JP3456972B2 - MONITOR MONITORING METHOD AND DEVICE - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for automatically determining a moving body and monitoring the moving state, and more particularly to a monitoring method and apparatus for automatically detecting a ship moving on water.

[0002]

2. Description of the Related Art In particular, many commercial ships such as freighters and fishing boats come and go in the harbor, the sea near the harbor or the gate.
It is necessary to monitor sailing vessels from the viewpoint of maritime traffic safety. For example, the person who manages the harbor monitors the vessels entering and leaving by using both eyes and TV cameras, but many vessels depart from the scheduled time, and it seems that there is no drop even when the number of vessels is extremely small. Since it needs to be monitored, the load on the administrator is heavy.

In order to reduce the burden on the administrator, it would be convenient if a monitoring device for automatically detecting a traveling vessel and issuing an alarm could be used. As a method of recognizing a moving body using a captured image, a method of creating a background image in which the moving body does not exist and detecting the moving body by taking a difference from the measurement image, or photographing at a certain time interval There is known a method of detecting and displaying a moved portion in an image by obtaining a difference between the images. However, the method using the background image works effectively in an environment with a constant background such as indoors, but it is not useful in an environment where the sunlight changes and the background scenery changes with time, so it detects moving vessels. Not available to do.

On the other hand, the method of obtaining the difference between images at time intervals can be used in the field where the environment changes. As an application of this method to ship monitoring, for example, Japanese Unexamined Patent Publication No.
In -65299, the observation area set for the water surface of the harbor or river is photographed with a TV camera, the video image data captured at appropriate time intervals is binarized, and then the difference processing is sequentially performed for each adjacent time. Further, there is described a water surface monitoring system in which a common part of the obtained difference images is logically ANDed next to each other to extract a common portion to detect a moving body and track the center of gravity of the moving body.

However, in the detection method using the image difference, if there is a thing whose brightness changes in the background, it becomes noise in the difference image and the moving body cannot be extracted correctly. Especially when monitoring a moving object on the water surface, the background noise with a large brightness difference that changes at high speed in the screen due to the reflection of the sun on the water surface and the wave surface of the water surface is generated. It is very difficult to extract from the image. The above-mentioned publication mentions that the photographing time interval is made as short as possible in order to suppress the generation of noise. This is because if the brightness noise is imprinted in the two continuously photographed images, it can be eliminated by cancellation. On the other hand, if the time interval is shortened, it becomes difficult to capture the movement of the ship. Therefore, it is said that the images are taken every 0.2 to 0.3 seconds as a result of considering the overlapping of the images of the ship. However, it is difficult to detect the relatively slow movement of ships such as commercial ships at such an imaging interval, especially when taking a picture of a ship navigating on the water surface from a long distance and monitoring it over a wide area such as port control. is there.

[0006]

SUMMARY OF THE INVENTION An object of the present invention is to provide a method for detecting a relatively slow moving object in an image in which a large amount of fast-changing noise is generated. In particular, it is to provide a monitoring method and a monitoring device for detecting a ship moving on the water surface.

[0007]

In order to solve the above problems, a method of monitoring a moving object according to the present invention is such that captured images of a target area are captured at a predetermined cycle, and images that are adjacent to each other are captured each time the captured images are captured. From the first difference image and the difference between the images separated from each other by a predetermined distance to obtain a second difference image.
A feature that the difference image is added when the same predetermined number of images are subtracted to form a composite image, and a notification is made when a region larger than a predetermined area is detected from the binarized image obtained by binarizing the composite image. And

The moving object monitoring method of the present invention uses simple image processing, paying attention to the fact that the change of the waves on the sea surface and the reflected state of sunlight is sufficiently fast as compared with the moving speed of a moving object such as a ship. The noise is canceled by the processing corresponding to the existing band pass filter to detect the moving body. If the second difference image is obtained by taking the image difference between images that are temporally distant among the images obtained at equal intervals, it is possible to detect a target that changes at low speed such as a ship. However, since the second difference image contains a noise component that changes at high speed such as a wave on the sea surface or a change in sunlight, a clear movement cannot be detected.

On the other hand, if the first difference image is obtained by taking the image difference between the images adjacent to each other among the images obtained at equal intervals, the noise component that changes at a high speed can be detected. However, since the sea level changes quickly and the occurrence position has not been determined, the noise in the first difference image does not overlap with the noise in the second difference image, and the process of subtracting the second difference image from the first difference image is performed. Even if you go, you cannot cancel the effects of noise.

However, since the noise is detected evenly in the area where the sea level changes, the noise generation area of the first difference image and the noise generation area of the second difference image are in the same degree of noise state when integrated a suitable number of times. Therefore, by subtracting the integrated image of the first differential image from the integrated image of the second differential image, it is possible to cancel the noise component in the noise occurrence region such as the sea surface and extract a moving body such as a ship. It goes without saying that the present method can be widely used not only for a ship traveling on water, but also for detecting a noise that changes at a high speed and a moving body that changes at a low speed by distinction.

Further, in order to solve the above-mentioned problems, the second method of monitoring a moving body according to the present invention takes in an image of a target area at a predetermined cycle, and sequentially picks up captured images of the target area from the latest one at regular intervals. Remember only the number of sheets. A first difference image, which is the difference between the stored latest captured image and the adjacent captured image, and a second difference image, which is the difference between the latest captured image and the captured images that are separated by a predetermined interval, are formed, and the intermediate image obtained previously is formed. The second difference image is added to the image and multiplied by the attenuation coefficient, and the first difference image is subtracted from the image to form a composite image as an intermediate image.
While repeating the same process after that, the obtained composite image is binarized using a predetermined threshold value to obtain a binarized image, which is notified when a region larger than a predetermined area is detected from the image of the binarized image. It is characterized by doing.

According to the second moving object monitoring method of the present invention, the second difference image obtained by calculating the image difference between the latest image and the image temporally separated among the images acquired at equal intervals is used to detect a low speed such as a ship. Detect changing objects. Also here, the noise component that changes at high speed is detected by the first difference image obtained by taking the image difference between the latest image and the image immediately before it among the images acquired at equal intervals, and the first difference image is displayed in the second difference image. The noise components generated in the noise generation area in the second difference image such as the sea surface are canceled by repeatedly integrating the operated results for an appropriate number of times, and the wave of the sea surface or the change in the sunlight included in the second difference image, etc. It is possible to detect a moving body such as a ship that changes at a low speed from the background.

It is preferable that the term to be integrated is multiplied by a damping coefficient to suppress the degree of influence of past events so that it is not influenced by past events. The attenuation coefficient determines the width of the window, and how much past events are considered can be adjusted by the size of the attenuation coefficient. Further, in the composite image, the presence of a moving object can be detected by the shape in which the moving portions of the low-speed moving object that appear as a difference between the two images that are temporally separated overlap each other for a predetermined time. The overlapping of the moving parts is a figure formed at the front and rear ends of the moving body and having a width of the moving body and a length in the moving direction. Due to the influence of the damping coefficient, the time when the most recent moving part becomes thicker passes. It is displayed so thinly that it can be easily recognized by humans.

In the method of the present invention, in order to mechanically extract the moving body, the composite image is converted into a binarized image using an appropriate threshold value, and then the overlap of the moving portion is extracted,
When the overlapping area has an appropriate size, it is recognized as a moving body. The threshold value used at this time may be determined by taking the density histogram of the captured image and using the result. Noise generation conditions such as sea surface reflection are greatly influenced by the environment. Under the normal lighting condition in which the sunlight is emitted from the rear of the camera, the entire screen is uniformly bright and the spread of the density histogram is small. Therefore, the threshold value is lowered so that the moving object can be accurately detected. However, under backlight conditions, the high-brightness part due to the reflection of sunlight on the sea surface increases the contrast with other areas, and the effect of sea-level changes becomes greater than in normal light, which reduces the detection performance of the moving object itself. Even so, it is preferable to set a high binarization threshold value to remove fine noise. Since the width of the histogram is widened under such a condition, the threshold can be determined using the distribution of the histogram as an index.

In order to solve the above problems, the present invention
In the moving object monitoring device, the image pickup device monitors the target area in a predetermined cycle.
Captured image, captured image memory acquired by the imaging device
Record a predetermined number of captured images at regular intervals from the latest one.
Remember, the image processing device can calculate the difference between the latest image and the adjacent image.
Generates the first difference image by taking the absolute value, and sets it as the latest image
The second difference image is calculated by taking the absolute value of the difference from the image that is separated by a distance.
Second difference between the intermediate image generated and recorded in the intermediate image memory
The first difference is obtained by adding the minute images and multiplying them by the attenuation coefficient.
Subtract the minute images to form a composite image.
Is stored in the intermediate image memory, and the composite image is binarized.
Label the binarized image obtained by
Alarm device alerts when it detects a larger area
It is characterized by

The binarization threshold value may be determined on the basis of a histogram relating to the lightness of the photographed image. Further, it is preferable to saturate each pixel of the intermediate image with the upper limit value or the lower limit value. Since the digital display pixel value immediately returns to the lower limit value when it exceeds the upper limit, and returns to the upper limit value when it exceeds the lower limit, it is preferable to saturate in each direction so that the calculation result does not become abnormal. Is.

The moving body monitoring apparatus according to the present invention implements the second moving body monitoring method according to the present invention, and cancels noise components that change at high speed such as waves on the surface of the sea and the reflected state of sunlight to change moving bodies at low speed such as a ship. Can be detected. Since the intermediate image obtained as a result of image calculation including integration may have a pixel value exceeding the physical limit, an upper limit value and a lower limit value are set. Is preferably prevented from appearing. In addition,
It goes without saying that the present device can be widely used not only for a ship traveling on water, but also for detecting a moving body that changes at a low speed in a noise component that changes at a high speed.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described in detail below with reference to examples.

[0019]

[Embodiment 1] FIG. 1 is a block diagram of a moving body monitoring apparatus of this embodiment, FIG. 2 is its configuration diagram, FIG. 3 is a conceptual diagram for explaining the detection principle of the present invention, and FIG. 4 is image processing in this embodiment. FIG. 5 is a process diagram schematically showing an image during the image processing process.

The mobile object monitoring apparatus of this embodiment comprises a monitoring camera 1, an arithmetic processing unit 2, a display 3 and a speaker 4. The surveillance camera 1 is installed at a fixed point overlooking the harbor and the barrier, and supplies the acquired image to the arithmetic processing device 2. When the arithmetic processing unit 2 performs image processing on the acquired image based on the method of the present invention to detect a moving ship, the operator is notified of the result through the display 3 or the speaker 4. The operator can visually confirm the movement of the ship after receiving the alarm that the moving ship has been found. This device eliminates the need for the monitor to watch the display at all times, so that other work can be performed in parallel with the monitoring work, and labor and manpower can be reduced in port monitoring, etc. .

The surveillance camera 1 is used with its field of view fixed in the surveillance direction. Whether or not a moving object exists in the visual field is detected by using the difference between frames. That is, FIG.
As shown in Figure 2, two images with a time interval (a)
Image difference is performed between (b) and the obtained image is binarized using an appropriate threshold value to generate a binarized image (c),
Since the stationary object is at the same position in both images, the difference is 0 and it does not appear in the binarized image, but the moving object has different positions in both images and the part corresponding to the moving part is 1
Therefore, it can be detected because it remains as a high brightness area. However, when the monitoring area is the sea surface, a wind causes a wave front or the sun reflects on the wave surface even when there is no moving object. Difficult to detect.

Therefore, in the apparatus of the present embodiment, the inter-frame difference method is subjected to noise suppression processing for suppressing the influence of sea level changes, and further storage processing for relatively improving the detection sensitivity of an object moving at a low speed such as a ship is added. Therefore, the low-speed moving object is detected with a high recognition rate. This embodiment focuses on the fact that the speed of light ray change due to the reflection of the waves and sunlight on the sea surface is very high compared to the moving speed of the ship, and an extremely simple image with a band-pass filter function for detecting a slow change. It is characterized by being realized by processing.

As shown in FIG. 1, the arithmetic processing unit 2 for executing the image arithmetic has a first image memory 11 for internally storing a picked-up image, two image difference units 12, 13, an image arithmetic unit 14, and an intermediate unit. A second image memory 15 for storing images,
The image processing device 16, the determination device 17, and the notification device 18 for operating the display 3 or the speaker 4 are provided. The images captured by the surveillance camera 1 are sequentially stored in the first image memory 11 at appropriate intervals. The first image memory 11 is a memory stack in which the oldest image is erased each time a new image is captured, and a predetermined number of images are always held from the latest image.

The image storage interval is determined by the time interval at which a noise component that changes at high speed can be detected.
It may be 300 milliseconds. Further, the number of image storage frames is determined from an advantageous interval for detecting a moving body moving at a low speed in the image, and depends on the position of the surveillance camera 1, but is set to about 10 frames, for example. You can

The processing contents of the arithmetic processing unit 2 are shown in FIG. First, in order to set the initial state, n images (here, 9 images) are captured and stored in the first image memory 11 (S1). Further, as the initial setting, the previous moving object detection image SI _t-1 stored in the second image memory 15 is set to 0 (S2). On top of that, captures a new image I _t, starts the normal moving object detection process (S3). This new image I _t, is imprinted spot due to for example the hull and waves as shown in FIG. 5 (a).

The difference between the frames of the image is obtained by combining two types. First, in order to detect optical noise on the sea surface,
The image I at the latest time t is displayed in the first image difference device 12.
The absolute value of the difference is calculated with respect to the value of each pixel by taking _t and the image I _t-1 one frame before and calculating the difference image DI ₁ (that is, DI ₁ = | I _t −I _t-1 |). Difference image DI ₁
For example, as shown in FIG. 5B, a light spot on the sea surface is detected, but the hull portion is almost canceled and does not appear on the image.
It should be noted that the image used here is not limited to the latest image, and an image intermediate between the latest image and the images separated by n can be used.

At the same time, in order to reliably detect the movement of the vessel, the image I _t at the latest time _t and the image I _tn n frames before are _loaded into the second image difference device 13 and the absolute value of the difference is _calculated for each pixel. And the difference image DI _n (that is, DI _n
= | I _t −I _tn |). (S4). Here, n is 9, for example. Difference image DI _n is as shown in the FIG. 5 example (c), in addition to the light spot on the sea surface, the tip portion and the rear end portion of the moving hull remains as high brightness point.

Next, the image processing unit 14 inputs the two difference images DI ₁ and DI _n and calculates the moving body detection image SI _t based on the following equation (S5). _{SI t = FIX {a (SI} t-1 + DI n) - DI 1} where, FIX is a saturated operator, the minimum value 0 for the minimum value, a value of 0 or less of the pixel values, the pixel values A maximum value FF is taken for a value equal to or greater than the maximum value FF (hexadecimal number), and an intermediate value x thereof is taken as the value x. In addition, a is a damping coefficient and is a positive number from 0 to 1 and SI _t-1 is the second
3 is a moving object detection image previously stored, which is stored in the image memory 15 of FIG. The calculation is performed for each pixel position over the entire surface of the image. The attenuation coefficient a sets a window that limits the calculation target in the time axis direction. In the above equation, (SI _t-1 + DI _n ) is multiplied by the damping coefficient a,
DI _n may not be multiplied by the reduction coefficient a.

According to the above equation, the attenuation coefficient a is obtained by adding the difference image DI _{n in} which the moving part of the ship and the optical noise on the water surface are visualized to the moving object detection image SI _t-1 obtained last time.
The subtraction of the difference image DI ₁ in which only the optical noise on the water surface is visualized is multiplied by to obtain the moving object detection image SI _t , and the calculation is repeated. That is,
The moving body detection image SI _t includes a moving body detection image obtained so far multiplied by an attenuation coefficient a and integrated, and is obtained by subtracting the integrated image of the differential image DI ₁ from the integrated image of the differential image DI _n. Corresponds to a windowed version of.

Since the generation position of the light spot on the water surface is not determined, noise cannot be eliminated in one processing cycle even if the difference image DI ₁ is subtracted from the difference image DI _n, but the processing is repeated to integrate the noise. In the meantime, noise components are equally generated and canceled on the entire water surface, resulting in overall suppression of noise. Further, since the movement of the ship is very slow compared to the image capturing interval, most of the moving parts of the difference image DI _n between the images separated by n sheets overlap each time the image processing is performed, but the light spot noise is random. Since they are generated in the water, the contrast between the moving part of the ship and the light spot noise on the water surface becomes extremely clear by integrating these.

The moving body detection image SI _t obtained by the image calculation device 14 has an attenuation coefficient a as shown in FIG.
The leading end portion and the trailing end portion of the moving body are overlapped by a time width determined by, and the moving locus of the moving body can be seen. Due to the action of the damping coefficient a, newer events are darker and older events are expressed lighter. Also, the damping coefficient a
Past events that have run out of the virtual window defined by are deleted from the image. Light noise on the surface of the water is 2
Since the two difference images DI ₁ and DI _n are accumulated almost to the same degree, they are canceled and erased. However, among the light spots on the water surface, a small amount of the light spots that are not canceled by the two difference images remain in the image. The moving body detection image SI _t is stored in the second image memory 15 (S6) and used for the next calculation.

Next, a binarized image in which the moving area is bright and the stationary area is dark is generated from the moving object detection image SI _t to visualize the moving state of the moving object to facilitate the determination. First,
By creating a density histogram of the latest captured image I _t to determine the appropriate binarization threshold lightness distribution on the screen (S
7). The moving body moving portion in the moving body detection image SI _t is displayed as the lightness difference between the background portion and the moving body multiplied by a predetermined attenuation rate and then integrated.

For example, under a normal light condition in which the sunlight is radiated from the back of the camera, the light ray state becomes calm on the entire screen, so that an image in which the spread of the density histogram is small is obtained.
At this time, since the difference in brightness between the background and the moving body is not so large, it is necessary to set the binarization threshold of the moving body detection image SI _t low. On the other hand, in the case of the backlight condition in which the sun illuminates from the front of the camera, the sunlight reflected on the wavefront becomes a high-brightness light point that changes at high speed and enters the camera, so the density histogram on the screen is greatly expanded. . At this time, 2
If the binarization threshold is raised, fine noise is picked up and advanced image processing is required. However, since the difference in brightness between the background and the moving body becomes large, it is possible to easily remove the fine noise by raising the binarization threshold. it can. Therefore, to determine a binarization threshold based on the spread of the density histogram of the captured image I _t. Note that the captured image used to determine the binarization threshold need not be the latest one.

It is determined whether or not the brightness value is larger than the binarization threshold value for each pixel in the moving object detection image SI _t , and based on the result, the binarized image BI _t displayed with 0 and 1 is generated ( S8). In the binarized image BI _t , the moving area is bright and the stationary area is dark. The moving area portion in the binarized image BI _t has a length proportional to the width and moving speed of the moving body. The binarized image BI _t may also include light spot noise. In the mobile monitoring device of the present embodiment, by determining the binarization threshold based on the spread of the density histogram of the captured image I _t, thereby enabling appropriate binarization any case frontlit, backlit .

Next, in the binarized image BI _t, continuous regions are detected and numbered by the labeling process (S9). Then, the number of pixels in the region having the same number is counted and used as a value representing the area S of the continuous region (S10). The noise component has a small area S, and the area S of the moving area of the moving body is larger than the noise. Therefore, in order to determine the moving body, each area S is compared with a predetermined threshold S _th (S
11).

If the area area S is larger than the threshold value S _th , it is determined that a moving object exists, and the notification device 18 notifies the operator via the display 3 or the speaker 4 (S12). When the operator receives the notification, the operator can perform the necessary work by confirming the moving body again by the photographed image or the direct view displayed on the monitor. If the area area S does not reach the threshold value S _th , it is determined that there is no moving body, the process returns to the image capturing step (S3) and the same process is repeated. As described above, since the mobile body monitoring apparatus according to the present embodiment does not have to constantly monitor, the operator can perform the monitoring work in parallel with other work.

[0037]

Second Embodiment FIG. 6 is a flow chart of image processing in the moving body monitoring apparatus of this embodiment, and FIG. 7 is a process diagram schematically showing an image during the image processing process.

Since the moving object monitoring apparatus of this embodiment is different from that of the first embodiment only in the image processing step, different points will be mainly described with reference to FIGS. 6 and 7. First, in order to prepare for arithmetic processing, n images are captured and stored in the first image memory 1
Store in 1. In addition, the previous integration image Σ
D _nt-1 and ΣD _1t-1 are set to 0 (S21). On top of that, captures a new image I _t, starts the normal moving object detection process (S22). This new image I _t will image as shown in (a) of FIG. Then, the latest image in the first image difference device 12 I _t and the previous frame image I
The absolute value of the difference is calculated for _t-1 , and the difference image DI ₁ (that is, DI ₁ = | I _t −I _t-1 |) is obtained. Difference image D
I ₁ is as shown in FIG.

Further, the second image difference device 13 calculates the absolute value of the difference between the latest image I _t and the image I _tn n frames before and calculates the difference image DI _n (that is, DI _n = | I _t −I
_tn |) is calculated. (S23). This difference image DI _n is as shown in FIG. Next, the image processing device 14
Input two difference images DI ₁ and DI _n, and enter 2 according to the formula below.
Two integrated images ΣD _1t and ΣD _nt are calculated (S24). ΣD _nt = DI _n + a ΣD _nt-1 ΣD _1t = DI ₁ + a ΣD _1t-1 where a is an attenuation coefficient.

As shown in FIG. 7D, the integrated image ΣD _1t becomes a screen in which the hull portion is white and the light spots on the water surface are evenly spread. The newer spots are darker and the older spots are thinner. Further, as shown in FIG. 7E, the integrated image ΣD _nt becomes a screen in which, in addition to the light spot on the water surface, a dark region having a width in the moving direction appears at the front end portion and the rear end portion of the hull. The width of the region that appears before and after the hull is affected by the moving speed and damping coefficient. Also, the moving parts of the hull become darker as they are newer and thinner as they are older.

The image calculation unit 14 further uses the integrated image ΣD.
_Subtracting ΣD _1t from _nt, the moving object detection image SI _t (SI _t = Σ
D _nt −ΣD _1t ) is calculated (S25). The moving object detection image SI _t thus obtained is as shown in FIG.
The moving object detection image obtained by the moving object monitoring apparatus according to the first embodiment is almost the same, and the moving point moving at low speed can be detected by canceling the light spot noise that changes at high speed. The subsequent processing is the same as that in the first embodiment, and an appropriate binarization threshold value is determined from the density histogram of the captured image I _t , and the moving object detection image SI _t is binarized based on this to be binarized image. B
To generate the I _t (S26). The binarized image BI _t is shown in FIG.
It becomes like (g).

Next, the binarized image BI _t is labeled (S27), and the area S _{i of} each independent region appearing in the image is obtained (S28). Furthermore, compared with the predetermined threshold value S _th each region area S _i (S29), determines that the mobile if the threshold S _{t h} a larger region area S _i is present, the operator notification by the notification device 18 Yes (S3
0). If there is no region area S _i that has reached the threshold value S _th , it is determined that there is no moving body, and the process returns to the image capturing step (S22) and the same process is repeated.

[0043]

As described above, the moving object monitoring method and apparatus of the present invention can detect a relatively low speed moving object in an image in which a large amount of fast-changing noise is generated. It can be used to automatically detect a ship moving on the water surface such as a ship or a gate monitoring, and can realize labor saving and labor saving of the monitoring work.

[Brief description of drawings]

FIG. 1 is a block diagram of a mobile body monitoring device showing a first embodiment of the present invention.

FIG. 2 is a configuration diagram of the present embodiment.

FIG. 3 is a conceptual diagram illustrating a detection principle in the present invention.

FIG. 4 is a flowchart of image processing in this embodiment.

FIG. 5 is a process diagram schematically showing an image during an image processing process in this example.

FIG. 6 is a flowchart of image processing according to the second embodiment of the present invention.

FIG. 7 is a process diagram schematically showing an image during an image processing process in this example.

[Explanation of symbols]

1 surveillance camera 2 Processor 3 display 4 speakers 11 First image memory 12, 13 Image difference device 14 Image processing device 15 Second image memory 16 Image processing device 17 Judgment device 18 Notification device

Front Page Continuation (72) Inventor Masayuki Mitsue 118 Futatsuka, Noda City, Chiba Prefecture Kawaju Techno Service Co., Ltd. Kanto Plant (56) References JP-A-2-214283 (JP, A) JP-A-5-64177 ( JP, A) JP-A-3-144797 (JP, A) JP-A-9-81752 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) H04N 7/18

Claims (6)

(57) [Claims] 1. A captured image of a target area is captured at a predetermined cycle, and each time the captured image is captured, the captured image is captured.
And a first difference image by taking the absolute value of the difference between adjacent captured images, also a second difference image by taking the absolute value of the difference between the captured image and the captured image predetermined distance taken the said first A predetermined number of two difference images are added to the first difference image and the same predetermined number of images are added to subtract the image to form a combined image, and the combined image is binarized by using a predetermined threshold value. A method of monitoring a moving object, characterized by obtaining a binarized image and notifying when a region larger than a predetermined area is detected from the image of the binarized image. 2. A captured image of a target area is fetched at a predetermined cycle, and a predetermined number of captured images of the target area are stored at regular intervals in order from the latest one, and a difference between the latest captured image and an adjacent captured image is stored. The first difference image that is an absolute value and the second difference image that is an absolute value of the difference between the latest captured image and the captured images that are separated by a predetermined interval.
A difference image is formed, and the first difference image is subtracted from the image obtained by adding the intermediate image obtained previously and the second difference image and multiplying it by the attenuation coefficient to form a combined image as an intermediate image,
After that, the same steps are repeated, and during this period, the composite image is binarized by using a predetermined threshold value to obtain a binarized image, and when a region larger than a predetermined area is detected from the image of the binarized image, a notification is given. A mobile object monitoring method characterized by the above. 3. The moving object monitoring method according to claim 1, wherein the binarization threshold value is determined on the basis of a histogram relating to brightness in a captured image of the target area. 4. An image pickup device, an image processing device, an image memory, and an alarm device, wherein the image pickup device fetches an image of a target area at a predetermined cycle, and the image memory includes a picked-up image memory, an intermediate image memory, and a composite image memory. provided by stores equidistant images acquired of the imaging device in the captured image memory by sequentially a predetermined number to the most recent one, the image processing apparatus is the absolute value of the difference between the image and the adjacent latest image A first difference image and a second difference image having an absolute value of a difference between the latest image and an image separated by a predetermined distance are formed, and the intermediate image and the second difference image recorded in the intermediate image memory are imaged. Binarization obtained by adding and multiplying the attenuation coefficient to subtract the first difference image to form a composite image, storing the composite image in the intermediate image memory, and binarizing the composite image. Label the image A mobile object monitoring device, wherein an alarm is issued via the alarm device when an area larger than a predetermined area is detected. 5. The moving object monitoring apparatus according to claim 4, wherein the binarization threshold value is determined based on a histogram relating to brightness in an image of the target area. 6. The moving body monitoring apparatus according to claim 4, wherein the intermediate image is saturated at an upper limit value or a lower limit value for each pixel.