JP3736989B2 - Image monitoring device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image monitoring apparatus that detects an intruder in a monitoring area by shooting the monitoring area and processing the captured image.
[0002]
[Prior art]
Conventionally, there is an image monitoring apparatus having a function of detecting an intruder in a monitoring area. In this image monitoring apparatus, the monitoring area is continuously captured by the imaging unit, the current image obtained by capturing the latest situation is compared with the reference image stored in advance by the image processing unit, and the change is made between the two images. Extract the region where there is. When the change area is larger than a predetermined size and moves a predetermined distance or more, it is determined that the change area between the images is caused by an intruder in the monitoring area.
[0003]
When detecting an intruder, the image monitoring device stores that an intrusion abnormality has occurred as security state information, and stores an image (abnormal image) at the time of occurrence of the abnormality. When there is a call from the outside such as a controller or a control center, the image monitoring apparatus outputs the stored state information and abnormal image to the outside.
A specific example of intruder detection logic in the image monitoring apparatus will be described with reference to FIG.
[0004]
(A) shows a reference image 11 in which a monitoring area is captured in advance by an imaging unit and stored in a memory. The reference image 11 includes a floor 12, a side wall 13, a front wall 14 and a window 15. (B) shows the latest image (current image) 16 taken by the imaging means, and the intruder 17 is shown. Further, the current image 16 in (B) is an image when the intruder 17 appears (first appearance) in the monitoring area for the first time. (C) is a difference image 18 obtained by taking a difference for each pixel between the reference image 11 and the current image 16, and the intruder 17 is taken out as a change area 19. The position of the change area 19 is set as the first appearance position 20.
[0005]
For example, the image monitoring apparatus captures a monitoring area at intervals of 0.5 seconds and acquires a current image. (D) is the current image 16 taken after a lapse of time from the current image 16 in (B), and the intruder 17 has moved. (E) shows a difference image 18 between the reference image 11 and the current image 16. The difference between the current position 22 and the first appearance position 20 of the current change area 19 is the movement distance 23.
[0006]
In the difference image 18 of (E), when the intruder 17 is in the monitoring area, the change area 19 appears in the current image 16 with a size larger than a predetermined size, and the change area 19 also moves as the intruder 17 moves. Move. Therefore, the image processing means of the image monitoring apparatus determines that the intruder 17 is present in the monitoring area when the size of the change area 19 is within a predetermined range and the moving distance 23 exceeds a predetermined value.
[0007]
In the intruder detection logic described above, when light enters the monitoring area, the portion where the light is inserted becomes a high-luminance area (hereinafter referred to as “light area”) on the captured image. When this light region is generated or disappears, the change region 19 is detected and detected as an intruder.
FIG. 2 is a diagram for explaining the reason why an optical area is erroneously determined as an intruder.
[0008]
2A shows the reference image 11, FIG. 2B shows the current image 16, and FIG. 2C shows the difference image 18 between the reference image 11 and the current image 16. In the current image 16, there is a light region 24 that is generated when sunlight enters through the window 5. When the reference image 11 and the current image 16 are compared, the light region 24 is detected as the change region 19 in the difference image 18. Further, when the sunlight disappears from the state in which the sunlight has been inserted, the change region 19 is similarly detected. In the conventional image monitoring device, if the change area 19 moves with the movement of the sun, this is detected as an intruder.
[0009]
Further, when a person passes outside the monitoring area and is projected onto the light area 24, and a shadow moves within the light area 24, this shadow is also detected as a moving change area. Therefore, also in this case, the shadow is detected as an intruder.
As described above, when light or a shadow is detected as an intruder, the image monitoring apparatus will falsely notify the controller or control center every time the light is inserted or a person passes outside the monitoring area. On the other hand, even in the conventional image monitoring apparatus, these lights or shadows and intruders are identified to prevent false alarms.
[0010]
If the incident light is light, a floor pattern appears in the light region 24 as shown in the current image 16 of FIG. Therefore, the correlation in the change area 19 between the reference image 11 and the current image 16 is high, or edge information is stored. Therefore, in the prior art, an intruder and light or shadow are identified based on the correlation value of the change area 19 or the storage state of the edge information, thereby preventing erroneous determination.
[0011]
[Problems to be solved by the invention]
In the conventional image monitoring apparatus, even if it is attempted to remove the influence of light (light removal) or the influence of shadow (shadow removal) by the correlation value of the change area or the storage state of the edge information as described above, the monitoring area If strong light such as sunlight is incident on the camera, a false alarm will occur.
[0012]
FIG. 3 illustrates the reason why a light or shadow is erroneously determined as an intruder when strong light such as sunlight is incident even if light removal and shadow removal are attempted according to the correlation value of the change area or the storage state of the edge information. It is a figure to do.
3A shows the reference image 11, (B) shows the current image 16, and (C) shows the difference image 18. When sunlight enters the monitoring area, a light area 24 with high brightness is generated on the current image 16. In addition, since the brightness of the light area 24 is very high compared to other areas of normal brightness, the pattern on the floor 12 in the reference image 11 cannot be detected. Therefore, the correlation value of the change area 19 between the reference image 11 and the current image 16 becomes low, and the edge information is not stored. Therefore, if the light region 24 moves, it will be detected as an intruder.
[0013]
Further, if a shadow moves in the light region 24 due to a person passing outside the window or the movement of clouds, the movement of the shadow is detected as an intruder and a false alarm is generated.
The present invention eliminates false detection caused by light or shadow in a monitoring area and accurately detects the intruder in an image monitoring apparatus that detects an intruder in the monitoring area by processing an image of the monitoring area. The purpose is to be able to.
[0014]
[Means for Solving the Problems]
The present invention has been made to achieve the above object.
The image monitoring apparatus of the present invention includes an imaging unit that captures a monitoring area, and an image processing unit that detects an intruder in the monitoring area by processing the captured image. extracting the light areas of high brightness from the image, as a logic for detecting an intruder to identify the light or shadow in the optical domain, using an intruder detection logic and different logic in the area of other normal intensity .
[0015]
The image monitoring apparatus of the present invention detects an intruder by comparing a reference image with a current image and extracting a change area by an image processing unit, as in the prior art. Here, when strong light such as sunlight is inserted into the monitoring area, the brightness of the light area corresponding to the portion where the light is inserted is markedly higher than the normal area in the captured image.
[0016]
The image monitoring apparatus of the present invention extracts in advance a light region having a predetermined luminance or higher by using the luminance difference in the captured image. Then, the intruder detection logic in the light region is made different from the intruder detection logic in other normal luminance regions. Accordingly, strong light such as sunlight entering from the window or a shadow caused by this light is identified as an intruder, and light or shadow is prevented from being detected as an intruder. Intruders are reliably detected by normal detection logic.
[0017]
In the image monitoring apparatus of the present invention, the light removal process is performed so that the light area is not detected as an intruder as the detection logic in the light area. The image processing means extracts a region where the luminance changes from the comparison between the captured images, and when this changed region occurs in the light region or includes the light region, the sensitivity for detecting an intruder is obtained. Is lower than the sensitivity in the normal luminance region. For example, an intruder is detected when the change area moves by a set distance, but the set movement distance in the light area is set longer than the set movement distance in the normal luminance area.
[0018]
In the present invention, as detection logic in the light region, shadow removal processing is performed in order not to erroneously detect a shadow in the light region as an intruder. The image processing means extracts a region where the luminance changes from the comparison of the captured images, and the changed region is in the light region, has a lower luminance than the light region, and is in contact with the extension of the light region The sensitivity for detecting an intruder is lower than the sensitivity in the normal luminance region.
[0019]
According to the present invention, in an image monitoring apparatus that detects an intruder in a monitoring area by processing a captured image, false detection due to light or shadow in the monitoring area is eliminated, and the intruder is accurately detected. be able to.
In fact, if there is an intruder and it is determined that it is light or shadow, it will be output to the outside with the security status set to normal even though there is an intrusion abnormality. Will occur. In order to prevent such a situation, it is possible to limit the change of the detection logic in the light region. If the light area is too large, the light removal and shadow removal logic can be prevented from operating.
[0020]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described with reference to the drawings.
FIG. 4 shows a configuration when the present invention is applied to an image sensor.
The image sensor 31 is controlled by a control unit 32 configured by a CPU or the like. The imaging means 33 is composed of an optical system and a CCD. The infrared light projecting means 34 projects infrared light onto the monitoring area under the control of the control means 32 when the monitoring area becomes dark at night or the like. The image processing unit 35 is configured by a CPU or the like, and processes the image acquired from the imaging unit 33 to determine whether there is an intruder. This determination logic will be described later. The contents of the storage means 36 will also be described later.
[0021]
The operation unit 37 is provided for setting or changing the setting of the image sensor 31. The output means 38 and the image output means 39 are connected to a controller (not shown) by wire or wirelessly, and the controller is connected to a control center (not shown) via a communication line. When the image sensor 31 detects an intruder, the abnormal state is stored in the storage means 36 as a security state, and an abnormal image at the time of occurrence of the abnormality is stored. The output unit 38 and the image output unit 39 output the stored abnormal state and abnormal image when a state is called from the controller or the control center via an input unit (not shown). The power source 40 converts AC power into DC power and supplies it to each part in the image sensor 31.
[0022]
FIG. 5 shows the contents of the storage means 36.
The storage means 36 includes a program area 51 for storing operation programs of the control means 32 and the image processing means 35, a parameter area 52 for storing various parameters, and a reference image storage for storing the reference image and the current image taken by the image pickup means 33. An area 53, a current image storage area 54, an abnormal image storage area 55 for storing an image when an intruder is detected, a work area 56 used for processing by the control means 32 and the image processing means 35, and a security state (normal or abnormal) A state storage area 57 for storing the image, a mode storage area 58 for storing the security mode (warning mode, release mode, maintenance patrol mode) of the image sensor, and a light area storage area 59 for storing the light area in the shooting screen.
[0023]
Next, an outline (principle) of logic for detecting an intruder, light removal, and shadow removal logic in the image processing unit 35 will be described.
The logic of intruder detection depends on the size of the change area 19 and the movement distance 23 as already described with reference to FIG.
The light removal logic will be described with reference to FIG.
[0024]
6A shows the reference image 11, FIG. 6B shows the current image 16, FIG. 6C shows the difference image 18 obtained by taking the difference between the reference image 11 and the current image 16, and FIG. 6D shows the light region 24 from the reference image 11. The extracted light region image 25 is represented. In the illustrated example, it is assumed that the current image 16 has a smaller light area 24 and a lower luminance than the reference image 11.
The image processing means 35 extracts the light region 24 having very high luminance from the luminance information of the reference image 11 and acquires the light region image 25 shown in (D). The optical area image 25 is stored in the optical area storage area 59 of the storage unit 36. Next, the reference image 11 and the current image 16 are compared to obtain a luminance change region 19. When the change area 19 is generated in the light area 24 or includes the light area 24, the sensitivity for detecting an intruder is lowered. As an example of reducing the sensitivity, if the change area 19 moves by a set distance, the intruder is determined. However, if the change area 19 includes the light area 24 or is included in the light area 24, the set movement distance is set to the normal luminance area. Set longer than the case of. This utilizes the fact that the change in luminance that appears due to the diurnal motion of sunlight, etc., has a shorter moving distance within a certain time than an intruder.
[0025]
In the example shown in FIG. 6, what is detected as the change area 19 is included in the light area 24. Therefore, if the determination is made according to the above procedure, the light area 24 is not detected as an intruder. If the light region 24 is very large, measures such as not operating this logic may be taken.
The shadow removal logic will be described with reference to FIG.
[0026]
7A shows the reference image 11, FIG. 7B shows the current image 16, FIG. 7C shows the difference image 18, and FIG. 7D shows the light region image 25 obtained by extracting the light region 24 from the reference image 11. In the example shown in the figure, a person passes outside the monitoring area, and the shadow 26 is projected into the light area 24, and there is no intruder in the monitoring area. In this example, when the reference image 11 and the current image 16 are compared, a shadow 26 appears in the difference image 18 and the shadow 26 moves.
[0027]
As the shadow removal logic, the light region 24 is extracted from the luminance information of the reference image 11, and the light region image 25 shown in (D) is acquired. The optical area image 25 is stored in the optical area storage area 59 of the storage unit 36. Next, the reference image 11 and the current image 16 are compared to obtain a change area 19. If this change area 19 is generated in (1) the light area 24, has a lower luminance than (2) the light area 24, and (3) touches the boundary of the light area 24, intruder detection is performed. For lower sensitivity. An example of lowering the sensitivity is the same as the example of light removal described above. When the conditions (1) to (3) are satisfied, the change area may be determined as a non-intruder.
[0028]
In the example shown in FIG. 7, the change area 19 satisfies the conditions (1) to (3) for shadow removal, and the shadow 26 moves only within the light area 24, so the moving distance is short. Therefore, the shadow 26 is not detected as an intruder. If the light region 24 is very large, measures such as not operating this logic may be taken.
Next, intruder detection processing of the image processing means 35 will be described in detail.
[0029]
8 and 9 show intruder detection logic. This logic includes light removal and shadow removal logic.
Prior to the start of the intruder detection logic, a monitoring area is imaged in advance by the imaging means 33 and stored in the reference image storage area 53 as the reference image 11. Further, an optical region image 25 obtained by extracting the optical region 24 is obtained from the reference image 11. This acquisition method will be described later with reference to FIG.
[0030]
This logic is repeatedly executed at intervals of 0.5 seconds, for example.
In step S <b> 11, the current monitoring area is captured by the imaging unit 33 to acquire the current image 16 and stored in the current image storage area 54.
In step S12, the difference between the reference image 11 and the current image 16 is obtained for each pixel and binarized. When there is no predetermined luminance difference, the pixel value is 0, and when there is a predetermined luminance difference, the pixel value is 1. A region where the pixels having the pixel value 1 are gathered, that is, a set of difference binarized regions is the change region 19.
[0031]
In step S13, moving object tracking processing is performed. This process tracks the same object on the image. For example, when there are a plurality of change areas 19, the past change areas 19 and the current change areas 19 are tracked as the same object by associating those having a short distance of movement. Note that the new change area 19 that is not associated with the past change area 19 is the first appearance moving object. Further, when the change area 19 detected in the past is not tracked for 3 seconds, for example, it is regarded as disappeared and the tracking process is terminated.
[0032]
In step S14, it is determined whether or not the change area 19 exists. If “NO” here, since there is no change between the reference image 11 and the current image 16 or the change is minute, the possibility of detecting an intruder is low, and the process proceeds to step S15 for terminating the process. If YES, since there is a possibility that an intruder has been detected, the process proceeds to step S21 to perform shadow removal processing.
[0033]
In step S15, the current image 16 is stored in the reference image storage area 53 as the reference image 11. As a result, the reference image 11 is updated. In step S16, the “normal state” is stored in the state storage area 57. In step S17, the light area 24 is updated, and the process is terminated. Although the update of the light area 24 in step S17 will be described in detail later, the light area 24 stored in the light area storage area 59 is updated to the light area of the current image 16 for each process. Go.
[0034]
In step S21, since the change area 19 exists, it is determined whether or not it satisfies the shadow condition. For the principle of shadow identification, refer to the description of FIG. Here, if NO, since there is a possibility that an intruder has been detected, the process proceeds to step S31, and it is determined whether or not the light area is detected. If “YES” in the step S21, the change area 19 is highly likely to be a shadow and is unlikely to be an intruder, so the process proceeds to the step S22 of the process for preventing the unreporting.
[0035]
In step S31, since the change area 19 is not a shadow, it is determined whether this is the light area 24 or not. For the principle of light identification, see the description of FIG. 6 above. If “NO” here, since there is a possibility that an intruder has been detected, the routine proceeds to step 41 for processing of intruder detection. If “YES” in the step S31, the change area 19 is highly likely to be the light area 24 and is unlikely to be an intruder, so the process proceeds to a step S32 of the process for preventing unreporting.
[0036]
In step S41, in order to determine whether or not the change area 19 is an intruder, it is determined whether or not the change area 19 has moved 1.5 m or more from the initial appearance position. That is, it is determined whether or not the moving distance 23 (see FIG. 1) between the first appearance position 20 in the reference image 11 and the current position 22 of the change area 19 in the current image 16 is 1.5 m or more.
[0037]
When the presence of the change area 19 is detected in the above-described step S14, the reference image 11 is not updated (step S17), so the first appearance position 20 of the change area 19 remains stored in the reference image 11. If the change area 19 has detected the intruder 17, the current position 22 of the change area 19 moves as the intruder 17 moves by repeating the processing of this flowchart. When the moving distance 23 between the first appearance position 20 and the current position 22 is 1.5 m or more, it is determined that the intruder 17 has been detected, and the process proceeds to step S42. It determines with it being in a normal state, and progresses to step S43.
[0038]
In step S42, since the intruder 17 is detected, “intrusion abnormality” is stored in the state storage area 57, and the current image 16 is stored in the abnormal image storage area 55 as an abnormal image. On the other hand, in step S 43, “normal state” is stored in the state storage area 57. Thereafter, the optical region 24 is updated in step S17, and the process is terminated.
If YES in step S31, the change area 19 is likely to be the light area 24, but may also be an intruder, so the change area 19 has moved from the first appearance position 20 by 3 m or more in step S32. It is determined whether or not. Thus, in the light removal process, the threshold value of the moving distance 23 that determines the determination sensitivity is set to be longer than 1.5 m for intruder detection. This makes it difficult to erroneously detect the light region 24 as an intruder 17. However, if the change area 19 is the intruder 17, it is detected as an intruder when moving 3 m or more from the first appearance position 20, and the process proceeds to step S 33, and if it is less than 3 m, it is determined to be in the normal state. Proceed to step S34.
[0039]
The process of step S33 is the same as that of step S42, and the process of step S34 is the same as that of step S43. Thereafter, the optical region 24 is updated in step S17, and the process is terminated.
If YES in step S21, the change area 19 is likely to be a shadow 26, but may also be an intruder 17, so the change area 19 has moved from the first appearance position 20 by 4 m or more in step S22. It is determined whether or not. Thus, in the shadow removal process, the threshold value of the moving distance 23 that determines the determination sensitivity is set to be longer than 3 m for detecting the light region. This makes it difficult for the shadow 26 to be erroneously detected as the intruder 17. However, if the change area 19 is the intruder 17, it is detected as an intruder when moving 4 m or more from the first appearance position 20, and the process proceeds to step S 23, and if it is less than 4 m, it is determined to be in a normal state. Proceed to step S24.
[0040]
The process of step S23 is the same as that of step S42, and the process of step S24 is the same as that of step S43. Thereafter, the optical region 24 is updated in step S17, and the process is terminated.
FIG. 10 shows the details of the process of updating the optical region 24 in step S17 in the intruder detection logic.
[0041]
In the update of the light region, a high-luminance pixel is detected for each pixel of the current image 16, and the light region image 25 is updated based on this result.
In step S51, the absolute luminance (actual luminance in the monitoring region) is obtained for one pixel of the current image 16 from the luminance value and the camera exposure control value (aperture, shutter speed).
[0042]
In step S52, it is determined whether or not the absolute luminance is greater than 10,000 lux. The absolute luminance of the place where normal sunlight is not directly irradiated is about several hundred to 1000 lux, but the absolute luminance of the place where sunlight is directly irradiated exceeds 10,000 lux. Therefore, if the absolute luminance obtained from the luminance information of the pixel in the current image 16 is greater than 10,000 lux, it is determined that the pixel is in the light region 24.
[0043]
If the absolute luminance exceeds 10,000 lux, a numerical value 32 is set to the value of the corresponding pixel of the light region image 25 in step S53. An arbitrary value can be used for this numerical value. If the absolute luminance is 10,000 lux or less, it is determined in step S54 whether or not the value of the corresponding pixel in the light region image 25 is zero. If the pixel value is 0, the process ends. If the pixel value exceeds 0, 1 is subtracted from the pixel value in step S55, and the process ends.
[0044]
The above steps S51 to S55 are executed for all the pixels of the light region image 25. Thereby, in the light region image 25, the value of the pixels included in the light region 24 is 1 or more, and the values of the pixels other than the light region 24 are 0. Note that the value of the pixel included in the light region 24 in the light region image 25 is set to 32 in order to prevent the light region 24 in the light region image 25 from changing due to an instantaneous change in luminance. Once it is determined that the pixel is in the light region 24, the pixel value is set to 32. Henceforth, until it is determined that the pixel is not in the light region continuously 32 times, the light region 24 is continued. That is, it has hysteresis.
[0045]
【The invention's effect】
According to the present invention, in an image monitoring apparatus that detects an intruder in a monitoring area by processing an image obtained by capturing the monitoring area, the false detection due to light or shadow in the monitoring area is removed, and the intruder is accurately detected. Can be detected.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating logic of intruder detection by an image monitoring apparatus.
FIG. 2 is a diagram for explaining the reason why an optical area is erroneously determined as an intruder in the prior art.
FIG. 3 is a diagram illustrating a reason for detecting light as an intruder according to a conventional technique (part 1);
FIG. 4 is a diagram showing a configuration of an image sensor to which the present invention is applied.
5 is a view showing the contents of storage means in the image sensor of FIG.
6 is a view showing the principle of light removal in the image sensor of FIG. 5;
7 is a diagram showing the principle of shadow removal in the image sensor of FIG.
FIG. 8 is a flowchart (part 1) showing processing of the image processing unit in FIG. 5;
FIG. 9 is a flowchart (No. 2) showing processing of the image processing means in FIG. 5;
FIG. 10 is a flowchart showing an optical region update operation in the operations of FIGS.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 11 ... Reference | standard image 15 ... Window 16 ... Current image 17 ... Intruder 18 ... Difference image 19 ... Change area 20 ... First appearance position 22 ... Current position 23 ... Movement distance 24 ... Light area 25 ... Light area image 26 ... Shadow 31 ... Image sensor 32 ... Control means 33 ... Imaging means 34 ... Infrared projection means 35 ... Image processing means 36 ... Storage means 37 ... Operation means 38 ... Output means 39 ... Image output means 40 ... Power source 51 ... Program area 52 ... Parameter area 53 Reference image storage area 54 Current image storage area 55 Abnormal image storage area 56 Work area 57 State storage area 58 Mode storage area 59 Optical area storage area

Claims (3)

In an image monitoring apparatus having an imaging means for capturing a monitoring area and an image processing means for detecting an intruder in the monitoring area by processing the captured image,
The image processing means extracts a high-luminance area from the captured image, and detects logic that detects an intruder in the high-luminance area, and logic that detects an intruder in other normal-luminance areas. be those used to extract a change region of luminance from the comparison of the captured image with each other, if this luminance change area generated in the high luminance region or, if the luminance change area includes the high luminance region as logic for detecting an intruder in the high luminance region, the image monitoring, characterized in that the sensitivity of detecting the intruder detects an intruder by using less logic than the sensitivity in the region of the normal intensity apparatus. In an image monitoring apparatus having an imaging means for capturing a monitoring area and an image processing means for detecting an intruder in the monitoring area by processing the captured image,
The image processing means extracts a high-luminance area from the captured image, and detects logic that detects an intruder in the high-luminance area, and logic that detects an intruder in other normal-luminance areas. be those used to extract the change in luminance from the comparison of the captured image with each other, the brightness change is in the said high luminance region, a low luminance than the luminance of the high luminance region, the high-brightness region If you have extension in contact, as logic for detecting an intruder in the high luminance region, the sensitivity of detecting the intruder detecting an intruder using less logic than the sensitivity in the region of the normal intensity A featured image monitoring device. In an image monitoring apparatus having an imaging means for capturing a monitoring area and an image processing means for detecting an intruder in the monitoring area by processing the captured image,
The image processing means extracts a high-luminance area from the captured image, and detects logic that detects an intruder in the high-luminance area, and logic that detects an intruder in other normal-luminance areas. be those used to extract the luminance change from the comparison of the captured image with each other, the brightness change is in the said high luminance region, a low luminance than the luminance of the high luminance region, extension of the high-brightness region An image monitoring apparatus characterized by being a non-intruder when in contact with the camera.

Images