JPH08171689A - Changed area detector - Google Patents

Abstract

PURPOSE: To provide a changed area detector which can accurately and stably detect only the part that is changed by an object originally to be detected out of a moving image. CONSTITUTION: This detection device is provided with a time luminance change detection means 23 which detects the time change caused between two sheets of images that are different from each other in terms of time, a background feature detection means 24 which detects the background feature of at least one of both images and generates a background feature image, a misdetection removing means 25 which removes the change occurred at the background part out of the change detected by the means 23 based on the background feature image generated by the means 24, and a changed part detection means 26 which detects a part having the change with time out of the image where the change occurred at the background part is removed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to, for example, an equipment abnormality monitoring device for monitoring equipment by a camera and automatically detecting an abnormal event. The present invention relates to a changed area detection device that detects an area.

[0002]

2. Description of the Related Art Conventionally, the difference method is most commonly used in this kind of change area detecting apparatus. FIG.
Is, for example, the third part of the document “Image Processing Handbook” (Edited by Image Processing Handbook, 1987, Shokoido).
It is a block diagram for demonstrating an example of the change area | region detection apparatus using the difference method disclosed by page 75 etc. In the figure, 1 is an image at time t, 2 is an object in image 1, 3 is an image at time t + 1, and 4 is an object in image 2. Further, 5 is a difference image between the image 3 and the image 1, 6 is a region where the density value in the difference image 5 is negative, and 7 is also a positive region.

Next, the operation will be described. In the changed area detection device shown in FIG. 13, the image 3 at time t + 1
The difference image 5 is obtained by subtracting the image 1 at the time t from. At this time, if there is a change between the two images 1 and 3,
In the difference image 5, a region 7 having a positive density value and a region 6 having a negative density value appear. In addition, the area where there is no change in the background portion is 0.
When the density of the changing area is higher than that of the background portion, the positive area 7 in the difference image 5 is the changing area at time t + 1, and the negative area 6 is the changing area at time t. An abnormal event is detected by binarizing the difference image 5 described above with an appropriate threshold value.

Although FIG. 13 shows the case where the background is flat, when there is a structure in the background, when the above difference method is applied to such an image, there is camera vibration or the like. Then, the change is detected from the background part which is not originally changed, and when it is used for the equipment abnormality monitoring device or the like, it causes an erroneous alarm and the reliability is lowered. Therefore, when there is camera vibration, it is common to measure the positional deviation between two images in advance, perform the alignment to correct the deviation, and then perform the difference. This alignment is also disclosed on page 304 of the above document.

Next, this alignment will be described. As shown in the above document, the correlation method or the sequential residual test method (hereinafter referred to as SSDA) may be used as the alignment method. Here, the operation of SSDA will be described.

FIG. 14 is an explanatory diagram showing the principle of SSDA. In the figure, 8 is a reference image W (i, j) having a size of M × N and 9 is a target image S (i, j) having a size of L × L. At this time, the reference image 141 is moved on the target image 142, and the residual R (m, n) is calculated by the following equation (1).

[0007]

[Equation 1]

This residual R (m, n) is defined as 1 ≦ m ≦ L−M +
The point (m, n) that minimizes R (m, n) is calculated as 1, 1 ≦ n ≦ L−N + 1. Residual R
Since (m, n) represents the degree of mismatch between the two images,
Speeding up can be achieved by stopping the calculation when the residual exceeds a certain threshold. However, there is a limit to speeding up because of matching at the original image level.

Further, in the image picked up by the light receiving element of the camera, the luminance information is taken out as a voltage, but in order to process this by a computer, analog / digital conversion (hereinafter,
It is necessary to convert to a digital image by A / D conversion). In this A / D conversion, the object at the same position may be temporally different due to the fluctuation of the synchronization of the camera signals. In such a case, there is no camera vibration or the position adjustment is performed. In some cases, the difference image may be detected as a change. This is shown in FIG.
15A shows an analog image taken out from the camera, FIG. 15B shows a digital image obtained by A / D converting the analog image, and FIG. 15C shows a difference image of the two digital images. . In the figure, 10 is a video signal of an analog image at a certain time, 11 is a video signal of an analog image at another time, and 12 is a sampling interval (δ _t ). Reference _numeral 13 denotes a digital image signal obtained by A / D converting the analog image video signal 10 at intervals of δ _t , 14 is a digital image signal obtained by A / D conversion of the video signal 12, and 15 is a signal thereof. It is the signal of the difference image. Thus, when the video signals 10 and 11 showing the same luminance change are temporally shifted as shown in FIG. 15A, the digital images are different from each other like signals 13 and 14 shown in FIG. 15B. When these differences are taken, a difference such as the signal 15 of the difference image shown in FIG. 7C is detected from the signals at the two times, which should originally have the same level.

The error due to this A / D conversion becomes relatively large in a location where a change in brightness is abrupt, such as an edge portion. FIG. 16 shows a change detection result when the background has an edge. In the figure, 1 is an image at a certain time t, 2 is an object in the image 1, 3 is an image at another time t + 1, 4
Is an object in the image 3, and 16 is an edge of the background in the images 1 and 3. Reference numeral 5 is a difference image between these two images 1 and 3. However, if there is a brightness fluctuation as described in FIG. 15, a brightness difference at the edge portion as shown by 17 may appear,
Other than changes due to abnormal events may be detected.

Further, in a liquid drop detecting device for detecting a drop of a drop such as an oil drop or a water drop, it is difficult to determine an optimum binarization level because the drop of the drop to be detected is an extremely minute change.
It cannot be distinguished from the error of the edge part, and sufficient detection accuracy cannot be obtained. Similarly, even in a smoke detection device that detects the occurrence of smoke, etc., it is difficult to determine the optimum binarization level because it is possible to detect fluctuation components with low contrast, and it is impossible to distinguish it from the error of the edge part. It is difficult to obtain sufficient detection accuracy.

[0012]

Since the conventional change area detecting device is constructed as described above, the alignment is required except when the fixed camera is used, and even if the fixed camera is used, the vibration of the camera itself does not occur. If the effect cannot be ignored, registration is necessary, and a considerable amount of calculation is required to perform the registration, and the combination of the difference method and registration is not suitable for real-time processing of moving images. was there.

Further, in digital image processing, an analog image signal captured by a light receiving element of a camera is sampled to be converted into a digital image. However, due to a sampling error at this time, luminances at the same position have different luminances at the same position. The error is not always the same, and this error becomes relatively large in a portion where the luminance changes abruptly, such as an edge portion. Therefore, a change may be detected in the edge portion even between images that are not originally misaligned, and even if alignment is performed between images that are misaligned, the edge portion that is not originally changed in the difference image is detected. Since there is a possibility that a change may be detected, there is a problem that further misdetection may occur even after performing the alignment.

Sufficient detection accuracy is required in the detection of extremely small changes such as detection of oil drops or water drops, and in the detection of fluctuation components with low contrast such as detection of smoke events. There was also a problem that it could not be obtained.

Further, since the conventional change area detecting device detects a change in the two images by the difference method, it detects such a change when a person enters the screen or a rapid illumination change. However, there is also a problem that the abnormal event that should be detected and the change factor due to the disturbance that should not be detected cannot be distinguished from each other.

The present invention has been made to solve the above problems, and a first object thereof is to provide a change area capable of accurately and stably detecting only a change area that should be originally detected in a moving image. To get the detector.

A second object is to obtain a changed area detecting device which does not erroneously detect a change in the background portion.

A third object of the present invention is to obtain a change area detecting device capable of easily detecting a minute change such as a drop of a droplet.

A fourth object of the present invention is to obtain a changed area detecting device which can easily detect fluctuation components such as smoke events.

A fifth object of the present invention is to obtain a change area detecting device capable of surely discriminating a change due to a detection target and a change due to a disturbance which should not be detected.

[0021]

According to a first aspect of the present invention, there is provided a change area detecting device which detects a time-brightness change detecting means for detecting a time change between two images which are temporally different from each other. The erroneous detection removing means for removing the change in the background portion based on the information from the background feature detecting means for detecting the background feature of at least one of the images.

In the change area detecting device according to the second aspect of the present invention, the background feature detecting means includes a spatial brightness change detecting means for detecting a spatial brightness change of one image and a detected image. It is provided with a binarizing means for binarizing.

Further, in the change area detecting device according to the invention described in claim 3, the background feature detecting means includes two spatial brightness change detecting means for detecting the spatial brightness change of each of the two images, respectively. It is provided with two binarizing means for binarizing the detected image and the feature synthesizing means for obtaining a synthesized image of the respective binary images.

Further, in the change area detecting apparatus according to the invention described in claim 4, the background feature detecting means detects the spatial brightness change detecting means for detecting the spatial brightness change of one image, and the detected image It is provided with a binarizing unit for binarizing and an image expanding unit for expanding the obtained binary image.

Further, in the change area detecting device according to the invention described in claim 5, the background feature detecting means has two spatial brightness change detecting means for detecting the spatial brightness change of each of the two images, and each of them. It is provided with two binarizing means for binarizing the detected image, image expanding means for expanding the respective binary images, and feature combining means for obtaining a composite image of the respective expanded images.

Further, in the change area detecting device according to the sixth aspect of the present invention, the erroneous detection removing means is provided with a logical product means for taking a logical product of the output of the time brightness changing means and the output of the background feature detecting means. It is a thing.

The change area detecting apparatus according to the invention of claim 7 is provided with an image emphasizing means for emphasizing two temporally different images in front of the temporal brightness change detecting means. Is.

Further, in the changed area detecting device according to the invention described in claim 8, the image from which the false detection factor is removed by the false detection removing means is accumulated between the false detection removing means and the changed portion detecting means. The image accumulating means is provided.

According to a ninth aspect of the present invention, there is provided a changing area detecting device, wherein the changing portion detecting means calculates an area of the changing portion, and a determination is made to determine a disturbance based on the obtained area. It is equipped with means.

Further, the change area detecting device according to the invention of claim 10 is provided with a disturbance determining means for removing a portion related to the disturbance determined based on the luminance distribution from the changed portion detected by the changed portion detecting means. It is a thing.

[0031]

According to the erroneous detection removing means in the invention described in claim 1, based on the background feature image from the background feature detecting means,
By removing the image change generated in the background feature part from the time change between images detected by the time-brightness change detection means, it is possible to detect only the changed part such as an abnormal event to be detected.

Further, the background feature detecting means in the present invention according to claim 2 detects the background feature by binarizing the detected image detected by the spatial luminance change detecting means by the binarizing means to obtain the background feature image. Stable detection is performed, and erroneous detection due to brightness fluctuation of the background part is suppressed.

Further, the background feature detecting means in the invention according to claim 3 is provided with a spatial luminance change detecting means and a binarizing means for each of the two images, and the respective binary images are combined by the synthesizing means. By synthesizing and obtaining the background feature image, the background feature is detected stably, and the erroneous detection due to the brightness fluctuation and the position shift of the background portion is suppressed.

In the background feature detecting means in the invention described in claim 4, the binary image detected by the spatial luminance change detecting means and binarized by the binarizing means is subjected to dilation processing by the image dilating means. By applying the image to obtain the background feature image, the background feature is stably detected, and erroneous detection due to the brightness fluctuation and the position shift of the background portion is suppressed.

Further, the background feature detecting means in the invention described in claim 5 is provided with a spatial luminance change detecting means, a binarizing means and an image expanding means for each of the two images, and the respective expanded images are By synthesizing by the synthesizing means to obtain the background feature image, the background feature is stably detected, and erroneous detection due to the brightness fluctuation of the background portion and the position shift is suppressed.

In the erroneous detection removing means according to the sixth aspect of the present invention, the logical product means takes the logical product of the output of the time-luminance change detecting means and the output of the background feature detecting means and occurs in the background portion. By excluding the change from the determination target, erroneous detection of the change in the background portion is suppressed.

Further, the image enhancing means in the invention of claim 7 enhances the two images inputted to the temporal brightness change detecting means to detect a minute change such as a drop of a liquid droplet. make it easier.

Further, the image accumulating means in the invention according to claim 8 accumulates the images from which the erroneous detection factors have been removed by the erroneous detection removing means and inputs the accumulated images to the change area detecting means so that a smoke event is generated. It facilitates detection of fluctuation components with low contrast.

Further, the changing portion detecting means in the invention described in claim 9 distinguishes an abnormal event from a disturbance based on the area of the changing portion, thereby eliminating an image change due to the disturbance and becoming a detection target. It makes it possible to easily detect only image changes.

Further, the disturbance determining means in the invention described in claim 10 determines the changed portion due to the disturbance based on the luminance distribution of the image from which the erroneous detection factor is removed, and uses the changed portion detected by the changed portion detecting means. By removing the part related to the disturbance before, the image change due to the disturbance is removed,
It is possible to easily detect only the image change to be detected.

[0041]

【Example】

Example 1. Embodiment 1 of the invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a change area detecting apparatus according to an embodiment of the present invention. In the figure, 21,2
Reference numeral 2 denotes two images that are temporally different, and is assumed to have been A / D converted in this embodiment. Reference numeral 23 is a time-luminance change detecting means for detecting a temporal change between these two images 21 and 22, and 24 is a background feature detecting means for detecting a background feature of the images 21 and 22.
Reference numeral 25 is an erroneous detection removing means for eliminating the erroneous detection of the image variation by removing the change occurring in the background feature portion based on the detection information of the temporal brightness change detecting means 23 and the detection information of the background feature detecting means 24. Is a changed portion detecting means for detecting a changed portion due to equipment abnormality or the like from the image from which the false detection factor is removed by the false detection removing means 25.

Next, the operation will be described. This Example 1
Then, in order to detect the equipment abnormality, the temporal luminance change detecting means 23 detects the temporal image change between the two images 21 and 22 which are temporally different. The image change detected by the time-luminance change detecting means 23 is caused by a positional shift between images, A / D, in addition to an abnormal event.
There are also fluctuations in brightness at the time of conversion, and even if there is no abnormal event, a change in the image may be detected, which causes erroneous detection. Such a change factor other than the abnormal phenomenon is likely to occur in the background portion where there is a large change in brightness as an image. Therefore, the background feature detecting means 24 detects the background features of the images 21 and 22 and outputs them together with the output of the temporal brightness change detecting means 23 and the erroneous detection removing means 25.
By inputting to and synthesizing, the change caused in the background is removed. In this way, the image from which the erroneous detection factor has been removed by the erroneous detection removing means 25 is input to the changed portion detecting means 26, and the area that has changed over time is detected as a changed portion due to equipment abnormality or the like.

As described above, of the changes occurring between two images which are temporally different from each other, the change in the background characteristic portion where erroneous detection is likely to occur is eliminated, so that the change originally to be detected in the moving image is detected. It is possible to realize a highly reliable changed area detection device that accurately and stably detects only an area and is free from erroneous detection.

In the above embodiment, the case in which two background images which are different in time are input to the background feature detecting means to detect the background feature is described. However, one of the two images is detected. Only the background feature may be input to the background feature detecting means to detect the background feature.

Example 2. Next, a second embodiment of the present invention will be described with reference to the drawings. FIG. 2 is a block diagram showing a background feature detecting means used in a change area detecting apparatus according to another embodiment of the present invention. In the figure, 21 is an image, and 24 is a background feature detecting means, which is equivalent to the one given the same reference numeral in FIG. Further, 31 is an edge detecting means as a spatial brightness change detecting means for detecting a spatial brightness change of the input image 21, and 32 is a binarizing means for binarizing the output of the edge detecting means 21. The background feature detecting means 24 is configured.

Next, the operation will be described. Example 2
Then, the background feature detecting means 24 is provided with the edge detecting means 31 and the image 21 is input thereto, and the spatial luminance change, that is, the edge of the input image 21 is detected. Then, the obtained edge image is input to the binarizing means 32, and only a portion of the edge image having an intensity equal to or higher than a predetermined threshold value is "1", and an intensity of the edge image is less than the threshold value. Is set to “0” and binarized to generate a background feature image.

As described above, by having the edge detecting means 31 and the binarizing means 32 as the background characteristic detecting means 24, a background characteristic image for excluding the vicinity of an edge where noise is likely to occur is not subject to change detection is generated. This can be used to eliminate erroneous detection of image changes due to background changes.

In the above embodiment, the case where the background feature is generated by using the edge image representing the spatial luminance change has been described. However, in addition to this, the background feature image is generated by using various statistics obtained from the image. It is also possible to do so.

Embodiment 3 FIG. Next, a third embodiment of the present invention will be described with reference to the drawings. FIG. 3 is a block diagram showing a background feature detecting means used in a change area detecting apparatus according to still another embodiment of the present invention. In the figure, 31a, 31b
Is the first and second equivalent to the edge detecting means in FIG.
Edge detecting means 32a, 32b are first and second binarizing means equivalent to the binarizing means 32, and 33 is the first
And the feature synthesizing means for synthesizing the binary images from the second binarizing means 32a and 32b.

Next, the operation will be described. This Example 3
Then, two images 2 input to the background feature detecting means 24
Of the images 1 and 22, the image 21 is the first edge detecting means 31.
a, the image 22 is input to the second edge detecting means respectively, and the first and second edge detecting means 31a, 31b are inputted.
Detects the edge of the input image 21 or 22, respectively. Next, the edge image from the first edge detecting means 31a is the first binarizing means 32a, and the edge image from the second edge detecting means 31b is the second binarizing means 32b.
Input to the first and second binarizing means 32a, 32
In b, a binary image is generated in which only edges having an intensity equal to or higher than a predetermined threshold value among the edges of those edge images are “1” and edges having an intensity less than the threshold value are “0”. In this way, the two binary images generated by the first and second binarizing means 32a and 32b are input to the feature synthesizing means 33, and the logical sum of the two is taken to obtain the two images. 21,
22 background features are combined and used as a background feature image.

The two images 21, 2 input in this way
By synthesizing each of the two background features, erroneous detection can be eliminated even for an image having a positional deviation of one pixel or more.

Example 4. Next, a fourth embodiment of the present invention will be described with reference to the drawings. FIG. 4 is a block diagram showing a background feature detecting means used in a change area detecting apparatus according to still another embodiment of the present invention. Corresponding parts are designated by the same reference numerals as those in FIG. 2 and their explanations are omitted. Reference numeral 34 in the drawing denotes an image expanding unit that expands the binary image generated by the binarizing unit 32.

Next, the operation will be described. This Example 4
Then, the edge of the image 21 input to the background feature detecting means 24 is detected by the edge detecting means 31, and among the edges of the detected edge image, an edge having an intensity equal to or higher than a predetermined threshold is “1”, The binarizing means 32 performs the binarizing process for setting the edge having the intensity less than "0". Next, the binary image is input to the image expanding unit 34, and by performing expansion processing on the binary image, the edge that is "1" of the binary image is expanded and emphasized, and the background feature Let it be an image.

After detecting the edge of the input image 21 as described above, expansion processing is performed to emphasize the background feature, thereby making it possible to eliminate erroneous detection due to a positional shift of about one pixel.

Example 5. Next, a fifth embodiment of the present invention will be described with reference to the drawings. FIG. 5 is a block diagram showing a background feature detecting means used in a change area detecting apparatus according to still another embodiment of the present invention. Corresponding parts are designated by the same reference numerals as those in FIG. 3 and their explanations are omitted. Incidentally, 34a and 34b in the figure
Are first and second image expanding means equivalent to the image expanding means 34 in FIG.

Next, the operation will be described. This Example 5
Then, two images 2 input to the background feature detecting means 24
The first and second edge detecting means 31a and 31b detect the edges of each of the first and second edges, and the edge images obtained next are first and second binarizing means 32.
Binarize at a and 32b. Further, for each of the binary images, first and second image expanding means 34a,
Expansion processing is performed by 34b. The two dilated images thus obtained are input to the feature synthesizing means 33, and the background features of the two images 21 and 22 are synthesized by taking the logical sum of the two, and used as the background feature image. .

Two images 21, 2 input in this way
By emphasizing the background features for each of the two, and combining them, it is possible to eliminate erroneous detection even for an image having a positional deviation of one pixel or more.

Example 6. Next, a sixth embodiment of the present invention will be described with reference to the drawings. FIG. 6 is a block diagram showing an erroneous detection removing means used in a change area detecting apparatus according to still another embodiment of the present invention. Corresponding parts are designated by the same reference numerals as those in FIG. 1 and their explanations are omitted. In the figure, reference numeral 41 indicates the logical product of the change area detected by the temporal brightness change detecting means 23 and the background feature area part detected by the background feature detecting means 24, and the background feature area is detected from the detected change area. It is a logical product means that excludes a part from detection targets. The false detection removing means 25 is formed by the logical product means 41.

Next, the operation will be described. The image obtained by the temporal brightness change detection means 23 may or may not include an abnormal event, but in both cases, it contains a change component due to brightness fluctuation near the edge of the background. Therefore, the background feature detection means 24 detects the background feature of the input image, and the image is detected as the temporal brightness change detection means 2
It is input to the logical product means 41 together with the image obtained in 3.
In the logical product means 41, the logical product of the inverted binary image from the background feature detection means 24 and the image from the temporal brightness change detection means 23 is calculated, and the change in the background edge portion is "0", and other than that. By leaving the change as it is, the change generated in the background portion is removed from the image detected by the temporal brightness change detection means 23.

In this way, by taking the logical product of the images from the time-luminance change detecting means and the background feature detecting means,
The change near the edge is removed from the image detected as the change, and the false detection near the background edge can be removed.

Example 7. Next, a seventh embodiment of the present invention will be described with reference to the drawings. FIG. 7 is a block diagram showing a main part of a change area detecting apparatus according to still another embodiment of the present invention. Corresponding parts are designated by the same reference numerals as those in FIG. 1 and their explanations are omitted. In the figure, 27 is the input images 21 and 22.
Is an image enhancing means for performing emphasis processing on each of them and inputting them to the temporal brightness change detecting means 23 and the background feature detecting means 24.

Next, the operation will be described. Usually, when a droplet such as an oil droplet or a water droplet is dropped, the droplet is photographed as a round or bright spot immediately after the dropping with a slow dropping speed, and the dropping speed increases after a while after the dropping. When interlaced scanning is performed, the image is divided into an odd field and an even field. If the circle is photographed in a certain size immediately after dropping,
Droplets are also detected in noise removal after differential binarization, but if they are bright spots, they may disappear due to noise removal. Further, the droplets photographed in the odd field and the even field may be lost by the noise removal. Therefore, each of the images 21 and 22 is input to the image enhancing means 27, and a maximum value filter having the maximum value in the vicinity of the pixel as the brightness value of the pixel is acted on each pixel to generate minute droplets. After emphasizing to a certain size, it is input to the temporal brightness change detecting means 23 and the background feature detecting means 24.

By inputting the input images 21 and 22 to the image enhancing means 27 and operating the maximum value filter in this way, the bright spots are enhanced and enlarged, and the bright spots are photographed separately in the odd field and the even field. Since the objects are collected into one, it is possible to reliably detect even a minute change such as a droplet.

In the above embodiment, the images 21 and 22 different in time are generated, and then the respective images are emphasized by the individual image emphasizing means 27. However, the images photographed by the camera are shown. It is also possible to generate two screens that are temporally different from the one that is emphasized.

Example 8. Next, an eighth embodiment of the present invention will be described with reference to the drawings. FIG. 8 is a block diagram showing a main part of a change area detecting device according to still another embodiment of the present invention. Corresponding parts are designated by the same reference numerals as those in FIG. 1 and their explanations are omitted. In the figure, 28 is an image accumulating means for temporally adding the outputs from the erroneous detection removing means 25 to create a cumulative image and inputting it to the changed portion detecting means 26.

Next, the operation will be described. Normally, when a fluctuation component such as a smoke event is present in the image, a difference in brightness that is considered abnormal is not seen only by the difference between the two images 21 and 22. However, by adding the images of this difference temporally by the image accumulating means 28, it is possible to detect only the component that fluctuates temporally. At this time, since the images from which the erroneous detection factors are removed by the erroneous detection removing means 25 are temporally accumulated,
The change component due to the brightness fluctuation near the edge of the background is not accumulated. The images accumulated by the image accumulating means 28 in this way are input to the changing portion detecting means 26, and it becomes possible to detect a component that fluctuates in time as a changing portion.

In this way, since the difference images after removing the luminance change generated in the background feature portion are accumulated, only the fluctuation component having a low contrast such as a smoke event is extracted from the image having a background with many edges. Good detection becomes easy.

Example 9. Next, a ninth embodiment of the present invention will be described with reference to the drawings. FIG. 9 is a block diagram showing a changed portion detecting means used in a changed area detecting apparatus according to still another embodiment of the present invention. Corresponding portions are designated by the same reference numerals as those in FIG. 1 and their explanations are omitted. In the figure, reference numeral 51 denotes a binarizing means for binarizing the image from which the erroneous detection removing means 25 has eliminated the erroneous detection factor. In the binary screen from the binarizing means 51, density values above a certain fixed value. Is an area calculating means for calculating the area of the area determined to be present, and 53 is the area calculating means 52.
The disturbance determining means removes the changed portion due to the disturbance based on the area value calculated in.

Next, the operation will be described. As described above, in the image obtained by the false detection removing unit 25, the change occurring in the background feature portion, that is, the false detection factor is removed. However, the temporal brightness change detection means 2
The change detected in 3 is due to a cause other than noise due to a positional shift between images and background characteristics such as brightness fluctuation at the time of A / D conversion, for example, when a person enters the screen, or when there is a sudden change in illumination. Even if a screen change occurs that is not caused by a facility abnormality, such as when it occurs, an alarm is issued. In order to prevent this, in the ninth embodiment, the image obtained by the erroneous detection removing means 25 is inputted to the binarizing means 51, and if the density value exceeds a predetermined value, it is "1", and the others are "0". After the binarization processing, the area is input to the area calculating means 52. The area calculating means 52 calculates the area of the changed portion having a density value equal to or higher than a predetermined value, and the disturbance determining means 53 calculates the area calculating means 5.
On the basis of the area calculated in step 2, it is discriminated whether it is a changed part due to equipment abnormality or a changed part due to other disturbance. That is, a range of the area of a region to be detected as an abnormal event is set in advance, and when the area of the changed portion calculated by the area calculation means 52 deviates from the set range, it is removed as a disturbance and caused by the abnormal event. Only detect.

By thus performing the determination based on the area after detecting the changed portion, it is possible to easily distinguish the change due to the detection target to be detected and the change due to the disturbance which should not be detected.

Example 10. Next, Example 10 of the present invention
Will be described with reference to FIG. FIG. 10 is a block diagram showing a disturbance determining means used in a change area detecting apparatus according to still another embodiment of the present invention. In the figure, 29 is the disturbance determining means and 25 is an erroneous detection removing means. Further, in the disturbance determining means 29, 61 is a luminance information detecting means for detecting the luminance distribution in the image from which the erroneous detection factor is removed by the erroneous detection removing means 25, and 62 is a change due to the disturbance based on the luminance distribution. It is a judging means for judging a part.

Next, the operation will be described. When the image in which the noise factor due to the background is removed by the erroneous detection removing unit 25 is input, the disturbance determining unit 29 measures the brightness distribution in the image by the brightness information detecting unit 61. Here, if the image is an image showing a temporal change between images at different times, it is possible to know how much change is occurring from the information of the luminance distribution. For example, when a person enters the screen, or when a disturbance occurs such as a sudden illumination change, a high brightness portion appears in the time-varying image. Further, when there is no change in the image, or when there is a change due to a low-contrast smoking event or the like, no high-luminance portion appears in the time-varying image. Therefore, when smoke detection or the like is performed, the brightness information detection unit 61 calculates the brightness distribution of the time-varying image in this way and determines the presence or absence of a high-luminance portion, thereby changing the image due to disturbance that should not be detected. Can be identified.

By thus detecting the image change and detecting the luminance information in the image from which the noise factor due to the background is removed, the change due to the low contrast detection target to be detected and the change due to the disturbance not to be detected are detected. Can be easily distinguished.

Example 11. Next, Example 11 of the present invention
Will be described with reference to FIG. FIG. 11 is a block diagram showing an embodiment in which the present invention is applied to a droplet detection device. In the figure, reference numeral 71 is A for A / D converting the image from the camera.
A / D conversion circuit, 72 is a maximum value filter for enhancing the A / D converted image, and 73 is an image memory in which the enhanced image is temporarily stored. Reference numeral 74 is a difference circuit for calculating the difference between the image at the current time directly input from the maximum value filter 72 and the image once before stored in the image memory 73, and 75 is read from the image memory 73. It is an edge detection circuit that detects an edge of an image. A logical product circuit 76 removes an erroneous detection factor based on the logical product of the difference image from the difference circuit 74 and the edge image from the edge detection circuit 75. Reference numeral 77 is a binarization circuit that binarizes the image from which the erroneous detection factor is removed, and 78 is a noise removal circuit that removes noise from the obtained binary image.
Reference numeral 79 is an area calculation circuit that calculates the area of the changed portion in the image from which noise has been removed, and reference numeral 80 is a disturbance determination circuit that removes the changed portion related to the disturbance based on the obtained area of the changed portion.

Next, the operation will be described. An analog luminance signal obtained from the CCD element of the camera (not shown) is input to the A / D conversion circuit 71 and converted into a digital luminance signal. A maximum value image is obtained by operating a maximum value filter 72 that outputs the maximum value in the vicinity of 3 × 3 of a certain pixel to this digital luminance signal. This maximum value image is once written in the image memory 73 and, at the same time, also input to the difference circuit 74. Further, in this image memory 73, the maximum value image from the maximum value filter 72 is written, and the image data of the maximum value image written one time before is also read out, and it is input to the difference circuit 74. It The difference circuit 74 receives the image data at the current time from the maximum value filter 72 and the image memory 73.
Based on the image data one hour before, a difference image between two temporally different images is generated and output to the AND circuit 76. Further, the image data read out from the image memory 73 is also input to the edge detection circuit 75, and the edge detection circuit 75 obtains an edge image from the edge detection circuit 75 and has an edge strength of a certain value or more in the edge image. Binarization is performed in which pixels are "1" and pixels having an intensity less than that are "0", and the binarized images are output to the AND circuit 76 as a background feature image. The AND circuit 76 performs an operation of setting the pixel which is “1” in the background feature image from the edge detection circuit 75 in the difference image from the difference circuit 74 to “0”, and changes in the background feature portion. Forced to "0". As a result, the erroneous detection factor is removed and only the change in the part other than the background feature is detected.

Further, the image from which the erroneous detection factor is removed by the above calculation is binarized by the binarization circuit 77, and then,
The noise removal circuit 78 removes isolated noise. In this way, a binary image is obtained, but if there is a change other than the background characteristic part, "1" is included in this binary image.
Area remains. In the area calculation circuit 79, this binary image is input, and "1" for each connected area in the remaining area.
Is counted, and the total sum is used as the area of the region. At this time, if the detection target is oil droplets or water droplets, the area of the changed portion is obviously different in size compared to when a person enters the image or when there is a sudden illumination change. Become. Therefore, the disturbance determination circuit 80 determines the area of this changed region, and if it is larger than a predetermined area, the changed portion is removed as a disturbance and the true value of oil drops, water drops, etc. is removed. Only the detection target of is detected.

As described above, after the maximum value filter 72 is operated, the erroneous detection factor due to the background variation is removed by the AND circuit 76, and the disturbance is removed based on the area of the changed portion calculated by the area calculation circuit 79. By doing so, it becomes possible to favorably detect a detection target such as a minute droplet.

Example 12. Next, Example 12 of the present invention
Will be described with reference to FIG. FIG. 12 is a block diagram showing an embodiment in the case where the present invention is applied to a smoke generation detecting device. Corresponding parts are designated by the same reference numerals as those in FIG. 11 and their explanations are omitted. In the figure, 75a is an edge detection circuit for detecting an edge of an image from the image memory 73, 75b is an edge detection circuit for detecting an edge of an image from the A / D conversion circuit 71, and 81 is both edge detection circuits 75a. , 75b
Is an OR circuit for generating an edge image by taking the logical OR of the outputs of the above and inputting it to the AND circuit 76. Reference numeral 82 denotes an image accumulation circuit for adding a predetermined number of images from which the erroneous detection factors have been removed by the AND circuit 76 and inputting the obtained accumulated image to the binarization circuit 77. Reference numeral 83 denotes the AND circuit 76. A histogram calculation circuit that measures the luminance distribution of the image from which the erroneous detection factor is removed, and 84 is a disturbance determination circuit that determines the changed portion due to the disturbance based on the obtained luminance distribution.

Next, the operation will be described. CC of camera
The analog luminance signal obtained from the D element is converted into a digital luminance signal by the A / D conversion circuit 71. The digital luminance signal is once written in the image memory 73, and at the same time, also input to the difference circuit 74 and the edge detection circuit 75b. Further, in the image memory 73, A /
While the image data is written by the D conversion circuit 71, the image data written one time before is also read out and input to the difference circuit 74 and the edge detection circuit 75a. The difference circuit 74 generates a difference image between two temporally different images based on the image data at the current time from the A / D conversion circuit 71 and the image data at the previous time from the image memory 73. Output to the AND circuit 76. On the other hand, the edge detection circuit 75a that receives the image data from the image memory 73 and the edge detection circuit 75b that receives the image data from the A / D conversion circuit 71 perform edge detection of the input image data and obtain the obtained edge data. In each edge image, a pixel having an edge strength of a certain value or more is “1”,
Binarization is performed in which pixels having an intensity less than that are set to “0”. The two binary edge images thus obtained are input to the logical sum circuit 81 to be combined, and are input to the logical product circuit 76 as a background feature image. The AND circuit 76 performs an operation of setting the pixel, which is “1” in the background feature image from the OR circuit 81, in the difference image from the difference circuit 74 to “0”, and changes in the background feature portion. Forced to "0". As a result, the erroneous detection factor is removed, and only the change in the part other than the background feature is detected.

Further, the images from which the erroneous detection factors have been removed by the above calculation are input to the image accumulating circuit 82 and added by the designated number. The image from which the erroneous detection factor has been removed is also input to the histogram calculation circuit 83 at the same time, and the luminance distribution of the difference image is measured. The image for which the designated number of images have been accumulated is sent to the binarization circuit 77 and binarized. In the normal case, the changed area is not detected, but the area having a change of a certain level or more is detected. When low-contrast changes such as smoke events are accumulated, the individual difference images do not detect large changes, but when a person enters the image or when there is a sudden change in illumination, A large change is also detected in the difference image of. The disturbance determination circuit 84 reads this change from the histogram of the luminance distribution measured by the histogram calculation circuit 83, excludes the changed portion with high luminance as the disturbance, and only the true detection target with low contrast such as a smoke event. To detect.

As described above, after the erroneous detection factor due to the background variation is removed by the AND circuit 76, the image accumulating circuit 8
By accumulating the difference images in 2, and removing the disturbance based on the luminance distribution measured by the histogram calculation circuit 83,
It is possible to satisfactorily detect a detection target such as a smoke event having a low contrast.

[0082]

As described above, according to the first aspect of the present invention, only the changes other than the background feature are detected from the outputs of the time-luminance change detecting means and the background feature detecting means. Therefore, an erroneous detection factor in the background portion can be easily removed, and a highly reliable changed area detection device can be obtained.

According to the second aspect of the present invention, as the background feature detecting means, the spatial brightness change detecting means and the binarizing means are combined, and a portion having a large brightness change in the image is used as the background feature. Since it is configured as described above, it is possible to stably detect the characteristics of the background, it is possible to effectively remove the brightness fluctuation of the edge portion, and it is possible to obtain a highly reliable change area detection device. There is.

According to the third aspect of the invention, as the background feature detecting means, the spatial luminance change detecting means and the binarizing means are provided for each of the two images, and the composite image of them is used as the background. I configured it for use as a feature,
The background feature can be detected stably, and the error due to the positional deviation as well as the edge brightness fluctuation can be effectively removed, and a highly reliable change area detection device can be obtained. .

Further, according to the invention described in claim 4, as the background feature detecting means, the image detected by the spatial luminance change detecting means and binarized by the binarizing means is expanded by the image expanding means, Since it is configured to use the part of the image where the brightness change is large as the background feature, it is possible to stably detect the background feature, and it is possible to effectively remove the error due to the position shift as well as the brightness fluctuation of the edge part. As a result, it is possible to obtain a highly reliable changed area detection device.

Further, according to the invention described in claim 5, as the background feature detecting means, a spatial luminance change detecting means, a binarizing means and an image expanding means are provided for each of the two images, and these are provided. Since the composite image is configured to be used as the background feature, it is possible to detect the background feature in a stable manner, and it is possible to effectively remove the error due to the positional deviation as well as the fluctuation of the brightness of the edge. There is an effect that a high change area detection device can be obtained.

Further, according to the invention described in claim 6, since the erroneous detection removing means is constituted so as to take the logical product of the image showing the background feature and the change detection image, the erroneous detection removing means can be formed in the background portion with a simple means constitution. It is possible to remove a changed portion that has not occurred as a detection target, and it is possible to obtain a highly reliable changed area detection device.

Further, according to the invention described in claim 7, since the two images which are temporally different from each other are subjected to the emphasis processing and inputted to the temporal luminance change detecting means, the droplets and the like are detected. There is an effect that it is possible to obtain a change area detection device that can reliably detect even minute changes.

According to the eighth aspect of the invention, since the images from which the erroneous detection factors due to the background noise are removed are accumulated and then input to the changed portion detecting means,
There is an effect that it is possible to obtain a changed area detection device that can detect a temporal fluctuation component that cannot be detected by only two images.

Further, according to the invention described in claim 9, the area calculating means and the judging means are combined as the changing portion detecting means to judge the area of the changing portion. Therefore, the area different from the detection object is detected. A change area having a value is processed as a disturbance, and there is an effect that a change area detecting device that does not cause erroneous detection can be obtained.

According to the invention described in claim 10,
Since it is configured to have a disturbance determining unit that removes a disturbance based on the luminance distribution measured by the luminance change detecting unit, a changed portion having a luminance change different from that of the detection target is processed as a disturbance to prevent false detection. There is an effect that it is possible to obtain a non-change area detection device.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a change area detection device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a background feature detecting means in Embodiment 2 of the present invention.

FIG. 3 is a block diagram showing a configuration of a background feature detecting means in embodiment 3 of the present invention.

FIG. 4 is a block diagram showing a configuration of a background feature detecting means in embodiment 4 of the present invention.

FIG. 5 is a block diagram showing a configuration of a background feature detecting means in embodiment 5 of the present invention.

FIG. 6 is a block diagram showing a configuration of an erroneous detection removing unit according to a sixth embodiment of the present invention.

FIG. 7 is a block diagram showing a main part of a change area detection device according to a seventh embodiment of the present invention.

FIG. 8 is a block diagram showing a main part of a change area detection device according to an eighth embodiment of the present invention.

FIG. 9 is a block diagram showing a configuration of a changed portion detecting means in embodiment 9 of the present invention.

FIG. 10 is a block diagram showing a configuration of a disturbance detecting means in Embodiment 10 of the present invention.

FIG. 11 is a block diagram showing a configuration of a droplet detection device according to an eleventh embodiment of the present invention.

FIG. 12 is a block diagram showing the structure of a smoke detection device according to a twelfth embodiment of the present invention.

FIG. 13 is a block diagram for explaining a conventional changed area detection device using a difference method.

FIG. 14 is an explanatory diagram showing the principle of alignment by SSDA.

FIG. 15 shows a fluctuation of synchronization of an analog image signal,
It is explanatory drawing which shows a mode that an error is detected in a digital image.

FIG. 16 is an explanatory diagram showing a detection result when an edge other than a moving object is in the background.

[Explanation of symbols]

21,22 images, 23-hour brightness change detecting means, 24
Background feature detecting means, 25 erroneous detection removing means, 26 changed portion detecting means, 27 image enhancing means, 28 image accumulating means, 29 disturbance determining means, 31 spatial brightness change detecting means (edge detecting means), 31a first spatial brightness Change detecting means (first edge detecting means), 31b second spatial luminance change detecting means (second edge detecting means), 32 binarizing means, 32a first binarizing means, 32b second binary Value conversion means, 33 feature synthesis means, 34 image expansion means,
34a 1st image expansion means, 34b 2nd image expansion means, 41 AND operation means, 52 Area calculation means, 53
Judgment means.

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Takashi Hirajo 8-1-1 Tsukaguchihonmachi, Amagasaki City Mitsubishi Electric Corporation Industrial Systems Research Institute (72) Inventor Akinobu Seki 8-1-1 Tsukaguchihonmachi, Amagasaki Mitsubishi Electric Corporation Industrial Systems Research Institute (72) Inventor Koichi Sasakawa 8-1-1 Tsukaguchi Honcho, Amagasaki City Mitsubishi Electric Corporation Industrial Systems Research Institute (72) Inventor Shinichi Kuroda 8-1-1 Tsukaguchi Honcho, Amagasaki Industrial Systems Research Center, Mitsubishi Electric Corporation

Claims (10)

[Claims] 1. A time-brightness change detecting means for detecting a time change between the two images, which is temporally different from each other, from two temporally different images.
A background feature detecting unit that detects a background feature of at least one of the two images to generate a background feature image, and a temporal brightness change detecting unit based on the background feature image generated by the background feature detecting unit. False detection removing means for removing the change occurring in the background portion from the detected changes, and the change portion that has changed over time from the image in which the change occurring in the background portion is removed by the false detection removing means. A changed area detecting device comprising a changed portion detecting means. 2. The background feature detecting means detects a spatial brightness change detecting means for detecting a spatial brightness change of one of the two images and an image obtained by the spatial brightness change detecting means with a predetermined threshold value. A binarization unit that generates the background feature image by binarizing the binarized image.
Change area detection device according to. 3. The background feature detecting means detects a spatial brightness change of one of the two images, and an image obtained by the first spatial brightness change detecting means. A first binarizing means for binarizing
Second spatial luminance change detecting means for detecting the other spatial luminance change of the two images, and second for binarizing the image obtained by the second spatial luminance change detecting means with a predetermined threshold value. A feature for generating the background feature image by synthesizing a binarization unit, an image binarized by the first binarization unit, and an image binarized by the second binarization unit The changing area detecting device according to claim 1, further comprising a synthesizing unit. 4. The background feature detecting means detects a spatial brightness change detecting means for detecting a spatial brightness change of one of the two images and an image obtained by the spatial brightness change detecting means with a predetermined threshold value. 2. A binarizing unit for binarizing, and an image dilating unit for dilating the image binarized by the binarizing unit to generate the background characteristic image. Change area detection device described. 5. The background feature detection means detects a spatial brightness change of one of the two images and a first spatial brightness change detection means and an image obtained by the first spatial brightness change detection means. A first binarizing means for binarizing
First image expanding means for expanding the image binarized by the first binarizing means, and second spatial brightness change detecting means for detecting the other spatial brightness change of the two images. When,
Second binarizing means for binarizing the image obtained by the second spatial luminance change detecting means with a predetermined threshold, and expansion processing of the image binarized by the second binarizing means. A characteristic for generating the background feature image by synthesizing the second image expanding means for performing The changing area detecting device according to claim 1, further comprising a synthesizing unit. 6. The erroneous detection removing means obtains by the temporal luminance change detecting means by taking a logical product of the image obtained by the background feature detecting means and the image obtained by the temporal luminance change detecting means. 2. The change area detecting apparatus according to claim 1, further comprising a logical product unit for removing a change occurring in a background portion of the captured image. 7. The change region detecting apparatus according to claim 1, further comprising image enhancing means for enhancing two temporally different images input to the temporal brightness variation detecting means. . 8. The erroneous detection removing means temporally adds and accumulates images obtained by removing the change occurring in the background portion from the changes detected by the temporal luminance change detecting means, and accumulates the images. 2. The changed area detecting device according to claim 1, further comprising image accumulating means for inputting to the changed portion detecting means. 9. An area calculating means for calculating the area of a time-varying changed portion in an image in which the change occurring in the background portion is removed by the erroneous detection removing means, the changed portion detecting means, and the calculated area calculating means. The changed area detection device according to claim 1, further comprising a determination unit that removes a changed portion due to a disturbance based on the area. 10. The luminance distribution in the image from which the change occurring in the background portion is removed by the false detection removing means is determined, the changed portion due to disturbance is determined based on the obtained luminance distribution, and the changed portion is determined. The change area detecting device according to claim 1, further comprising a disturbance determining means for removing a change portion due to the disturbance from a change portion detected by the detecting means.