JP2889082B2 - Method and apparatus for predicting entry into prohibited areas - 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 an apparatus for predicting entry into a prohibited area, and more particularly to a method and apparatus for predicting the entry of a moving object into the prohibited area by monitoring an image of a monitored area including the prohibited area. .

[0002]

2. Description of the Related Art FIG. 7 shows an example of a conventional method for monitoring the intrusion of a prohibited area using an image. The image I of the monitored area including the prohibited area Z is sequentially captured by the image input unit 50 such as a TV camera,
Reference image 52 and image I of the monitored area in a normal state prepared in advance
Is extracted as a difference image, the difference image is binarized based on a predetermined threshold, and the monitoring target is identified from the shape characteristics of each connected component of the binarized image. Detecting the overlap of identified by the abnormality determining unit 55 connected component and the prohibited area image I _z, a malfunction occurs when overlapping the judges.

Further, the present inventors have disclosed an intrusion monitoring method based on a motion vector of an image I in Japanese Patent Application No. 3-306445. To the extent necessary for understanding the present invention, a monitoring method using a motion vector will be described with reference to FIG. Motion vector image forming means 8 1, divides the image I _t of the surveillance area image example 9 image input unit 5 is continuously generated blocks B fixed number of pixels, the current image B in _t of each block the image replacing the previous image B _t-1 of the block to the motion vector of the current image B _t, and outputs the vector image I _v in FIG.
Here before image B _t-1, the luminance difference between the previous image I _t-1 a the block of the same number of pixels of the image in and the current image B _t is smallest. That is, the motion vector corresponds to the moving amount of each block from the previous image to the current image. After the binarization of the image I _{v by} the binarization unit 2, the identification unit 3 identifies the monitoring target from the shape feature and the motion feature of each connected component,
For example outputs the image I _g of FIG. 11. The abnormality determination means 8
The number of overlapping blocks between the image _Ig and the prohibited area image _Iz shown in FIG. 12 is counted, and when the number of overlapping blocks exceeds a preset value, it is determined that there is an abnormality.

FIG. 13 and FIG. 14 show a method of predicting the intrusion of a monitored object using a motion vector. The connected component of the output image _Ig of the identification unit 3 is converted into a motion vector for each block by the conversion unit 30, and the first prediction processing unit 34a in the prediction state storage unit 32 stores the current position of each block (the hatched block in FIG. 14). ) And the motion vector to predict the connected component position after a predetermined time (τ). The prediction result is stored in the storage means 30, and the image I after a predetermined time (τ) is
_g, and the number N of consecutive hits is stored in the counter 39. Second prediction processing means 34b of abnormality determination means 40
The predicted position of the connected component after a time (ατ) obtained by multiplying the predetermined time (τ) by an integer (α) equal to or smaller than the continuous medium number N, similarly to the prediction processing of the first prediction processing means 34a (FIG. 14). , And a duplication detection means 42 detects duplication between the predicted position and the prohibited area image I _z (FIG. 12). 13 and 14
Is based on the assumption that, if the prediction is successful consecutively N times in the past, the same movement will be performed N times thereafter, and the position predicted by the current position of each block by N times the current motion vector is predicted. Position. In FIG. 14, the connected component shape between the current position and the predicted position is different.
This is because the prediction is performed for each block.

[0005]

In any of the conventional intrusion monitoring methods, an abnormality is determined from the overlap between the image of the monitored object and the image of the prohibited area. Therefore, the abnormality is detected only after the monitored object enters the prohibited area. Is detected. However, if the banned area is a dangerous area such as a construction site that is life-threatening, detection of an abnormality after intrusion may be too late. Therefore, it is necessary to detect an abnormality before the monitored object enters the prohibited area.

[0006] The intrusion prediction method using the motion vector is as follows.
Certainly, the movement of the monitored object is tracked, and the intrusion into the prohibited area is predicted from the past trajectory. However, since prediction / confirmation processing is performed in block units for intrusion prediction, a large amount of image processing is required. In many cases, the object to be monitored is a person or the like having a complicated movement, and its prediction requires many recognition processes such as meandering, sudden stop and restart of movement, shape deformation, and overlapping of objects. The complication causes a further increase in image processing. Since there is a limit to the prediction processing time, a certain limit is imposed on the prediction probability when too much image processing and recognition processing are required.

It is an object of the present invention to provide an intrusion prediction method and apparatus for predicting the intrusion of a monitored object into a prohibited area with a small amount of image processing.

[0008]

Referring to FIG. 1, the method for predicting entry into a prohibited area according to the present invention is based on the monitoring of an image I of a monitored area including the prohibited area Z, and the movement object 9 is prohibited.
A plurality of detection areas E surrounding the prohibited area Z in a layered manner around the prohibited area Z, and the outermost detection areas for the prohibited area Z and each detection area E are set. An area number N _i (N _i is a natural number) that increases by one as the vehicle approaches the prohibited area Z is set to 1, and a counter 17 is provided for each detection area E to store the continuous penetration C (N _i ) of the detection area E. . The degree of continuous penetration C in the detection area E (area number N _a ) where the object 9 was detected at the time of detection of the object 9
_t (N _a) adding a 1 to the outer adjacent detection area E (area number N _a-1) the previous consecutive penetration stored in the counter 17 of C _t-1 of the detection area E (N _{a-1) (C t (N a) =} C t-1 (N
_a-1 ) +1) and stores it in the counter 17 of the detection area E (area number N _a ).
This continuous penetration depth C _t (N _a) because the prediction penetration area number what area number, the largest of the detection area E of penetration of the object 9 is predicted P of the area number N _a and the detection area E ( calculated as N _a), to predict entry into prohibited area Z of the object 9 by comparison with the area number N _max of the predicted penetration area number P (N _a) and prohibited area Z.

Preferably, the current continuous penetration degree C _t (N _i ) of the detection area E (area number N _i , i ≠ a) other than the detection area E in which the object 9 was detected is set to zero, Detection area E where the continuous penetration degree C _t (N _i ) was detected this time as zero.
Counter 1 of the detection area E (area number N _i , i ≠ a) other than
Remember in 7. More preferably, the previous continuous penetration time is calculated continuously penetration C _t (N _a) is stored in the counter 17 of the detection area E in there is detected the detection area E (area number N _a) C _{t- 1} if (N _a) less than replaced by this continuous penetration C _t (N _a) of the previous continuous penetration _{C t-1 (N a)} .

Referring to FIG. 1, the apparatus for predicting entry into a prohibited area according to the present invention includes an image change identifying means 10 for identifying a moving object 9 from an image I of a monitored area including a prohibited area Z; A plurality of detection areas E surrounding the prohibited area Z in a layered manner are set around the area, and the outermost end is set to 1 for the prohibited area Z and each detection area E, and the area number N _i (N _i is a natural number) and the monitored area storage means 12 stores the position and area number of the prohibited area Z and the detection area E in the image I; the position of the object 9 and the prohibited area Z
And a position comparing means 11 for comparing the position of the detection area E with the position of the detection area E to obtain the detection area E where the object 9 has been detected; and having a counter 17 for storing the continuous penetration degree C (N _i ) for each detection area E. The degree of continuous penetration C _t (N _a ) of the detection area E (area number N _a ) in which detection has been performed is stored in the counter 17 of the detection area E adjacent to the detection area E (area number N _a-1 ). Adding 1 to the previous continuous penetration degree _Ct-1 (Na _-1 )
_{t (N a) = C t} -1 (N a-1) +1) is calculated by and stored in the counter 17 of the detection area E, area number had detected the detection area E N _a and the detection area E Prediction processing means 16 for obtaining, as the predicted intrusion area number P (N _a ), the detection area E in which the intrusion of the object 9 is predicted from the current continuous intrusion degree C _t (N _a ) from the current detection area E; and it becomes comprise final determination means 18 for outputting an intrusion prediction signal by comparing the area number N _max of the predicted penetration area number P (N _a) and prohibited area Z. Preferably, the prediction processing means 16 _obtains the current continuous penetration degrees C _t (N _i ) of all the detection areas E (area number N _i ) and stores them in the counter 17 of each detection area E.

[0011]

FIG. 2 is a diagram showing a detection area consisting of eight layers around the prohibited area Z.
Divided into rear E, each detection area E continues from the outside to the inside
Area number N_i(= 1, 2, ...)
I have identified. That is, the outermost detection area E is set to the area number.
No. 1 and prohibited area Z is designated as area number 9 (hereinafter referred to as prohibited area).
N is the area number of Z_maxIn some cases). However
Each detection area E is an area surrounding the prohibited area Z in a layered manner.
The shape and number are not limited to the illustrated examples.
No. Position and area number of each detection area E and prohibited area Z
The signal is stored in the monitored area storage means 12. Also counter 17
Are provided corresponding to the prohibited zone Z and each detection area E.
Degree of continuous penetration C obtained by the procedure described below _t(N_i)
You. FIG. 2A shows the state of each counter 17 in the initial state (t = 0).
Content C_t(N_i) Is zero. In FIG. 2, continuous
In order to facilitate calculation of the intrusion degree, all detection areas E
The outside is defined as area number 0, which corresponds to area number 0.
A counter 17 is provided.

The image change identification means 10 (FIG. 1) detects the position of the detection target 9 from the continuously generated image I, and the position comparison means 11 (FIG. 1) determines the position of the detection target 9 and the area to be monitored. Storage means 12
Is compared with the position of each detection area E to determine the detection area where the detection was made (area number N _a , hereinafter sometimes referred to as the detection area number). FIG. 2B shows that the area number detected at t = 1 is 1. This detection target 9
Is detected for the first time in the monitored area, so that the current continuous penetration degree C ₁ (1) of area number 1 is set to 1. In this case, by adding 1 to the content of the counter 17 outside the detection area (area number 0) which is always zero, the continuous penetration degree C
_{1 It} may be considered that (1) is calculated. FIG. 2C shows that the detection area number at t = 2 is 2. This detection target 9 can be regarded as the detection target 9 detected at area number 1 at the previous point in time (t = 1) having entered beyond the area boundary. Therefore, this time continuous penetration degree C of area number 2
₂ (2) is calculated by adding 1 to the previous continuous penetration degree C ₁ (1) of area number 1 (C ₂ (2) = C ₁ (1) +1).
FIG. 3D shows that the detection area number at t = 3 is 3, which is regarded as a result of the intrusion of the detection target 9 detected by the area number 2 at the previous time point (t = 2) in the same manner as described above. Therefore, this time continuous penetration degree C ₃ (3) is obtained by adding 1 to the previous continuous penetration degree C ₂ (2) of area number 2 (C ₃ (3) = C ₂ (2) +
It is determined by 1). In short, the continuous intrusion degree means the number of continuous times that the detection target 9 crosses the boundary of the detection area from outside to inside, and indicates that the target 9 continuously approaches the prohibited zone Z. Detection area N _a of this continuous penetration C _t (N _a) can be generally obtained by the following equation.

C _t (N _a ) = C _t-1 (N _a-1 ) +1 (1)

If it is assumed that when the vehicle continuously approaches the prohibited area in the past, the vehicle will approach in the future, the future intrusion degree can be predicted from the above continuous intrusion degree C _t (N _a ). For example, as shown in the following equation, this continuous penetration C _t (N _a) the prediction coefficient k obtained by multiplying the product and the area number of the detection area (N _a)
It is possible to calculate the predicted penetration area number P (N _a) by adding and.

[0015] _{P (N a) = kC t} (N a) + N a ......... (2) However, 0 ≦ k ≦ 1

Here, the prediction coefficient k adjusts the weight of the current intrusion degree C _t (N _a ) in the prediction. When the prediction coefficient k is close to 1, the prediction is large, and when the value is 0, the prediction is not included.
Prediction penetration area number P (N _a) is when the above area number N _max of prohibited zone Z likely from entering the prohibited area Z, it is possible to predict the penetration, for example, by the following equation.

P (N _a ) ≧ N _max : Intrusion is predicted... (3) P (N _a ) <N _max : Intrusion is not predicted... (4)

FIG. 2 (E) is detected area number N _a is 5, indicating that this continuous penetration was calculated in the detection area C _t (N _a) is 5. In this case, the prediction coefficient k
The calculated and predicted penetration area number P (N _a) is 9 (= 0.8 × 5 + 5) in 0.8 Tosureba area number 5 by the equation (2),
Since it is equal to the area number 9 of the prohibited area Z, it can be predicted from equation (3) that there is a possibility of intrusion. This time continuous penetration degree C
_t (N _a), after calculating the predicted penetration area number P (N _a),
It is stored in the counter 17 of the detection area (area number N _a ) detected for the next prediction process.

According to the present invention, since the intrusion prediction is performed based on the detection of the object 9 in each detection area E around the prohibited area Z instead of following the detection object 9 itself, the motion trajectory and the motion of the detection object 9 are determined. and without requiring a complicated image processing and recognition processing on the time variation of the shape, essentially important subject by continuous penetration C _t (N _i) simple conditional judgment processing counter 17 that stores the detection area N _i The approach of the object 9 to the prohibited zone Z can be continuously predicted.

Accordingly, the object of the present invention is to provide "an intrusion prediction method and apparatus for predicting the intrusion of a monitored object into a prohibited area with a small amount of image processing".

Preferably, the current continuous penetration degree C _t (N _i ) of the detection area E (area number N _i , i ≠ a) other than the detection area E in which the detection was performed is set to zero. That is, the present continuous intrusion degree in each detection area E is obtained by the following equation.

C _t (N _a ) = C _t-1 (N _a-1 ) +1: When an object is detected = 0: When no object is detected (5)

As described above, by setting the current continuous penetration degree C _t (N _i ) of the detection area E in which the object 9 is not detected to zero,
Stopping the accumulation of successive penetration depth when the detection object 9 moves away the boundary detection area from prohibition zone Z across from the inside to the outside, this continuous penetration depth C _t of had detected detection area N _a (N _a ) Can be returned to 1. Further, the current continuous penetration degree C _t (N _i ), which is set to zero, is stored in the counter 17 to prepare for the next and subsequent prediction processing. For example, as shown in FIG. 3A, when the object 9 is detected while repeatedly approaching / separating from the prohibited area Z, the continuous penetration degree C _t (6) in the detection area 6 where the detection was made is calculated. , The continuous intrusion degree 3 from the detection area 4 where the continuous intrusion in the prohibited area Z direction has started to the detection area 6 can be obtained.

[0024] In this case, more preferably, this calculated in had detected detection area N _a continuous penetration depth C _t
(N _a) is the previous continuous penetration depth of the detection area N _a C _t-1 (N
_a) it is smaller than, replaces this continuous penetration C _t a (N _a) by a previous continuous penetration C _t-1 (N _a). That is, the degree of continuous penetration C _t (N _a ) is obtained by the following equation.

C _t (N _a ) = C _t-1 (N _a-1 ) +1: When detection is performed and C _t (N _a )> C _t-1 (N _a ) = C _t-1 (N _a) ): Detected and when C _t (N _a ) ≦ C _t-1 (N _a ) = 0: When not detected ……… (6)

This is because, when the width of the detection area E is large as shown in FIGS. N _a and the current detection area N _a is a measure of the case is the same. Before time, continuous intrusion of C _t-1 of a detection detection area N _a (N _a) before last consecutive penetration stored in the detection area N _a-1 of counter 17 is C _t-2 (N _{a- 1} ), the detection area Na _-1 does not detect the target 9, and therefore the continuous penetration _Ct-1 (Na _-1 ) becomes zero. At this time,
Since zero is recorded in the counter 17 of the detection area N _a-1, the continuous penetration degree C _t (N _a ) of the detection area N _a this time is calculated by the equation (5).
Calculate from this to be 1. Therefore, the previous continuous penetration depth stored in the counter 17 of the detection area N _a C
With _t-1 (N _a) was held as it is this continuous penetration C _t (N _a), to avoid the loss of consecutive penetration is approaching continuation information of the object 9.

[0027]

FIG. 1 shows an embodiment of an intrusion prediction apparatus according to the present invention. An image I around the prohibited area is acquired by an image input unit 5 such as a camera, and a change in the image I to be monitored that is different from normal is identified and extracted by an image change identification unit 10. FIG.
The image change identifying means 10 is composed of, for example, the motion vector forming means 1, the binarizing means 2, and the identifying means 3 in FIG. However, an apparatus for identifying a change from the difference image of the luminance change shown in FIG. 7 may be used.

The monitored area storage means 12 shown in FIG. 1 is connected to a detection area setting means 13 having a display section 14 and an operation section 15, and displays an image I of the image input means 5 on the display section 14, for example. The detection area E can be set by setting means such as a mouse and a mouse. Further, as shown in FIG. 4, it is also possible to set only the prohibited area Z from the operation unit 15 and automatically set the detection area E around the prohibited area Z at predetermined intervals by the detection area setting means 13 as appropriate. The finally set position and area number of the detection area E are stored in the monitored area storage means.
Stored in 12.

The position comparing means 11 is provided to the image change identifying means 10.
The position of the object 9 and the information stored in the monitored area storage means 12
The position of the detection area E is compared with, for example, the detection of the object 9
Area number N_aOutput as Prediction processing
The means 16 has a counter 17 for each detection area E,
Area number N from position comparing means 11_aAnd the formula (6) and the formula
Based on (2), the calculating means 16a calculates the predicted intrusion area number P
(N_a) Is calculated. Figure 5 shows the flow of calculating the predicted intrusion area number.
An example of the drawing is shown. In step 501, the target 9 is invaded.
Is detected in the detection area where the
Detection area N_aIf so, proceed to step 502, this time
Continuous penetration C_t(N_a) Is calculated based on equation (6). Request
This time continuous penetration degree C_t(N_a) Indicates the same detection area N_aof
Previous continuous penetration C_t-1(N_a) And step 504,
Previous continuous penetration C_t-1(N_aIf) is larger, continue this time
Penetration C_t(N_a) To the previous continuous penetration degree C_t-1(N_aReplace with)
(Step 505). If the detection area E is not detected,
Go to 503, this time continuous penetration degree C_t(N_a) Is always zero
It is. Predicted intrusion area number in step 506 based on equation (2)
P (N_a) Is calculated, and the above cycle is applied to all the detection areas E.
Repeat for Prediction processing for all detection areas E is completed
After this, the degree of continuous penetration C of this detection area E _t(N_a)
Each is stored in the corresponding counter 17, for example, each detection error
The predicted intrusion area number P (N_a) Maximum
The value is output to the final determination means 18.

FIG. 6 shows an example of a flowchart of the final judgment. First the area number N _a detection area where there is detected at step 601 and the area number N _max of prohibited zone Z are compared, if they are equal to output a warning sound or message signal has intrusion (step 603). In step 602, the prediction processing means 16
Prediction penetration area number P determined (N _a) and prohibited area Z area number N _max of are compared, predicted penetration area number P (N _a)
If is larger, for example, an alarm sound or a message signal with intrusion prediction different from the alarm sound with intrusion is output (step 604).

[0031]

As described above in detail, the method and apparatus for predicting entry into a prohibited area according to the present invention sets a plurality of detection areas surrounding the prohibited area in a layered manner around the prohibited area, and intrudes into each detection area. Since the intrusion of the monitored object into the dangerous area is predicted from the detection result, the following remarkable effects are obtained.

(1) Since complicated image processing and recognition processing such as prediction of the motion trajectory of the detection target are not required, it is possible to predict intrusion into the prohibited area with a small amount of image processing. (2) The approach to the prohibited area can be predicted from the detection in each detection area around the prohibited area without tracking the detection target itself. (3) The approach to the prohibited area can be continuously predicted by the counter for each detection area and the simple condition judgment processing. (4) It is possible to freely adjust the shape and number of surrounding detection areas according to the degree of danger and the shape of the prohibited area. (5) By using a computer, it is possible to automatically predict entry into a prohibited area.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram showing a configuration of the present invention.

FIG. 2 is an explanatory diagram showing the principle of the present invention.

FIG. 3 is another explanatory diagram showing the principle of the present invention.

FIG. 4 is an explanatory diagram showing a detection area according to the present invention.

FIG. 5 is a flowchart of a prediction process according to the present invention.

FIG. 6 is a flowchart of the final judgment of the present invention.

FIG. 7 is an explanatory diagram of a conventional monitoring device.

FIG. 8 is an explanatory diagram of a conventional monitoring apparatus using a motion vector.

FIG. 9 is an explanatory diagram of a monitored area image I.

FIG. 10 is an explanatory diagram of a motion vector image _Iv .

FIG. 11 is an explanatory diagram of a binary detection image _Ig .

FIG. 12 is an explanatory diagram of a prohibited area image _Iz .

FIG. 13 is an explanatory diagram of a conventional monitoring device with prediction.

FIG. 14 is an explanatory diagram of a conventional prediction process.

[Explanation of symbols]

1 motion vector image forming means 2
Binarization means 3 Identification means 5 Image input means 7
Prohibited area storage means 8 Abnormality judgment means 9 Detection target 10
Image change identification means 11 Position comparison means 12 Monitored area storage means 13
Detection area setting means 14 Display section 15 Operation section 16
Prediction processing means 16a Calculation means 17 Counter 18
Final judgment means 20 Shrink diffusion means 22 Labeling means 24
Shape feature extraction means 26 Motion feature extraction means 28 Comparison means 30 Motion vector conversion means 32
Prediction state storage means 34a, 34b Prediction processing means 36 Storage means 38
Prediction match judgment means 40 Abnormality judgment means 42 Duplication detection means 50
Image input unit 51 Difference image extraction unit 52 Reference image 53
Brightness change binarization unit 54 Object identification unit 55 Abnormality judgment unit 56
Prohibited area storage.

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Jinichi Kikuchi 3-20-8 Narita Nishi, Suginami-ku, Tokyo Inside Okura Electric Co., Ltd. (72) Inventor Yutaka Maeda 3--20 Narita Nishi, Suginami-ku, Tokyo No. 8 Okura Electric Co., Ltd. (56) References JP-A-5-143737 (JP, A) JP-A-5-14892 (JP, A) JP-A-62-136988 (JP, A) (58) Survey Field (Int.Cl. ⁶ , DB name) G06T 7/20

Claims (6)

(57) [Claims] 1. A method for predicting the intrusion of a moving object into a prohibited area by monitoring an image of a monitored area including a prohibited area, wherein a plurality of detection areas surrounding the prohibited area in a layered manner around the prohibited area. The outermost detection area is set to 1 for the prohibited area and each detection area, and an area number N _i (N _i is a natural number) that increases by one as approaching the prohibited area is assigned. A counter that stores the continuous penetration degree C (N _i ) of the detection area is provided, and the current continuous penetration degree C _t (N _a ) of the detection area (area number N _a ) where the object was detected at the time of detection of the target object is used as the counter. Adding 1 to the previous continuous penetration degree C _t-1 (N _a-1 ) stored in the counter of the detection area adjacent to the outside of the detection area (area number N _a-1 ) (C _t (N _a ) = C _t-1 (N _a-1 ) +1) and the counter of the detection area And the area number N _a of the detection area where the intrusion of the object is predicted from the area number N _a of the detection area where the detection has been performed and the current continuous penetration degree C _t (N _a ) of the detection area. the determined as the prediction penetration area number P (N _a) those of the predicted penetration area number P (N
_a ) A method for predicting entry into a prohibited area by comparing the object with the area number _Nmax of the prohibited area by predicting entry of the object into the prohibited area. 2. The intrusion prediction method according to claim 1, wherein the predicted intrusion area number P (N _a ) is the area number N _a of the detection area in which the detection was performed and the current continuous intrusion degree C _t (N _a) to (adding and N _a) (P (N _a) prediction coefficient k obtained by multiplying the product _{kC t = kC t (N a} ) + N a) penetration method of predicting the prohibited zone formed by 5101. 3. The intrusion prediction method according to claim 1 or 2, wherein the current intrusion degree of a detection area (area number _Ni, i ≠ a) other than the detection area in which the object was detected when the object was detected. C _t (N _i ) is set to zero, and the continuous penetration degree C is set to zero this time.
_t (N _i ) is the detection area other than the detection area where the detection was made
A method of predicting entry into a prohibited area stored in the counter of (area number _Ni, i ≠ a). 4. A penetration prediction method of claim 3, wherein the detection there was detection area (the area number N _a) this continuous penetration C _t calculated in (N _a) is the counter for storing the detection area If has been last continuous penetration C _t-1 (N _a) smaller than said to this continuous penetration C _t (N _a) the previous continuous penetration C _t-1 (N _a) by replacing and becomes prohibited area Intrusion prediction method. 5. An image change identification unit for identifying a moving object from an image of a monitored area including a prohibited area; a plurality of detection areas surrounding the prohibited area in a layered manner around the prohibited area; And an area number N _i (N _i is a natural number) that is incremented by one as the outermost end is set to 1 for each detection area and approaches the prohibited area, and the position and area number of the prohibited area and each detection area in the image Monitored area storage means for comparing the position of the object with the positions of the prohibited area and each of the detection areas to obtain _a detection area (area number N _a ) where the object was detected A counter for storing the continuous penetration degree C (N _i ) for each of the detection areas, and the present continuous penetration degree C _t (N _a ) of the detection area (area number N _a ) where the detection has been made is used as the detection area. Outside adjacency detection Adding 1 to the previous continuous penetration degree C _t-1 (N _a-1 ) stored in the counter of the rear (area number N _a-1 ) (C _t (N _a ) = C _t-1 (N _{a -1)} +1) is calculated by and stored in the counter of the detection area, the object from this continuous penetration C _t (N _a) of area number N _a and the detection area of the detection area where there is the detection and predicted invasion area number P (N _a) and the area number of the prohibited area; predictive processing unit area number of the detection area intrusion is expected seek the maximum of those predicted penetration area number P (N _a) A device for predicting intrusion into a forbidden area, comprising a final determination means for outputting an intrusion prediction signal by comparing with _Nmax . 6. The intrusion prediction device according to claim 5, wherein said prediction processing means is provided for all said detection areas (area number N _i ).
A device for predicting intrusion into a forbidden area, which is obtained by calculating the degree of continuous intrusion C _t (N _i ) this time and storing it in the counter of each detection area.