JPH03138596A - Moving body detecting device - Google Patents

Abstract

PURPOSE:To enable accurate detection without reference to a monitor place and monitor environment by providing plural kinds of processing conditions of brightness difference outputs, processing the brightness difference outputs in parallel according to the respective processing conditions, and detecting a moving body CONSTITUTION:A gradation difference (j) which is the threshold value of a brightness differ ence, a mask area M which limits the monitor area, and the alarm lower-limit number (k) of picture elements which is the threshold value of the number of picture elements are set under conditions I and II individually. Then the gradation differences C between the latest digital image signal B and picture element units of an image which is stored 4 and a specific time before are calculated and the gradation differences j1 and j2 are compared with each other to determine picture elements whose brightness variation is recognized eventually in mask areas M1 and M2. Then the generation of an alarm signal AL is controlled 8 according to whether or not the number of picture elements which are counted 6 exceeds the lower-limit numbers k1 and k2 of picture elements and when the signal AL is outputted based upon either one and the counted value exceeds the specific number l of continuous times, a moving body detection alarm is sent. Consequently, fine conditions can be set according to the monitor place and monitor state, so misdetection and a detection omission are evaded.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to a moving object detection device, and more specifically, it detects a moving object based on changes in brightness at predetermined time intervals of images captured by a two-dimensional image sensor such as a television camera. The present invention relates to a moving object detection device.

<Conventional technology> Various types of moving object detection devices are used in security and surveillance systems, but conventional ones mainly use one-dimensional sensors such as infrared photoelectric sensors, loop switches, and vibration switches. The detection area of the sensor is small, and moving objects can only be detected in a narrow linear or two-dimensional area on a point or line.

By detecting moving objects in a wide three-dimensional area, which is difficult to do with such -dimensional sensors, this sensor is used in surveillance systems that can provide security and surveillance over a wide area by bringing the situation closer to what is being monitored with the human eye. As a method for identifying moving objects from captured images using a two-dimensional image sensor such as a television camera, a method is being considered.

However, this type of moving object detection requires either a manned system in which a security guard constantly monitors images captured by a television camera to detect moving objects, or a system in which such images are temporarily recorded with a video tape recorder. It was a system that later discovered moving objects from repeated images. here,
The image may be a moving image, or it may be a still image taken at predetermined time intervals for economic reasons, but in either case, the human eye will ultimately determine the presence or absence of a moving object. It is supposed to be.

For this reason, especially when detecting the presence or absence of a moving object in real time, it is necessary to constantly monitor the monitor, which increases the burden on the monitor, especially when monitoring the monitors of cameras installed in multiple locations at the same time. There was also the risk of missing the presence of moving objects such as intruders.

A moving object detection device that solves this problem converts an analog image signal captured by a two-dimensional image sensor into a digital image signal, and compares the current digital image signal with the one at a predetermined time ago. , a method for automatically detecting a moving object based on the occurrence of a brightness difference between two image signals was previously proposed (Japanese Patent Laid-Open No. 61654).
(See Publication No. 86, etc.)

In other words, if there is no moving object in the captured image, an image that is between the road at the current moment and a predetermined time ago should be obtained, and a large brightness change will not occur. When a person enters a space, a large difference in brightness occurs between the image data before and after the person enters the space, so the presence or absence of a moving object can be detected by capturing the occurrence of such a brightness difference. .

<Problems to be Solved by the Invention> By the way, in an unmanned monitoring system using a two-dimensional image sensor as described above, detecting a slight change in brightness or a change in brightness over a small area may cause a false alarm. For example, in addition to setting a part of the image that is excluded from the detection area (mask area), it is also possible to set a part of the image that is excluded from the detection area (mask area), and also to ensure that a brightness difference is not recognized as occurring unless the area in the detection area where the brightness changes exceeds a certain value. Various detection conditions are set so that only the moving object that is truly desired to be detected can be detected.

However, for example, when trying to detect the same moving object at a near distance and a far distance from a television camera used as a two-dimensional image sensor, if the moving object is at a short distance from the camera, the moving object will be imaged in a large size. Therefore, even if the moving speed and comparison time are the same, the area where the brightness difference occurs is wider than when the moving object is far away (therefore, when the moving object moves close to the camera, there is a relatively large brightness change. It is desirable to detect the area and make it invalid for brightness changes over a small area, but conversely, when a moving object moves a long distance from the camera, the moving object is imaged small, so The area where the brightness difference occurs also becomes smaller, and if the brightness change area is determined to be relatively large so that it is suitable for close distances and is invalidated, objects that move over long distances will not be detected, and vice versa. If the brightness change area that is invalidated in accordance with the distance is set to a small value, the movement of an object smaller than the object to be detected will be detected at a short distance from the camera, causing false alarms.

Furthermore, even if there are both bright and dark areas in the detected image, it is only possible to set conditions that correspond to the brightness of one of them, making it difficult to ensure detection accuracy in both bright and dark areas. Met.

In this way, in the past, it was not possible to set detailed conditions according to the monitoring location and monitoring situation, especially when trying to detect a moving object in a wide range of distance from the camera as described above, or when trying to detect a moving object in a wide range of distance from the camera, When there is both, there is a problem that it is not possible to achieve both detection accuracy in far and near or bright and dark.

In addition, in systems that simply monitor the presence or absence of stationary objects, such as systems that monitor theft of exhibits, for example, images taken of exhibits are stored and retained, and then updated with the latest images captured at regular intervals. , the stored image is compared with the stored image to monitor the presence or absence of the exhibit, and the difference in brightness between the compared images continues to occur depending on the presence or absence of the exhibit. A method for detecting the presence or absence of a stationary object is also being considered.

However, in the above case, an image from a long time ago is compared with an image at the present moment. When the ambient brightness changes, slow changes in ambient brightness are compared over long time intervals, resulting in large brightness differences between the compared images. There was a risk that a false alarm could be issued by misunderstanding the situation, and there was a risk that detection accuracy could not be ensured depending on the monitoring location and monitoring conditions.

The present invention has been made in view of the above-mentioned problems, and when comparing the brightness of an image a predetermined time ago with an image at the present moment, it is possible to process a brightness difference output corresponding to the monitoring location and monitoring situation. In addition to doing so,
The purpose is to prevent the generation of false alarms due to changes in the surrounding environment when detecting moving objects by comparing images at long intervals, and to enable highly accurate detection regardless of the monitoring location or monitoring environment.

<Means for Solving the Problems> Therefore, the present invention includes a two-dimensional imaging means, and compares the image signal at the present moment taken by the two-dimensional imaging means with the image signal at a time a predetermined time ago, and determines the brightness. A moving object detection device that extracts a difference output and detects a moving object in a captured image based on the brightness difference output has multiple types of brightness difference output processing conditions and outputs the brightness difference based on each processing condition. The system was configured to perform parallel processing to detect moving objects.

Here, it is preferable to configure the luminance difference output to be output as a digital signal for each pixel.

The processing conditions include a brightness difference threshold for identifying the occurrence of a brightness difference, a pixel count threshold for determining the invalid number of pixels for which a brightness difference has been identified, and a It may also include at least a mask area map indicating the portion to be excluded.

On the other hand, as an image signal at a predetermined time before which is compared with the current image signal, there are an image signal that is stored and updated at regular intervals, and an image signal that is stored and retained from a predetermined time. A moving object detection signal is output only when a moving object is detected from the brightness difference output obtained by comparing the image signal and the current image signal.

According to this configuration, the current image signal captured by the two-dimensional imaging means and the image signal captured by the two-dimensional imaging means at a predetermined time before are compared, and a luminance difference output is extracted. . Here, a plurality of processing conditions for processing the luminance difference output are provided, processing based on each processing condition is executed in parallel, and a moving object is detected based on the respective processing results. In other words, the luminance difference output obtained by comparing the current image signal and the image signal at a predetermined time point is not processed based on only one processing condition, but is processed based on multiple different processing conditions. By performing processing based on 0, processing corresponding to various detection conditions can be performed in parallel, and it is possible to avoid false detections and ensure detection accuracy.

Here, when the brightness difference output based on the image signal comparison is configured to be output as a digital signal for each pixel, a brightness difference threshold value is used to invalidate the occurrence of minute brightness differences and identify the brightness difference occurrence. , a threshold value for the number of pixels is specified to invalidate the luminance change in a minute area, and a luminance difference is output by excluding the pixels from the detection area in the image. A mask area map that specifies areas to be invalidated as processing conditions, and pixels in which a brightness difference exceeding the brightness difference threshold has occurred in a detection area other than the detection exclusion area specified by the mask area map are set to brightness. The occurrence of such a luminance difference (luminance change) is identified as a difference occurrence.
The number of pixels that exceeds the pixel number threshold from the number of specified pixels is treated as the effective number of pixels for which luminance difference occurrence has been specified.

In addition, as an image signal at a time a predetermined time 1 before which is compared with the current image signal, an image signal that is stored and updated at regular time intervals and an image signal that is stored and retained from the predetermined time are provided. , these image signals are compared with the current image signal to obtain a luminance difference output, and moving object detection is performed based on the respective luminance difference outputs. A moving object detection signal is output only when a moving object is detected from both of the luminance difference outputs, and when a moving object is detected from only one of the luminance difference outputs, such moving object detection is invalidated and the detection signal is output. is not output.

<Example> The present invention will be explained in detail below.

FIG. 1, which shows the first embodiment, shows a monitoring system that warns the outside of the presence of a moving object at a specific monitoring location. The two-dimensional image sensor 1 is installed facing the specific monitoring location, and photographs the monitoring location and outputs an analog image signal.

2 The analog image signal output from the two-dimensional image sensor 1 is clamped by the clamp circuit 2 and then passed to the A/D converter 3.
For example, the luminance of each pixel is quantized into 256 gradations expressed by 8 bits.

The latest (current) digital image signal B is compared pixel by pixel with the image A stored in the video memory 4 at a predetermined time ΔT before, and the luminance gradation difference C(
luminance difference output) is calculated. Here, if a pixel is represented by X and Y coordinates, the gradation difference (brightness difference) C is c
(x, y) = A (X, Y) - B (X, Y)l
becomes.

When one image stored in the video memory 4 is compared with the latest image, the latest image is stored in the video memory 4 as a new stored image, and the image stored in the video memory 4 is It is updated every predetermined time ΔT.

Once the pixel-by-pixel gradation difference C (X, Y) between the latest image (comparison image) and the image at a predetermined time period ΔT (reference image) stored in the video memory 4 is calculated, a The gradation difference j, which is the threshold value of the luminance difference to be set, and the gradation difference C(X, Y) are compared for each pixel, and the gradation difference j
Select only the pixels whose brightness has changed by the above amount, and classify all pixels into pixels whose brightness has changed (1 is set as data) and pixels whose brightness has not changed (O is set as data). Let it be binarized data D (X, Y). Therefore, even if the brightness changes during a predetermined period of time ΔT, pixels for which the change is less than the gradation difference j are treated as if there was no change in brightness, and are unrelated to moving objects caused by natural phenomena or other factors. Prevent false detection of brightness changes.

When all pixels are binarized based on the presence or absence of a luminance change within a predetermined time ΔT as described above, next, the mask area map M stored in the mask memory 5, which is also arbitrarily set from the outside, is processed. This disables brightness change detection in areas that are not monitored. Specifically, for example, pixels with a brightness change are set to 1, pixels with no brightness change are set to 0, and masked pixels that are not to be monitored are set to 1, and non-masked pixels to be monitored are set to 0. In this case, the logical negation of M (X, Y), which indicates the masked area, and D (X, Y), which indicates the presence or absence of brightness change detection, are performed to detect the brightness change only in the area to be monitored. Pixel E (X, Y
) = D (X, Y) ・rYT'7) is extracted. By setting the mask area as described above, an area in which movement of an object other than the detection target is expected to be detected is excluded from the monitoring target, thereby avoiding issuing a false alarm.

Then, the error pixel counter 6 is counted as a pixel in the monitoring target area that has a luminance change of 3 or more gradation difference, that is, E
The number of pixels En where (X, Y) = 1 = the number of counted up pixels En (number of error pixels) is the alarm lower limit number of pixels k (threshold value of the number of pixels) which is arbitrarily set from the outside.
The alarm signal A is generated only when the counted up pixel number En exceeds the alarm lower limit pixel number k.
Output L. In other words, when there are few pixels in the monitored area that have a luminance change of 5 or more than the gradation difference j, in other words,
When the area on the screen where the gradation change occurs is small, for example, the movement of an object smaller than the moving object to be detected is detected, so if this is detected as the desired moving object, it will result in a false alarm. Therefore, this is invalidated for luminance changes over a small area. This results in
Luminance changes unrelated to a moving object such as video noise can be removed, and even if the luminance change pixel takes a non-zero value when there is no moving object in the image, processing can be performed so that an alarm signal AL is not transmitted.

When the counted-up pixel number En exceeds the alarm lower limit pixel number k and an alarm signal AI is output, the alarm counter 7 counts this alarm signal A L for a consecutive number of times that can be arbitrarily set from the outside. 1 and the count value of the alarm signal AL, and if the alarm signal AL is output for two or more consecutive times, an alarm is issued to the outside to notify that a moving object has been detected.

In this way, instead of issuing an external alarm for each alarm signal AL, by issuing an external alarm only when a predetermined number of consecutive alarms are issued, it is possible to prevent one-off occurrences, such as turning lights on and off at a monitoring location. It is possible to avoid outputting alarms to the outside due to such phenomena.

Furthermore, when the alarm signal AL is interrupted for a predetermined time or longer, the count number is reset to zero, and the interrupted alarm signal AL is counted up to prevent the result from being output to the outside as a detection of a moving object. That's what I do.

Furthermore, the latest image to be compared with the stored image on the video memory 4 is captured at a comparison interval i that is arbitrarily set from the outside.
For example, the field frequency can be variably controlled by
For example, when complying with the NTSC system of Hz,
The minimum unit is a frame image every 3m5, and the shortest is 33
The image is compared every m5, and the comparison time is set to be longer by a multiple of 33m5, and when the speed of an object moving in the screen is fast, the comparison interval is shortened, and when it is slow, the comparison interval is lengthened.

Note that the update timing of the stored image in the video memory 4 may be synchronized with the capture cycle of the latest image, and the previously captured image and the latest image may be compared from time to time. The image on the video memory 4 may be updated once every several times of image capture, so that the image on the video memory 4 is compared with the latest image several times after that, or the image signal captured at a predetermined time may be updated It is also possible to store the stored image signal only until it is reset after that, and to compare the stored image signal with the latest image signal each time without updating it.

By the way, as mentioned above, the gradation difference j, the mask area M on the mask memory 5, and the alarm lower limit pixel number k.

Consecutive times! , comparison interval j are detection conditions (processing conditions for brightness difference output) that can be arbitrarily set from the outside, and are displayed as numerical values or areas on a monitor attached to the device, and can be set according to the monitoring location and monitoring situation. However, in this embodiment, two systems are provided in parallel with a system configuration of 7 8 from obtaining the brightness difference output to outputting the alarm signal by comparing it with the alarm lower limit pixel number. , two patterns can be set as processing conditions consisting of the detection conditions for the credit union excluding the comparison interval i,
It is configured so that moving object detection can be performed by processing images in parallel according to respective patterns.

Detection patterns in moving object detection under these two processing conditions are specifically shown in FIGS. 2 and 3, and will be described with reference to FIG. 1.

In addition, in FIG. 2 and FIG. 3, in order to simplify the explanation,
The number of pixels and gradations are shown reduced compared to the actual numbers.

In Fig. 2, when the gradation difference j is set as condition ■, j+=89
In condition setting ■, j2-4, mask area M is stored in the mask memory 5a.In condition setting I (Ml), the left half of the screen is stored in condition setting II (Ml), which is stored in mask memory 5b.
In the right half of the screen, the alarm lower limit pixel count (lower limit of error pixel count) k is kl in condition setting I + kz in condition setting ■.
(>k,).

9 Here, assuming that the larger the value of the gradation indicating the brightness of each pixel, the brighter it is, the reference image stored in the video memory 4 is an image where the right half is relatively bright; On the other hand, the latest comparison image is an image whose overall brightness has been increased by increasing the gradation of each pixel by five gradations. Therefore, each pixel is represented by five gradations as a difference image (brightness difference output).

Next, determine whether the detected brightness change is greater than or equal to a predetermined value.
The discrimination will be based on the gradation difference j, but as mentioned above, two processing conditions are set in parallel,
Processing based on each condition is performed in parallel.

In other words, in condition setting ■, the gradation difference j is set to 8, so unless there is a change of 8 or more gradations, it is not recognized that there has been a change in brightness, and as in this case, each pixel is uniformly displayed on the 5th floor. If only the tone has changed, it is assumed that there has been a change in brightness throughout the screen, and each pixel is set to zero, indicating that there has been no change in brightness.

On the other hand, in condition setting ■, the gradation difference j2 is set to 4, which is 0, so if each pixel uniformly changes in brightness by 5 gradations like this time, it is recognized that there has been a brightness change for all pixels. 1 is set in each pixel to indicate that there has been a change in brightness.

In this way, when the brightness difference image (brightness difference output) is converted into a binarized image indicating brightness change and no brightness change based on the gradation difference JI+Jz, next, the mask area M to be excluded from the monitoring area is This is a process based on (mask area map), but in condition setting ■, mask area M+ is set to the left half of the screen, but as mentioned above, there are no pixels where brightness changes are observed on the entire screen. The masking process is ineffective, and there will be no error pixels whose brightness changes are finally recognized.

In addition, in condition setting ■, the mask area M2 is set in the right half of the screen, so even if a luminance change is detected for the pixels included in this area, it is invalidated, and the remaining pixels in the left half are finally A change in brightness was observed, and the results and conditions were set.■
In this case, 1 is set to indicate that there has been a change in brightness for the pixels on the left half of the screen.

When the pixels (error pixels) in which a luminance change is finally recognized are determined in this way, the number of such error pixels is calculated by the error pixel number counter 6a for each image.
6b, and generation of the alarm signal AL is controlled depending on whether the counted number exceeds the alarm lower limit pixel number (lower limit for the number of error pixels) kl, which is 2 or not. Since no error pixels were recognized in the processing in condition setting ■, the alarm signal AL was not sent, and the alarm lower limit pixel count (lower limit for the number of error pixels) set in condition setting ■
If 6, which is the number of error pixels here, is larger than k2, an alarm signal AL is transmitted.

The alarm signal (ON/OFF signal) that is the result of processing the luminance difference output based on each processing condition is OR
As a result of performing two processes in parallel based on two processing conditions, if an alarm signal is output from at least one of them, this alarm signal is input to the circuit 8. When the counter 7 counts and the count exceeds the continuous number of times, an alarm is sent to the outside as a moving object detection signal.

In Fig. 1, the number l of consecutive alarm signals AL is commonly used, but alarm counters 7 are provided for each condition so that alarms can be transmitted to the outside completely independently. For °C, by changing the numerical value between condition setting ■ and condition setting ■, an alarm is output to the outside for the number of consecutive screens (continuous alarm signal) for which the number of pixels with brightness change is determined to be greater than a predetermined value for each condition setting. The sharpness can be adjusted. Depending on the monitoring conditions, the alarm signal AL may be output only when a moving object is detected under both processing conditions.
It may be configured to output an alarm signal only when a moving object is detected from only one side.

In the case shown in FIG. 3, the same condition setting I as above is used.

2) and the comparison image that is different from the reference image. The comparison image shown in FIG. 3 is a case where the brightness of the pixels in the center row other than the left and right pixel rows is increased by 10 gradations. Therefore, in the difference image, a brightness change of 10 gradations is set for the central two vertical pixel columns.

Then, in the same way as described above, this difference image is processed in parallel by condition setting ■ and condition setting ■, but this time the brightness change is 10 gradations, which exceeds the gradation difference j of condition setting I and condition setting ■. Therefore, in the gradation difference binarization process using the gradation difference j as the threshold value, 1 is set for each pixel in which a 10-level gradation change is observed in condition setting I and condition setting ■. With condition setting I and condition setting (2), the same binarized image in which 1 is set for the two vertical pixel columns in the center is obtained.

In addition, the mask area is the left half of the screen in condition setting I, and the right half of the screen is in condition setting ■, so the masked screen is
In this case, the left pixel column of the two vertical pixel columns 4 in the center is masked, and only the right pixel column is finally set as the pixel in which a brightness change has been detected. The right pixel row of the pixel rows is masked, and only the left pixel row is finally set as the pixel in which a brightness change has been detected.

Therefore, although the pixels remaining after mask processing are different,
In both Condition Setting ■ and Condition Setting ■, there are 3 pixels in which a change in brightness is observed in the monitoring area, but in Condition Setting I, these 3 pixels exceed the alarm lower limit pixel number (lower limit number of error pixels) kI. Therefore, the alarm signal AL is transmitted, but in the condition setting (3), the three pixels are below the alarm lower limit pixel number (lower limit of the number of error pixels) k2, so the alarm signal AL is not transmitted. In this way, even if the same number of error pixels are detected, the alarm lower limit pixel number may be different because the gradation difference j is different between condition setting ■ and condition setting ■ as in the case of Figure 3. Under processing conditions in which only large luminance changes are detected, the alarm signal AL is transmitted with a small number of pixels, and in processing conditions in which even relatively small gradation differences are detected, a larger number of pixels are not detected. This is to prevent the alarm signal AL from being transmitted.

By providing two types of differential image processing conditions as described above and executing detection processing based on each condition in parallel, detailed condition settings can be made depending on the monitoring location and monitoring situation, and false detections can be avoided. Accurate moving object detection can be performed by avoiding detection failures.

For example, when trying to detect both the same moving object moving far from sensor 1 and moving relatively close to sensor 1, the moving object will appear small at long distances, and the moving object will appear small at short distances. Because moving objects appear larger,
The processing conditions for long-distance detection are such that the monitoring area is narrowed so that an alarm is sent even if a relatively small number of pixels within the area show a change in brightness, and the processing conditions for short-distance detection are such that the monitoring area is narrowed. so that an alarm is sent only when the brightness of a relatively large number of pixels changes.
If two conditions are provided, one for long-distance detection and one for short-distance detection 5 6 , moving objects can be detected accurately both at far and near locations. Furthermore, even when the screen is divided into bright and dark areas, by masking the bright or dark areas using a mask area map, processing conditions corresponding to each area can be set, and moving objects can be detected accurately in both bright and dark areas.

In addition, in the above embodiment, two processing conditions were described in which the mask area is reversed and different areas are used as monitoring areas, but conditions are set respectively for a relatively large monitoring area and a relatively small monitoring area included in this. Processing may be performed by changing the processing conditions, and further, different processing conditions may be set for each of the partially overlapping monitoring areas.

In the above embodiment, the number of pixels is assumed to be about 16 to 20 to simplify the explanation, but in reality, a solid-state image sensor with opposing pixels is used. However, the number of pixels that can balance between ensuring detection accuracy and avoiding false alarms is 1.
Experiments have shown that a number of pixels of about 5,000 to 16,000 is appropriate.

In addition, in the above embodiment, the masked area/non-masked area is divided into the left and right or top and bottom of the screen, but it is also possible to set multiple areas as the masked area or the non-masked area. The area is not limited to a rectangle, but can be specified by surrounding it with a continuous curved line using pixels as the minimum unit. Furthermore, it is also possible to specify multiple masked areas and non-masked areas.

Furthermore, once the detection/processing conditions are set, the conditions will not be changed frequently, so the condition setting section, which includes a monitor, and the detection section, which detects moving objects based on the set conditions, are divided into two parts. After setting the detection conditions, only the detection unit may be installed together with the sensor 1 (television camera) at the monitoring location.

In this case, after installation, the condition settings cannot be changed unless a condition setting section is prepared, which prevents damage to the reliability of the monitoring device due to careless condition changes, and also allows the condition setting section to be used for multiple detection It is possible to reduce the cost by sharing the same with other parts, and also to reduce the size of the installed detection device.

8 When setting the detection conditions by connecting the condition setting section,
The configuration is configured so that the detection conditions that are variably set by button operations etc. are displayed as numerical values on the monitor, while the actual detection results (number of error pixels, etc.) are also displayed as numerical values on the monitor together with the screen of the monitoring location. By doing so, you can easily set conditions according to the monitoring location and monitoring situation.

In addition, in the above embodiment, two types of detection conditions are set and two types of processing are performed in parallel based on these.
It is clear that the detection conditions are not limited to two types, and three or more types of processing conditions may be provided, although this increases the hardware configuration or the burden of calculation processing by software.

In FIG. 4 showing a second embodiment of the present invention, two video memories 4a are provided to store two types of reference images.

4b, the current image signal is compared with the image at a predetermined time ago stored in the two video memories 4a and 4b, and two luminance difference outputs are outputted respectively. Then, the luminance difference output is
Processing is performed based on each processing condition (gradation difference j, mask area M, alarm lower limit pixel number k), and finally, as mentioned above, if the error pixel number En exceeds the alarm lower limit pixel number k, an alarm is issued. The signal is output, but the processing result based on the two luminance difference outputs (alarm signal ON/OFF
) is input to the AND circuit 10,
As a result of processing the two luminance difference outputs, an alarm signal is output to the alarm counter 7 only when both alarm signals are output.

Here, the reference image stored on the video memory 4a side is
For example, the reference image stored in the video memory 4b is updated every minute time (fixed time) by being updated every read cycle of the latest image determined by the comparison interval i. is for storing and holding an image signal captured at a predetermined time in response to a trigger signal from the timer 9.

That is, on the video memory 4a side, 90 image signals are compared at regular time intervals, but on the video memory 4b side, the time intervals are lengthened as time passes, and the image signals are compared. For example, an image of an unoccupied space to be monitored or an image of an exhibit can be stored on the side, and if there is an intruder in the space or an exhibit is removed from its installation location due to theft, etc. By comparing the image signal in the normal state with an abnormal image in which an intruder is captured or no exhibit is present, a brightness difference output depending on the presence or absence of an intruder or exhibit is obtained, and an alarm is issued.

By the way, when comparing an image from a long time ago with an image at the present time as described above, even slow changes in natural environmental conditions, such as the shading of the sun, may be interpreted as large changes. Since there is a habit of issuing false alarms, the above-mentioned false alarms can be prevented by using image comparison at minute intervals, which is performed by updating the stored image (reference image) on the video memory 4a side at regular intervals. That's what I do.

If slow changes in ambient brightness are monitored at long intervals, large changes in brightness will be detected, but if images are compared at minute intervals,
The effects of slow changes in environmental conditions as described above can be eliminated by the brightness difference j, etc., so image comparisons made at minute time intervals are more sensitive to the presence or absence of intruders or exhibits than image comparisons made at long time intervals. Although it is difficult to generate a clear luminance difference output due to the method, it is not affected by slow environmental changes.

Therefore, if the configuration is configured so that it is recognized that a true moving object has been detected only when the two detection results match and indicate that a moving object has been detected, the environment in which images are compared over a long time interval can be improved. It is possible to eliminate the effects of change.

FIGS. 5 and 6 show the details of processing in moving object detection using the two reference images as described above at the pixel level of the images.

First, in FIG. 5, the luminance level of the comparison image (current image signal) has increased by 10 gradations as a whole with respect to the reference image a that has been captured and stored at a predetermined time. On the other hand, the reference image, which is captured and stored a predetermined minute time before the second comparison image, has almost no change from the comparison image. For this reason, the brightness difference output obtained by comparing the two reference images and the latest comparison image is that when comparing with reference image a, all pixels are set to 10 gradations, whereas when comparing with reference image a, all pixels are set to 10 gradations, whereas Then, the luminance difference output on the side to be compared is entirely set to zero.

Here, if we set 8 gradations on the reference image a side and 4 gradations on the reference image side as the gradation difference j for identifying the brightness change, then the luminance on the side compared with the reference image a 1 indicating that there has been a change is set for all pixels, but on the side compared to the reference image, O indicating that there has been no change in brightness is set for all pixels, indicating whether or not there is a change in brightness. It is converted into a binarized image showing .

In addition, the mask images (mask areas) for masking areas that are not included in the monitoring area from the binarized image are both set to the left half of the screen, and only the right half of the screen is set. Since this is a monitoring area, when mask processing is applied, the binarized image obtained by comparing with the reference image a becomes an image in which a change in brightness is observed in the right half of the screen, and compared with the reference image a. In the binarized image, since no change in brightness was originally observed over the entire screen, it is assumed that there was no change in brightness over the entire screen.

Here, the number of pixels where a brightness change was observed (number of error pixels)
When the number of error pixels obtained by comparing with reference image a exceeds the invalid number of , an alarm signal is output, but in the luminance difference output based on one reference image A, the number of error pixels is zero. Because there is, no alarm signal is output,
The output of the AND circuit 10 to which these results are input will not issue an alarm signal.

The example shown in FIG. 5 is for a case where the current brightness as a whole is higher than the surrounding brightness at the time when the reference image a is stored, and there is no moving object. A slow change in brightness is captured and an alarm signal is output in detection based on the reference image a, but when comparing images at minute time intervals based on the reference image a, there is almost no change in brightness and a moving object is recognized. Therefore, it is assumed that the alarm signal on the reference image a side captures a change in the surrounding environment as described above, and in this case, it is not recognized that a moving object has been detected, and a false alarm can be avoided. Ru.

In addition, in the case shown in Fig. 6, for the same reference image a, the comparison image has a luminance change of 10 gradations in the pixel columns at both the left and right ends of the screen, and a luminance change of 20 gradations in the two vertical columns of pixels in the center. It shows a change in brightness. On the other hand, in a comparison between the reference image and the comparison image, there is no change in brightness in the pixel columns at both the left and right ends of the screen, while the example of pixels in the two vertical columns in the center shows a change in brightness of 10 gradations.

Therefore, the gradation difference j is set to 8.0 on the reference image a side. If it is set to 4 on the side of the reference image a, the error pixels will cover the entire screen on the side of the reference image a, and the error pixels on the side of the reference image a will only be in the central two vertical columns.

Then, when masking the left half of the screen similar to that shown in FIG. Pixel example 5 Only the right pixel column remains as an error pixel.

Here, the number of error pixels remaining after the masking process as described above is equal to the predetermined alarm lower limit pixel number, and 2.
If it exceeds , alarm signals are output from both to the AND circuit 10, and the AND circuit 10 outputs the alarm signal to the alarm counter 7.

That is, in the example shown in FIG. 6, although there is an increase in brightness across the screen due to changes in ambient brightness when comparing images over a long time interval, such brightness changes also include brightness changes due to moving objects. , the brightness changes due to such moving objects are captured by image comparison at minute time intervals,
In this case, the presence of a moving object is detected by comparing images at minute time intervals that are not affected by slow environmental changes, and since changes in brightness are also observed in image comparisons at long time intervals, the movement of the monitored object is detected. It is difficult to recognize that this is just a change in brightness in the surrounding environment, etc., unrelated to the movement (presence or absence) of the object, and it is assumed that a moving object has truly been detected.

6 In image comparison over a long time interval, if the ambient brightness changes as described above, this change in the surrounding environment will be detected as a moving object, and it may be due to a moving object or a change in the surrounding environment. Although it is difficult to clearly distinguish between images, it is difficult to clearly distinguish between images at short time intervals. By comparing images at short intervals, it is possible to determine whether the change is due only to a change in the surrounding environment or includes a change in brightness due to detection of a moving object.

Note that even when two types of reference images are provided as described above, two types of processing conditions are provided and each is executed in parallel. For example, when there is a moving object such as an intruder, comparison with reference image a is performed. Since it is possible to obtain a brightness difference output even when there is no movement of the object by comparing it with an image of a moving object that has not been captured at all, conditions such as gradation difference j can be adjusted compared to the processing conditions for the reference image. On the reference image a side, the setting can be made more strict.

Here, the state of the difference image with respect to the actual image when the reference image is updated at regular time intervals and when the reference image at a predetermined time is retained will be explained with reference to FIGS. 7 and 8.

Figure 7 shows the case where the reference image is updated every time the image is compared, and when someone enters an unoccupied room (A) (B), the intruder enters as a difference image between the unoccupied state and the state in which the intruder is imaged. person is extracted and an alarm signal is output. Furthermore, when the intruder moves within the room (C), differential images are generated before and after the intruder moves, and an alarm signal is output. However, if the intruder stops moving in the room (D), no alarm signal will be output because there will be no difference in brightness at short intervals, but if the intruder raises his arm on the spot (D), the alarm signal will not be output.
E) A difference image is generated between the lowered arm and the raised arm, and an alarm signal is output. Furthermore, when the intruder leaves the room, a difference image is generated by comparing the image in which the intruder was present and the image in the unoccupied state, and an alarm signal is output.

FIG. 8 shows a case where the reference image is stored and retained without being updated, and the reference image is a rectangular object placed in an unoccupied room.

For such a reference image, if someone enters the room (A
), a difference image is generated by capturing an image of an intruder who is not in the reference image, and an alarm signal is output. (B) to (D) continue until the intruder leaves the room.

Furthermore, when the intruder moves the rectangular object, a difference image is also generated due to the movement of this object, and when the intruder leaves the room with the object moved from its original position (D), the object becomes the reference image. A differential image is output as the object moves and an alarm signal is continuously output until the object is returned to its original position stored as .

Therefore, in the cases shown in FIGS. 7 and 8, if the intruder moves or enters or exits the room, an alarm signal will be output from both, and in the detection shown in FIG. There is a possibility that the presence or absence of an intruder or the movement of an object may be falsely reported due to changes in ambient brightness, etc.9 However, such detection is performed by updating the reference image shown in Figure 7, and is compared with images that are not easily affected by environmental changes. If it is captured as the movement of an object, the results of image comparison performed with a fixed reference image can be identified as including those caused by the presence of an intruder or the movement of the object, which will affect the environmental changes in detection based on the fixed reference image. The reliability of the detection device can be improved by compensating for the weakness that it is easily detected.

Incidentally, the updating of the stored image in the video memory 4b is as follows.
In addition to specifying the time using the timer 9 as described above, it may also be configured to update at any time using an operation switch.

<Effects of the Invention> As explained above, according to the present invention, a plurality of processing conditions are provided for outputting a luminance difference obtained by comparing the luminance of an image signal at the present moment and an image signal at a point a predetermined time before, and each By configuring the system to detect a moving object by processing the brightness difference output in parallel based on the processing conditions of Even when the moving object to be detected moves both near and far, or when the screen is divided into bright and dark areas, by setting corresponding processing conditions and performing parallel processing, accuracy can be achieved in both distance and brightness. can perform good detection.

Further, the luminance difference output is configured to be output as a digital signal for each pixel, and the processing conditions include a luminance difference threshold for identifying the occurrence of a luminance difference, and a pixel for which the occurrence of a luminance difference is identified based on the threshold. By including at least a threshold value for the number of pixels that specifies the invalid number of pixels, and a mask area map that indicates the part of the image to be excluded from the detection area, false detection can be prevented by setting multiple parameters. This can be avoided with high accuracy, and the threshold value of the processing conditions can be quantified and clarified, making it possible to easily set the conditions.

Furthermore, it has an image signal that is stored and updated at regular intervals and an image signal that is stored and retained from a predetermined time point, and compares these image signals with the current image signal to obtain a brightness difference output. By configuring to output a moving object detection signal only when a moving object is detected from both luminance difference outputs, the occurrence of false alarms in the detection processing based on one stored image can be compensated for by the other, and especially , the presence or absence of an object can be detected with high accuracy.

[Brief explanation of the drawing]

FIG. 1 is a system block diagram showing a first embodiment of the present invention, and FIGS. 2 and 3 are diagrams for explaining processing contents based on two processing conditions in the first embodiment, respectively.
FIG. 4 is a system block diagram showing a second embodiment of the present invention, and FIGS. 5 and 6 are diagrams for explaining processing contents based on two reference images in the second embodiment, respectively.
FIG. 1 and FIG. 8 are diagrams each showing a specific example of differential image output by image comparison. 1... Two-dimensional image sensor 3... A/D converter 4
a, 4b... Video memory 5a, 5b... Mask memory 6a, 6b... Error pixel counter 7.
... Alarm counter, i... Gradation difference (brightness difference)
1 2 M...Mask area map k...Alarm lower limit number of pixels! ...Continuous number of times i...Comparison interval

Claims (1)

[Scope of Claims] (1) A two-dimensional imaging means is provided, and a luminance difference output is extracted by comparing an image signal at the current moment captured by the two-dimensional imaging means with an image signal at a time a predetermined time before; A moving object detection device that detects a moving object in a captured image based on the brightness difference output, comprising a plurality of types of processing conditions for the brightness difference output, and processes the brightness difference output in parallel based on each processing condition. A moving object detection device characterized in that it is configured to detect a moving object. (2) The moving object detection device according to claim 1, wherein the luminance difference output is configured to be output as a digital signal for each pixel. (3) The processing conditions include a brightness difference threshold that specifies the occurrence of a brightness difference, a pixel count threshold that specifies the invalid number of pixels for which a brightness difference has been identified, and a detection area in the image. 3. The moving object detection apparatus according to claim 2, further comprising at least a mask area map indicating a portion to be excluded from the moving object detection apparatus. (4) A two-dimensional imaging means is provided, and a luminance difference output is extracted by comparing the current image signal captured by the two-dimensional imaging means with an image signal at a predetermined time before, and based on the luminance difference output. In a moving object detection device that detects a moving object in an image captured by a user, an image signal stored and updated at regular time intervals as an image signal at a point in time before the predetermined time, and an image signal stored and retained from a predetermined time point onwards. image signals, and is configured to output a moving object detection signal only when a moving object is detected from the brightness difference output obtained by comparing these image signals with the current image signal. Characteristic moving object detection device.