquake has occurred. The earthquake detection means 220 extracts a pixel having a difference between both images, and detects the variation if the distribution range of the extracted pixel is equal to or greater than a predetermined ratio (for example, 90% or more) of the entire image. Moreover, the earthquake detection means 220 performs the update which replaces a previous image with the present image at each time for the said image comparison.

  When the monitoring image is input, the distortion correction unit 221 performs correction using the distortion correction coefficient 212 on the monitoring image, and outputs the corrected monitoring image.

  When a monitoring image is input, the line segment extraction unit 222 extracts line segment elements from the monitoring image and outputs information on the extracted line segment elements (line segment information). Moreover, the line segment extraction means 222 measures the length (line segment length) of each extracted line segment element as necessary, and outputs line segment information including the measured line segment length.

  Specifically, the line segment extraction unit 222 generates an edge image by applying an edge detection operator to the input monitoring image, labels the generated edge image, and performs a known Hough transform on the label to thereby generate a line. Extract minute elements. As the edge detection operator, a Canny filter, a Sobel filter, a Prewitt filter, a Laplacian, or the like can be used.

  Here, as the distance from the center of the monitoring image increases due to the influence of lens distortion, a portion that is originally a straight line is likely to be captured as a curve, and a portion that is originally a curved line is likely to be captured as a straight line. Therefore, the line segment extraction unit 222 extracts line segment elements from the monitoring image corrected by the distortion correction unit 221. This eliminates the influence of lens distortion and improves the accuracy of extracting a true line segment element, thereby improving the accuracy of damage determination.

  The damage determination means 223 determines the number of line segment elements between the pre-detection image captured before the earthquake detection and the monitoring image (hereinafter also referred to as the post-detection image) captured after the earthquake is no longer detected. The amount of change is calculated, and the occurrence of a disaster is determined when the calculated amount of change exceeds a preset reference amount. In the present embodiment, the determination result is expressed in two stages: no damage and damage.

  Line segment elements are easily extracted from artifacts, but are difficult to extract from natural objects such as humans. For this reason, determining the presence or absence of damage based on changes in the number of line segment elements suppresses the impact of natural objects on the determination, so it is possible to accurately determine the presence or absence of damage even in a monitoring space where natural objects such as humans exist. it can.

  Hereinafter, the damage determination based on the change in the number of line segment elements will be described more specifically.

  The line segment elements before and after the detection of the earthquake are extracted by inputting the monitoring image at each time point to the line segment extracting means 222. In addition, in order to make it possible to refer to the pre-earthquake monitoring image after the earthquake, the damage determination unit 223 uses the pre-detection image stored as the past image 211 when the earthquake detection unit 220 has not detected an earthquake. Updates to be replaced with newly captured monitoring images are sequentially performed. When the earthquake detection unit 220 detects an earthquake, the update is stopped.

  FIG. 3 is an example of a timing chart. Assuming that the time when the occurrence of the earthquake is detected by the earthquake detection means 220 is t0, and the time when the earthquake is no longer detected is t1, the damage determination means 223 includes the pre-detection image and time taken at time (t0-1). The post-detection images captured at t1 are compared.

  The damage determination unit 223 counts the line segment elements extracted from the pre-detection image and the post-detection image, and normalizes by dividing the absolute value of the difference between the count results by the count result for the pre-detection image. The value is calculated as the change amount, and the occurrence of the disaster is determined by comparing the calculated change amount with the reference amount. When an earthquake occurs, the surveillance space tends to be messy due to scattering, overturning, destruction, etc., and the number of line segment elements tends to increase, so the amount of change exceeds the reference amount and the occurrence of damage is determined. The Note that the normalization has an effect of suppressing variation in determination due to a difference in the monitoring space.

  Here, although a line segment element is difficult to extract from a natural object, a short line segment element may be extracted from a natural object. Therefore, the damage determination means 223 calculates the amount of change only for line segment elements having a length equal to or greater than a predetermined threshold value TL. Specifically, the damage determination unit 223 refers to the line segment length measured by the line segment extraction unit 222, counts only the line segment elements whose line segment length is TL or more, and calculates the change amount. Do. By doing so, short line elements that are also extracted from natural objects are removed, so that damage can be detected more accurately in a monitoring space where people are present.

The safety determination means 224 detects a person image by comparing a monitoring image captured before detecting the occurrence of an earthquake with a background image, extracts image features such as a color histogram from the detected person image, and extracts the image. A feature is stored (registered) in the storage unit 21 for each person, and a monitoring image captured after the end of the earthquake is detected is searched to determine whether or not the registered image feature exists. The presence of a person or the like under the household property or equipment is not determined and the safety determination result is “unknown”, and the presence of a safe person is determined to be “safe”, and the safety determination result is “safe”.
The background image is initialized with the monitoring image immediately after startup, and is updated as needed by combining the partial images of the monitoring image in which no human image has been detected.

[Operation of image monitoring device]
The operation of the image monitoring apparatus 1 will be described with reference to FIG.

  When the apparatus is turned on and each unit and each unit is initialized, the imaging unit 20 images the monitoring space at each time, and the monitoring images are sequentially input to the control unit 22 (S1). The control unit 22 repeats the processes in steps S2 to S13 each time a new monitoring image is input.

  First, the earthquake detection means 220 of the control unit 22 detects an earthquake, that is, a shake due to the earthquake (S2). The earthquake detection means 220 performs an inter-frame difference process between the current image input in step S1 and the immediately preceding image stored as the past image 211, and if a change over the entire screen is detected, an earthquake is detected. Detect, otherwise do not detect earthquakes. It should be noted that for the sake of convenience, no earthquake is detected immediately after the start-up in which the previous image has not yet been stored. Further, in preparation for processing at the next time, the earthquake detection means 220 replaces the immediately preceding image with the current image and causes the storage unit 21 to store the detection result.

  Next, the control unit 22 determines the processing stage by referring to the current time and the earthquake detection result one hour before stored in the storage unit 21 and the setting of the safety determination period (S3). That is, the control unit 22 determines that “the stage before the earthquake detection” if an earthquake is not detected at the current time and the safety determination period is not set. If the earthquake is detected at the current time, “the earthquake is occurring” If no earthquake was detected at the current time and no earthquake was detected at the current time, it is determined as “the stage immediately after the earthquake has ended”, and there is no earthquake detected at the current time. If the period is set, the processing stage is determined as “safety determination stage”.

  Subsequently, the control unit 22 branches the process according to the determined processing stage (S4).

  If the processing stage is “stage before earthquake detection”, the safety determination means 224 of the control unit 22 extracts the person image of the occupant from the current image by background difference processing and registers the image feature of the person image. In step S5, the disaster determination unit 223 of the control unit 22 updates the pre-detection image stored as the past image 211 with the current image (S6). When the process of step S6 ends, the control unit 22 returns the process to step S1.

  If the processing stage is “the stage during the occurrence of an earthquake”, the control unit 22 returns the process to step S1 and waits for the end of the earthquake.

  If the processing stage is “stage immediately after the end of the earthquake”, the damage determination means 223 of the control unit 22 determines the presence or absence of the damage (S7).

  Details of the damage determination processing in step S7 will be described with reference to FIG. Note that at this stage, the current image corresponds to the post-detection image.

  First, the damage determination unit 223 of the control unit 22 reads the pre-detection image from the storage unit 21 and inputs it to the distortion correction unit 221 of the control unit 22, and further inputs the current image to the distortion correction unit 221. The distortion correction unit 221 performs distortion correction processing based on the distortion correction coefficient 212 on each input image and outputs a corrected image (S700).

  Next, the damage determination unit 223 that has received the output of the corrected image inputs the corrected pre-detection image to the line segment extraction unit 222 of the control unit 22, and further outputs the corrected current image to the line segment extraction unit 222. input. The line segment extraction unit 222 performs edge extraction and Hough transform on each input image to extract line segment elements (S710), measures the length of each extracted line segment element (S720), and Line segment information including the start point coordinates, end point coordinates, and line segment length of each line segment element is output. The line segment length can be calculated as the distance between the start point coordinate and the end point coordinate.

Subsequently, the damage determination unit 223 that has received the line segment information outputs the line segment element whose line segment length is less than the threshold TL from among the line segment information regarding the pre-detection image and the line segment information regarding the current image. Pruning to delete data is performed (S730), the number of line segment elements is counted in each piece of line segment information after pruning, and the line segment element number NA0 in the pre-detection image and the line segment element number NA1 in the current image, that is, the post-detection image, are obtained. The change amount is calculated by applying to the following equation (S740).
Change amount = | NA1-NA0 | / NA0 (1)

  Subsequently, the damage determination unit 223 compares the calculated change amount with the reference amount (S750), and determines that “there is a damage” if the change amount is larger than the reference amount (YES in S760 → S770), and the change amount is If it is less than the reference amount, it is determined that there is no damage (NO in S760 → S780). After the determination, the process proceeds to step S8 in FIG.

  FIG. 6 illustrates an example of simulation of the damage determination process. An image 90 represents an edge image generated from an image before detection, and an image 91 represents an edge image generated from an image after detection. The image 92 is an image illustrating a line segment element extracted from the edge image 90 and having a line segment length of TL or more, and the image 93 is an image illustrating a line segment element extracted from the edge image 91 and having a line segment length of TL or more. is there. 117 and 132 line segment elements were extracted from the pre-detection image and the post-detection image, respectively, and the amount of change was calculated to be 0.11. In other words, there was an 11% change. Since the amount of change is larger than a preset reference amount of 0.10, it is determined that “there is a disaster”. Based on the results of previous experiments, a length of 7 pixels (urbanization distance) was set for the TL on the monitoring image.

  Returning to FIG. 4, the control unit 22 refers to the result of the damage determination to determine whether or not to perform the safety determination (S8). When the result of the damage determination is “with damage”, the control unit 22 sets the safety determination period by setting a counter or timer for measuring the safety determination period (several minutes) (YES in S8 → S9). And if the safety determination period is set, the control part 22 will advance a process to step S10. On the other hand, when the result of the damage determination is “no damage” (NO in S8), control unit 22 returns the process to step S1 without setting the safety determination period. Note that when the number of registered persons is zero, the setting of the safety determination period may be omitted.

  When the processing stage is “safety judgment stage” in step S4 described above, or when the process flows from step S9, the safety judgment means 224 of the control unit 22 performs a safety judgment process (S10). The safety determination means 224 determines whether or not the person image registered in step S5 exists by searching the current image. The process of step S10 is repeatedly performed during the safety determination period, and the safety determination unit 224 stores a determination result “safe” in the storage unit 21 for a registered person whose presence has been determined even once during the safety determination period. Let

  When the process of step S10 ends, the safety determination means 224 confirms whether the safety determination period has expired (S11). If not completed (NO in S11), the control unit 22 returns the process to step S1.

  On the other hand, when the safety determination period expires, control unit 22 outputs monitoring information to controller 3 (YES in S11 → S12). That is, the damage determination means 223 outputs the damage determination result, and the safety determination means 224 outputs the number of registered persons and the number of registered persons whose “safe” is stored in the storage unit 21 as the safety determination result.

  Upon receiving the monitoring information, the controller 3 converts the monitoring information into a readable text format, adds the address of the receiving device 5 that is the notification destination, and sends it to the communication network 4. For example, when the damage determination result is “with damage”, the number of registered persons is N, and the number of registered persons with “safe” is n, the content of the text is “has been damaged. “N / N names are unknown”.

  When the monitoring information is output, the control unit 22 cancels the setting of the safety determination period (S13), and returns the process to step S1.

<Second Embodiment>
The surveillance space affected by the disaster is in a messy state with articles falling or scattered or destroyed. In a messy surveillance space, artifacts lined up in various directions are observed. On the other hand, the surveillance space that is not affected by the disaster is relatively orderly, and many line segment elements of artifacts facing the vertical direction are observed. These mean that a remarkable change occurs in the number of line segment elements oriented vertically before and after the disaster.

  Hereinafter, as a second embodiment of the present invention, an image monitoring apparatus 1 that determines a disaster by paying attention to a line segment element oriented in a substantially vertical direction will be described. In the following, description will be made by omitting portions common to the first embodiment.

[Configuration of image monitoring device]
The configuration of the image monitoring apparatus 1 according to the second embodiment of the present invention will be described with reference to FIG. 1 common to the first embodiment.

  In the first embodiment, the line segment extraction unit 222 extracts line segment elements from the monitoring image, measures the length of each line segment element, and outputs line segment information. In the second embodiment, the line segment extraction unit 222 further measures the inclination of each extracted line segment element, and outputs line segment information including the measured inclination.

  When the imaging unit 20 is installed at a relatively small depression angle, for example, when a monitoring image as shown in FIG. 6 is captured, the vertical direction of the monitoring image can be approximated to the vertical direction of the monitoring space. In this case, the line segment extraction unit 222 sets a horizontal line in the left-right direction of the monitoring image, and measures an angle formed by the line segment element and the horizontal line as an inclination.

  In another embodiment in which the imaging unit 20 is installed with its optical axis directed vertically downward, the direction of a straight line extending radially from the center of the image is the vertical direction of the monitoring space. In this case, the line segment extraction unit 222 calculates a straight line connecting the center of the image and the midpoint of the line segment element, and calculates an angle obtained by subtracting an angle formed by the calculated straight line and the line segment element from 90 degrees. Measured as the slope of.

  In the first embodiment, the damage determination unit 223 determines the damage based on the amount of change in the total number of line segment elements. In the second embodiment, the damage determination unit 223 determines the amount of change in the number of line segment elements for the line segment elements whose inclination is substantially vertical in the monitoring space among the line segment elements extracted by the line segment extraction unit 222. calculate.

  More specifically, the damage determination unit 223 refers to the line segment information extracted from each of the pre-detection image and the post-detection image, and detects the line segment element having an inclination within a reference range that is preset around 90 degrees. The amount of change is calculated by counting the number and normalizing by dividing the absolute value of the difference between the number counted for the post-detection image and the number counted for the pre-detection image by the number counted for the pre-detection image. . Then, the damage determination unit 223 compares the calculated change amount with a preset reference value and determines that there is a damage if the change amount exceeds the reference value, and determines that there is no damage otherwise. Note that the normalization has an effect of suppressing variation in determination due to a difference in the monitoring space.

  The rest of the configuration in the second embodiment is the same as in the first embodiment.

[Operation of image monitoring device]
Of the operations of the image monitoring apparatus 1 in the second embodiment, the overall flow of the image monitoring process is the same as that of the first embodiment described with reference to FIG. The second embodiment and the first embodiment differ in operation in the details of the damage determination processing in step S7.

  Hereinafter, the damage determination process in the second embodiment will be described with reference to FIG. In FIG. 7, the steps common to FIG. 5 are denoted by the same reference numerals as in FIG.

  First, similarly to the first embodiment, distortion correction, line segment element extraction, and line segment length measurement are performed on the pre-detection image and the current image that is the post-detection image (S700 to S720). Further, the line segment extraction means 222 of the control unit 22 calculates the inclination of the extracted line segment element (S722), and obtains line segment information including the start point coordinates, end point coordinates, line segment length and inclination of each line segment element. Output.

Subsequently, the damage determination means 223 prunes the data of the line segment elements as in the first embodiment (S730). Then, the damage determination unit 223 counts the number of line segment elements in the substantially vertical direction whose inclination is within the reference range in each line segment information after pruning (S732), and the number of line segment elements NV0 in the substantially vertical direction in the pre-detection image. Then, the amount of change is calculated by applying the number of segment elements NV1 in the substantially vertical direction in the current image, that is, the post-detection image, to the following equation (S740).
Change amount = 1.0− | NV1−NV0 | / NV0 Formula (2)

  Subsequently, the damage determination unit 223 compares the amount of change with the reference amount in the same manner as in the first embodiment, and determines that there is a damage if the amount of change is greater than the reference amount. It is determined that there is no damage (S750 to S780). After the determination, the process proceeds to step S8 in FIG.

  FIG. 8 shows a process when the damage determination process according to the second embodiment is applied to the simulation example of FIG.

  A graph 94 represents the slope distribution of the line segment element extracted from the pre-detection image. In this example, the data is counted every 12 degrees. It can be seen that the line segment elements are unevenly distributed at the inclination of 0 degree and the inclination of 85 to 96 degrees.

  On the other hand, the graph 95 represents the slope distribution of line segment elements extracted from the post-detection image. It can be seen that the distribution is gentle overall except for the inclination of 0 degree, and the strong peak with the inclination of 85 to 96 degrees disappears.

  The reference range of inclination is set to 85 to 96 degrees, the number of line segment elements NV0 in the substantially vertical direction in the pre-detection image is 31 and the number of line segment elements NV1 in the substantially vertical direction in the post-detection image is 14 and counted. The change amount is calculated as 0.45. In other words, there was a 45% change. Since the amount of change is larger than a preset reference amount 0.4, it is determined that “there is a disaster”.

  While the amount of change in the first embodiment was 11%, the amount of change in the second embodiment was 45%, indicating that the damage could be determined with higher sensitivity.

  As described above, in the second embodiment, it is possible to determine the damage with higher accuracy by calculating the amount of change in the number of line segment elements for the line segment elements whose inclination is substantially in the vertical direction in the monitoring space. Note that the reliability of the calculated amount of change is high because there are relatively many line segment elements in the substantially vertical direction in the normal monitoring space.

<Third embodiment>
Large artifacts such as shelves and tables are generally small in number, but are less likely to be moved by people in the room, so if there is a change, there is a high probability that a disaster has occurred. On the other hand, small artifacts such as books are likely to be moved by people in the room, but generally there are a large number of them, and these are moved all together during an earthquake. Therefore, even if it is a small artifact, if there are many changes, the probability that a disaster has occurred is high. These mean that the length distribution of the line segment element changes before and after the disaster, and that the accuracy of the damage determination increases when the sensitivity to the long line segment element is increased.

  Hereinafter, as a third embodiment of the present invention, an image monitoring apparatus 1 that determines damage by paying attention to the length distribution of line segment elements will be described. In the following, description will be made by omitting portions common to the first embodiment.

[Configuration of image monitoring device]
The configuration of the image monitoring apparatus 1 according to the third embodiment of the present invention will be described with reference to FIG. 1 common to the first embodiment.

  In the first embodiment, the damage determination unit 223 determines the damage based on the amount of change in the total number of line segment elements. In the third embodiment, the damage determination means 223 aggregates the number of line segment elements for each predetermined length section, and sets the weight of the section change amount for each length section to be larger as the length section is longer. The amount of change is calculated by accumulating. By doing so, the sensitivity to the long line segment element that is less likely to be moved by the occupant increases, so that it is possible to determine the damage with higher accuracy.

  More specifically, the damage determination means 223 counts the number of line segment information extracted from each of the pre-detection image and the post-detection image for each length section, and the number counted for the post-detection image in each section The interval change amount is calculated by dividing the absolute value of the difference counted for the pre-detection image by the number counted for the pre-detection image and normalizing, and the weighted sum of the interval change amounts is divided by the sum of the weights. To calculate the amount of change. Then, the damage determination unit 223 compares the calculated change amount with a preset reference value and determines that there is a damage if the change amount exceeds the reference value, and determines that there is no damage otherwise. Note that the normalization has an effect of suppressing variation in determination due to a difference in the monitoring space.

  Here, if the length section is set to a constant width, the reliability of the section change amount in the section having a small number is lowered, and the reliability of the change amount that is the sum of the section change amounts is lowered. Therefore, the damage determination unit 223 sets the length section so that the number of line segment elements extracted from the pre-detection image in each length section is equal to or greater than a preset lower limit number TN. By doing so, the reliability of the calculated change amount is ensured, and it is possible to determine the damage with high accuracy.

  The rest of the configuration in the third embodiment is the same as in the first embodiment.

[Operation of image monitoring device]
Of the operations of the image monitoring apparatus 1 in the third embodiment, the overall flow of the image monitoring process is the same as that of the first embodiment described with reference to FIG. The operation of the third embodiment differs from that of the first embodiment in the details of the damage determination processing in step S7.

  Hereinafter, the damage determination process in the third embodiment will be described with reference to FIGS. 9 and 10. FIG. 9 is a flowchart of the damage determination process, and steps common to FIG. FIG. 10 shows a state of processing when the damage determination processing according to the third embodiment is applied to the simulation example of FIG.

  First, as in the first embodiment, distortion correction, line segment element extraction, line segment length measurement, and line segment element pruning are performed on the pre-detection image and the current image that is the post-detection image, Line segment information about each image is generated (S700 to S730).

  Subsequently, the damage determination unit 223 sorts the line segment information regarding the pre-detection image with respect to the line segment length, and sets a length section in which the number of line segment elements equal to or greater than the lower limit TN can be secured in order from the data with the longest line segment length ( S734). The reason why the line segments are long in order is to ensure a long section useful for damage determination. In the shortest length section, the number of line segment elements equal to or greater than the lower limit value TN cannot be secured. However, the length section may be excluded from the calculation of the change amount, or the second shortest length section and You may integrate. In any case, a length section in which the number of line segment elements equal to or greater than the lower limit value TN is not set is not set.

  In the example of FIG. 10 where TN = 20, the length of the line segment is 26 or more, the length of the line is 20 or more and less than 26, the length of the line is 16 or more and less than 19, and the line A length section having a length of 11 or more and less than 15 and a length section having a line segment length of 9 or more and less than 10 were set. Length segments with a line segment length of less than 9 were excluded because they did not exceed 20 data.

  Subsequently, the damage determination means 223 counts the number of line segment elements related to the pre-detection image for each set length section, counts the number of line segment elements related to the post-detection image for each length section, and before and after the earthquake. Are aggregated (S736).

  A graph 96 in FIG. 10 shows the length distribution aggregated for the pre-detection image. The number of line segment elements in each section is leveled to 20 to 25 by the setting of the length section performed in step S734. Incidentally, the reason why the number of lines does not become 20 is that there are a plurality of line segment elements having the same line segment length. On the other hand, the graph 97 shows the length distribution collected for the post-detection image. Changes in the distribution are observed centering on the length section of the line segment length of 16-19.

  Subsequently, the damage determination means 223 sets the weight of each length section (S738). Specifically, the lower end of each length section is set as the weight of the length section. In the example of FIG. 10, 26, 20, 16, 11, and 9 are set in order from the long length section. In another embodiment, it may be set as a linear function such as 5, 4, 3, 2, 1 or may be set as a step function as 2, 2, 2, 1, 1.

Subsequently, the damage determination unit 223 calculates the section change amount Δm in each section m by applying the count result Nm0 of the pre-detection image and the count result Nm1 of the post-detection image in each section m to the following equation (3). The calculated change amount is calculated by applying the calculated section change amount Δm and the set weight Wm to the following equation (4) (S740). Note that Σ represents an operation for summing up m.
Δm = | Nm1−Nm0 | / Nm0 (3)
Change amount = Σ (Δm × Wm) / ΣWm Equation (4)

  Subsequently, the damage determination unit 223 compares the amount of change with the reference amount in the same manner as in the first embodiment, and determines that there is a damage if the amount of change is greater than the reference amount. It is determined that there is no damage (S750 to S780). After the determination, the process proceeds to step S8 in FIG.

  In the example of FIG. 10, the amount of change is calculated as 0.16, and since it is larger than a preset reference amount 0.10, “damaged” is determined. While the amount of change in the first embodiment was 11%, the amount of change in the third embodiment is equivalent to 16%, indicating that the damage can be determined with higher sensitivity.

  As described above, in the third embodiment, the change in the number of line segment elements for each line segment length is calculated by calculating the amount of change accumulated with higher sensitivity as the line segment length is longer, thereby determining damage with higher accuracy. be able to.

<Modification>
In the above embodiment, the earthquake detection means 220 detects an earthquake based on the image change. In another embodiment, the earthquake detection means 220 may be an acceleration sensor earthquake sensor or a pendulum type earthquake sensor that is connected to the control unit 22 and outputs a detection signal when a shake is detected. In this case, in step S2 of FIG. 4, the control unit 22 detects the occurrence and end of an earthquake by checking the presence or absence of a detection signal instead of the presence or absence of fluctuations throughout the image. In another embodiment, the earthquake detection means 220 detects an earthquake by receiving an emergency earthquake bulletin from outside via the communication network 4. In this case, the earthquake detection means 220 performs an output corresponding to the detection of the occurrence of an earthquake at the time when the earthquake early warning is received, and the earthquake is terminated when a sufficient time (for example, 30 seconds) has elapsed since the reception of the earthquake early warning. Output corresponding to detection.

  In the first embodiment, the damage determination means 223 determines the presence or absence of damage based on the amount of change calculated from the total number of line segment elements. In the second embodiment, the damage determination means 223 determines the number of line segment elements in a substantially vertical direction. The damage determination unit 223 determines the presence or absence of the damage based on the change amount calculated from the length distribution of the line segment elements in the third embodiment. In another embodiment, the damage determination means 223 determines the presence or absence of a damage by combining these two or three determination results. That is, the damage determination means 223 determines “disaster” when any two change amounts exceed each reference amount, or when three change amounts exceed each reference amount, and “no damage” otherwise. judge. This further improves the accuracy of damage determination.

  In the above embodiment, the damage determination means 223 determines the damage in two stages. In another embodiment, the damage determination means 223 determines the damage at three or more stages, where the presence of a damage is divided into a plurality of stages. In this case, based on prior studies, multiple reference amounts are set, such as a reference amount when the damage is small and a reference amount when the damage is larger, and the amount of change is sequentially compared with these reference amounts to determine whether there is any damage. At the same time, the magnitude of the disaster is determined.

DESCRIPTION OF SYMBOLS 1 ... Image monitoring device 2 ... Image sensor 3 ... Controller 4 ... Communication network 5 ... Receiving device 20 ... Imaging part 21 ... Memory | storage part 22 ... Control part 23. ..Communication unit 211 ... Past image 212 ... Strain correction coefficient 220 ... Earthquake detection means 221 ... Strain correction means 222 ... Line segment extraction means 223 ... Damage determination means 224 ... Safety judgment means

Claims (8)

An imaging unit that sequentially images the surveillance space;
A storage unit for storing an image captured by the imaging unit;
An earthquake detection means for detecting the occurrence of an earthquake;
Line segment extraction means for extracting line segment elements from the image captured by the imaging unit;
A change amount of the number of line segment elements is calculated between a pre-detection image captured before the earthquake detection and a post-detection image captured after the earthquake is no longer detected, and the change amount is preset. A damage determination means for determining the occurrence of a damage when exceeding a specified reference amount;
An image monitoring apparatus comprising: The line segment extraction means further measures the inclination of each line segment element,
The image monitoring apparatus according to claim 1, wherein the damage determination unit calculates the amount of change for a line segment element whose inclination is a substantially vertical direction in the monitoring space. The line segment extraction means further measures the length of each line segment element,
The image monitoring apparatus according to claim 1, wherein the damage determination unit calculates the amount of change for a line segment element whose length is equal to or greater than a preset threshold value. The line segment extraction means further measures the length of each line segment element,
The disaster determination means totals the number of line segment elements for each predetermined length section, and accumulates the section change amount for each length section with a larger weight as the length section is longer. The image monitoring apparatus according to claim 1, wherein the amount of change is calculated.   The image monitoring apparatus according to claim 4, wherein the damage determination unit calculates the amount of change for a line segment element whose length is equal to or greater than a preset threshold length.   The said damage determination means sets the said length section so that the number of the line segment elements extracted from the said image before a detection in each of the said length section may become more than the preset minimum number. Image monitoring device.   The image monitoring apparatus according to claim 1, wherein the damage determination unit normalizes the change amount based on a number of line segment elements extracted from the pre-detection image. Distortion correction means for performing correction to cancel distortion caused by lens characteristics of the imaging unit with respect to the image captured by the imaging unit;
The image monitoring apparatus according to claim 1, wherein the line segment extraction unit extracts a line segment element from the image corrected by the distortion correction unit.