JPH08317373A - System for detecting and extracting object in video monitoring device - Google Patents

Abstract

PURPOSE: To easily identify only the image of an object appearing on a screen by preserving prescribed positive/negative match information, performing division corresponding to the match and binarizing a differential value corresponding to each positive/negative match. CONSTITUTION: An image signal provided by picking up the image of a monitor zone with a television camera 121 is inputted through an input I/F123 to a data bus 122 and stored in an image memory 124. Besides, a CPU126 analyzes the image in a program memory 125 according to the program stored in the memory 125 and executes object detecting processing and corresponding to the result, a warning lamp 130 is turned on through an output I/F128. In this case, when the processing of the CPU126 is started, the image signals picked up by the camera 121 are fetched as image data of a prescribed number of picture elements. Next, match information Sn is found, the difference in the luminance value of picture elements between an input image and a background image is calculated, and positive/negative match information is preserved. Further, processing is divided into plural binarizing processing steps corresponding to that sign and binarization is performed corresponding to each positive/ negative match.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video surveillance device using a television camera, and particularly to a video surveillance device suitable for extremely high reliability such as detection of an object entering a dangerous area. Object detection method in.

[0002]

2. Description of the Related Art Video surveillance devices using television cameras have been widely and widely used in recent years. However, in such surveillance systems, it is not merely manned surveillance by an image monitor but within the surveillance field of view. There is a demand for a system capable of automatically detecting an intruding object such as a human being or an automobile from an image signal to obtain a predetermined notification or alarm treatment. Therefore, it is necessary to detect an object by signal processing.

By the way, as a conventional technique of a method of detecting an object by processing such an image signal, the latest image of successively input images is compared with an old image to obtain a difference in luminance value for each pixel, A method of detecting an area having a large difference value as an object is known and widely used.

FIG. 8 shows the basic flow of the system according to this conventional technique. That is, in this conventional technique, first, the process 8 is performed.
In step 1, the image is captured, then in step 82, the difference in the brightness value is obtained, the difference is binarized in step 83, and finally, in step 84, the binarized data is judged with a predetermined threshold value. It is designed to detect objects.

However, if the difference is simply obtained as in this prior art, a partial missing portion 91 is generated in the object 90 to be detected, as shown in FIG. There is.

However, the above-mentioned problems of the prior art can be solved by applying an expansion / contraction operation.

As an application example of such a moving object detection method by difference processing, for example, issued in July 1994,
Ueda described in "O plus E" No.176, pp.122-136
There is a paper entitled "Intelligent Video Handling Using Image Recognition Technology" by others 1.

The moving object detection method according to this paper is designed to detect an object from three temporally consecutive images. As shown in FIG. , 112 and 113 as input images, the brightness difference between the image 111 and the image 112 and the brightness difference between the image 112 and the image 113 are calculated, binarized, and then the expansion / contraction operation is performed to obtain the images 114 and 115. obtain. Next, AND these binary images 114 and 115.
The common part of each image is obtained by processing, and the object image 116
To get.

However, even with this method, FIG.
(B) False detection 93 of the background 92 due to illumination fluctuation. (c) If the entering object is emitting light, erroneous detection 95 by the area 94 where the light is reflected. (d) False detection 97 due to the shadow area 96 of the entering object. However, as a result, there is a possibility that an object may be erroneously detected as being present even though the object is not actually present in the image visual field.

[0010]

As described above, the prior art cannot be said to give sufficient consideration to the erroneous detection of an object, and as shown in FIG. There is a problem of false detection 93 due to the background, and further, FIG. 9 (c)
As shown in Fig. 9, when an object emits light like a headlight of a car, the problem of false detection 95 due to the area 94 where the light is reflected, and as shown in Fig. 9 (d), There was a problem of false detection 97 due to the shadow area 96 of the object.

These erroneous detection problems can be almost fatal, especially when applied to the monitoring of objects such as pedestrians and vehicles entering the level crossing.
There was a drawback that the application range was limited.

A first object of the present invention is to make it possible to easily identify only an image of an object existing within a screen of an image signal and to always reliably detect an object that has entered the field of view being monitored. An object of the present invention is to provide an object detection method in the video monitoring device.

Further, a second object of the present invention is to detect an object in a video monitoring device which can surely detect when a part having a locally different temperature appears in the field of view being monitored. To provide a method.

[0014]

First, according to the first aspect of the present invention, the first object is to perform the basic processing of the object detecting method according to the prior art shown in FIG. First, a code information storage step 13, a judgment processing step 14 for branching according to the code, and a first binarization processing step 1
This is achieved by adding 5 and the second binarization processing step 16 and making the threshold values of the first binarization processing step 15 and the second binarization processing step 16 different from each other. The numbers shown in parentheses indicate the steps of the same processing content in other figures.

Next, according to the second aspect of the present invention, the first purpose is, as shown in FIG. 2, in the configuration of the first aspect of the present invention shown in FIG. And the threshold value judgment by adding the information of the degree of similarity to the information of the difference value
And the second binarization processing steps 26 and 27 are added.

Further, according to the third aspect of the present invention, the first purpose is, as shown in FIG. 3, a judgment processing step 36 for classifying into one of the two types of regions 37 and 38 according to the degree of similarity. Is added, and the areas 37 and 38 are distinguished and detected.

Next, according to the fourth aspect of the present invention, the second object is to use the infrared image pickup device as the image pickup means in the configuration of the second aspect of the present invention shown in FIG. This is achieved by adding an evaluation processing step 67 of the similarity between the local luminance distribution shapes of the input image and the background image, as shown in FIG.

Further, according to the fifth aspect of the present invention, the second object is to use the infrared image pickup device as the image pickup means in the configuration of the second aspect of the present invention shown in FIG. As the processing in this case, as shown in FIG. 7, it is achieved by adding an evaluation processing step 77 of the similarity between the local luminance distribution shapes of the input image and the background image.

[0019]

In FIG. 1, when an image is captured from the television camera in the image input processing step 11, in the difference processing step 12, the difference in the brightness value between the captured input image and the background image created in advance. Is calculated. Next, in the code information storage processing step 13, the code of the brightness difference value is stored. Then, in the judgment processing step 14, the processing is branched according to this sign, and either the first binarization processing step 15 or the second binarization processing step 16 is executed according to the positive or negative sign. It

These first and second binarization processing steps 1
5 and 16 have different threshold values, and as a result, accurate binarization processing can be obtained even when the change in the luminance value is positive and negative with respect to the background image and the property is different, and reliable object detection 1
7 is possible.

That is, the change in the input image with respect to the background image becomes positive when the image becomes brighter, for example, from cloudy weather to sunny weather or illuminated by a light. Shadows on objects,
When the image becomes dark and becomes negative,
The degree of change is different, and simply binarizing the absolute value of the difference as in the prior art cannot deal with the degree of change.

However, in the method of FIG. 1, since the threshold value is changed depending on the direction of change, that is, whether it is positive or negative, the degree of change amount can be flexibly dealt with and accurate detection can be performed.

In FIG. 2, a processing step 24 of calculating the similarity of the local luminance distribution shapes of the input image and the background image is added to the processing of FIG. 1, and the first binarization processing step 26 and the second binary processing step are added. Since the degree of similarity is included in the condition for binarization in the binarization processing step 27, the degree of change can be more flexibly dealt with and more accurate detection can be performed.

Here, the similarity indicates the correlation between the shapes of the local luminance distributions in the input image and the background image. Now, in FIG. 10, the background image 101 indicates the environment. Even if the input image becomes dark as a whole as indicated by 102 due to the change in illuminance, the background image 10
Patterns such as white lines and trajectories in 1 remain as they are, and therefore local regions of the background image 101, for example, the region 104.
And the local area 105 in the input image 102 has a high degree of similarity.

On the other hand, as indicated by 103 in the input image,
In the area where the object (vehicle) 106 exists, this object 1
Since 06 blocks the background, there are no patterns such as white lines or trajectories in the background image, and therefore the area 104
And the region 106 have a low degree of similarity.

Therefore, in FIG. 3, after the binarization processing step 35, the area evaluation step 36 based on the similarity is performed.
By providing, it is possible to separate two regions having different properties, such as a region in which the brightness of the background image has changed and a region in which an object exists.

Next, in FIG. 4, in the process of FIG. 2, when the sign becomes positive, a judgment processing step 47 for evaluating the similarity of the local luminance distribution shape of the input image and the background image is added. As a result, as in the case of FIG. 9C, the area 94 that has been brightened by light from a light or the like
If the object exists, the object detection 49 is obtained separately from it, so that the object can be detected more accurately.

That is, the area brightened by light or the like is, as shown by the area 94 in FIG. 9C,
Although the brightness value is high, the road surface pattern remains, so that the similarity is high as in the case of the local regions 104 and 105 in FIG. Therefore, by using this degree of similarity, it is possible to reliably discriminate the area 94 and the object lightened by the light.

Next, in FIG. 5, in the processing of FIG. 2, when the sign becomes negative, a judgment processing step 57 for evaluating the similarity of the local luminance distribution shape of the input image and the background image is added. As a result, as in the case of FIG. 9D, when there is a region 96 darkened by a shadow, the object detection 59 is obtained by distinguishing it from the area 96, so that the object can be detected more accurately.

That is, the area darkened by the shadow is
As shown by a region 96 in FIG. 9D, the road surface pattern remains although the luminance value is low, so that the similarity is high as in the case of the local regions 104 and 105 in FIG.

Therefore, by using this similarity,
The region 96 darkened by the shadow and the object can be reliably discriminated.

In FIG. 6, a region in which the temperature locally rises can be detected by adding a processing step 67 for evaluating the similarity between the local luminance distribution shapes of the input image and the background image.

That is, in a region where the temperature locally rises, a difference in luminance distribution occurs between that region and a region where the temperature does not rise, so that the similarity of the luminance distribution shape becomes low. For example, in the case of application to pyroclastic flow detection, the temperature at which the lava dome is formed by the pyroclastic flow rapidly rises locally, so the degree of similarity is low, but on the other hand, in clouds heated by the pyroclastic flow, The degree of similarity increases because the temperature rises uniformly over almost the entire area.

Therefore, by providing the similarity evaluation processing step 67, the heated cloud portion and the pyroclastic flow portion can be separated.

In FIG. 7, a region in which the temperature is locally lowered can be detected by adding a processing 77 for evaluating the similarity of the local luminance distribution shapes of the input image and the background image.

That is, in the region where the temperature is locally decreased, the difference in the brightness distribution is generated between that region and the region where the temperature is not decreased, so that the similarity of the brightness distribution shape is low. For example, when applied to pyroclastic flow detection, the part where the lava dome collapses falls because the heat source falls downward, so the temperature drops sharply, so the similarity becomes low,
On the other hand, in a cloud or the like warmed by a pyroclastic flow, even if the heating by the heat source is reduced, the whole is cooled almost uniformly and the temperature drops, so the degree of similarity increases.

Therefore, these can be detected separately by the similarity evaluation processing step 77.

[0038]

EXAMPLE An object detection system according to the present invention will be described below.
This will be described in detail with reference to the illustrated embodiment.

The embodiment described below is an application of the present invention to an image monitoring apparatus for detecting an object such as a pedestrian or an automobile that has entered the railroad crossing, and the television camera is located at a position where the entire railroad crossing can be monitored. Is installed and an image signal obtained thereby is processed to detect an object.

FIG. 12 shows an example of a video surveillance apparatus to which an embodiment of the present invention is applied, which is the television camera 1 described above.
The image signal obtained by imaging the surveillance area (surveillance field of view) at 21 is
The data is input to the data bus 122 via the input I / F 123 and stored in the image memory 124.

On the other hand, the CPU 126 analyzes the image in the work memory 125 according to the program stored in the program memory 125, executes the object detection process, and outputs the alarm lamp 1 via the output I / F 128 according to the result.
30 is turned on, and an image is displayed on the monitor 131 via the image output I / F 129.

FIG. 13 shows an embodiment of the object detection processing by the CPU 126. The processing according to this embodiment is a combination of the processing shown in FIGS. 4 and 5 and a background image update processing added thereto. is there.

When the processing of FIG. 13 is started, first, in the image input processing step 141, the image signal picked up by the camera 121 is image data f (x, 320 × 240 pixels).
y) is included. Next, in determination step 143, it is determined whether or not a new background image has been created. The background image is created by the CPU 126 executing the processes shown in FIG. 14 in parallel. This FIG.
In the processing step 161, first, in processing step 161, the input image is captured and saved for 0.2 frames at intervals of 20 frames.
(Remember.

Next, in processing step 162, for each of the stored 20 frame images, for each pixel,
The median is calculated by the equation (1), and the background image data r
Create (x, y).

[0045]

[Equation 1]

Here, med {} represents a median, and in the case of this expression (1), the same pixels are arranged in the order of luminance value for the image of 20 frames, and the tenth value (intermediate value) ) Pixel is extracted and the background image data r
(x, y), which gives a new background image 163. Therefore, the background image 1
63 = image data r (x, y).

Returning to the processing of FIG. 13, when the background image has been created, the background image up to that point is updated to the newly created background image in the processing step 143 before the next processing. If the processing has not been completed yet, the procedure proceeds to the difference processing step 144, where the input image is divided into blocks of 2 × 2 pixels, and the entire block region f of 160 × 120 is formed.
The brightness average Af of each block area is obtained from _n (i, j). Here, n is a block number, and i and j represent pixel positions in the block.

Similarly, for the background image data r (x, y), 160 × 120 blocked regions r _n (i,
j), the luminance average Ar of each blocked area is obtained, and the difference value ε _n is obtained from the equation (2).

[0049]

[Equation 2]

In the subsequent step 145 of storing the code information, the code information S _n is obtained by the equation (3) and stored.

[0051]

(Equation 3)

In the similarity calculation step 146, the similarity γ _n is calculated for each block by the equation (4).
Here, D _n represents the neighborhood area of the n-th block, and [] represents the number of elements.

[0053]

[Equation 4]

Next, in the judgment processing step 147, the first binarization processing step 148 is executed according to the code information S _n.
Alternatively, the process branches to the second binarization processing step 152.

These processes are thresholding binarization processes.
First, the threshold when the sign is positive, that is, processing step 1
The threshold value th + at 48 is calculated by the equation (5). Here, K + is an adjustable parameter, and R + is a region where the sign is positive and no object exists.

[0056]

(Equation 5)

Where K + is an adjustable parameter,
R + is an area where the sign is positive and no object exists.

The threshold when the sign is negative, that is,
The threshold value th ~ in the processing step 152 is similarly obtained by the equation (6).

[0059]

(Equation 6)

Here, K ~ is an adjustable parameter,
R ~ is an area where the sign is negative and no object exists.

Similarity determination processing steps 149 and 15
In 3, the adjustable parameters T + and T ~ are set as thresholds for each, and the similarity γ _n is determined. That is, first in the judgment processing step 149, the similarity γ _n is compared with the threshold value T +, and when it exceeds the threshold value T +, it is shown as a light-reflecting area 94 in FIG. 9C. As described above, the region 150 brightened by the light is ignored as detected, and is classified as the object detection 151 only when the region 150 becomes equal to or less than the threshold value T +.

Similarly, in the judgment processing step 153, the similarity γ _n is compared with the threshold value T ~, and when it exceeds the threshold value T ~, it is shown as a shadow area 96 in FIG. 9 (d). If such a region 154 darkened by the shadow is detected, this is also ignored, and only when the region 154 is equal to or less than the threshold value T ~, it is classified as the object detection 155.

Then, the CPU 126 detects the object 15
When the classification of 1 is obtained and the detection 151 of the object is obtained, the alarm lamp 1 is output through the output I / F 128.
The fact that an object has been detected is notified by turning on 30 or by displaying an image on the monitor 131 via the image output I / F 129.

Therefore, according to this embodiment, it is possible to always reliably detect only the object from the imaging visual field.
For example, in FIG. 15, the input image 171 shows a case where there is a vehicle 171A stopped outside the railroad crossing at dusk, and a shadow 171B of the vehicle 171A extends to the inside of the railroad crossing. In this case, this input If the image 171 is determined by the conventional method, as shown in the image 173, it is recognized that the object 173A that has entered the railroad crossing exists.

On the other hand, according to the embodiment shown in FIG. 13, first, on the other hand, the background image 17 is processed by the processing shown in FIG.
2 is obtained and updated to a new background image about every 4 seconds. Next, in this case, since it is the shadow 171B, the code information S _n is negative, so the process proceeds to the determination processing step 153, where the degree of similarity between the background image 172 and the input image 171 is determined.

In the area where the vehicle 171A is present, a part of the background is hidden, and a high degree of similarity cannot be obtained. Therefore, the area is actually classified as the object 154 in FIG. 13, and this area actually exists. On the other hand, it can be distinguished from the object, and in the area of the shadow 171B, the degree of similarity is high because the background is not hidden, and as a result, it is classified as the area 154 of FIG.

Thus, according to this embodiment, the image 174
1, a vehicle 174A and a shadow 174B are shown in FIG.
3 objects 155 and areas 154 can be clearly classified, and it is possible to reliably detect only objects that have entered the railroad crossing.
In this case, it can be determined that there is no entering object within the railroad crossing despite the presence of the shadow 174B.

Next, FIG. 16 shows the case where the input image 181 shows the case where the illuminated portion 181B by the light of the motorcycle 181A before entering the railroad crossing has entered the road surface inside the railroad crossing. If the input image 181 is determined by the conventional method, it is recognized that the object 183A that has entered the railroad crossing still exists as shown in the image 183.

On the other hand, according to the embodiment shown in FIG. 13, first, on the other hand, the background image 17 is processed by the processing shown in FIG.
2 is obtained and updated to a new background image about every 4 seconds. Next, in this case, since it is the light irradiation portion 181B, the code information S _n becomes positive this time, and the process proceeds to the determination processing step 149, where the background image 172 is generated.
And the similarity of the input image 181 is determined.

In the area where the motorcycle 181A is present, a part of the background is hidden, and a high degree of similarity cannot be obtained. Therefore, the area is actually classified as the object 151 in FIG. It is possible to discriminate that there is an object that has entered the area. On the other hand, in the area of the illuminated portion 181B by the light, the degree of similarity is high because the background is not hidden. Will be classified.

Thus, according to this embodiment, image 184
As shown in FIG. 13, the motorcycle 174A and its light irradiation portion 184B can be clearly classified into the object 151 and the region 150 in FIG. 13, and it is possible to reliably detect only the object that has entered the railroad crossing. In this case as well, it can be determined that there is no entering object within the railroad crossing despite the presence of the light irradiation portion 184B.

Next, another embodiment of the present invention will be described. In this embodiment, the present invention is applied to the monitoring of pyroclastic flow in an active volcano, and an infrared television camera is used as the camera 121 in the image monitoring apparatus shown in FIG.
(Infrared imaging device) is used to obtain an input image having high brightness in a high temperature part of the subject, and then the CPU 126 is configured to execute the processing shown in FIG. is there.

The processing steps 191 to 198 of FIG. 17 and the processing step 201 are respectively shown in parentheses in the processing step 1 of FIG.
41 to processing step 148, and processing step 1
Using the same routine as 52, the process branches to either the similarity determination processing step 199 or the similarity determination processing step 202 depending on the brightness of the input image and the background image, that is, the sign of the temperature difference. , The adjustable parameters T + and T ~ are used as threshold values for determination, and when the similarity is less than or equal to the threshold value, the object 200 whose temperature is increasing,
The object 203 is descending.

The operation of this embodiment will be described with reference to FIG. 18. In this figure, an image at the moment when a pyroclastic flow occurs is shown as an input image 211, and a background image is 212.
It is assumed to be as shown in.

Then, according to this embodiment, the image 213 is obtained as the processing result, and the portion where the temperature is high due to the existence of the pyroclastic flow and the portion where the temperature is lowered due to the flow of the pyroclastic flow are discriminated. , Can be confirmed.

That is, in a region where the temperature locally rises, a difference in luminance distribution occurs between that region and a region where the temperature does not rise, so that the similarity of the luminance distribution shape becomes low. For this reason, the temperature at which the lava dome is formed by the pyroclastic flow rises rapidly locally, and the degree of similarity is low.On the other hand, the temperature of the clouds warmed by the pyroclastic flow is almost uniform throughout. Therefore, the similarity is high.

Therefore, the similarity evaluation processing step 199
By providing the above, the pyroclastic flow portion can be clearly separated as the area 200 in which the temperature has risen, even if there is a portion due to a warmed cloud or the like.

On the other hand, in the region where the temperature is locally decreased, the difference in the brightness distribution is generated between that region and the region where the temperature is not decreased, so that the similarity of the brightness distribution shape is low. As a result, the heat source drops below the part where the lava dome collapses, and the temperature drops sharply, resulting in a low degree of similarity.On the other hand, in the case of clouds heated by pyroclastic flow, the heat source Even if the heating due to is reduced, the whole is cooled almost uniformly and the temperature is lowered, so that the similarity is high.

Therefore, the similarity evaluation processing step 202
By providing, even if there is a part such as a cloud whose temperature has dropped, the part where the pyroclastic flow has flowed away can be clearly separated and detected as the region 203 where the temperature has dropped.

The present invention is not limited to the above application example,
It goes without saying that it can also be applied to a monitoring system for an intruder in a restricted area such as plant equipment.

[0081]

According to the present invention, since the sign of the difference value is used for the problem of erroneous detection due to the change of the background due to the change of the illuminance, the influence of the environmental change such as the change of the illuminance of the monitoring visual field is used. Since it is possible to accurately detect the object by reducing the risk of being affected by the object, and to determine the preservation of the pattern using the similarity of the brightness distribution shape, the change in illuminance that does not depend on the existence of the object. Even if there is a region due to, the object can always be reliably detected.

Therefore, according to the present invention, even in the case where extremely high reliability such as railroad crossing is required, the present invention can be applied without anxiety, and the applicable range of the video surveillance device can be sufficiently expanded.

[Brief description of drawings]

FIG. 1 is a flowchart for explaining the operating principle of the first present invention.

FIG. 2 is a flowchart for explaining the operation principle of the second invention.

FIG. 3 is a flowchart for explaining the operation principle of the third invention.

FIG. 4 is a flowchart for explaining the operation principle of the fourth invention.

FIG. 5 is a flowchart for explaining the operating principle of the fifth invention.

FIG. 6 is a flow chart for explaining the operation principle of the sixth invention.

FIG. 7 is a flow chart for explaining the operation principle of the seventh invention.

FIG. 8 is a flowchart for explaining the operation principle of the conventional technique.

FIG. 9 is an explanatory diagram showing a problem of the conventional technique.

FIG. 10 is an explanatory diagram of the degree of similarity in the present invention.

FIG. 11 is an explanatory diagram of object detection using a temporal difference according to the related art.

FIG. 12 is a configuration diagram showing an example of a video surveillance device to which an embodiment of the present invention is applied.

FIG. 13 is a flowchart showing an example in which the present invention is applied to detection of an entering object at a railroad crossing.

FIG. 14 is a flowchart showing background image creation processing according to an embodiment of the present invention.

FIG. 15 is an explanatory diagram showing an example of an object detecting operation according to an embodiment of the present invention.

FIG. 16 is an explanatory diagram showing an example of an object detecting operation according to an embodiment of the present invention.

FIG. 17 is a flow chart showing an embodiment in which the present invention is applied to pyroclastic flow generation monitoring.

FIG. 18 is an explanatory diagram showing an example of a detection operation according to an embodiment when the present invention is applied to pyroclastic flow generation monitoring.

[Explanation of symbols]

 121 television camera (video input device) 122 data bus 123 input I / F 124 image memory 125 work memory 126 CPU 127 program memory 128 output I / F 129 image output I / F 130 warning lamp 131 warning display monitor

Claims (14)

[Claims] 1. A method of detecting an object entering a visual field by a computer using an input image of a camera, wherein a luminance value of a pixel between an image input step and an input image and a background image, that is, an image in which no object exists. Of the difference value, the step of storing the positive / negative sign information of the difference value, the step of branching according to the sign, and the step of binarizing the difference value for each positive / negative sign. Object detection method characterized by. 2. A method of detecting an object invading in a visual field by using a computer input image, wherein an image input step and a step of calculating a difference in luminance value of pixels between the input image and the background image, The step of storing the positive / negative sign information of the difference value, the step of obtaining the similarity of the local luminance value distribution shape of the input image and the background image, the step of branching according to the sign of the difference value, and the positive / negative sign An object detection method comprising a step of performing binarization according to a difference value and similarity for each. 3. The method according to claim 1 or 2, further comprising a step of determining a region having a high similarity and a region having a low similarity in local luminance value distribution shapes of the input image and the background image as different regions. Object extraction method characterized by being added. 4. The invention according to any one of claims 1 and 2, wherein an area in which the input image and the background image have a high similarity in local luminance value distribution shape in the area where the sign information of the difference value is positive, For example, an object extraction method characterized by adding a step of determining an area brightened by a light in the visual field. 5. The invention according to any one of claims 1 and 2, wherein an area where the sign information of the difference value is negative and the similarity between the local luminance value distribution shapes of the input image and the background image is high, For example, an object extraction method characterized by adding a step of determining as an area darkened by a shadow appearing in the visual field of the object. 6. The invention according to claim 1, 2 or 3, wherein the image input device is an infrared camera, and the local luminance value distribution of the input image and the background image is in an area where the sign information of the difference value is positive. An object detection method characterized by adding a step of judging a region having a low shape similarity as a region where the temperature is remarkably increased due to a pyroclastic flow. 7. The invention according to claim 1, 2 or 3, wherein the image input device is an infrared camera, and the local luminance value distribution of the input image and the background image is in an area where the sign information of the difference value is negative. An object detection method characterized by adding a step of judging a region having a low degree of shape similarity as a region where the temperature has significantly decreased locally due to, for example, a pyroclastic flow flowing down. 8. An image signal which is sequentially captured in frame units from a predetermined field of view by a predetermined image pickup means,
The past image signal and the latest image signal are compared for each pixel or for each area composed of a plurality of pixels, and the difference between the brightness values thereof is binarized to be within the predetermined visual field based on the obtained image data. In the object detection method in the video monitoring device of the method of detecting an incoming object, the image signal of a predetermined number of frames until the latest image signal is taken in is taken, and the background image signal is created by the median calculation thereof. Background image signal generating means, compares the background image signal and the latest image signal for each pixel or for each region consisting of a plurality of pixels, and detects the difference in their brightness values, including their positive and negative signs. And a code storing means for storing the stored value, a code determining means for determining the sign of the difference between the brightness values, and first and second binarizing means having different thresholds. When the result is positive, the object is detected by using the first binarizing means, and when the result is negative, the object is detected by using the second binarizing means. Object detection method for video surveillance. 9. An image signal which is sequentially captured in frame units from a predetermined field of view by a predetermined image pickup means,
An object that enters the predetermined visual field based on the image data obtained by comparing the past image signal and the latest image signal for each area composed of a plurality of pixels and binarizing the difference in the brightness values thereof. In the object detection method in the video monitoring apparatus of the method of detecting the background image signal, a background image signal is created by taking in image signals of a predetermined number of frames until the latest image signal is taken in and calculating their median. Means, comparing the background image signal and the latest image signal for each area consisting of a plurality of pixels, the difference between the brightness values, a code storage means for detecting and storing the positive and negative signs, Code determination means for determining the sign of the difference between the brightness values, first and second binarization means having different thresholds, and the background image signal and the current image signal each consisting of a plurality of pixels. And a similarity detection means for detecting the similarity of the luminance value distribution shapes, and detecting the object by using the first binarization means when the determination result by the code determination means is positive. When negative, an object is detected using the second binarizing means, and according to the similarity detected by the similarity detecting means,
An object detection method in a video surveillance device, characterized in that the authenticity of an object detected by either one of the first binarization means and the second binarization means is determined. 10. A comparison between a past image signal and a latest image signal for each area composed of a plurality of pixels by using image signals sequentially taken in frame units from a predetermined visual field by a predetermined image pickup means. Then, in the object detection method in the video monitoring apparatus of the method of detecting an object coming into the predetermined visual field based on the image data obtained by binarizing the difference between the brightness values, the latest image signal is A background image signal creating means for creating a background image signal by taking in image signals of a predetermined number of frames until it is taken in, and the background image signal and the latest image signal are composed of a plurality of pixels. Code storage means for comparing each area and detecting and storing the difference between the brightness values including the positive and negative signs, and a code determining means for determining the sign of the brightness value difference. , The first and second binarizing means having different thresholds, the background image signal and the current image signal are compared for each area including a plurality of pixels, and the similarity of the luminance value distribution shape is calculated. A similarity detection unit for detecting is provided, and when the determination result by the sign determination unit is positive, the first binarizing unit is used to detect the object, and when the determination result is negative, the second binarizing unit. While detecting the object using, according to the similarity detected by the similarity detection means,
An object detection method in an image monitoring device, characterized in that the areas detected by either the first binarizing means or the second binarizing means are detected separately. 11. The invention according to claim 10, wherein when the sign of the difference value is determined to be positive and the area determined to have the similarity exceeding a predetermined value is detected, the area is an object. An object detection method in a video monitoring device, characterized in that it is configured not to detect the presence but to detect the local illuminance increase in the field of view. 12. The invention according to claim 10, wherein when the sign of the difference value is determined to be negative and the area determined to have the similarity exceeding a predetermined value is detected, the area is an object. An object detection method in a video monitoring device, characterized in that it is configured to determine that it is not a detection based on the presence but a detection based on a local decrease in illuminance. 13. The invention according to claim 10, wherein an infrared image pickup device is used as the image pickup means, and a region in which the sign of the difference value is determined to be positive and the similarity is determined to be a predetermined value or less is detected. In this case, the object detection method in the video monitoring device is configured so that it is determined that there is a locally high temperature portion in the visual field. 14. The invention according to claim 10, wherein an infrared image pickup device is used as the image pickup means, and a region in which the sign of the difference value is determined to be negative and the similarity is determined to be a predetermined value or less is detected. In this case, the object detection method in the video monitoring device is configured so that it is determined that there is a locally low temperature portion in the visual field.