JP3263311B2 - Object detection device, object detection method, and object monitoring system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an object detection device, an object detection method, and an object monitoring system for detecting an object such as a flame or an intruder (for example, an intruder).

[0002]

2. Description of the Related Art Conventionally, as an object detecting device for detecting a flame generated in a fire or the like, a device using an ultraviolet type or infrared type sensor (sensor) is generally known. Here, the UV-type sensor senses light in the ultraviolet region emitted from the flame using a discharge tube such as a UVtron having a high sensitivity characteristic to ultraviolet light, and responds to the amount of light in the ultraviolet region. It has a function to output a number of discharge pulses, and a flame detection device using an ultraviolet detector detects whether or not it is a flame based on the number of discharge pulses output from the ultraviolet detector. Are doing. In addition, the infrared sensor is
It has a function to detect light in the infrared region emitted from the flame, especially light with a wavelength of about 4.3 microns of the resonance radiation of carbon dioxide. In a flame detection device using an infrared detector,
The detection of the presence or absence of a flame is performed based on the amount of light in the infrared region detected by the infrared sensor and its fluctuation.

[0003]

However, ultraviolet detectors are also sensitive to light emitted from an arc discharge such as welding, and therefore, in a flame detector using an ultraviolet detector, When an operation such as welding is performed, erroneous detection may be performed. Infrared detectors are also sensitive to sunlight that includes the infrared region,
In a flame detection device using an infrared detector, for example, sunlight reflected by a rippled water surface (fluctuation
(Infrared light with flickering) may cause erroneous detection. In particular, an infrared flame detector using a pyroelectric element as a sensor is susceptible to various noises because it uses an amplifier with a high amplification degree to detect flicker of the flame. In particular, popcorn noise is generated by electronic components such as operational amplifiers (operational amplifiers), capacitors, and pyroelectric sensors due to sudden temperature changes and mechanical stress. Popcorn noise generated by the mold element has the greatest effect.

[0004] Some flame detectors use a thermal imager. In a flame detector using a thermal imager,
The number of pixels (pixel number, that is, area) having a pixel value equal to or greater than a predetermined threshold value among the pixels captured as a thermal image is obtained, and it is detected whether or not there is a flame based on this pixel number (area). ing. However, in a flame detection device using a thermal imaging device, since the thermal imaging device has a high sensitivity characteristic to a high-temperature object, an image of a high-temperature object other than the flame may be taken and erroneous detection may be performed. is there.

[0005] The above-mentioned problems in the flame detection device are as follows.
This also occurs in an intruder detection device (for example, an intruder detection device) used in a security system. That is, when an infrared heat sensor module (pyroelectric element) that detects the difference between the emitted energy due to the background temperature and the emitted energy due to the human body or the like is used as the intruder detection device, false alarms due to popcorn noise specific to the pyroelectric element, Small animal
(Cats, rats, etc.) and false reports due to rapid temperature changes.

SUMMARY OF THE INVENTION The present invention provides an object detection device which can prevent erroneous detection due to factors other than a flame or a predetermined intruder (for example, an intruder), and can reliably detect an object such as a flame or an intruder. It is an object to provide an object detection method and an object monitoring system.

Another object of the present invention is to provide an object detection device and an object monitoring system capable of detecting both a flame and an intruder (for example, an intruder) with a single device having a simple configuration. It is an object.

[0008]

In order to achieve the above object, according to the present invention, an image pickup means and an object image area are extracted from an image picked up by the image pickup means. Image processing means for calculating information on the target object based on the pixel information of the image processing means, and determination means for determining whether or not the target object is an intruder based on the information on the target object determined by the image processing means. The image processing means may include, as the information, at least one of a shape similarity ratio of an object image region, a circularity of the object image region, a size of the object image region, and a moving speed of the object image region. It has become a split out of Suyo, and also, in the judgment means
Is the shape similarity ratio of the object image area, the object image area
Circularity of object, size of object image area,
Shape similarity ratio, circularity, size corresponding to each moving speed
Now, the membership is set with the moving speed as a parameter.
Loop function is set in advance, and the determining means
The shape similarity ratio of the object image area by the image processing means, the object
Circularity of image area, size of object image area, object image
When each parameter value of the moving speed of the image area is calculated
And the confidence for each parameter value determined
Calculated by the membership function corresponding to
It is characterized in that it is determined whether or not the target object is an intruder based on the certainty factor obtained for the meter value .

[0009]

According to a second aspect of the present invention, in the object detecting apparatus according to the first aspect, one CCD having sensitivity to light in a visible region is used as the imaging means.

According to a third aspect of the present invention, an image pickup means, and an object image area is extracted from the image picked up by the image pickup means, and based on the pixel information of the object image area,
An object detection apparatus, comprising: an image processing unit that determines information on an object; and a determination unit that determines whether the object is a predetermined object based on information on the object determined by the image processing unit. And has both a flame detecting function and an intruding object detecting function. The processing function of the imaging means, the image processing means, and the judging means is detected by flame detection depending on whether the detection target is a flame or an intruding object. Function or
Switching control means for performing control for switching to an intruder detection function is further provided, and the switching control of the switching control means causes the target detection device to function as a flame detection device or to function as an intrusion detection device. It is characterized by being made to let.

According to a fourth aspect of the present invention, in the object detection device of the third aspect , the switching control means performs flame monitoring (or intrusion monitoring) while performing intrusion monitoring (or flame monitoring). It is characterized in that the state of the target detection device is switched and controlled so as to perform the operation.

According to a fifth aspect of the present invention, in the object detection device of the third aspect , the switching control means performs switching control by setting an intruder and a flame as equal monitoring targets. It is characterized by having.

According to a sixth aspect of the present invention, in the object detection device of the third aspect , the switching control means causes the object detection device to function as, for example, a flame detection device in a normal state, and a monitoring state. Is characterized in that switching control is performed so as to function as an intruder / flame detection device that detects both flames and intruders.

[0015] The invention of claim 7, wherein the claim 3乃
7. The object detection apparatus according to claim 6 , wherein the switching control means is configured to be capable of manually switching control between intruder monitoring and flame monitoring.

According to an eighth aspect of the present invention, in the object detection apparatus according to the third aspect, one CCD having sensitivity to light in a visible region is used as the imaging means, and the switching control means includes: In the intruding object monitoring state, the light entering the CCD is controlled so that the light in the visible region enters the CCD as it is, while in the flame monitoring state, the visible light cut light enters the CCD. It is characterized by.

[0017]

Further, an invention according to claim 9, wherein, intruder unique feature is preset membership functions and parameter information to be well represented, by imaging the object
When information that uniquely expresses the unique characteristics of the intruder is determined from the image data obtained, the degree of certainty that the target is an intruder is determined from the membership function using the determined information as a parameter value. It is characterized in that it is determined whether or not the target is an intruder based on the obtained certainty factor.

The invention according to claim 10 is the first invention.
An object detection device according to any one of claims 1 to 8 , wherein the object detection device is connected to a receiver via a predetermined transmission path, and the object detection device detects an object when called from the receiver. It is characterized by transmitting the result.

The invention according to claim 11 is an object monitoring system in which an object detection device for detecting a predetermined object is connected to a receiver via a predetermined transmission path, wherein the object detection device is , Equipped with both a flame detection function and an intruder detection function, and in the receiver, switching between functioning the target detection device as a flame detection device or functioning as an intrusion detection device Switching control means for performing control is provided, and according to a switching control command from the switching control means of the receiver, the target detection device is caused to function as a flame detection device or to function as an intruder detection device. It is characterized by being.

[0021]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a configuration example of a target detection device according to the present invention. Referring to FIG. 1, the object detection device extracts an object image region from an image captured by an image capturing unit 1 and an image captured by the image capturing unit 1, and outputs information about the object based on pixel information of the object image region. An image processing unit 2 to be determined, and a determination unit 3 that determines whether or not this is a predetermined target, that is, a detection target (a flame or an intruder), based on information on the target determined by the image processing unit 2.
And

Here, the image pickup unit 1 has an optical system 11 and a photoelectric conversion unit 12 that receives light from the optical system 11 and converts the light into an electric signal. In the example of FIG. 1, the optical system 11 includes a lens system 13 having an angle of view (viewing angle) corresponding to a predetermined target monitoring area, and a spectrum (wavelength) of light entering the photoelectric conversion unit 12 from the lens system 13. And an optical filter 14 that restricts The photoelectric conversion unit 12 includes a photoelectric conversion element having a predetermined number of pixels, for example, a two-dimensional CC.
D is used.

The angle of view (viewing angle) of the lens system 13 is monitored
(Flame and / or invaders) size and photoelectric conversion element
It is set appropriately based on the resolution (total number of pixels) of (CCD). Specifically, the wider the angle of the lens system 13 is, the larger the viewing angle becomes, and the lens system 13 can cover a wider monitoring area. In this case, the photoelectric conversion element (C
In CD), the size of the monitoring area corresponding to one pixel is also wide, and when the size of the target is small, information on the target can be obtained from only one pixel of the photoelectric conversion element, and the information amount on the target can be obtained. Is significantly reduced. Therefore, the viewing angle of the lens system 13, that is, the size of the monitoring area corresponding to one pixel of the photoelectric conversion element is determined by the number of pixels necessary to sufficiently obtain information on the object even when the size of the object is small. (Preferably multiple pixels)
It is preferable that the object can be imaged (monitored).

For example, if the object to be detected is a flame and the object detecting device is a flame detector, the standard for approval of the type of the flame detector according to the Fire Service Law specifies that a flame test of a flame detector should be 33 × 3.
Using a 3 cm ² area fire tray (33 cm square fire tray), normal heptane is put in the fire tray and ignited, and detection performance within 30 seconds is required. The combustion flame of normal heptane in this grate (hereinafter referred to as the flame of a 3.3 cm square grate) has a width,
Since the heights are about 50 cm and about 1.5 m, respectively, the viewing angle of the lens system 13 (that is, 1
As the size of the monitoring area corresponding to the pixel), the number of pixels that can effectively capture the fluctuation of the combustion flame from a position separated by a predetermined distance from the combustion flame, and has a size that can image (monitor) the flame Good thing.

As the photoelectric conversion element of the photoelectric conversion unit 12, a photoelectric conversion element having a sensitivity characteristic according to a detection target (a flame or an intruding object) can be selected and used. For example, when the detection target is a flame, an infrared CCD having high sensitivity to infrared light emitted from the flame, particularly light having a wavelength of about 4.3 microns of carbon dioxide resonance radiation is used. Most preferably, a normal CC sensitive to light in the visible region
D (even a normal CCD, which has a certain sensitivity in the near infrared region) can also be used. On the other hand, when the detection target is an intruder, a normal CCD having sensitivity in the visible region is preferably used.

When an infrared CCD is used as a photoelectric conversion element for detecting a flame, the optical filter 14 of the optical system 11 cuts light in a wavelength band other than the wavelength band of around 4.3 μm. It is preferable to use a bandpass filter having a characteristic of transmitting light in a wavelength band around 3 microns. When a normal CCD having sensitivity to light in the visible region is used as the photoelectric conversion element for detecting a flame, the optical filter 14 may be a near-infrared low-pass filter (for example, a wavelength of 8500 ° or less). Cut the light,
The visible light region may be cut using a filter that transmits light having a wavelength of 8500 ° or more. That is, the optical filter 14 having a characteristic according to the spectral sensitivity characteristic of the photoelectric conversion element is used.

When the object to be detected is, for example, a flame, the total number of pixels of the CCD is about 50 cm and about 1.
A lens system 1 from a position where a flame having a size of 5 m is separated by a predetermined distance.
When the image is taken through the camera 3, it is necessary that the image area corresponding to the flame of this size has a resolution such that it covers a plurality of pixels. FIG. 2 shows a pixel configuration example of a two-dimensional CCD. In the example of FIG.
It has N and M pixels (total number of pixels N × M) in the x and y directions, respectively. When a flame of a predetermined size is imaged as a target object at a predetermined distance, the flame The corresponding image area (object image area) OBJ extends over five pixels.

The photoelectric conversion unit 12 of the image pickup unit 1 picks up an image at a predetermined time interval and supplies the image to the image processing unit 2. The image data as shown in FIG. 2 taken at predetermined time intervals is sequentially captured, and predetermined image processing necessary for object detection (flame detection or intruder detection (for example, intruder detection)) is performed. To obtain the information.

More specifically, the image processing unit 2 includes the photoelectric conversion unit 1
An A / D conversion function is provided for performing, for example, binarization processing on image data from two (two-dimensional analog image data) and converting analog image data into binary image data. Here, the threshold value a _{th for} binarization generally uses a certain value between the highest luminance value and the lowest luminance value in one image data, but the image data obtained every time at a predetermined time interval is used. In this case, the highest luminance value and the lowest luminance value are not always the same value and generally change. Therefore, it is preferable to change the threshold value for binarization in accordance with the change in the highest luminance value and the lowest luminance value. As a result, it is possible to remove the influence of the instantaneous change at a certain point in time. Further, the image processing unit 2 performs feedback control on the photoelectric conversion unit 12 so that a predetermined dynamic range can be obtained without saturating the image data having the highest luminance in the A / D conversion processing, thereby performing automatic gain adjustment control of the image data. It may have a function to perform.

In any case, the image processing unit 2 includes the imaging unit 1
In the analog image data a _x for each pixel position that is captured (x, y), with respect to _y, for example, a _x, when _y ≧ a _th pixel values p _x of the pixel position (x, y), _{y is set} to “1” (black pixel), and when a _x , _y <a _th , the pixel position
(x, y) pixel values p _x of the _y "0" can be carried out binarization process of the (white pixel). Further, in order to extract the contour of the object, a differentiation process can be performed on the binarized image data. Specifically, the differential processing on the binarized image data is performed at one pixel position (x, y) of the binarized image data.
When attention is focused on, the value p _x of the pixel, _y is for example p _{x, y} ≠
_{p x-1, y-1} , p x, y ≠ p x-1, y, p x, y ≠ p x-1, y + 1 or _{p x,, y ≠ p x} , in the case of _y-1 Only by setting p _x , _y = 1 (black pixel), otherwise, by setting p _x , _y = 0 (white pixel).

As described above, in the image processing unit 2, the imaging unit 1
In the image taken at every predetermined time interval,
A / D conversion processing (binarization processing) is performed, and object detection is performed based on binarized image data obtained at predetermined time intervals.
Image processing necessary for (flame detection and intrusion detection) is performed.

As image processing necessary for object detection, first, candidates of a detection object (a flame and / or an intruder (for example, an intruder)) are extracted from the binarized image data obtained at predetermined time intervals. It is necessary to perform a process of extracting a region. For this reason, the image processing unit 2 compares, for example, the binary image data at a certain time with the binary image data at the immediately preceding time, and as a result of this comparison, When there is a change in the pixel value of a predetermined number of pixels in a certain image area (for example, an area where black pixels are connected) between data, the image area in which this change has occurred
(Connected area) is extracted as the object image area (the object image area is specified).

[0033] Specifically, for example, binary image data at time t ₀ is one as shown in FIG. 3 (a), 2 the next time t ₁
When the binarized image data is as shown in FIG. 3B, the difference image data becomes as shown in FIG. 3C, and the pixel values of five pixels change in the image area OBJ. since it is found that (because some object at time t ₁ is found to have occurred), it is possible to identify this region OBJ as the object image area.

The present invention focuses on the fact that a flame or an intruder (for example, an intruder) has characteristics (information) different from factors other than the flame or the intruder, and specifies the object specified as described above. Based on the pixel information of the image area OBJ, information (feature information) on the target is determined, and based on the information on the determined target, whether the target is a predetermined target, that is, whether or not the detection target (flame or intruder) I try to judge.

More specifically, when the object to be detected is a flame, the center of gravity of the object image area OBJ, the main axis, The change rate and the shape similarity rate can be obtained. When the detection target is an intruder, the shape similarity between the target object image area OBJ and a standard human body model (pattern) can be obtained.

Here, the center-of-gravity mobility, the principal axis change rate, and the shape similarity rate of the object image area OBJ can be obtained as follows. That is, as described above, the predetermined time interval Δ
object image area O from the binarized image data obtained for each
When the BJ is specified, the moment of inertia M _pq of the following equation (Equation 1) is first considered as a characteristic expression of the object image area OBJ. In Equation 1, x and y are not coordinate axes of a captured image having N and M pixels as shown in FIG.
As shown in FIG. 4, when an object image area OBJ is specified from the captured image of FIG. 2, it is assumed that the coordinates are orthogonal coordinate axes in contact with the object image area OBJ.

[0037]

(Equation 1)

In the equation of the moment of inertia M _pq , p
When = q = 0, M _pq , that is, M ₀₀ , represents the area as in the following equation (Equation 2).

[0039]

(Equation 2)

In the equation of the moment of inertia M _pq of the equation (1), if p = 1, q = 0; p = 0, q = 1, M
_pq , that is, M ₁₀ and M ₀₁ are represented by the following equation (Equation 3).

[0041]

(Equation 3)

Therefore, as shown in the following equation (Equation 4), M ₁₀ /
Coordinates I _c and J _c obtained by M ₀₀ , M ₀₁ / M ₀₀ represent the center of gravity G in the x and y directions of the object image area OBJ, respectively. Note that the coordinates I _c and J _c are obtained by normalizing M ₁₀ and M ₀₁ with M _00, that is, the area, respectively, and thus have an invariant scale that is not affected by the size of the object image area OBJ.

[0043]

I _c = M ₁₀ / M ₀₀ J _c = M ₀₁ / M ₀₀

[0044] Further, the center of gravity _{G = (I c, J c} ) as described above inertial moment M _f around is expressed by the following equation (5).

[0045]

(Equation 5)

A straight line passing through the center of gravity as described above, y = x
The moment of inertia M _θ of P _{x, y} for tan _θ is
(Equation 6)

[0047]

(Equation 6)

Note that with respect to Equations 5 and 6, x and y are
It is assumed that the coordinates are coordinates in the coordinate system when the centers of gravity I _c and J _c are set as the origin.

In Equation 6, the angle at which M _θ is the minimum is θ
_When it is set to ₀ , y = xtan θ ₀ is the principal axis of inertia of P _{x, y} . From Equation 6, M _θ = M ₂₀ sin ² θ−2M ₁₁ sin θcos
Since θ + M ₀₂ cos ² θ is established, in order to minimize M _theta, 0 and when put which was differentiated by theta, the following equation (7) is obtained.

[0050]

Tan2θ ₀ = 2M ₁₁ / (M ₂₀ −M ₀₂ )

From this, the direction θ ₀ of the main axis of the object image area OBJ is expressed by the following equation (Equation 8).

[0052]

[Equation 8] θ ₀ = (1/2) tan ⁻¹ [2M ₁₁ / (M ₀₂ −M ₂₀ )]

FIG. 4 shows the center of gravity G of the object image area OBJ.
The direction θ ₀ of the principal axis of the object image area OBJ obtained as described above is shown together with = (I _c , J _c ).

As described above, the object image area OBJ
Of the center of gravity G = (I _c , J _c ) and the direction θ _{0 of the} main axis are obtained, and based on these and the contour of the object image area OBJ extracted by the above-described differential processing, the current (current screen)
Can be obtained. That is, the distance r from the center of gravity G = (I _c , J _c ) to the contour is calculated by a function r of the angle.
When plotted as (θ), as shown in FIG. 5, a 2π periodic waveform peculiar to the object is drawn. In FIG. 5, θ ₀ is the direction of the main axis.

As described above, since the distance function r (θ) characterizes the object image area OBJ, that is, the shape of the object, using this distance function r (θ)
The similarity of the shape of the object image area OBJ can be obtained.

For example, if the object detection device in FIG. 1 is a flame detection device that detects a flame as an object, it is difficult to preset a standard pattern (standard distance function) for the shape of the flame. In this case, the shape (distance function r ₁ ) of the object image area OBJ at the current time (current screen) relative to the shape (distance function r ₂ (θ)) of the object image area OBJ at the previous time point (previous screen) (θ)) can be obtained as a shape similarity ratio.

FIGS. 6A and 6B show examples of the object image area OBJ and the distance function r (θ) at the previous time point (previous screen) and the current time point (current screen), respectively. When obtaining these shapes similarity rate, in order to eliminate the impact of the direction theta ₀ of the main shaft between them changes, respectively, a reference point of the distance function r (theta), in the direction theta ₀ of the main shaft at that time Match. Then, for example, each distance function r (θ) is sampled at a predetermined angle Δθ (for example, 5 °) up to 360 °. When Δθ is 5 °, 72 samples r (θ _i ) (i = 1 to 72) are obtained respectively. The distance function of the object image area OBJ at the current time (current screen) is represented by r ₁ (θ _i )
(i = 1 to 72), and when the distance function of the object image area OBJ at the previous time point (previous screen) is r ₂ (θ _i ) (i = 1 to 72), these shape similarities f , For example given by:

[0058]

(Equation 9)

Here, in order to eliminate the influence on the shape similarity f due to the size, mobility, etc. of the object, r ₁ (θ _i ), r ₂ (θ
_i ) needs to be standardized. For example, let the distance function be r
(θ) can be normalized to “1” at θ = θ ₀ (that is, r (θ ₀ ) = 1). Thus, r ₁ (θ _i = θ ₀ ) = 1 and r ₂ (θ _i = θ ₀ ) = 1, and the ratio between r ₁ (θ _i ) and r ₂ (θ _i ) is obtained. At this time, it is possible to eliminate the influence of the change in the size of the object, the mobility, and the like. Also, in this case, F is such that r ₂ (θ _i ) is r
When it is smaller than ₁ (θ _i ), F (r ₂ (θ _i ) / r ₁ (θ _i ))
= R ₂ (θ _i ) / r ₁ (θ _i ), and when r ₂ (θ _i ) is larger than r ₁ (θ _i ), F (r ₂ (θ _i ) / r ₁ (θ _i )) = R ₁ (θ _i ) / r
₂ (θ _i ). Therefore, the shape similarity f is 0
≦ f ≦ 1. As described above, the similarity (shape) between the distance function r ₁ (θ) of the object image area OBJ at the current time (current screen) and the distance function r ₂ (θ) of the object image area OBJ at the previous time (previous screen) is obtained. (Similarity ratio) f can be used as information in which the unique features of the flame are well expressed.

In the above example, r ₁ (θ _i ) and r ₂ (θ _i )
The shape similarity ratio f was obtained by calculating the ratio of
For example, based on the difference between r ₁ (θ _i ) and r ₂ (θ _i ) (for example,
Based on the square of these differences), the shape similarity ratio f can also be determined. In this case as well, r ₁ (θ _i ), r ₁ (θ _i ),
It is necessary to standardize r ₂ (θ _i ). That is, for example, by setting r ₁ (θ _i = θ ₀ ) = 1 and r ₂ (θ _i = θ ₀ ) = 1, a difference between r ₁ (θ _i ) and r ₂ (θ _i ) is obtained. At this time, the influence of the size, mobility, and the like of the object can be eliminated.

As described above, when the object to be detected is a flame, the information representing the unique characteristics of the flame is expressed as:
The shape similarity of the object image area OBJ can be used, and by determining whether or not the shape similarity of the object is equal to or greater than a predetermined threshold value, it is possible to reliably detect whether or not this object is a flame. Can be.

Also, for example, when the object detection device in FIG. 1 is an intruder detection device that detects an intruder as an object, the shape of the intruder can be, for example, standard human body models (some A standard human body shape pattern distance function r ₀ (θ)) is provided in advance, and these standard human body shape patterns (several distance functions r ₀ (θ)) and the target (current screen) The shape (distance function r ₁ (θ)) of the object image area OBJ is subjected to pattern matching, and the one having the largest value (similarity) can be obtained as the shape similarity. At this time, matching with one standard pattern is performed, for example, in the same manner as in the case of flame detection, for example, r ₁ (θ _i )
Or r ₀ (θ _i ), or r
_This can be done based on the difference between ₁ (θ _i ) and r ₀ (θ _i ) (eg, based on the square of these differences). In addition,
Distance function r for some standard body shape patterns
₀ The (theta), for example, can be prepared such as the distance function r ₀ of the standard pattern of the human body standing front (theta), the distance of the standard pattern of the human standing side function r ₀ (theta).

As described above, the standard human body shape pattern (distance function r ₀ (θ)) and the shape of the object image area OBJ at the current time (current screen) (distance function r ₁ (θ)) are obtained. Even when the shape similarity is obtained by pattern matching, r ₀ (θ _i ) and r ₁ (θ _i ) are standardized in order to eliminate the influence on the shape similarity f due to the size and mobility of the object. Need to be For example, the distance functions r ₀ (θ) and r ₁ (θ _i ) can be normalized to “1” at θ = θ ₀ (ie,
r ₁ (θ _i = θ ₀ ) = 1 and r ₂ (θ _i = θ ₀ ) = 1). Thus, when performing pattern matching between r ₀ (θ _i ) and r ₁ (θ _i ), it is possible to eliminate the influence of the change in the size of the target object, the mobility, and the like.

As described above, when the detection target is an intruder, the shape similarity of the object image area OBJ can be used as information in which the unique characteristics of the intruder are well expressed. By determining whether or not the shape similarity ratio is equal to or greater than a predetermined threshold value, it is possible to reliably detect whether or not this object is an intruder.

As described above, the similarity of the shape based on the distance function r (θ) from the center of gravity G = (I _c , J _c ) of the object image area OBJ to the contour is determined by the object (flame or intruder). It can also be used as information in which unique features are well expressed.

When the object to be detected is a flame, information representing the characteristic of the flame in a good manner includes not only the shape similarity of the object image area but also the center of gravity mobility and the principal axis change rate. Can also be used. Here, the center-of-gravity mobility and the main-axis change rate can be obtained as follows. In other words, the center-of-gravity mobility g of the object image area OBJ is a plurality (n) separated by a predetermined time interval Δt from the time point t ₁ at which the object is specified (for example, including this time point t ₁ ). time t ₁ of, t _2, ... t n pieces each object as determined respectively on the basis of the binarized image data in the _n image region OBJ _1, OB
J _2, ..., the center of gravity G of the _{OBJ n (t 1), G} (t 2), ..., G
The center of gravity G (t ₁ ), G (t ₂ ), with respect to the average value <G> of (t _n )
, G (t _n ) can be obtained as the following equation (Equation 10) as a variation (standard deviation value of variance).

[0067]

(Equation 10)

The center of gravity G is represented by x, such as (I _c , J _c ).
Since it is composed of the coordinates and the y-coordinate, the arithmetic operation of Expression 10 is performed on I _c , and independently of this, J _c
, And the calculation result of I _{c and} the calculation result of J _c are combined, for example, to obtain the center of gravity mobility g.

The principal axis change rate h of the object image area is
From the time t ₁ was subjected to a particular object (e.g., the time t
₁ (including ₁ ), a plurality (n
Time t _1, t _2, ... n pieces each object as determined respectively on the basis of the binary image data at t _n image areas OBJ ₁ of the individual),
OBJ _2, ..., the direction of the main axis of the _{OBJ n θ 0 (t 1)} , θ
_{0 (t 2), ...,} θ 0 mean value of (t _n) <θ _0> direction of the main shaft with respect to _{θ 0 (t 1), θ} 0 (t 2), ..., θ variation of ₀ (t _n) ( (Standard deviation value of variance) can be obtained as follows.

[0070]

[Equation 11]

Therefore, when the detection target is a flame, the image processing unit 2 of FIG. 1 specifies the target object image area OBJ from the binarized image data obtained at every predetermined time interval Δt. As information in which the unique features of the flame are well represented, the center of gravity mobility and / or the principal axis change rate and / or the shape similarity rate can be obtained.

When the predetermined object is a flame,
In addition to the above-mentioned shape information (center-of-gravity mobility, principal axis change rate, shape similarity rate), it is also possible to use predetermined time-varying characteristics as information that expresses the unique characteristics of the flame. By judging whether or not the object image area OBJ has a predetermined time-varying characteristic peculiar to a flame, it is possible to reliably detect whether or not the object is a flame.

As the temporal change characteristics, the inventor of the present application has proposed a temporal fluctuation characteristic of the size (number of pixels; area) of the object image region and the size (number of pixels;
Particular attention was paid to the regularity of the time variation of the area).

Therefore, when the detection target is a flame, the image processing unit 2 in FIG. 1 specifies the target object image area OBJ from the binarized image data obtained at every predetermined time interval Δt. As information in which the unique feature of the flame is well expressed, the temporal fluctuation characteristic of the size of the object image area OBJ and / or the regularity of the temporal fluctuation can be obtained. More specifically, the temporal fluctuation characteristics of the size (the number of pixels) of the object image area OBJ
It can be obtained as a variance ratio (standard deviation value of variance) with respect to the temporal average value of the size (number of pixels) of J, and the regularity of the temporal variation of the size (number of pixels) of the object image area OBJ. Can be obtained as an autocorrelation with respect to a temporal change in the size (the number of pixels) of the object image area OBJ.

As described above, when the detection target is a flame, the image processing unit 2 obtains at least one of the shape information, the variance, and the autocorrelation as information in which the characteristic unique to the flame is well expressed. Processing can be performed. In this case, the imaging time interval Δt needs to be an appropriate one (for example, a time interval of about 0.5 second to 2 seconds) so that the temporal fluctuation characteristics of the flame can be reliably detected. There is.

More specifically, the temporal fluctuation characteristic of the object image area, that is, the variance (standard deviation value of variance) d with respect to the temporal average value of the size (number of pixels) of the object image area is determined by the following equation. From the time point t ₁ at which the identification was performed (including the time point t _1, for example), a plurality (n) separated by a predetermined time interval Δt
Time t ₁ of, t _2, ... t n pieces each object as determined respectively on the basis of the binarized image data in the _n image region OBJ _1, O
BJ _2, ..., the number of black pixels in _{OBJ n w (t 1),} w
(t ₂ ),..., w (t _n ), the average value <w> of black pixels w (t ₁ ), w (t ₂ ),..., w (t _n ) ) Can be obtained as in the following equation.

[0077]

(Equation 12)

The autocorrelation of the time change of the size (number of pixels) of the object image area is calculated from the number of black pixels in the object image area at a plurality of time points separated by a predetermined time interval Δt. It can be determined as follows.

[0079]

(Equation 13)

The autocorrelation cor (k) used by the judging unit 3 to actually judge whether or not it is a flame is obtained by time-averaging the autocorrelation function cor _m (k) with respect to the time parameter m. Is used. That is, the image processing unit 2 actually calculates the autocorrelation value c given by the following equation:
or (k) is given to the judgment unit 3.

[0081]

[Equation 14]

Here, m can be an arbitrary integer value, but m can be set to 0. In the case where m is set to 0, Expression 14 is calculated at time t _i
And the time (t _i + t _k ) is a well-known autocorrelation function of

[0083]

(Equation 15)

As described above, in the image processing unit 2, the dispersion ratio (standard deviation value of dispersion) with respect to the temporal average value of the shape information of the target image region and the size (number of pixels) of the target image region.
d, when at least one of the autocorrelation cor (k) regarding the temporal change of the size (the number of pixels) of the object image area is determined, the determination unit 3 determines, based on these processing results,
It is determined whether or not the target image area is a predetermined target (flame).

The inventor of the present application has described that an object (light source)
A flame, a rotating light, and an incandescent light bulb of a 3.3 cm square fire plate were imaged from a predetermined imaging distance, and the temporal change in the number of black pixels in the object image area in the binary image data of each of these objects was measured.

FIG. 7 shows the measurement results of the temporal change in the number of pixels (the number of black pixels) in the object image area corresponding to the flame in the binarized image data obtained by imaging the flame of a 3.3 cm square fire plate as the object. FIG. 8 is a diagram showing the measurement of the temporal change in the number of pixels (the number of black pixels) in the object image area corresponding to the rotating light in the binarized image data obtained by imaging the rotating light as the object. FIG. 9 is a diagram showing the results, and FIG. 9 is a graph showing the temporal change of the number of pixels (the number of black pixels) in the object image region corresponding to the incandescent lamp in the binary image data obtained by imaging the incandescent lamp as the object. It is a figure showing a measurement result.

Further, based on the measurement result of the temporal change of the number of pixels (the number of black pixels) of each object in FIGS. 7, 8, and 9, the inventor of the present application provides a dispersion ratio d for each object. And the autocorrelation function cor _m (k). In this measurement, the imaging time interval Δt was set to 1 second.

FIG. 10 is a diagram showing the dispersion ratio (temporal dispersion ratio) d of the flame, the rotating lamp, and the incandescent lamp of the 3.3 cm square fire plate obtained as described above while changing the imaging distance. FIG.
At 0, the dispersion ratio d of the stove flame when the stove flame is imaged as the object is also shown.

FIG. 10 shows that when the object is a flame of a 3.3 cm square fire dish or a flame of a stove, the dispersion ratio d is about 0.3 to
When the object is a rotating lamp, the dispersion ratio d is approximately in the range of 0.9 to 1.2, and when the object is an incandescent lamp, the dispersion ratio d is , Which is almost zero. From this, it is possible to clearly distinguish the flame, the rotating lamp and the incandescent lamp based on the dispersion ratio d. In particular,
Since the dispersion ratio (temporal dispersion ratio) d does not change much even if the imaging distance changes, even if an image is taken from an arbitrary imaging distance, based on the dispersion ratio d, the object is a flame or a rotating lamp. Or an incandescent light bulb can be stably and easily identified. Conventionally, in order to detect the fluctuation of the flame, for example, a spectrum by a fast Fourier transform is obtained. In this case, there is a problem that a complicated operation is required. Rate d
Can easily find this without the need for complex operations such as the fast Fourier transform, and the temporal dispersion d is, as described above, The dispersion ratio d is particularly useful information for judging whether or not it is a flame because it can be stably and clearly identified.

FIG. 11 is a diagram showing the autocorrelation cor _m (k) of the rotating lamp obtained based on the measurement results of FIG. FIG. 11 shows the autocorrelation cor _m (k) of the rotating lamp when the imaging distance is 1 m, 3 m, and 5 m, respectively.

FIG. 12 is a diagram showing the autocorrelation cor _m (k) of the flame of the 3.3 cm square fire tray obtained based on the measurement results of FIG. In FIG. 12, the imaging distance is 4 m, 6
autocorrelation cor of the flame at m, 8m, 10m
_m (k) are shown.

In FIGS. 11 and 12, m is set to "0".

When comparing FIG. 11 and FIG. 12, the autocorrelation cor _m (k) of the rotating lamp of FIG. 11 periodically changes with time, whereas the flame of the 3.3 cm square fire tray of FIG. It can be seen that the autocorrelation cor _m (k) is substantially constant in time. In the rotating lamp, the emission direction of light rotates at a constant rotation speed, and therefore, the amount of light received by the imaging unit 1 changes periodically with this rotation, and the autocorrelation c
or _m (k) also changes periodically with time as shown in FIG. On the other hand, the amount of light emitted from the flame does not change periodically like a rotating lamp, and the autocorrelation cor _m (k) also changes as shown in FIG.
Does not change periodically.

As described above, the autocorrelation makes it possible to clearly discriminate the artificial periodic noise such as a rotating lamp from the flame.

As described above, the dispersion ratio d and the autocorrelation cor
Is different for each object, that is, for a flame, a rotating light, and an incandescent light bulb, and is useful information for determining whether or not it is a flame.

As described above, in the object detection device of FIG. 1, when the object to be detected is a flame (that is, when the object detection device is used for flame detection), the image processing unit 2 performs the operation at every predetermined time interval Δt. When a flame candidate is specified as an object image area OBJ from the obtained binarized image data, the center of gravity mobility g of the object image area OBJ is obtained, and / or
Alternatively, the main axis change rate h of the object image area is obtained, and / or the shape similarity rate f of the object image area is obtained, and / or the temporal fluctuation of the size (number of pixels) of the object image area OBJ. As a characteristic, a dispersion ratio (standard deviation value of dispersion) with respect to a temporal average value of the size (number of pixels) of the object image region OBJ is obtained, and / or the object image region OBJ is obtained.
As the regularity of the temporal variation of the size (number of pixels) of J, an autocorrelation with respect to the temporal change of the size (number of pixels) of the object image area OBJ is obtained. When one is determined, the determination unit 3 determines whether or not the target object image area is a flame based on these processing results.

Accordingly, the determination unit 3 determines the variance d, the autocorrelation cor, the center of gravity mobility g, the principal axis change rate h, and the shape similarity f
Can be determined based on at least one of the following. That is, it can be determined whether or not it is a flame based on the dispersion ratio d, or it can be determined whether or not it is a flame based on the autocorrelation cor, or the center of gravity mobility g To determine if it is a flame, or
Whether or not it is a flame can be determined based on the main axis change rate h, or whether or not it is a flame can be determined based on the shape similarity rate f. In combination, it can be determined whether or not it is a flame.

For example, the judgment unit 3 determines, for example, the variance d, the autocorrelation cor, the center of gravity mobility g,
When the principal axis change rate h and the shape similarity rate f are obtained, the variance d, the autocorrelation cor, the center of gravity mobility g, the principal axis change rate h, and the shape similarity rate f are used as parameters, and fuzzy theory is used.
A determination can be made as to whether it is a flame.

More specifically, for example, for each parameter of the shape similarity f, the center of gravity mobility g, the principal axis change rate h, the variance d, and the autocorrelation cor, the membership function MEM (f ), MEM (g), MEM (h), M
EM (d) and MEM (cor) are set in advance, and the image processing unit 2
, The shape similarity f, the center of gravity mobility g, the spindle change rate h,
When the dispersion ratio d and the autocorrelation cor are obtained, the corresponding membership functions MEM (f), MEM (g), M
From EM (h), MEM (d), and MEM (cor), the confidences o (f), o (g), o (h), o (d),
o (cor), and these confidences o (f),
It is possible to finally detect whether or not it is a flame from o (g), o (h), o (d), o (cor) or by combining these certainty factors.

FIG. 13 shows the parameters (0
FIG. 14 is a diagram showing an example of setting a membership function MEM (f) that expresses the degree of certainty that the flame is fired, and FIG. 14 shows parameters (0 to 1) of the center of gravity mobility g.
Membership function MEM expressing certainty of flame
FIG. 15 is a diagram illustrating a setting example of (g), and FIG. 15 is a diagram illustrating a setting example of a membership function MEM (h) expressing a certainty factor of a flame with respect to a parameter (0 to 1) of a main axis change rate h. FIG.

FIG. 16 is a diagram showing an example of setting a membership function MEM (d) expressing the degree of certainty of a flame with respect to the parameters (0 to 1) of the temporal dispersion ratio d. 17 is a parameter (0 to 0) of the autocorrelation cor (k).
It is a figure which shows the example of a setting of the membership function MEM (cor) expressing the certainty factor which is flame about 1). In addition,
For a revolving light, the dispersion ratio d may be greater than 1.0, but for values greater than 1.0, this is converted to a parameter value of 1.0 and the membership function MEM (d )It was set.

As can be seen from FIGS. 13 to 17, each membership function MEM (f), MEM (g), MEM
(h), MEM (d) and MEM (cor) are the shape similarity f,
Center of gravity mobility g, principal axis change rate h, variance d, autocorrelation co
When the parameter value of r is most likely to be a flame, the certainty (weighting) of “1” is given, and when the probability of being a flame is the lowest, the certainty (weighting) of “0” is given. And such a membership function is
It can be set by human judgment, and thus, by using such a membership function, it is possible to make a flame detection judgment close to human judgment. Further, each information of the shape similarity f, the center of gravity mobility g, the principal axis change rate h, the variance d, and the autocorrelation cor is well adapted as a parameter of such a membership function of fuzzy theory. I have.

Now, for example, the object is a flame, and the variance d, the autocorrelation cor, the center of gravity mobility g, the principal axis change rate h, and the shape similarity f of the object obtained by the image processing unit 2 are obtained.
Are “0.3” and “0.9”, respectively.
5 "," 0.9 "," 0.8 "," 0.7 "
From the membership functions of FIGS. 13 to 17, the confidences o (0.3), o (0.95), o (0.9),
o (0.8) and o (0.7) are “1.0” and “1.
0 "," 1.0 "," 1.0 ", and" 1.0 ".

When the object is a rotating lamp, the rotating lamp does not appear in the image unless it is synchronized with the shutter speed of the imaging unit. Therefore, the center of gravity mobility g, the principal axis change rate h, and the shape similarity rate f are measured. It is impossible. For example, the dispersion ratio d and the autocorrelation c of the object (rotating light) obtained by the image processing unit 2
or, the center-of-gravity mobility g, the principal axis change rate h, and the shape similarity rate f are “1.0”, “0.5”, “measurement impossible”, “measurement impossible”, “measurement impossible”, respectively. At one time, from the membership functions of FIGS. 13 to 17, the confidences o (1.0), o (0.5), o (measurable), o
(Measurement impossible) and o (measurement impossible) are “0.0”, “0.0”,
"Measurement impossible", "measurement impossible", "measurement impossible".

The object is an incandescent lamp (a general light source whose position is fixed), and the dispersion ratio d, the autocorrelation cor, the center-of-gravity mobility g,
The parameter values of the main axis change rate h and the shape similarity rate f are “0.1”, “1.0”, “1.0”, “1.0”,
When it is “1.0”, the certainty factor o (0.1), o of each parameter value is obtained from the membership functions shown in FIGS.
(1.0), o (1.0), o (1.0), o (1.0) are "0.
0 "," 0.0 "," 0.0 "," 0.0 "," 0.0 ".

The object is an artificial light flicker, and the variance d, the autocorrelation cor, the center-of-gravity mobility g, the principal axis change rate h, and the shape similarity ratio of the object obtained by the image processing unit 2 are obtained. Each parameter value of f is "0.
4 "," 0.8 "," 0.5 "," 0.2 ", and" 0.5 ", the degree of certainty o (0) of each parameter value is obtained from the membership functions shown in FIGS. .4), o (0.8), o
(0.5), o (0.2) and o (0.5) are "1.0", "1.
0 "," 0.0 "," 0.0 ", and" 0.5 ".

As described above, when the certainty factors o (d), o (cor), o (g), o (h), and o (f) for each parameter value are obtained, the judgment unit 3 Average value of each confidence (o
(d) + o (cor) + o (g) + o (h) + o (f)) / 5 is determined, and when this average value is equal to or more than a predetermined threshold value, for example, “0.8”, it is determined that the flame is present. When it is less than "0.8", it can be judged that it is not a flame.

In the above example, the object is a flame, and the certainty factor o (0.3), o (0.95), o (0.9) of each parameter value.
9), o (0.8), o (0.7) are "1.0", "1.
0 "," 1.0 "," 1.0 "," 1.0 "
This average value is “1.0” and is “0.8” or more, so it can be determined that the target object is a flame.

When the object is a rotating lamp, the confidences o (1.0) and o (0.5) of each parameter value are “0.0”,
In the case of "0.0", "measurement impossible", "measurement impossible", "measurement impossible", since the average value cannot be measured, it can be determined that the object is not a flame.

Further, when the object is an incandescent light bulb, certainty factors o (0.1), o (1.0), o (1.0),
o (1.0) are “0.0”, “0.0”, “0.0”,
If the values are “0.0” and “0.0”, the average value is “0.0”.
0 "and not more than" 0.8 ", it can be determined that the object is not a flame.

Further, the object is an artificial light flicker, and certainty factors o (0.4) and o (0.4.
8), o (0.5), o (0.2), o (0.5) are "1.
0, "1.0", "0.0", "0.0", and "0.5", the average value is "0.5", which is less than "0.8". It can be determined that the object is not a flame.

As described above, according to the object detection apparatus of the present invention, when this is used for flame detection, the unique feature of flame generated at the time of fire or the like (other Information that can be clearly distinguished from the factors) is determined, and based on this information it is determined whether or not the target object is a flame. Can be significantly improved.

Therefore, the object detection device of the present invention can be used as a fire detection device and applied to a fire alarm device.

FIG. 18 is a diagram showing a configuration example of a fire alarm device to which the object detection device of the present invention is applied. Referring to FIG. 18, this fire alarm device is configured such that the image pickup unit 1 is set so as to satisfactorily image a flame as a detection target in the target detection device shown in FIG. ,
That is, an alarm output unit 4 for judging an actual fire and outputting a fire alarm (abnormal alarm) when the judgment result in the judgment unit 3 is flame is further provided.

FIG. 19 is a flowchart for explaining the processing operation of the fire alarm device of FIG. Prior to actually operating this fire alarm device, for example, membership functions for each parameter (center of gravity mobility, main axis change rate, shape similarity rate, variance rate, autocorrelation) are, for example, shown in FIGS. The determination is made as shown in FIG. 17, and this is set in the determination unit 3 in advance (step S1).

After performing such initial settings, the object is imaged by the imaging unit 1, and the image is processed by the image processing unit 2.
When the target is detected as the target image area, the image processing unit 2 then calculates the center of gravity mobility, the principal axis change rate, the shape similarity rate, the dispersion rate, and the self-center based on the pixel information of the target image area. Each piece of correlation information (each parameter value) is determined as described above (step S2).

When each piece of information (each parameter value) is calculated in the image processing section 2, the judgment section 3 uses the membership function corresponding to each piece of information (each parameter value) for each piece of information (each parameter value). , The confidence that the object is a flame
(Fuzzy value) is obtained (step S3). Then, based on the certainty factor obtained for each parameter value (for example, by checking whether or not the average value of each certainty factor is greater than a predetermined threshold value (for example, “0.8”)), the target object is a flame It is determined whether or not the object is a flame (step S4). If it is determined that the object is a flame, a signal for outputting an alarm is output to the alarm output unit 4 (step S5). Thus, the alarm output unit 4 can output an alarm only when the determination unit 3 determines that a flame has occurred.

As described above, in the fire alarm device of FIG.
In the flame detection device, using the imaged image data of an object with a large amount of information, information that uniquely expresses the characteristic of the flame that occurs at the time of fire (a characteristic that can be clearly distinguished from other non-fire alarm factors) Can be determined, and based on this information, it can be determined whether or not the target object is a flame, so that the reliability of the fire determination can be significantly improved. In other words,
In a conventional fire alarm device (self-fire alarm system) using an ultraviolet or infrared sensor (sensor), false alarms may be output due to various non-fire alarm factors.
In the fire alarm device of FIG. 18, it is possible to extremely effectively prevent the occurrence of false alarms due to non-fire alarm factors.

Further, the object detection device of the present invention can be used as an intruder detection device (for example, an intruder detection device) and applied to a security system (security alarm device).

FIG. 20 is a diagram showing a configuration example of a security alarm device to which the object detection device of the present invention is applied. Referring to FIG. 20, in the security alarm device, in the target detection device shown in FIG. 1, the imaging unit 1 is set so as to image an intruder (for example, an intruder) as a detection target. An alarm output unit 4 for outputting an alarm (abnormal alarm) when the output from the unit 3, that is, when the judgment result in the judgment unit 3 is an intruder, is further provided.

FIG. 21 is a flowchart for explaining the processing operation of the security alarm device of FIG. Prior to actually operating the security alarm device, for example, a distance function r of some standard human body shape patterns is used.
₀ (θ) is set in advance (step S11).

After performing such initial settings, the object is imaged by the imaging unit 1, and the image is processed by the image processing unit 2.
When the target is detected as the target image area, the image processing unit 2 then uses the distance function as described above as shape information of the target image area based on the pixel information of the target image area. r ₁ (θ) (Step S1
2). That is, in the image processing unit 2, the object image area O
Distance function r ₁ (θ _i ) from the center of gravity to the contour of BJ
Is determined as described above. And this distance function r
₁ (θ _i ) and a pattern function of distance functions r ₀ (θ _i ) of some standard human body shape patterns prepared in advance,
Among them, the one having the largest value (similarity ratio) is determined as the shape similarity ratio (the degree of coincidence of the target object pattern with the standard human body shape pattern) (step S13).

As described above, when the shape similarity ratio is calculated by the image processing unit 2, the judgment unit 3 determines whether or not this shape similarity ratio is larger than a predetermined threshold (for example, “0.8”). By checking, it is determined whether or not the target is an intruder (step S14). As a result, when it is determined that the target object is not an intruder, the processing ends. On the other hand, when it is determined in step S14 that the target is an intruder, a signal for outputting an alarm is output to the alarm output unit 4 (step S15). Thus, the alarm output unit 4 can output an alarm only when the determination unit 3 determines that the person is an intruder.

As described above, in the security alarm device of FIG.
In the intruder detection device, using the captured image data of the object with a large amount of information, the information that uniquely expresses the intruder's unique characteristics (the characteristics that can be clearly distinguished from other factors other than the intruder) Since it is possible to determine and determine whether or not the target object is an intruder based on this information, the reliability of the intruder determination can be significantly improved.

In the above example, the detection target is an intruder.
(E.g., an intruder), the shape similarity ratio based on the distance function r ₁ (θ) from the center of gravity to the contour of the object image area is used as the information in which the unique characteristics of the intruder are well expressed. , As information in which the unique characteristics of the intruder are well expressed, along with the shape similarity based on the distance function r ₁ (θ),
Alternatively, the circularity can be used instead.

More specifically, the circularity e of the object image area OBJ in the binary image data at a certain point in time is determined by the number of pixels (the area of the object image area OBJ) Corresponding) u and this object image area O
The number of pixels forming the outline of the BJ (corresponding to the length of the outline of the object image area OBJ) v is obtained by the following equation.

[0127]

E = (4π · u) / v ²

Here, the number u (area) of pixels constituting the entire object image area OBJ is equal to the number of pixels (area) u in the binarized image data when each pixel of the object image area OBJ is represented as a black pixel. It can be obtained by counting the number of black pixels constituting the entire area OBJ. The number of pixels (contour length) v forming the contour of the object image area OBJ is determined by, for example, further performing a differentiation process on the binarized image data. Can be obtained by counting the number of.

For example, the binarized image data at a certain point in time is as shown in FIG.
In (a), when the object image area is OBJ, the number of pixels (area) u constituting the entire object image area can be obtained as the total number of black pixels “5” in the object image area OBJ.

Further, when differentiation processing is performed on the binarized image data of FIG. 22A (that is, focusing on one pixel position (x, y) of the binarized image data of FIG. 22A) and when,
The values p _x and _y of this pixel are, for example, p _x , _y yp _{x-1, y-1} , p _x ,
_{y ≠ p x-1, y} , p x, y ≠ p x-1, y + 1 or p _{x,, y} ≠ p
Only when _{x, y-1} is p _x , _y = 1 (black pixel), otherwise, processing is performed to set p _x , _y = 0 (white pixel)),
The differential image data as shown in FIG.
In the differential image data of (b), the number of pixels (contour length) v that form the contour of the object is the total number of black pixels “4” in the portion corresponding to the object image area OBJ in FIG. You can ask.

If the object image area OBJ has a circular shape and the radius is assumed to be r, the number u of pixels constituting the entire object image area is proportional to “πr ² ”. The number of pixels v forming the outline of the image area is “2π
Therefore, the circularity e is “1.0” because the circularity e is proportional to “r”. In other words, the circularity e of the target image region is “1” when the shape of the target image region is almost circular. .0.0 "and the more complicated the shape, the closer to" 0.0 ".When the target is an intruder, the circularity e is, for example, 0.4 to 0.4.
If it is about 0.6, the circularity e is, for example, 0.4 to 0.4.
It can be determined whether or not the target object is an intruder based on whether or not it is within the range of 0.6.

As described above, the circularity of the object image area OBJ is determined only by the binarized image data at a certain time point, for example, the binarized image data at the time point t ₁ when the object image area OBJ is specified. Can be obtained as a circularity e on the basis of the following formula. However, in order to further enhance the reliability, a predetermined time interval Δt is used from the time point t ₁ (for example, including the time point t ₁ ) when the target object image area is specified. time t ₁ of the plurality of (n) that are separated by, t _2, ... t n-number of objects respectively obtained based on the binary image data in the _n image region OBJ _1, OBJ _2,
..., circularity e (t ₁₎ of the _{OBJ n, e (t 2)} , ..., the average value of e (t _n) as <e>, can also be obtained as follows.

[0133]

[Equation 17]

As described above, the information in which the unique characteristics of the intruder are well expressed, together with the shape similarity ratio f,
Alternatively, the circularity e (<e>) can be used instead.

Further, as information in which the unique characteristics of the intruder are well expressed, in addition to the above-mentioned shape information (f, e) or instead of the above-mentioned shape information (f, e), , The size S of the target object, the moving speed mv, and the like can also be used. Here, the size S of the target object can be obtained as the total number of pixels of the target object image area OBJ, and the moving speed mv of the target object is different from the target object image area at a certain time and the previous time. It can be obtained by comparing with the object image area. When the size S of the object and the moving speed mv are obtained, whether the object is an intruder depends on whether the size S of the object is appropriate for the size of the human body. Can be determined, and the moving speed mv of the object
Can be determined whether or not the target object is an intruder, based on whether or not is an appropriate speed for human movement.

As described above, in the object detection device of FIG. 1, when the detection target is an intruder (for example, an intruder) (that is, when the object detection device is used for intruder detection),
The image processing unit 2 converts an intruder candidate from the binarized image data obtained at each predetermined time interval Δt into the object image area OBJ.
, The shape similarity f of the object image area OBJ is obtained, and / or the object image area O
The circularity e of the BJ can be determined, and / or the size S of the object image area OBJ can be determined, and / or the moving speed mv of the object image area OBJ can be determined, and at least one of these can be determined. When determined, the determination unit 3 determines whether the target object image area is an intruder based on these processing results.
(E.g., an intruder).

Therefore, the determination unit 3 determines whether or not the target is an intruder (for example, an intruder) based on at least one of the shape similarity f, the circularity e, the size S, and the moving speed mv. You can judge. That is, it can be determined whether or not the object is an intruder (intruder) based on the shape similarity ratio f, or whether or not the object is an intruder (intruder) based on the circularity e. Can be determined based on the size S, and whether the object is an intruder (intruder) can be determined based on the moving speed mv. You can decide
Furthermore, by appropriately combining them, it can be determined whether or not the object is an intruder (intruder).

For example, when the image processing unit 2 obtains, for example, the shape similarity f, the circularity e, the size S, and the moving speed mv, the determining unit 3 determines the shape similarity f, the circularity e, the size It is also possible to determine whether or not the vehicle is an intruder (intruder) using fuzzy logic with S and the moving speed mv as parameters.

Specifically, for example, for each parameter of the shape similarity f, the circularity e, the size S, and the moving speed mv,
Membership functions MEM (f), MEM (e), MEM (S), MEM (m
v) is set in advance, and when the shape similarity f, the circularity e, the size S, and the moving speed mv are obtained in the image processing unit 2, the membership functions MEM (f), M
From EM (e), MEM (S) and MEM (mv),
It is possible to obtain certainty factors o (f), o (e), o (S), o (mv) which are flames, and these certainty factors o (f), o (e), o
From (S), o (mv), or a combination of these certainty factors, it can be finally detected whether or not it is an intruder (intruder).

As described above, when the object detection device of FIG. 1 is used as an intruder detection device, it is possible to reliably determine whether or not an intruder (intruder).

In each of the above configuration examples, the analog image data from the imaging section 1 (photoelectric conversion section 12) is subjected to binarization processing as A / D conversion processing in the image processing section 2 to obtain analog image data. Is converted into binary image data, and processing necessary for flame detection and intrusion detection is performed. The analog image data from the imaging unit 1 (photoelectric conversion unit 12) is converted by the image processing unit 2 It is also possible to perform a multi-level conversion process as an A / D conversion process, convert analog image data into multi-level digital image data, and then perform the same processing necessary for flame detection and intrusion detection as described above.

Further, in each of the above configuration examples, there is one target object, and one target image region (connected region) is obtained from the captured image.
Although the case where the OBJ is specified has been described, the present invention can be similarly applied to a case where a plurality of objects exist and a plurality of object image regions (a plurality of connected regions) are included in the captured image. In this case, by performing the above-described processing for each of the plurality of target object image areas, it is possible to reliably determine whether each target is a predetermined detection target (flame or intruder).

For example, two object image areas from a captured image
(Two connected areas) When OBJ ₁ and OBJ ₂ are specified,
The image processing unit 2 includes two object image areas OBJ ₁ and OBJ
The pixels in the J ₂ can perform labeling process subjecting the label number. For example, each pixel of the object image region OBJ ₁ are denoted by the same label number "1", In each pixel of the object image region OBJ ₂ are denoted by the same label number "2", Then, the same image processing as described above is performed for each pixel region having the same label number, and a determination process can be performed based on the image processing result.

[0144] That is, in the above example, as the label number "1" predetermined image processing for the object image region OBJ _1, labeled with is made to acquire the first image processing result, also as a label number " predetermined image processing is performed on assigned an object image region OBJ ₂ of 2 ", the second
Is obtained, and it is determined whether or not the target with the label number “1” is a predetermined target based on the first image processing result, and based on the second image processing result. It can be determined whether or not the object with the label number “2” is a predetermined object.

In the above description, the alarm output unit 4 is provided in the object detection device of FIG. 1, and when it is determined that the object is a flame or an intruder, an alarm is output from the alarm output unit 4. Instead of performing such an alarm output from the alarm output unit 4 or together with such an alarm output, it is also possible to transmit the judgment result from the judgment unit 3 to the outside of the object detection device. For example, as shown by a broken line in FIG. 1, this object detection device is connected to the receiver 100 via a predetermined transmission path, and the object detection device is connected to the receiver 1
When a call is made from 00 (for example, when address polling is performed), the determination result from the determination unit 3 of the target detection device is returned to the receiver 100, and the receiver 100 performs alarm processing and the like. Is also possible.

In the above description, the object detection device of FIG. 1 is configured as a flame detection device for detecting a flame or an intruder detection device for detecting an intruder. The detection device may have both functions of a flame detection device and an intruder detection device. That is,
As described above, regardless of whether the target is a flame or an intruder, the image processing unit 2 of the target detection device in FIG. In addition, since the judgment unit 3 has basically the same judgment processing function for these, the object detection device of FIG.
When both the flame detection function and the intruder detection function are provided, at least the image processing unit 2 and the determination unit 3 can be used for both flame detection and intruder detection. Specifically, the image processing unit 2 and the determination unit 3
In the case of one processor and RAM, ROM, etc.,
This one processor unit can have both image processing and judgment processing functions for flame detection and intruder detection. For example, an image processing program and a judgment processing program for detecting an intruder are stored in a RAM or ROM together with an image processing program and a judgment processing program for detecting a flame. The flame detection processing can be performed by the image processing program and the judgment processing program, and the intruder detection processing can be performed by the image processing program and the judgment processing program for intruder detection.

Further, the image pickup section 1 can be configured to be used for both flame detection and intruder detection. That is, as the photoelectric conversion unit 12 of the imaging unit 1, 1
It can be configured to detect both flames and intruders using only one sensor (eg, a normal visible light CCD (CCD sensitive in the visible region)).

However, when monitoring an intruder, the visible light region is used, while when monitoring a flame, for example, the near-infrared region is used. Because the area is different,
In a case where both flame detection and intruder detection are performed using only one sensor (for example, a normal visible light CCD), when intruder detection is performed, visible light is passed as it is and light enters this sensor. At the time of flame detection, it is desirable to perform switching control so that the visible region is cut and the cut light in the visible region enters this sensor.

At the time of intruder detection, visible light is allowed to pass through to enter this sensor, while at the time of flame detection, the visible region is cut, and the cut light of the visible region is incident on this sensor. The switching control and the switching control of causing the image processing unit 2 and the judging unit 3 to perform the flame detection or the intruder detection can be realized as the switching control means by, for example, the processor unit described above.

FIG. 23 is a diagram showing a configuration example of an object detection device provided with such a switching control means 7. By providing such a switching control means 7, FIG.
The function of detecting both a flame and an intruder (function as an intruder / flame detection device) can be provided in various forms with only one target detection device.

For example, the switching control means 7 of the object detection device shown in FIG. 23 initially makes the image pickup section 1 have a function as an intruder sensor, and makes the object detection device monitor the intruder, for example. When the object is detected, the determination unit 3 determines whether the object is an intruder.
When it is determined that the user is not an intruder, the imaging unit 1
The function of the imaging unit 1 is switched so as to have a function as a flame sensor, and the function is switched so as to have a function as a flame sensor.
, It is possible to perform switching control for determining whether or not it is a flame.

FIG. 24 is a flow chart for explaining the processing operation of the object detection device of FIG. Referring to FIG. 24, the switching control unit 7 initially sets the photoelectric conversion unit 12 (one sensor) of the imaging unit 1 to directly enter visible light, and in this state, for example, performs brightness adjustment or the like. After the adjustment and the like are performed, the target detection device starts the intruder monitoring processing operation (step S21).

Thus, in the image processing unit 2, the imaging unit 1
For example, the image processing for intruder detection as described above is performed on the target imaged in the above step, and the determination unit 3 performs, for example, the above-described processing based on the image processing result for intruder detection from the image processing unit 2 as described above. In this way, it is determined whether or not the person is an intruder (step S22). Note that the process of determining whether or not the user is an intruder is started, for example, by triggering when a difference between the current image data and the previous image data is extracted (a connected component of a difference pixel is extracted at a predetermined threshold or more).

If it is determined in step S22 that the user is an intruder, for example, an intruder alarm is sounded.
(Step S26). On the other hand, if it is determined in step S22 that the user is not an intruder, a difference (difference data) between the current image data and the previous image data is further extracted, and it is determined whether or not difference data exists. (Step S23). As a result, if there is difference data, the process returns to step S21 again to determine whether or not the user is an intruder.

In this manner, the intruder determination process is performed. If it is determined in step S22 that the intruder is not an intruder, and if it is determined in step S23 that there is no difference data, the switching control unit 7 performs image capturing. The state of the imaging unit 1 is switched so that the light whose visible region has been cut is incident on the photoelectric conversion unit 12 (one sensor) of the unit 1, and in this state, for example, brightness adjustment is performed, and brightness adjustment and the like are performed. After
The target detecting device performs a flame monitoring process operation (step S24).

Thus, in the image processing unit 2, the imaging unit 1
For example, the above-described image processing for flame detection is performed on the object imaged in the above. The determination unit 3 uses the image processing result for flame detection from the image processing unit 2 in the above-described manner, for example. Then, it is determined whether or not it is a flame (step S25). As a result, if it is determined that it is a flame, for example, a fire alarm is sounded (step S26).

On the other hand, if it is determined in step S25 that the flame is not a flame, the switching control means 7 again takes the image so that the visible light enters the photoelectric conversion unit 12 (one sensor) of the imaging unit 1 as it is. The state of the unit 1 is switched, and in this state, for example, brightness adjustment and the like are performed. After the brightness adjustment and the like are performed, the target detection device starts the intruder monitoring operation (step S21).

In this way, flame monitoring (flame detection) can be performed while monitoring an intruder (intruder detection) by a single target detection device having a simple configuration.

In the above example, flame monitoring is performed while monitoring an intruder. However, conversely, intruder monitoring may be performed while performing flame monitoring.
That is, in this case, the switching control means 7
Initially, the imaging unit 1 is provided with a function as a flame sensor, and the object detection device performs flame monitoring. When an object is detected, the determination unit 3 determines whether or not the object is a flame. When the determination unit 3 determines that it is not a flame, the function of the imaging unit 1 is switched so that the imaging unit 1 has a function as an intruder sensor, and the target detection device performs intruder monitoring. Switching control is performed in the unit 3 so as to determine whether or not the user is an intruder.

In the above example, either the intruder or the flame is set as a normal monitoring target, and while monitoring the intruder (or flame) (normal monitoring), the flame (or intruder) is monitored. However, monitoring control (switching control) can also be performed with an intruder and a flame as equal monitoring targets. In other words, in the object detection device of FIG. 23, the switching control means 7 alternately switches the imaging unit 1 between a function as an intruder sensor and a function as a flame sensor, and synchronizes with the image processing unit 1 2, judgment part 3
Can be switched between the intruder process and the flame process alternately.

In this case, for example, at one time point t ₁ , the imaging unit 1 is switched to the function as an intruder sensor.
(Switched to a state where visible light is directly incident on the photoelectric conversion unit 12 (one sensor)), the image processing unit 2, the determination unit 3
Then, image processing for intruder detection is performed on the image captured in this state, and based on the result of the image processing,
Performs determination processing of the intruder, in the next time t _2, the imaging unit 1 is switched to function as a flame sensor (photoelectric conversion section 1
2 (one sensor is switched to a state where visible light cut light is incident), and the image processing unit 2 and the determination unit 3 perform image processing for flame detection on the image captured in this state. Based on the result of the image processing, a flame determination process is performed, and at the next time point t ₃ , the imaging unit 1 is switched to the function as the intruder sensor (the photoelectric conversion unit 12 (one sensor)). The image processing unit 2 and the determination unit 3 perform image processing for intruder detection on the image captured in this state, and based on the result of the image processing. Thus, the monitoring of the intruder and the monitoring of the flame can be performed alternately, such as performing the intruder determination process.

Further, the object detecting device shown in FIG.
In the normal state, control can be performed so as to function as a flame detection device, and in the monitoring state, the flame can be controlled to function as an intruder / flame detection device that detects both intruders.

In each of the above examples, the switching control means 7
Automatically controls switching between intruder monitoring and flame monitoring, but manually controls intruder monitoring and flame monitoring.
It is also possible to make the switching controllable (for example, by a manual switch).

In this case, when the operator sets, for example, a manual switch on the intruder monitoring side, the target detection device in FIG. 23 functions as an intruder detection device. That is, the imaging unit 1 is switched to a function as an intruder sensor (switched to a state where visible light is directly incident on the photoelectric conversion unit 12 (one sensor)), and the image processing unit 2 and the determination unit 3
Image processing for intruder detection is performed on the image captured in this state, and intruder determination processing is performed based on the result of the image processing.

When the operator sets, for example, a manual switch to the flame monitoring side, the object detection device in FIG. 23 functions as a flame detection device. That is, the imaging unit 1 is switched to a function as a flame sensor (switched to a state where visible light cut light enters the photoelectric conversion unit 12 (one sensor)), and the image processing unit 2 and the determination unit 3 Image processing for flame detection is performed on the image captured in this state, and flame determination processing is performed based on the result of the image processing.

As described above, according to the configuration of the switching control means 7, it is possible to monitor and control various kinds of objects such as an intruder and a flame with a single object detection device. That is, the function of detecting both a flame and an intruder (function as an intruder / flame detection device) can be provided in various forms by only one target detection device in FIG.

In the switching control means 7, the image pickup unit 1
For example, a mechanical shutter or the like can be used to switch between the function as the intruder sensor and the function as the flame sensor.

FIG. 25 is a diagram showing a configuration example of a mechanical shutter. In the example of FIG. 25, the mechanical shutter 10
Reference numeral 1 is installed in front of the lens system 13 of the imaging unit 1, and inserts and removes a visible light cut filter on the optical path of the object monitoring area according to a control signal from the shutter control line 102.
(Close or open). 26A shows a state in which the visible light cut filter 103 is closed (closed), and FIG. 26B shows a state in which the visible light cut filter 103 is open (open). ing.

When such a mechanical shutter 101 is used, the switching control unit 7 controls the mechanical shutter 101 when the imaging unit 1 functions as an intruder sensor.
Is controlled so as to be in an open state as shown in FIG. 26 (b), and when the imaging unit 1 functions as a flame sensor, the mechanical shutter 101 is controlled so as to be in a closed state as shown in FIG. 26 (a). I do.

Further, in the object detecting device of FIG.
1, an alarm output unit 4 may be provided. In this case, an intruder alarm and a flame alarm can be output according to the intruder monitoring state and the flame monitoring state. it can.

As described above, the object detection apparatus shown in FIG. 23 has a function of monitoring and detecting both an intruder and a flame by itself, despite its simple configuration. Therefore, if this object detection device is used,
There is no need to separately arrange the intruder detection device and the flame detection device, so that wiring and the like can be reduced, and installation space and cost can be reduced.

In the above description, the alarm output unit 4 is provided in the object detection device of FIG. 23, and when it is determined that the object is a flame or an intruder, an alarm is output from the alarm output unit 4. Instead of performing such an alarm output from the alarm output unit 4 or together with such an alarm output, it is also possible to transmit the judgment result from the judgment unit 3 to the outside of the object detection device. For example, as shown by a broken line in FIG. 23, this target detection device is connected to the receiver 100 via a predetermined transmission path, and when the target detection device is called from the receiver 100 (for example, when address polling is performed). In this case, the determination result from the determination unit 3 of the target detection device may be returned to the receiver 100, and the receiver 100 may perform an alarm process or the like.

In this case, as in the case of the object detection device shown in FIG. 23, when a single device has a function of monitoring and detecting both an intruder and a flame, the object monitoring device using this object detection device Only one system (receiver) is required, and the intruder sensor function and the flame sensor function
In this one receiver, only one address needs to be assigned (since only one address (polling address) needs to be assigned to the target detection device in FIG. 23), both intruder monitoring and flame monitoring are performed. , It is possible to avoid complicated management.

That is, when the intruder detection device and the flame detection device are arranged separately from each other,
Even if the intruder detection device and the flame detection device are monitored by one receiver, it is necessary to assign separate addresses (for example, polling addresses) to the intruder detection device and the flame detection device. However, there is a problem that management becomes complicated, but in the object detection device of FIG. 23, when this is connected to a receiver to perform both intruder monitoring and flame monitoring, only one receiver is required. ,Also,
Since only one address is required, management and the like can be performed very easily.

Further, in the configuration example of FIG. 23, the switching control means 7 is provided in the object detection device. However, the object detection device is connected to the receiver 100 via a predetermined transmission path, and is used as an object monitoring system. When configured (when the target detection device is called from the receiver 100 (for example, when address polling is performed), the determination result from the determination unit 3 of the target detection device is returned to the receiver 100. In the case where an alarm process or the like is performed on the receiver 100 side), the switching control means 100 is provided in the receiver 100 as shown in FIG. It is also possible to switch the target detection device to the state of the intruder detection device or the state of the flame detection device by a switching control command from the receiver 100.

In the configuration examples of FIGS. 23 and 27,
The image processing unit 2 and the determination unit 3 of the target detection device can perform image processing and determination processing for intruder detection and image processing and determination processing for flame detection in the manner described above. The configuration example of FIG. 27 is not limited to the above-described method as long as the image processing unit 2 and the determination unit 3 can be used for intruder detection and flame detection.
Image processing and determination processing for detecting an intruder and image processing and determination processing for detecting an arbitrary flame can be performed.

In the above example, when the image processing and the judgment processing are performed by one processor unit, the image processing and the judgment processing for intruder detection (program) and the image processing and the judgment processing for flame detection (program) ) To RAM
Or ROM, but the image processing and determination processing (program) for intruder detection and the image processing and determination processing (program) for flame detection can be held in any manner. For example, image processing and judgment processing (program) for intruder detection and image processing and judgment processing (program) for flame detection are performed on a floppy disk or C
It is recorded in advance on a portable storage medium such as a D-ROM, and when the target detection device is started, this recording medium is mounted on the target detection device, and image processing for intruder detection recorded on this recording medium is performed. Also, the judgment processing (program), the image processing for flame detection and the judgment processing (program) can be loaded into a RAM or the like and used.

[0178]

As described above, according to the present invention, according to claim 1 or 請
According to the second and ninth aspects of the present invention, the unique features of the intruder (features that can be clearly distinguished from other factors) are satisfactorily expressed by using the captured image data of the subject having a large amount of information. Since the information to be obtained is determined and whether or not the target object is an intruder is determined based on this information, the reliability of determining whether or not the target object is an intruder can be significantly improved.

In particular, according to the first and ninth aspects of the present invention, a membership function is set in advance using information in which characteristics unique to an intruder are well expressed as parameters.
When information that uniquely expresses the unique characteristics of an intruder is determined, the confidence that the target is an intruder is obtained from the membership function using the determined information as a parameter value, and the confidence obtained. , It is determined whether or not the target object is an intruder, so that a more reliable intruder detection determination close to human determination can be performed.

According to the third to eighth aspects of the present invention, the image pickup means and the object image area are extracted from the image picked up by the image pickup means, and the object image area is extracted based on the pixel information of the object image area. An image processing means for determining information about an object, and a target detection device including a determination means for determining whether the target is a flame based on information about the target determined by the image processing means, It is configured to be able to detect both the flame and the intruder, and switches the state of the imaging means depending on whether the detection target is a flame or an intruder,
Switching control means for controlling the switching of the processing functions of the image processing means and the determining means is provided, and the switching control of the switching control means allows the object detection device to function as a flame detection device, Since it is made to function as a detection device, it is possible to monitor (detect) both the flame and the intruder with one target detection device despite its simple and compact configuration.

In particular, according to the invention of claim 8, since the CCD having sensitivity to light in the visible region is used as the imaging means, a low-cost product can be provided.

According to the tenth aspect of the present invention, the object detecting device according to any one of the first to eighth aspects is connected to a receiver via a predetermined transmission path, and When the target detection device is called from the receiver, the target detection result is transmitted to the receiver.Therefore, based on the target detection result from the target detection device, the receiver side should perform an alarm process etc. Can be.

According to the eleventh aspect of the present invention, when the object detection device has both the intruder detection function and the flame detection function, the state of the object detection device can be changed by the switching control command from the receiver. Switching control can be performed to either the person detection device or the flame detection device.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration example of an object detection device according to the present invention.

FIG. 2 is a diagram illustrating a pixel configuration example of a two-dimensional CCD.

FIG. 3 is a diagram illustrating a process of specifying a target object image area from a captured image.

FIG. 4 shows the center of gravity G of the object image area OBJ and the direction θ of the main axis.
₀ is a diagram for explaining a distance function r (θ). FIG.

FIG. 5 is a diagram illustrating an example of a distance function r (θ) of an object image area OBJ.

FIG. 6 is a diagram for explaining a shape similarity ratio f of an object image area OBJ.

FIG. 7 shows the number of pixels in the object image area corresponding to the flame in the binarized image data obtained by imaging the flame of the 3.3 cm square fire plate.
FIG. 9 is a diagram illustrating a measurement result of a temporal change of (the number of black pixels).

FIG. 8 shows the number of pixels in an object image area corresponding to a rotating light in binarized image data obtained by imaging the rotating light as an object.
FIG. 9 is a diagram illustrating a measurement result of a temporal change of (the number of black pixels).

FIG. 9 is a diagram showing a measurement result of a temporal change in the number of pixels (the number of black pixels) in an object image region corresponding to an incandescent light bulb in binary image data obtained by imaging an incandescent light bulb as an object.

FIG. 10 is a diagram showing a measurement result of a temporal dispersion ratio.

FIG. 11 is a diagram showing an autocorrelation of a rotating lamp obtained based on the measurement result of FIG. 8;

FIG. 12 is 3.3 cm obtained based on the measurement result of FIG. 7;
It is a figure which shows the autocorrelation of the flame of a square fire.

FIG. 13 is a diagram illustrating a setting example of a membership function expressing a certainty factor of a flame with respect to a parameter of a shape similarity ratio f.

FIG. 14 is a diagram illustrating a setting example of a membership function expressing a certainty factor of a flame with respect to a parameter of a center of gravity mobility g.

FIG. 15 is a diagram illustrating a setting example of a membership function expressing a certainty factor of a flame with respect to a parameter of a main axis change rate h.

FIG. 16 is a diagram showing a setting example of a membership function expressing a certainty factor of a flame with respect to a parameter of a temporal dispersion ratio d.

FIG. 17 shows parameters of autocorrelation cor (k).
It is a figure showing the example of setting of the membership function which expresses certainty which is flame.

FIG. 18 is a diagram showing a configuration example of a fire alarm device to which the object detection device of the present invention is applied.

FIG. 19 is a flowchart illustrating an example of a processing operation of the fire alarm device of FIG. 18;

FIG. 20 is a diagram showing a configuration example of a security alarm device to which the object detection device of the present invention is applied.

FIG. 21 is a flowchart illustrating an example of a processing operation of the security alarm device of FIG. 20;

FIG. 22 is a diagram for explaining a process of calculating the circularity of an object.

FIG. 23 is a diagram showing another configuration example of the object detection device according to the present invention.

FIG. 24 is a flowchart illustrating a processing operation of the object detection device in FIG. 23;

FIG. 25 is a diagram illustrating a configuration example of a mechanical shutter.

26 is a diagram for explaining the operation of the mechanical shutter in FIG.

FIG. 27 is a diagram showing a configuration example of a target monitoring system according to the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image pickup part 2 Image processing part 3 Judgment part 4 Alarm output part 7 Switching control means 11 Optical system 12 Photoelectric conversion part 13 Lens system 14 Optical filter 100 Receiver 101 Mechanical shutter

Claims (11)

(57) [Claims] An image processing unit that extracts an object image region from an image captured by the imaging unit and calculates information on the object based on pixel information of the object image region; Determining means for determining whether or not the object is an intruder, based on the information on the object determined by the means, wherein the image processing means comprises:
As the information, the shape similarity rate of the object image region, circularity of the object image region, the object image area size, to exit dividing at least one of a moving speed of the object image area
And the determining means includes the object
Image area shape similarity, circularity of target image area, target
Each of the size of the object image area and the moving speed of the object image area
Shape similarity, circularity, size, and moving speed
Membership functions are set in advance as parameters.
Is determined, and the determining unit is configured to
Shape similarity ratio of object image area, circular shape of object image area
Degree, size of object image area, moving speed of object image area
When each parameter value of degree is calculated,
The confidence factor for each parameter value is calculated using the
Calculated by the membership function, and for each parameter value
An object detection device that determines whether an object is an intruder based on the obtained degree of certainty . 2. The object detection device according to claim 1, wherein
An object detection apparatus, wherein one CCD having sensitivity to light in a visible region is used as the imaging unit. 3. An image pickup means, an image processing means for extracting an object image area from an image picked up by the image pickup means, and calculating information on the object based on pixel information of the object image area; An object detection device comprising: a determination unit configured to determine whether an object is a predetermined object, based on information on the object determined by the means. And a control for switching the processing function of the imaging means, the image processing means and the determination means to the flame detection function or the intrusion detection function depending on whether the detection target is a flame or an intruder. And a switching control unit for performing the switching control by the switching control unit such that the target detecting device functions as a flame detecting device or an intruding object detecting device. An object detection device characterized in that: 4. The object detection device according to claim 3 , wherein
The switching control means, while performing intrusion monitoring (or flame monitoring), to perform flame monitoring (or intrusion monitoring),
An object detection device characterized in that the state of the object detection device is switched and controlled. 5. The object detection device according to claim 3 , wherein
An object detection device, wherein the switching control means performs switching control by setting an intruder and a flame as equal monitoring targets. 6. The object detection device according to claim 3 , wherein
The switching control means causes the target detection device to function as, for example, a flame detection device in a normal state, and to function as an intruder / flame detection device that detects both a flame and an intruder in a monitoring state. An object detection device characterized in that switching control is performed. 7. The object detection device according to claim 3 , wherein the switching control means is configured to be capable of manually switching control between intruder monitoring and flame monitoring. An object detection device characterized by the above-mentioned. 8. The object detection device according to claim 3 , wherein
One CCD having sensitivity to light in the visible region is used for the imaging unit, and the switching control unit allows the light in the visible region to enter the CCD as it is in an intruder monitoring state.
On the other hand, in the flame monitoring state, the object detection device controls light incident on the CCD so that visible light-cut light enters the CCD. 9. A membership function is set in advance by using a parameter that is a parameter of information that favorably expresses a unique feature of an intruder, and a unique feature of the intruder is favorably represented from image data obtained by imaging an object. When the information is determined, the degree of certainty that the object is an intruder is determined from the membership function using the determined information as a parameter value,
Based on the obtained confidence, target detection method characterized by the object to determine whether the intruder. 10. The object detection device according to claim 1 , wherein the object detection device is connected to a receiver via a predetermined transmission path, and the object detection device is called from the receiver. A target monitoring result transmitted to a receiver when the target is detected. 11. An object monitoring system in which an object detection device for detecting a predetermined object is connected to a receiver via a predetermined transmission path, wherein the object detection device has a flame detection function and an intruder detection. And a switching control means for performing switching control of whether the target detection device functions as a flame detection device or functions as an intruding object detection device. According to a switching control command from a switching control unit of the receiver, the target detection device is caused to function as a flame detection device, or to function as an intruder detection device. Target monitoring system to do.