JP2865442B2 - Method of extracting change area of surveillance image - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a surveillance image change area extracting method for extracting a surveillance image change area in an image monitoring apparatus for detecting an intruding object in a surveillance area.

[0002]

2. Description of the Related Art FIGS. 7 and 8 show an example of this type of image monitoring apparatus. That is, I
The TV camera (imaging means) 1 captures an image in the monitoring area 8 and converts it into an electric signal. The image picked up by the ITV camera 1 is sent to a processing device 3, a display device 5, and a video tape recorder (hereinafter simply abbreviated as VTR) 7 via a transmission path 2. The display device 5 displays a monitoring image captured by the ITV camera 1, and the VTR 7
Surveillance images captured by the ITV camera 1 are continuously
Alternatively, it is recorded when an intruding object is detected.

On the other hand, the image signal sent to the processing device 3 is
The signal is converted into a digital signal by an A / D converter 11 by a sampling pulse of a predetermined cycle output from the sampling pulse generating circuit 10 and stored in the image memory 12 as image data at time t = ti, for example.

Then, in the difference binarization circuit 14, the image data at the time t = ti is stored in the image memory 12 before the image data to extract the change area, as described later.
A difference image is obtained by performing a pixel-to-pixel difference operation on the image data at time t = ti-1 that has been captured in the image data, and the changed pixel is further converted to a difference binary image represented by "1". It is converted and stored in the difference binary image memory 15.

[0005] The CPU 13 has a differential binarized image memory 15.
The analysis of the change area is performed using the difference binarized image in
If it is determined that the change area is an intruding object, the alarm device 6 is sounded, and if the VTR 7 is not operating, it is operated to record a monitoring image. If it is determined that the object is not an intruding object, the image is continuously acquired and
The above process is repeated.

[0006] Incidentally, as an inter-pixel operation for extracting a change area, generally, a difference from a background image when there is no intruding object, or a time t = ti-1 before a predetermined time Δ is used.
Either of the time-series differences from the image of FIG. With the difference from the background image, it can be detected even if the intruding object is stopped, but if the environment such as brightness changes, such as outdoors, the intruding object will not exist unless the background image is updated. Also has a characteristic that a change region is generated.

In the case of a time-series difference, since the difference is generally made at relatively close time intervals, the follow-up to environmental changes is excellent. However, when the intruding object stops, it cannot be detected. Even if the intruding object is moving, the difference between the two screens is characterized in that both the disappeared portion and the generated portion of the intruding object are detected.

In order to eliminate the latter disadvantage of the time series difference, instead of the two screens of t = ti and t = ti−1, t
= Ti + 1, the difference calculation is performed on the previous two screens and the rear two screens, respectively, and the intrusion object in the image of t = ti is obtained by the logical product of the obtained two difference images. In some cases.

Further, when the CPU 13 analyzes a changed area from a difference binary image, a value of "" is used as a basic method of image processing in order to obtain a predetermined integrated area.
A serial number is assigned to a closed area which is 1 "(labeling), and the area and distance of each partial area are extracted to collectively extract the area.
In many cases, the area that is 1 "is calculated and analyzed.
Performing this operation on the entire image area takes a very long time.

Therefore, for example, Japanese Patent Application Laid-Open No. 62-14789
As disclosed in Japanese Patent Application Publication No. 1-101, there has been proposed a method of roughly classifying a changed area from a differential binary image. This is because, as shown in FIG. 9, projection histograms are obtained in the X and Y axis directions, respectively, in the binary difference image, and the heights equal to or larger than the predetermined values Xth and Yth are determined to be the predetermined width Xw. , Yw or more, and a method of analyzing the inside of the divided area.

That is, only the divided areas A11 to A22 in FIG. 9 are examined to analyze whether or not the changed area is an intruding object. At the time of analysis of the changing area, analysis using labeling can be performed as described above. Since the sectioning area is often a circumscribed rectangle of the intruding object, as in the case of the sectioning areas A11 and A22 in FIG. When the segmented area has a predetermined size, analysis may be performed by a method such as directly inspecting the shape or calculating the area of a portion having a value of "1" without performing processing such as labeling. .

The method of image difference and the method of analyzing an intruding object have been described above. However, since the captured image contains a noise component, it is general to perform spatial filtering or logical filtering.

[0013]

What is important in the case of an image monitoring apparatus is to detect a changed area and determine whether the detected changed area is an intruding object.
For this purpose, it is necessary to extract a change area that is likely to be an intruding object from all monitoring areas for detecting an intruding object, and determine whether or not each extracted change area is an intruding object.

In general, image processing takes a long time, so that these processes must be performed at a high speed in order to operate on an actual device. From this point, considering the above method, the method of obtaining and analyzing the segmented area from the histogram of the difference binary image in the X and Y axis directions often requires less time than inspecting the entire image area. If the number of intruding objects is small and the noise is small as shown in FIG. 9, it can be said that the segmented area becomes a circumscribed rectangle of the intruding object, and the intruding object can be easily analyzed.

However, the change area is rarely extracted as a united area as shown in FIG. 9, and as shown in FIG. 10, the change area is divided and extracted by many areas. If the number of intruding objects increases, the circumscribed rectangle of each intruding object does not match the intruding object. And it takes time.

Therefore, according to the present invention, the number of areas in which it is necessary to analyze whether an object is an intruder is reduced, and the number of cases where the extracted area becomes a circumscribed rectangle of the intruder is increased. It is an object of the present invention to provide a method for extracting a change region of a monitoring image, which enables an object to be analyzed at high speed.

[0017]

According to the present invention, there is provided a method for extracting a change region of a monitoring image, wherein an image in the monitoring region is captured, and the captured image is continuously captured at predetermined time intervals and digitized. In an image monitoring apparatus that monitors an intruding object in the monitoring area by arithmetically processing a plurality of digitized images, a difference image between continuously captured and digitized images is calculated. A difference binarized image obtained by binarizing the calculated difference image is generated, and a histogram obtained by integrating the generated difference binarized image in the X and Y-axis directions is calculated. A first divided area is obtained by combining areas in which sections larger than the predetermined value continue for a predetermined section or more, and the obtained first divided areas are again multiplied in the X and Y axis directions, respectively. Histogram calculated was, in each histogram this calculation, the second by combining lasted region above the predetermined interval is continuously larger interval than the predetermined value
Is obtained, and an intruding object is analyzed in the obtained second divided area.

[0018]

In the first divided area, by further obtaining the second divided area, the area which must be analyzed to determine whether or not the object is an intruding object is reduced, and the extracted area is intruded. Since the number of circumscribed rectangles of the object increases, the analysis of the subsequent intruding object can be performed at high speed.

[0019]

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 shows an example of a differential binarized image when a plurality of intruding objects occur, as an example for explaining the present invention. This figure is an explanatory diagram for obtaining a first divided area. Hx1 and Hy1 are histograms in the X and Y-axis directions, respectively, and Xth1 and Yth1 are first histograms Hx1 and Hy1 respectively. Threshold values Xw1 and Yw1 are width threshold values for obtaining the first segmented area, respectively. In this figure, Hx1, Hy1 and X
From the values of th1, Yth1, Xw1, and Yw1, the boundary values X1s1, X1e1, X2s1, X2e1, Y1s1, and Y1e1 of the first partitioned area are obtained.
, Y2s1, Y2e1 are obtained.

This is the same as the conventional divided area. In the figure, in the area A11, one intruding object is included, but the area is wide due to the influence of the intruding objects in the areas A12 and A21. Since it has been extracted and further includes a noise portion, in order to detect this intruding object, a process such as labeling for separating in the area A11 must be performed. The same applies to the area A12. In the area A21,
Although two invading objects are included, they are extracted as one area under the influence of the area A11, and it is necessary to perform processing such as labeling in the same manner.

In FIG. 2, histograms in the X and Y axis directions are obtained again in each of the divided areas A11 to A22.
FIG. 3 to FIG. 6 show the second section in each first section area.
FIG. 9 is an explanatory diagram for obtaining a divided area of FIG. In the figure, Hxi2, Hyi2 (i = 1, 2, 3, 4) are histograms in the X and Y directions in each of the divided areas A11 to A22, and Xth2, Yth2 are divided by Hxi2, Hyi2, respectively. The height thresholds Xw2 and Yw2 for obtaining the area are respectively the width thresholds for obtaining the divided area. In this figure, from Hxi2, Hyi2 and Xth2, Yth2, Xw2, Yw2, the boundary values Xijs2, Xije2, Yijs2,
Yije2 is found.

FIG. 3 is a view in which the section area A11 is sectioned.
The second divided area is smaller than the first divided area, the area to be analyzed is smaller, and the second divided area is a rectangle circumscribing the intruding object. The same applies to FIG. In FIG. 5, in the first sectioned area, the two intruded objects connected in the X-axis direction are cut off because the area A11 is no longer affected by the area A11. Is a circumscribed rectangle of the intruding object. In FIG. 6, since there is no intruding object, X4js2 and X4je2 cannot be obtained, and it can be seen that there is no intruding object in this area, and it becomes unnecessary to analyze.

That is, by finding the second divided area in each of the first divided areas, all the intruding objects in FIG. 2 are extracted in narrow areas, respectively, and in some of them, the intruding object is It can be seen that it is a circumscribed square. The method of determining an intruding object in the second sectioned area is the same as the conventional method. Labeling or the like is performed as necessary, and whether or not the object is an intruding object is determined based on the shape and area. That is, by obtaining the second segmented area, an intruding object can be detected at a higher speed than in the related art.

Next, the procedure of the monitoring process of an intruding object to which the method of the present invention is applied will be described with reference to the image monitoring apparatus shown in FIGS. 7 and 8 and the flowchart shown in FIG.
Here, a method of obtaining a change area using three consecutive images will be described. However, other methods have no problem at all, and similar effects can be obtained. First, a detection sensitivity level for detecting an intruding object is set. The detection sensitivity level is
This is a threshold value for obtaining a difference binary image representing a changed portion later from the time-series difference image.

With the above, the preparation is completed.
The A / D converter 11 converts the image signal from the TV camera 1 into an A / D signal.
D conversion is performed, and at a predetermined time interval, the time t is stored in the image memory 12.
= Ti. And the difference 2
The binarization circuit 14 calculates the difference between pixels between the image data at time t = ti and the image data at time t = ti-1 which was stored in the image memory 12 before the time interval Δt. As a result, a difference image representing the change is obtained, and further, this difference image is binarized by a preset detection sensitivity level to generate a difference binary image Bi.

Since this image includes a portion where the change from t = ti-1 to t = ti has disappeared and a portion where the change has occurred, the image is further captured at time t = ti + 1 after a predetermined time interval Δt. Similarly, from the difference image at t = ti and t = ti + 1, t
Binary image B representing a change from t = ti to t = ti + 1
get i + 1. The time t = ti is applied to the differential binary image Bi + 1.
Since the disappearance part and the occurrence part of the change from t to t + 1 are included, the binary image of the intruding object at time t = ti is obtained by taking the logical product of Bi and Bi + 1. Pi is obtained. This image is stored in the difference binarization memory 15.
The CPU 13 detects the presence or absence of a change area in the monitoring area 8 by referring to the difference binarization memory 15.

In detecting the presence or absence of a change area, the CPU 1
Reference numeral 3 denotes a projection histogram obtained in the X- and Y-axis directions in the difference binary image Pi, and a height equal to or more than a predetermined value Xth1 or Yth1 in the obtained histogram is a predetermined width Xw1 or Yw.
Find a first segmented area that is greater than or equal to w1. If the first segmented area cannot be obtained, it is determined that no change has occurred in the monitoring area 8 at time t = ti, and the image Bj (j is an integer) is continuously obtained at a predetermined capture time interval Δt.
Is obtained, a first section area is obtained, and the presence or absence of a change area is checked.

When the first divided area is obtained, there is a possibility that the first divided area is an intruding object. Therefore, in each of the first divided areas obtained from the differential binarized image Pi, the X and Y axes are again obtained. The histogram integrated in the direction is calculated, and the predetermined values Xth2, Y
A section in which the height equal to or larger than th2 is equal to or larger than the predetermined widths Xw2, Yw2 is obtained, and a second sectioned area is obtained by combining the sections, and an intruding object is analyzed in this area.

That is, labeling is performed on the difference binarized image Pi, the area and distance of each partial area are obtained, a grouped area is extracted, and the shape and area are calculated and the intruding object is calculated in this area. Is determined. Here, if the second segmented area is about the same size as the circumscribed rectangle of the intruding object to be predicted, it is determined whether the intruding object is obtained by calculating the shape and area without performing processing such as labeling. It is also possible to determine whether or not. The calculation time of the second segmented region of the intruding object is short because the second segmented region is generally smaller than the original image including the region of the mass change such as the intruding object. In the analysis of this intruding object,
If it is determined that the object is an intruding object, a command is issued to the alarm device 6, whereby the alarm device 6 issues an alarm for abnormality detection.

[0031]

As described above in detail, according to the present invention, within the first segmented region obtained from the differential binarized image,
Further, by obtaining the second segmented area, the area that must be analyzed for an intruding object is reduced,
In addition, since the number of cases where the extracted region becomes a circumscribed rectangle of the intruding object increases, it is possible to provide a surveillance image changing region extracting method that enables high-speed subsequent analysis of the intruding object.

[Brief description of the drawings]

FIG. 1 is a flowchart for explaining an intruding object monitoring processing procedure to which a monitoring image changing area extraction method according to an embodiment of the present invention is applied.

FIG. 2 is a diagram illustrating an example of a differential binarized image when a plurality of intruding objects occur.

FIG. 3 is an explanatory diagram for obtaining a second divided area.

FIG. 4 is an explanatory diagram for obtaining a second divided area.

FIG. 5 is an explanatory diagram for obtaining a second divided area.

FIG. 6 is an explanatory diagram for obtaining a second divided area.

FIG. 7 is a configuration diagram illustrating an example of an image monitoring device.

FIG. 8 is a block diagram illustrating a configuration of an image monitoring device.

FIG. 9 is a view showing an example of image data for explaining a conventional method of extracting a change area of a monitoring image.

FIG. 10 is a view showing an example of image data for explaining a conventional method of extracting a change area of a monitoring image.

[Explanation of symbols]

1 ... ITV camera (imaging means), 2 ... Transmission path, 3 ...
... Processing device, 4 ... Light pen, 5 ... Display device, 6 ...
... alarm device, 7 ... VTR, 8 ... monitoring area, 11 ...
A / D converter, 12 image memory, 13 CPU
14 ... Difference binarization circuit, 15 ... Difference binarization image memory.

Continuation of front page (72) Inventor Shozo Abe 70, Yanagicho, Saiwai-ku, Kawasaki-shi, Kanagawa Prefecture Inside the Toshiba Yanagimachi Plant (56) References JP-A-62-147891 (JP, A) JP-A-3-41582 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) H04N 7/18 G06T 7/20 G08B 13/196

Claims (1)

(57) [Claims] An image in a monitoring area is captured, the captured image is continuously captured at predetermined time intervals and digitized, and a plurality of digitized images are processed to perform the monitoring. In an image monitoring apparatus for monitoring an intruding object in a region, a difference image between continuously captured and digitized images is calculated, and the calculated difference image is binarized into a difference binary image. Is generated, and X and Y are generated by the generated difference binarized image.
Histograms integrated in the axial direction are calculated, and in each of the calculated histograms, a region in which a section larger than a predetermined value is continuously continued for a predetermined section or more is obtained to obtain a first divided area. In each of the first divided regions, a histogram integrated in the X and Y-axis directions is calculated again, and in each of the calculated histograms, a region in which a section larger than a predetermined value continuously continues for a predetermined section or more is combined. A method of extracting a changed area of a monitoring image, comprising: obtaining a second divided area; and analyzing an intruding object in the obtained second divided area.