JP6239376B2 - Image monitoring device and elevator monitoring device - Google Patents

Description

  The present invention relates to an image monitoring device and an elevator monitoring device that monitors the inside of an elevator car, and more particularly, to an image monitoring device and an elevator monitoring device that determine the presence or absence of a person in a monitoring area such as an elevator car.

  As background art of this technical field, in Patent Document 1, a prohibited area floating in the air is set as an entire area in an image, and an area corresponding to the ground or floor immediately below the prohibited area is set as a foot area and detected. The portion of the object image body that is closest to the foot area is extracted as the human image foot area, and the object image body overlaps the entire area and the person image foot area overlaps the foot area. Is described.

  Patent Document 2 discloses a predetermined area (area excluding the door portion) between an image signal obtained by imaging with an imaging device and an image signal when a passenger is not on the car. It is described that it is determined that there are passengers in the car with the difference of.

JP 7-21476 A JP 2012-184080 A

  However, the techniques described in Patent Documents 1 and 2 have the following problems.

  In this specification, a person (an intruder or an intruder (hereinafter, both are collectively referred to as an intruder)) and a three-dimensional object other than a person are collectively referred to as a person. Further, determining that the person is unmanned even though the person is in the predetermined area is referred to as unreported, and determining that the person is manned even though the person is not in the predetermined area is referred to as false information.

  First, as disclosed in Patent Document 1, the technique that uses the portion closest to the detected foot area of the human body as a human image foot is based on the premise that the whole body of the human body can always be detected. If only the upper part of the main body is detected instead of the whole body, the location where the person's image foot is located is mistakenly determined. In spite of not having it, it has the subject that the false report that it has invaded arises.

  Moreover, in patent document 2, it determines with a person having a difference with the image at the time of unattended in the floor area | region in a cage | basket | car. And if the door is open, there will be a big difference from the unattended image regardless of the presence or absence of people, so the area between the entrance and exit will be masked in advance and the difference between images will not be taken into account. The difference in floor area is evaluated. Here, in many cases, the interior of the car is illuminated from above by ceiling illumination, so that the vicinity of the floor tends to be darker than the human head. For this reason, performing a process focusing on only the floor area as in Patent Document 2 ignores information that is easy to detect at the upper part of the person, which makes it difficult to detect a stable person. Further, there is a problem that if parameters are adjusted in order to stabilize the person detection, there is a possibility that a false report that the manned determination is made despite the absence of the person is likely to occur.

  The present invention has been made in view of such problems, and an object of the present invention is to provide an image monitoring apparatus capable of suppressing misreporting and misreporting when determining the presence or absence of a person in a monitoring area such as an elevator car. It is to provide an elevator monitoring device.

  Therefore, in the image monitoring apparatus of the present invention, for example, a closed region on a certain surface is a space first region, a space vertically above the space first region is a space second region, and the space second region is within the space second region. In an image monitoring apparatus that processes an image obtained by an imaging unit installed so as to overlook the first space area, when the second space area shown in the image is the second image area, the first image area is displayed. A first feature detection unit that detects a first feature that indicates a part of a person in two regions, and the first image region when the first spatial region shown in the image is an image first region. A second feature detection unit for detecting a second feature that indicates shielding or shadow or reflection of the first region of the space caused by the person, and the first feature detection unit and the second feature detection. Based on the detection result of the part in the space second region It is determined whether the reporter is manned or whether the person is not present, and at least the first feature and the second feature are detected as a necessary condition. And a determination output unit.

  Moreover, the elevator monitoring apparatus of the present invention includes, for example, the image monitoring apparatus, the imaging unit, and an elevator control unit that receives the determination result from the determination output unit and controls the elevator, and the second space area. Is a space in the elevator car, the first space area is a floor of the elevator car, and the imaging unit is a surveillance camera provided in the elevator car.

  According to the present invention, for example, when it is determined whether or not there is a person in a monitoring area such as in an elevator car, it is possible to keep the false alarm and false alarm low.

It is the example which shows the installation aspect of the imaging part of Example 1, and the example of the image obtained by an imaging part. It is the perspective view in case a person exists in the inside and outside of the area | region 1a, and the example of the image obtained by an imaging part. It is an example of the image in the elevator car by an imaging part. This is an example of when there is a person in the elevator car and when there is no person. It is an example of 1 structure of an elevator monitoring apparatus. It is an example in the case where the consistency of the generation | occurrence | production location of a 1st characteristic part and a 2nd characteristic part is confirmed in a determination output part. This is a determination example in which no person exists in the elevator car. It is an example of the area | region setting in an elevator car. It is an example of the image acquired by installing an imaging part from a ceiling corner and a little center. It is an example of the flowchart of operation | movement of an elevator monitoring apparatus. It is an example of a cage structure map. It is an example of a 1st characteristic part detection. This is an example when the processing kernel is 3 × 3. It is an acquisition example of a template. 10 is a configuration example of Example 1 to Example 6. FIG. This is an example in which the relationship between the region 1a, the region 2a, the region 1b, the region 2b, the first feature portion 6 and the second feature portion 7 is summarized.

  Embodiments of the present invention will be described with reference to the drawings. In each drawing and each embodiment, the same or similar components are denoted by the same reference numerals, and description thereof is omitted.

[Principle explanation]
Hereinafter, an embodiment of an image monitoring apparatus of the present invention and an elevator monitoring apparatus using the same will be described. FIG. 1 illustrates an example of an installation mode of the imaging unit according to the first embodiment and an example of an image obtained by the imaging unit. FIG. 1A is a perspective view illustrating an installation mode of the imaging unit 3 according to the first embodiment. In FIG. 1A, a closed region 1a (space first region) on a certain surface such as a floor is indicated by a broken line. The region 1a (space first region) may be a flat surface or a curved surface. In addition, in the three-dimensional space, a space that is vertically above the region 1a is indicated by hatching as a region 2a (space second region). And the imaging part 3 in the space of the area | region 2a is installed so that the area | region 1a may be looked down on.

  FIG. 1B is an example of an image obtained by the imaging unit 3. A region 1b indicated by a broken line is a region where the region 1a is shown. The hatched area 2b is an area where the area 2a is shown. In this image, it is assumed that the region 2b includes the region 1b. The area 2b is called an image second area, and the area 1b is also called an image first area.

  FIG. 2 is a perspective view when a person is inside and outside the area 1a and an example of an image obtained by the imaging unit. FIG. 2A1 is a perspective view when the person 5 is outside the area 1a, and FIG. 2B1 is a perspective view when the person 5 is inside the area 1a. 2A2 is an image obtained by photographing the situation of FIG. 2A1 with the imaging unit 3, and FIG. 2B2 is an image obtained by photographing the situation of FIG. 2B1 with the imaging unit 3. . As already described, in this specification, the person 5 includes a person and a three-dimensional object other than the person, and FIG. 2 shows the case of a person. Reference numeral 6 indicated by a rectangle indicates a state in which a characteristic part indicating a part of the person 5 such as a human head is detected and held in the region 2 b, and is denoted as a first characteristic part 6. Even if the person 5 is a person other than a person, if it is a three-dimensional object, an outline having a curved line or a broken line is generated. In this case, this is the first feature section 6. 7 shown by the curve is a partial occlusion or shadow or reflection of the area 1a by the legs of the person 5 that occurs because the person 5 is in the area 1a in the area 1b. This is referred to as a second feature 7 generated by the person 5 in the region 1b. And when the 1st characteristic part 6 and the 2nd characteristic part 7 are detected (the 1st characteristic part 6 and the 2nd characteristic part 7 exist simultaneously), the person 5 exists in the area | region 1a (area | region 2a). It can be determined that it exists. Thus, if it is measured whether the 1st characteristic part 6 and the 2nd characteristic part 7 generate | occur | produce, even if it cannot connect and cut out the whole body from the head of the person 5 to a foot | leg, it is the area | region 1a (area | region) It is possible to accurately determine whether or not a person exists in 2a).

  Here, FIG. 16 summarizes the relationship among the region 1a, the region 2a, the region 1b, the region 2b, the first feature portion 6, and the second feature portion 7. The region 1a is a region existing in a three-dimensional space, and is also referred to as a space first region, and corresponds to a floor or the ground. The area 1a captured by the imaging unit 3 and converted into a two-dimensional space is the area 1b, which is also referred to as the first image area. Moreover, the 2nd characteristic part 7 is the outline which the shielding, the shadow, and reflection which a person or a solid thing produces on the area | region 1b forms. On the other hand, the region 2a is a region existing in a three-dimensional space, which is also referred to as a space second region, and corresponds to a space above the floor or the ground. The area 2a taken by the imaging unit 3 and converted into a two-dimensional space is an area 1b, which is also referred to as an image second area. Moreover, the 1st characteristic part 6 has the curve or broken line-shaped outline which a person's head and other solid objects form in the area | region 2b.

  FIG. 3 is an example of an image in the elevator car by the imaging unit. FIG. 3 is an image obtained by the imaging unit 3 that is installed with the floor 8 of the elevator car as the area 1a, the space in the elevator car as the area 2a, and the area 1a overlooking the area 2a. As the imaging unit 3, for example, a surveillance camera is used. The elevator car is composed of a floor 8, a ceiling, and four walls 9 to a wall 12. The wall 10 has a door 16 as an entrance. Here, an intersection line between the wall 9 and the wall 10 is 13, an intersection line between the wall 10 and the wall 11 is 14, and an intersection line between the wall 11 and the wall 12 is 15. The intersection line and ceiling formed by the walls 9 and 12 are outside the field of view of the image shown in FIG. In real space, the door 16 is rectangular, and the left and right sides extending in the vertical direction of the door 16 and the intersection lines 13 to 15 are in the longitudinal direction of the elevator hoistway (hereinafter referred to as hoistway direction). Although it is a parallel line segment that extends, the image shown in FIG. 3 does not appear parallel due to perspective transformation. However, the line segments described above are parallel to each other and coincide with the hoistway direction in real space. Therefore, on the image in FIG. 3, for example, a point 17 that intersects the intersection line 13 and the intersection line 14 is an infinite point in the hoistway direction and a vanishing point 17. In this specification, in the image obtained by the imaging unit 3, the vanishing point 17 is referred to as the vanishing point 17 in the longitudinal direction of the elevator hoistway.

  The region 2b may be the entire image obtained by the imaging unit 3. However, as another example, when an arbitrary half line is drawn from the vanishing point 17 toward the floor 8, a range in which the half line can pass (that is, FIG. 3) (the area excluding the wall 9 and the wall 12 from the image 3) may be the area 2b. The reason is as follows. First, the upward direction of the elevator car is restricted by the ceiling. Here, if the ceiling is the same quadrangle as the floor 8 and the imaging unit 3 is installed in one of the four corners of the ceiling, the floor corner outside the field of view is directly below the vanishing point 17 and the imaging unit. Matches above. In this case, a person standing vertically on the floor 8 including a blind spot area cannot appear in the range of the wall 9 and the wall 12 on the image shown in FIG. Therefore, in this case, the image second region (region 2b) may be a region in which the wall 9 and the wall 12 are excluded from the image of FIG. Thereby, it is possible to prevent the possibility of erroneous determination due to noise appearing on the walls 9 and 12. The first image area (area 1b) is an area of the floor 8 excluding the blind spot. As already described, the second image area (area 2b) includes the area of the floor 8.

  FIG. 4 is an example of the case where there is a person in the elevator car and the case where there is no person. 4A is an image when the person 5 is in the elevator car, and FIG. 4B is an image when the person 5 is not in the elevator car. In any image, the person 5 is present in the image, and the first feature 6 is also present in the image. However, in FIG. 4A, the second feature 7 is generated on the floor 8 because the foot of the person 5 blocks a part of the floor 8, casts a shadow on the floor 8, or reflects on the floor 8. ing. Therefore, in this case, it can be determined that the person 5 exists in the elevator car. On the other hand, in FIG. 4B, since the second feature 7 is not present on the floor 8, it can be determined that the person 5 exists outside the elevator car.

[Description of device configuration]
FIG. 5 shows a configuration example of the elevator monitoring apparatus. The elevator monitoring apparatus includes an imaging unit 3, an image processing unit 18, and a region setting tool 19. The image processing unit 18 functions as an image monitoring device, and includes a first feature detection unit 20, an in-car structure map holding unit 21, a second feature detection unit 22, and a determination output unit 23. Yes. Note that the area setting tool 19 is not always required, and when the image pickup unit 3 is installed, the car structure map may be created by connecting to the image pickup unit 3 and held in the car structure map holding unit 21. Alternatively, a memory for temporarily storing an image is provided in the image processing unit 18 to store the image acquired by the imaging unit 3, and then the region setting tool 19 is connected to the image processing unit 18 to create a car interior structure map. May be.

  Here, the car interior structure map is an image in which the range of the first image area (area 1b) and the second image area (area 2b) is determined in the image acquired by the imaging unit 3.

  In the monitoring operation of the elevator monitoring device, the image in the elevator car obtained by the imaging unit 3 is analyzed by the first feature detection unit 20 and the second feature detection unit 22.

  The first feature detection unit 20 corresponds to the region 2 b, that is, the wall 10 including the floor 8, the door 16, and the wall 11 with reference to the car internal structure map held by the car internal structure map holding unit 21. In the region, the first feature portion 6 indicating a part of the person 5 is detected, and the change in the position of the first feature portion 6 is tracked in the continuous image series sequentially sent from the imaging unit 3. As the first characteristic part 6 of the person 5, a curved line part or a broken line part of a contour line can be used. This is because most of the outlines of the light and the door edge to be opened and closed are formed by simple straight lines, whereas the outline of the person 5 which is a three-dimensional object on the floor 8 includes a curved part or a broken line part. As the first feature, for example, a part of the person 5 such as the head and shoulders may be detected, and it is not necessary to detect the whole body. Therefore, the detection accuracy is improved as compared with the method that must always detect the whole body. be able to.

  The second feature detection unit 22 also refers to the in-car structure map, and the area 1b, that is, a trace corresponding to the person 5 on the floor 8, for example, the floor generated when the person 5 exists on the floor 8 is displayed. 8. Detect second feature 7 indicating 8 partial occlusions or shadows or reflections. More specifically, the difference between the luminance distribution or the contour of the image obtained by the imaging unit 3 having the image of the floor 8 taken as a reference when it is guaranteed that the person 5 is not present is obtained. The second feature 7 is assumed. In addition, when a contour is used as the second feature portion 7, it is difficult to be affected by a change in luminance of the image. It can be said that the 2nd characteristic part 7 shows traces, such as a leg of the person 5 in the area | region 1b. Since the second feature section 7 does not need to be detected as a continuous area with the first feature section 6, the detection accuracy can be increased as compared with a method in which the whole body must be detected.

  The car structure map holding unit 21 may be incorporated in the first feature detection unit 20 or the second feature detection unit 22.

  The determination output unit 23 confirms the presence or absence of the first feature 6 of the person 5 detected by the first feature detector 20 and the second feature 7 detected by the second feature detector 22. If there is a person at the same time, it is determined that there is a person in the elevator car, and if there is no person at the same time, it is determined that there is no person in the elevator car. Then, a manned signal or an unmanned signal is output as a determination result.

  In addition to the condition that the first feature 6 and the second feature 7 are detected (that is, both exist simultaneously), the determination output unit 23 further confirms the consistency of these occurrence locations. You may make it perform.

  FIG. 6 is an example in the case where the determination output unit confirms the consistency between the occurrence locations of the first feature portion and the second feature portion. In the determination output unit 23, the confirmation of the consistency of the occurrence locations of the first feature unit 6 and the second feature unit 7 is performed as follows. For example, as shown in FIG. 6A and FIG. 6B, the second feature within the range of a line segment 24 connecting the vanishing point 17 from the center of the first feature 6 by the first feature detector 20. When the second feature 7 by the feature detector 22 is present, it can be said that the second feature 7 corresponds to the first feature 6, and the positions of both are matched. It can be determined that there is a person in the elevator car. Here, the line segment 24 is a line segment that indicates the direction in which the hoistway extends from the first feature 6, that is, the vertical direction in the real space. Here, the line segment 24 connecting the vanishing point 17 from the center of the first feature portion 6 is assumed here, but the vanishing point is not limited to the center of the first feature portion 6 but from any place of the first feature portion 6. You may judge using the line segment 24 which connected 17. FIG.

  FIG. 7 is a determination example in which a person does not exist in the elevator car. In the example of FIG. 7A, a reflection 26 caused by illumination 25 such as solar radiation is generated outside the range of a line segment 24 connecting the first feature 6 and the vanishing point 17, and this reflection 26 is used as the second feature 26. Detected. In this case, since the first feature portion 6 and the second feature portion 26 do not correspond and are not matched, it can be determined that the person 5 does not exist in the car. In the example of FIG. 7B, even if the wall 11 is locally brightened by the illumination 27 such as solar radiation and the first feature portion 28 is detected, the second feature portion 7 is not in the range of the line segment 24. Therefore, since the features detected by the first feature detection unit 20 and the second feature detection unit 22 do not correspond to each other, they do not match. Therefore, it can be determined that the person 5 does not exist in the elevator car.

  In the case of an elevator that does not have solar radiation from outside into the elevator car, confirmation of consistency by the determination output unit 23 can be omitted. Under such circumstances, when the second characteristic portion 7 is detected, it can be determined with certainty that the person is in the elevator car. In this case, the determination as to whether or not it is within the range of the line segment 24 as described above is omitted, the first feature 6 is detected, and the second feature 7 is present in the area of the floor 8. For example, it can be determined that the person 5 exists in the elevator car. That is, if the first feature 6 and the second feature 7 are detected, it is determined that the person 5 exists in the elevator car regardless of the positional relationship between the first feature 6 and the second feature 7. Can do.

  In this case, the indicator of the presence / absence of the second feature 7 is calculated by calculating the luminance difference for each pixel between the reference image and the current image in the area of the floor 8, and the difference is greater than a predetermined threshold value. When the number of certain pixels is counted and the count value exceeds a predetermined threshold value, it can be determined that the second feature portion 7 is present.

  Further, as an indicator of the presence or absence of the other second feature portion 7, the second feature is obtained by counting the amount of the contour detected from the current image and not in the reference image, and exceeding a predetermined threshold value. It can be assumed that there is a part 7. The amount of the contour is, for example, a region corresponding to the edge by binarizing after emphasizing the edge (also called the contour) by a difference or differentiation between pixels such as a Sobel filter with respect to the reference image and the current image. Thus, the number of pixels exceeding the binarization threshold can be used as the contour amount.

  Thus, the elevator monitoring apparatus of the present invention is necessary to determine that the person 5 exists in the elevator car when the first feature 6 exists and the second feature 7 occurs. As a condition. For this reason, since the detection method of the second feature part 7 can be simplified and easier to detect than the method of simply determining the presence or absence of a person based on whether or not a foot is detected on the floor, the false alarm can be reduced, It can be operated well. In addition, as a result of simplifying the condition for identifying the foot candidate as described above using the number of pixels of the edge as an index, an error of the second feature portion 7 with respect to noise such as slight reflection on the floor 8 is obtained. Even if there is detection (overdetection), the final determination is made together with the first feature section 6, so that misreporting due to erroneous detection (overdetection) of the person 5 can be suppressed.

  Further, as described above, if the first feature portion 6 and the second feature portion 7 are positioned with respect to the vanishing point 17, misreporting can be suppressed even for strong reflection due to solar radiation or the like. .

  Further, as shown in FIG. 15, the elevator monitoring apparatus according to the first embodiment may have an elevator control unit (not shown) that receives the determination result from the determination output unit 23 and controls the elevator.

  Next, the region setting tool 19 having the point instruction unit 29 and the map creation unit 30 shown in FIG. 5 will be described. The area setting tool 19 need not always be incorporated in the elevator monitoring apparatus of the present invention. For example, a monitor (not shown) such as a handy personal computer and an input interface (not shown) such as a key input device or a pointing device. ), An arithmetic unit (not shown), and a memory (not shown).

  FIG. 8 is an example of area setting in the elevator car. After installing the imaging unit 3 in the elevator car, the area setting tool 19 is connected to the imaging unit 3 at least once. Then, for example, an image in the elevator car as shown in FIG. 8 is obtained and displayed on the monitor of the area setting tool 19. Then, using the input interface of the point instruction unit 29, the points 31 to 34, which are the four corners of the door 16, are sequentially specified, and the coordinate values of each point in the image are acquired in the memory of the map creation unit 30. Here, although the door 16 is originally rectangular, it is perspective-transformed by the imaging action of the imaging unit 3, and the width becomes narrower toward the lower part of the hoistway as described above. Next, an arithmetic unit of the map creation unit 30 calculates an intersection of a straight line connecting the point 31 and the point 34 and a straight line connecting the point 32 and the point 33, and this is the vanishing point 17.

  FIG. 9 is an example of an image obtained by installing an imaging unit slightly from the ceiling corner and slightly in the center.

  In the example of FIG. 8, the ceiling of the elevator is square or rectangular, and the imaging unit 3 is installed at a corner formed by the wall 12 facing the wall 10 with the door 16, the adjacent wall 9, and the ceiling. . In such a case, since at least three corners of the floor having the same shape as the ceiling are in the field of view of the image, these three points are set as points 35 to 37, respectively, and sequentially using the input interface of the point indicating unit 29. Instruct and get to memory. Thus, in the image acquired by installing the imaging unit 3 at the ceiling corner of the rectangular parallelepiped elevator car, the corner of the floor that is out of the image field of view coincides with the vanishing point 17. This is shown in FIG. That is, the fourth corner deviating from the image field coincides with the vanishing point 17. A region 40 indicated by a dot pattern is a blind spot region 40 of the imaging unit 3. The blind spot area 40 is enlarged more than actual by the perspective projection conversion. Further, a point where a straight line connecting the vanishing point 17 and the point 35 intersects with the frame of the image field of view is defined as 41, and a point where a straight line connecting the vanishing point 17 and the point 37 intersects with the frame of the image field of view is denoted as 42. By the above setting process, the area inside the corner formed by the point 41, the vanishing point 17, and the point 42 and excluding the blind spot area 40 outside the image becomes the area 2b, and the area of the floor 8 becomes the area 1b. The area 2b includes the area of the floor 8. The area of the floor 8 is an area excluding the blind spot area 40 outside the image in the square formed by the vanishing point 17 and the points 35 to 37. As described above, the wall 9 and the wall 12 are out of the region 2b.

  Further, when an image is obtained by installing the imaging unit 3 at a position moved in the central direction from the corner of the ceiling, an image is obtained as shown in FIG. In this case, the vanishing point 17 calculated from the points 31 to 34 does not coincide with one of the floor corners in the blind spot area 40. Therefore, it is necessary to make it possible to configure the position of the point 43 indicating one of the floor corners in the blind spot area 40 outside the field of view of the image. In this case, the position of the point 43 must be adjusted so that the line segment connecting the point 35 and the point 43 matches the floor edge 38 and the line segment connecting the point 37 and the point 43 must match the floor edge 39. . Then, similarly to the case described with reference to FIG. 9A, the region 2b and the region 1b are determined. However, in this case, it is difficult to exclude the area of the wall 9 and the wall 12 using the vanishing point 17, so the area 2b is the entire screen.

  In the actual imaging unit 3, the drawing as shown in FIG. 3 or FIG. 9 does not completely match due to barrel distortion of the lens. However, such inconsistencies can be eliminated by performing image processing after correcting barrel distortion. However, in many cases used in the elevator monitoring apparatus of the present embodiment, the influence of the deviation from the drawing caused by the barrel distortion on the detection accuracy is not so great, so there is no problem even if the drawing is slightly shifted due to the barrel distortion. .

  A car internal structure map is created by the action of the area setting tool 19 described above. After creating the car structure map, the area setting tool 19 can be temporarily connected to the car structure map holding unit 21 and stored. For the first feature detection unit 20 and the second feature detection unit 22, regions other than the region 2b and the region 1b are regions that are not processed. Therefore, even if a contour based on the incident light is generated in the wall 9 and the wall 12 that do not belong to any of these, the first feature detection unit 20, the second feature detection unit 22, and the determination output unit 23 are not affected. And no false detection occurs.

[Description of processing flow and example of car interior structure map]
FIG. 10 is an example of a flowchart of the operation of the elevator monitoring apparatus. The operation of the elevator monitoring apparatus of the present embodiment having the above-described configuration will be described with reference to the flowchart of FIG. An example of the elevator monitoring apparatus of the present embodiment starts processing upon power-on and performs memory initialization processing (step s40). Next, the car internal structure map is created (step s41). This is done by downloading the coordinate data of each point (points 31 to 37, vanishing point 17, point 41, point 42) created by the area setting tool 19. The in-car structure map may be created based on the above, or the in-car structure map created by the area setting tool 19 may be downloaded. The car structure map can be, for example, a digital image shown in FIG. Here, for example, 0 is assigned to the plain white area, 196 is assigned to the hatched area, and 255 is assigned to the dot area. The 255 area is the area 1b, and 196 or more areas (255 (Including the region) is the region 2b.

  The above steps are pre-processing steps, and in actual operation, the operations from step s42 to step s48 are repeated for each moving image cycle, and the process ends only when a power off or end interrupt is detected in step s49. Otherwise, return to image acquisition (step s42). The first feature detection (step s43) and the second feature detection (step s44) are performed by the first feature detection unit 20 and the second feature detection unit 22, respectively. Either can be designed to be performed first. After the first feature part detection and the second feature part detection are finished, the judgment output unit 23 comprehensively determines whether the person is manned based on the detection results of the first feature part 6 and the second feature part 7, and the first feature 'True' if both the part 6 and the second characteristic part 7 are detected, or if it is determined to be manned by verifying the positional consistency of both using the vanishing point 17 as necessary. As a manned determination (step s47). On the other hand, if the first feature 6 or the second feature 7 is not detected, or if both are detected as necessary and the positional alignment is not verified and determined to be unmatched, it is determined to be unattended An unattended determination is made (step s46). The determination output unit 23 outputs the result of the previous step s47 or step s46 in the determination output (step s48).

[Detection example of the first feature and the second feature]
Next, an example of detecting a person in the first feature detection unit 20 and holding the first feature will be described. FIG. 12 is an example of first feature detection, and is a conceptual diagram illustrating an example of person detection. 12 (a1), FIG. 12 (a2), and FIG. 12 (a3) are examples of time-series digital images when persons 44A, 44B, and 44C board the elevator entrance, and are illustrated in FIG. 12 (a1). Time elapses toward 12 (a3). Although the person appearing in FIG. 12A1 to FIG. 12A3 is the same person, the luminance pattern changes with the passage of time, so that they are named 44A, 44B, and 44C for convenience of explanation.

  FIG. 12B1 is an image composed of the absolute value of the luminance difference for each pixel of the image (also referred to as a frame) in FIG. 12A1 and the image in FIG. 12A2, and is a so-called interframe difference image. A relationship between a person 44A indicated by a broken line, a person 44B indicated by a solid line, and an inter-frame difference area (lattice pattern area) is shown. Similarly, FIG. 12 (b2) is an inter-frame difference image between the image of FIG. 12 (a2) and the image of FIG. 12 (a3), and the inter-frame difference between the person 44B indicated by a broken line and the person 44C indicated by a solid line. The relationship of the area | region (grid pattern area | region) is shown. 12 (b1) and 12 (b2) may be gray scale images, or after binarization with an appropriate threshold value. As an appropriate threshold value, the person shown in FIGS. 12 (b1) and 12 (b2) does not necessarily need to be detected by connecting the whole body. However, it is better to set the threshold value as a value that can be detected by solidifying to the extent that the contour of the curve or broken line of the head, shoulders, waist, etc. is too divided to be recognized as a curved line or broken line. . Hereinafter, this is expressed as a threshold value that is detected without being subdivided inappropriately. If the difference in brightness is poor in the pattern inside the person's outline, a hole may be formed in the binarized patterns of FIG. 12 (b1) and FIG. 12 (b2). There is no effect on the detection of.

  The extraction pattern 44D in FIG. 12 (c1) and the extraction pattern 44E in FIG. 12 (c2) are respectively expanded or extracted into the extraction patterns (areas indicated by the lattice pattern areas) in FIG. 12 (b1) and FIG. 12 (b2). The maximum value processing is added. If FIG. 12B1 and FIG. 12B2 are after binarization, expansion processing is performed, and in the case of a grayscale image, maximum value processing is performed.

  FIG. 13 shows an example when the processing kernel is 3 × 3. The expansion process is performed with reference to adjacent pixels using a processing kernel as shown in FIG. 13, for example. FIG. 13 shows a case where the processing kernel is 3 × 3, and a pixel adjacent to the target pixel C (pixels assigned 1 to 8) is referred to the original image and converted according to these values. The value of the target pixel C in the image is determined. In the case of so-called four-connected dilation processing, the target pixel C is 0 (non-object), and any one of the pixels numbered 2, 4, 6, or 8 adjacent to the target pixel C is 255 (object). If there is, there is a process of replacing the target pixel C with 255 (object). In the case of so-called 8-connected expansion processing, the target pixel C is 0 (non-object), and any one of 1, 2, 3, 4, 5, 6, 7, or 8 adjacent thereto is assigned. If at least one pixel has 255 (object), the target pixel C is replaced with 255 (object).

  The maximum value processing is also performed with reference to adjacent pixels as shown in FIG. In the case of so-called four-connected maximum value processing, conversion processing is performed in which the maximum luminance value of the pixel of interest C and any of the pixels numbered 2, 4, 6, or 8 adjacent thereto is the value of the pixel of interest C. . Further, in the case of so-called 8-connected maximum value processing, attention is paid to the maximum value among the target pixel C and pixels numbered 1, 2, 3, 4, 5, 6, 7, or 8 adjacent thereto. This is a conversion process for setting the value of the pixel C. The brightness value of the object and its surroundings is greatly expanded by the expansion process and the maximum value process.

  Note that the pixel 255 (object) means a pixel where some object exists, and is a person candidate.

  Subsequent to expansion processing or maximum value processing, inter-image logical product processing or inter-image minimum value processing is performed.

  When the images shown in FIGS. 12 (c1) and 12 (c2) are binarized, inter-image logical product processing is performed. When the images are grayscale images, inter-image minimum value processing is performed. The inter-image logical product is defined as 255 (object) only when the corresponding pixels in FIGS. 12 (c1) and 12 (c2) are both 255 (object), and 0 (non-object) otherwise. In the process, the output image is a binary image. The inter-image minimum value process is a process of converting with the smaller value of the luminance of the corresponding pixel in FIGS. 12C1 and 12C2, and the output image is grayscale. In this case, it is further necessary to binarize with an appropriate threshold value. Here, the appropriate threshold value may be a threshold value obtained by a discriminant analysis method, or may select a threshold value that is detected without being subdivided to an inappropriate degree as described above.

  The lattice pattern region in FIG. 12D1 is a binarized image after inter-image logical product processing or inter-image minimum value processing of the extraction pattern 44D in FIG. 12C1 and the extraction pattern 44E in FIG. . An appropriate threshold value for creating the lattice pattern region of FIG. 12 (d1) is such that the object shown in FIG. 12 (d1) is not subdivided to an inappropriate level and is too low to be buried in noise. It is preferable to use a value, for example, it may be selected by discriminant analysis. With the above processing, the output is a binary image, and then label processing is performed.

  The label processing is processing for a binarized image, and is processing for attaching one label by regarding a pixel group satisfying the 4-connected or 8-connected condition as one object. A circumscribed rectangle circumscribing the object (person) can be drawn by the label processing. A rectangle 45 in FIG. 12D2 is a circumscribed rectangle of the object (person) detected as a result of the series of processes described above. As shown in FIG. 12D3, the detected object (person) is the object shown in FIG. 12A2 that is an intermediate image among the three time-series images shown in FIGS. 12A1 to 12A3. It almost coincides with the position of (person).

  As described above, an example in which the first feature detection unit 20 detects the human candidate having the first feature has been described. Next, a part having a curved line or a polygonal outline such as a head, a shoulder, and a waist is detected from the detected person candidates. Further, in order to hold the first feature portion over the progressing time-series image, a pattern matching technique using the first feature portion as a template is used. As a pattern matching technique, there are a technique using a well-known correlation calculation (normalized correlation method) and a technique for searching so that the sum of absolute differences of corresponding pixel values is minimized.

  FIG. 14 is an example of obtaining a template. FIG. 14A shows an example in which a rectangle 46 (template acquisition range), which is a region to be detected and held as a first feature portion, is detected in a person having a circumscribed rectangle 45. An image acquired as a template from the time-series image shown in FIG. 14B based on the coordinates of the rectangle 46 is a partial image shown by a broken-line rectangle 46 in FIG. The method of determining the rectangle 46 may be determined in advance as a predetermined area of the upper portion of the circumscribed rectangle 45 in order to detect the head, for example. Or you may make it select the area | region where the outline of a person concentrates densely. The person 44B in FIG. 14B is shown as a lattice pattern for convenience of expression, but is actually a grayscale image due to unevenness and color of the subject. Therefore, when a well-known contour detection process is performed, for example, a contour having many curved portions or broken line portions as shown in FIG. 14C is extracted. You may make it acquire the image of the magnitude | size decided beforehand centering on the part with this dense outline as a template. The images in the elevator car are generally taken from above with the imaging unit 3 and illuminated from above, so that the head is generally finer and has better contrast than the feet, and thus the contour is also dense. Easy to extract. As a filter for detecting the contour, for example, there is a Sobel filter shown in FIG. When these filters are sequentially applied and the logical sum of the obtained contours is taken, the contours of the grayscale image can be obtained as eight approximate contours in 360-degree directions at 45-degree intervals.

  Since the rectangle 46 corresponds to the first feature part, the first feature part detection unit 20 displays position information such as the presence or absence of the detection of the first feature part 46 indicating a part of the person and the coordinates of the first feature part 46. Output. The specific example of the process performed by the first feature detection unit 20 has been described above by taking the person who is about to board as an example.

  As for the processing of the second feature detection unit 22, the image first region (region 1b), that is, the range of the floor 8, by using, for example, a Sobel filter or the like, by shielding, shadowing, reflection, etc. caused by the presence of a person. Detect the contour that appears. This is the second feature. The second feature detection unit 22 outputs position information such as the presence or absence of detection of the second feature and the coordinates of the second feature.

  Next, a second embodiment of the elevator monitoring device will be described with reference to FIG. The elevator monitoring apparatus according to the second embodiment includes an imaging unit 3, an image processing unit 18, an elevator control unit 47, and an in-car operation panel 48.

  There are cases where the user wants to contact the elevator manager directly due to an unexpected situation such as an elevator failure or an earthquake. In such a case, the user in the elevator car can talk directly with the elevator manager by pressing the emergency call button on the operation panel 48 in the car. However, in many cases, the emergency call button is often caused by a user's erroneous operation. A user who accidentally presses an emergency call button often goes out of the elevator car as it is. Then, the elevator manager must call into the elevator car for a certain period of time even though the elevator car is unattended. This significantly reduces the work efficiency of the elevator manager.

  Thus, in the second embodiment, when the elevator control unit 47 obtains an unmanned determination output from the image processing unit 18 after the emergency call button is pressed, the elevator control unit 47 is unmanned in the elevator car. This is to notify the elevator manager.

  With such a configuration, it is possible to immediately cancel the erroneous operation of the emergency call button, and it is possible to prevent the work efficiency of the elevator manager from being lowered.

  Next, an elevator monitoring apparatus according to a third embodiment will be described with reference to FIG. The elevator monitoring apparatus according to the third embodiment includes an imaging unit 3, an image processing unit 18, an elevator control unit 47, an in-car operation panel 48, and a diagnostic operation management unit 49.

  Generally, an elevator having a remote diagnosis function has a portion corresponding to the diagnostic operation management unit 49 and a portion corresponding to the operation panel 48 in the car, and there is no button press operation from the elevator car for, for example, 30 minutes or more. In addition, it performs automatic driving for diagnosis, collects diagnostic information by diagnosing brake performance and measuring vibration. However, an elevator with sporadic users regardless of day and night, such as a residential building in Tokyo, has a problem that automatic operation for diagnosis cannot be performed.

  Therefore, in the third embodiment, when the elevator control unit 47 obtains an unmanned signal in the car from the image processing unit 18, the diagnosis operation management unit is in a period in which automatic operation for diagnosis is in accordance with the schedule. 49 is configured to perform an automatic operation for diagnosis in accordance with an automatic diagnosis operation procedure held by 49.

  By adopting such a configuration, it is possible to perform diagnostic operation by effectively using an unmanned period in the car. Also, if there is any operation on the car operation panel 48 at the moment of entering the automatic driving for diagnosis or during the automatic driving, it means that there is a person in the elevator car. It can also be configured to stop immediately.

  Next, a fourth embodiment of the elevator monitoring device will be described with reference to FIG. The elevator monitoring apparatus according to the fourth embodiment includes an imaging unit 3, an image processing unit 18, and an elevator control unit 47.

  When a user gets off at a destination floor, a general elevator closes a door after a predetermined time such as 15 seconds, and responds to calls from other floors.

  Therefore, in the fourth embodiment, when the elevator control unit 47 obtains a signal from the image processing unit 18 that the elevator car is unmanned, the elevator control unit 47 promptly closes the door without waiting for a predetermined period of time. If there is a floor call, it responds immediately and moves to the destination floor.

  With such a configuration, when the user gets off and the elevator car becomes unattended, it is possible to immediately respond to the next call, which can contribute to increasing the efficiency of elevator operation.

  Further, when the elevator control unit 47 obtains an unmanned signal from the image processing unit 18, the elevator car is unmanned and it is not necessary to consider the ride comfort of the passenger. You may make it brake and go to the destination floor. As a result, there is an effect that a call can be answered promptly.

  Next, an elevator monitoring apparatus according to a fifth embodiment will be described with reference to FIG. The elevator monitoring apparatus according to the fifth embodiment includes an imaging unit 3, an image processing unit 18, and an elevator control unit 47.

  In general elevators, when a user gets off at the destination floor, the door is closed after a predetermined time such as 15 seconds, and then the elevator car is turned off if there is no call from the user such as for 3 minutes or more. To save power.

  Therefore, in the fifth embodiment, when the elevator control unit 47 receives an unmanned signal from the image processing unit 18, the elevator control unit 47 can control to turn off the lights in the elevator car without waiting for a predetermined time. By performing such an operation, useless illumination energy can be saved.

  Next, an elevator monitoring apparatus according to a sixth embodiment will be described with reference to FIG. The elevator monitoring apparatus according to the sixth embodiment includes an imaging unit 3, an image processing unit 18, an elevator control unit 47, an in-car operation panel 48, and a destination floor registration holding unit 50.

  When a user sets a plurality of destination floors by pressing a button on the in-car operation panel 48 and then gets off halfway, a general elevator will use the destination floor set on the in-car operation panel 48 to answer the call. It stops at and opens and closes the door.

  Therefore, in the fifth embodiment, when the elevator control unit 47 receives the unmanned signal from the image processing unit 18, all the destination floors registered in the destination floor registration holding unit 50 set by the operation of the in-car operation panel 48 are displayed. Cancel the process. With this operation, it is possible to quickly respond to a call response by reducing the time required for stopping at an unnecessary floor and opening and closing the door.

  As mentioned above, although the Example of this invention has been described, the structure demonstrated by each Example so far is an example to the last, and this invention can be suitably changed within the range which does not deviate from a technical idea. Further, the configurations described in the respective embodiments may be used in combination as long as they do not contradict each other.

1a ... area 1a (space first area)
2a ... area 2a (space second area)
3 ... Imaging unit 1b ... Area 1b (first image area)
2b ... area 2b (second image area)
5 ... Person 6 ... First feature 7 ... Second feature 8 ... Floor 9-12 ... Walls 13-15 ... Intersection 16 ... Door 17 ... Vanishing point 18 ... Image processor 19 ... Area setting tool 20 ... First Feature part detection unit 21 ... Car interior structure map holding part 22 ... Second feature part detection part 23 ... Determination output part 24 ... Line segment 25 ... Illumination such as solar radiation 26 ... Reflection, second feature part 27 ... Illumination 28 such as solar radiation ... first feature part, solar radiation 29 ... point instruction part 30 ... map creation parts 31 to 34 ... door four corner points 35 to 37 ... floor points 38 and 39 ... floor edge 40 ... blind spot areas 41 and 42 ... intersect with the frame of the image field of view Point 43 ... Points 44A, 44B, 44C assumed outside the field of view ... People 44D, 44E ... Extraction pattern 45 ... circumscribed rectangle 46 ... Template acquisition range (first feature)
47 ... Elevator control unit 48 ... In-car operation panel 49 ... Diagnostic operation management unit 50 ... Destination floor registration holding unit

Claims (12)

A closed area on a certain plane is defined as a first space area, a space vertically above the first space area is defined as a second space area, and the first space area is installed in the second space area. In an image monitoring apparatus that processes an image obtained by the imaging unit,
A first feature detection unit that detects a first feature that indicates a part of a person in the second image region when the second space region reflected in the image is the second image region;
When the first space area shown in the image is the first image area, the second feature portion indicates shielding or shadow or reflection of the first space area caused by the person in the first image area. A second feature detection unit for detecting
Based on the detection results of the first feature detection unit and the second feature detection unit, it is determined whether the person is present in the second space area or whether the person is unattended. An image monitoring apparatus comprising: a determination output unit that determines that a person is manned on the condition that at least the first feature unit and the second feature unit are detected. The image monitoring apparatus according to claim 1, the imaging unit, and an elevator control unit that inputs an evaluation result in the determination output unit and controls the elevator,
The elevator, wherein the second space area is a space in an elevator car, the first space area is a floor of the elevator car, and the imaging unit is a monitoring camera provided in the elevator car. Monitoring device.   The elevator monitoring apparatus according to claim 2, wherein the determination output unit determines that the interior of the elevator car is unmanned when the first feature unit or the second feature unit is not detected.   In the case where the first feature portion and the second feature portion are detected, the determination output portion is configured so that the inside of the elevator car is not related to the positional relationship between the first feature portion and the second feature portion. The elevator monitoring device according to claim 2, wherein the elevator monitoring device is determined to be manned. The decision output unit, said a first feature and the second feature is detected and the position of the first feature, extending in the longitudinal direction of the hoistway of the elevator, mutually parallel in the real space If the second feature portion is within the range of the line segment connecting the vanishing point, which is a point where the continuous line segment intersects on the image , determining that the elevator car is manned The elevator monitoring apparatus according to claim 2 or 3, characterized by the above. The decision output unit, wherein at the first feature and the second feature is detected, the position of the first feature, extending in the longitudinal direction of the hoistway of the elevator, mutually parallel in the real space If the second feature does not exist within the range of the line segment connecting the vanishing point, which is a point that intersects the image , the elevator car is determined to be unmanned. The elevator monitoring apparatus according to claim 5, wherein In the image obtained by the imaging section, extending in the longitudinal direction of the hoistway of the elevator, parallel line segments to each other in the real space is toward the image first region from the vanishing point is the point of intersection on said image The elevator monitoring device according to any one of claims 2 to 6, wherein when the arbitrary half line is drawn, a range in which the half line can pass is set as the second image area. The elevator car has an emergency operation panel with an emergency call button,
When the elevator user gets off after pressing the emergency call button and the determination output unit determines that the elevator car is unmanned, the elevator car is unmanned. The elevator monitoring apparatus according to any one of claims 2 to 7, wherein the elevator administrator is notified of the fact that A diagnostic operation management unit for the elevator,
The elevator control unit, when the determination output unit determines that the elevator car is unmanned, performs an automatic operation for diagnosis according to an automatic diagnosis operation procedure held by the diagnosis operation management unit. The elevator monitoring device according to any one of claims 2 to 8.   If the determination output unit determines that the elevator car is unmanned, the elevator control unit promptly closes the elevator door and responds to a call from another floor to the destination floor. The elevator monitoring apparatus according to any one of claims 2 to 9, wherein the elevator monitoring apparatus moves.   The said elevator control part turns off the illumination in the said elevator car, when the said determination output part determines that the inside of the said elevator car is unattended. Elevator monitoring device. An in-car operation panel provided in the elevator car, and a destination floor registration holding unit in which a destination floor set by operation of the in-car operation panel is registered,
When the determination output unit determines that the elevator car is unmanned, the elevator control unit determines a destination floor set by operating the in-car operation panel registered in the destination floor registration holding unit. The elevator monitoring device according to any one of claims 2 to 11, wherein all of them are canceled.