JP2002099909A - Method and device for processing infiltrating object detection image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent erroneous detection due to a virtical image appearing on a background image, when an infiltrating object is stopped in the conventional background image updating process, carried out on the basis of weighted average computing of an inputted image for detecting an invading object. SOLUTION: For preventing the appearance of the virtical image, which is formed by updating of the background image and causes error detection of the object to be detected on the background image, updating of the background image is stopped in an existence area of the detected object and an area around the object receiving influence of a brightness value change of the inputted image, that is, the influence of random noise, while updating is carried out in the area where the object does not exist. In this way, misdetections can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image processing system for automatically detecting an intruding object from an image taken by an imaging device such as a camera.

[0002]

2. Description of the Related Art In a video surveillance system using an image pickup device such as a camera, an object such as a human or a car (vehicle) that intrudes into a surveillance area is automatically detected from a video and an intrusion in which an alarm is taken is performed. For an object detection device, a difference method that calculates a difference in luminance value for each pixel between an input image obtained from an imaging device and a background image in which an intruding object is not reflected, and detects a portion having a large difference value as an intruding object, is a conventional method. Widely used from.

The processing of the difference method will be described with reference to FIG. FIG. 1 is a conceptual diagram illustrating conventional image processing. In the figure, 102 is an input image, 101 is a background image, 1
03 denotes a difference image, 601 denotes a binarized image, 110 denotes a subtraction process, and 602 denotes a binarization process. A subtraction process 110 calculates a difference in luminance value between the input image 102 and the background image 101 for each pixel, outputs a difference image 103, and performs a binarization process 602.
Is 0 if the luminance value of each pixel of the difference image 103 is less than the predetermined threshold value, and 255 is the luminance value which is equal to or greater than the threshold value.
To obtain a binarized image 601. As a result, the object 120 such as an intruder shown in the input image 102 (the figure shows a person) is the portion 603 where the difference is generated by the subtractor 110.
Further, the image 604 is formed by a binarization process 602 as an image 604 having a luminance value 255 equal to or higher than a threshold value.
Is detected as

FIG. 5 is a flowchart of the difference processing. First, in step 201 for creating a background image, a background image 101 (see FIG. 6) is created by performing several weighted averaging operations on an input image captured at regular intervals. In the image input step 202, the captured image is subjected to processing such as correction and processing, if necessary, to obtain the input image 102 (see FIG. 6). In the difference processing step 203, a difference between the background image and the input image is calculated to obtain a difference image 103 (see FIG. 6). In the binarization processing step 204, a binarized image 601 (see FIG. 6) is created from the difference image 103, and in a step 205 for determining an intruding object, the numbering and size of the intruding object are determined based on the binarized image 601. If there is an intruding object, alarm processing 207 is performed as necessary in the conditional branch of step 206. In the background image update step 208, a weighted average calculation is performed on the input image and the background image to update the background image.

FIG. 7 shows a method of updating the background image.
An image obtained by multiplying the input image 102 by the coefficient a and a background image 1
01 is multiplied by a coefficient b, and the sum of the coefficients (a +
The image 106 divided by b) is used as an updated background image. With this method, the intruding object 120 appears as a virtual image 121 in the updated background image 106. Here, when the intruding object 120 is stopped, the virtual image 121 of the background image gradually updated becomes clear by the repetitive processing of this updating method, and finally, the intruding object 12
0 becomes the background image. Therefore, in such a case, the detection accuracy of the intruding object is significantly adversely affected.

FIG. 8 shows the state with the lapse of time. A is the input image 102, B is the background image 101,
In C, the difference images 103 are arranged, and reference numerals 1 to 5 represent respective arithmetic processes with the passage of time. The difference image C1 between the input image A1 and the background image B1 has no intruding object, and is a normal detection operation. An intruding object appears in the input image A2, and a virtual image starts to be formed in the background image B2. However, an intruding object appears in the difference image C2, and a normal detection operation is performed.
When this state continues for a long time, that is, when the intruding object stops for a long time, the virtual image clearly appears in the background image B3,
When the difference C3 from the input image A3 is less than the threshold value, detection becomes impossible, resulting in erroneous detection. At this time, the intruder 120
It is determined that there is no intruding object in spite of the presence of. Thereafter, when the intruding object leaves, the background image B4 still has a clear virtual image, but the input image A
No. 4 already has no intruding object, and a virtual image is detected in the differential image C4, resulting in erroneous detection. That is, the intruding object 120
It is determined that there is an intruding object even though is not present. Then, as the intruding object leaves, the virtual image gradually disappears, and the input image A5, the background image B5, and the difference image C
5 is back to normal.

[0007]

As described above, in the case of detecting an intruding object by the conventional weighted average calculation method, an intruding object is not detected when it should be detected. Alternatively, erroneous detection such as detection when detection should not be performed is performed. An object of the present invention is to reduce such erroneous detection and perform a more normal operation of detecting an intruding object.

[0008]

As described above, in order to prevent an object to be detected, which is a cause of such a problem, from appearing as a virtual image in the background image due to the background image update processing, the present invention provides a method for detecting the object. The updating process of the background image is stopped in the range where the object exists, and the updating process of the background image is performed in the range where the object does not exist, thereby reducing erroneous detection.

In other words, the present invention provides a method for obtaining an image of an intruding object in an image pickup area of the TV camera from A / D conversion of a video signal obtained by imaging with the TV camera into digital image data. An intruding object detection image processing method for detecting an intruding object that has entered an imaging area of the TV camera, that is, by comparing the background image with image data newly captured by the TV camera. In the above, when the intruder is detected, updating of the background image of at least a portion where the intruder is detected is stopped by software processing.

More specifically, when the intruder is detected, the process of updating the background image of the surrounding area including the portion where the intruder is detected is stopped, and the background image of the imaging area other than the area is updated. The processing is performed continuously.

The present invention may be constituted by hardware instead of software processing. That is, the present invention has a TV camera for detecting an intruder, forms an image of a specific area such as an area where an intruder cannot enter, through a lens, and converts the obtained video signal into an image by A / D conversion. Means for converting data into data, means for creating a background image for obtaining the image of the intruder by weighted averaging from the image data obtained by the image data converting means, and storing the image data obtained by the creating means And an intruder detection image processing apparatus having a means for detecting the intruder by obtaining a difference image from the reference image obtained by the creating means and an input image for detecting the intruder, Means for stopping the updating of the background image in such a case.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described with reference to the image processing explanatory diagram of FIG. 1 and the flowchart of FIG. FIG. 1 schematically illustrates the operation of the present embodiment based on processing and an image at that time, and FIG. 2 is a flowchart corresponding thereto.

Steps 201 to 206 of the flowchart are the same as the conventional method. That is, the object 120 to be detected (see FIG. 1) is recognized, and if an intruding object is present, the alarm 207 is performed as before.

However, in this embodiment, the alarm processing 20
7 is different. Next, this operation will be described in detail with reference to FIGS. In the binarization processing step 220, based on the difference image 103 obtained by the subtraction processing 110 in the difference processing step 203, a threshold value smaller than the threshold value used for object detection in the binarization processing step 204 is used. The binarization process 112 is performed. This allows
A binarized image 104 is obtained.

Next, in a background image holding step 221, a logical product operation processing 114 is performed on the binary image 104 obtained in the binary processing step 220 and the background image 101, and the detected object of the background image 101 is An image 107 holding only the existing area 124 is obtained. AND operation processing 1 here
Reference numeral 14 denotes an image in which the area 122 corresponding to the luminance value 255 of the binarized image 104 is left from the background image 101, and the other area (corresponding to the luminance value 0 of the binarized image 104 of the background image 101) is set to the luminance value 0. There is to get. Next, in the background image update processing step 208, the weighted average calculation processing 1
In step 11, an image 106 obtained by updating the background image is obtained. If the background image is held in the branching step 222, the process proceeds to the masking step 223 of the updated image. Here, a mask image 108 is created from the binarized image 105 and the updated background image 106. This binarized image 105 is obtained by binarizing process step 2
The binarized image 104 obtained in step 20 is inverted and produced by an inversion operation processing 113. Logical operation processing 115 is performed on the binarized image 105 and the image 106 obtained by updating the background image obtained in the background image update processing step 208 to obtain a mask image 108. As a result, in the mask image 108, the area 120 where the detected object is located is the background image update processing step 2
The weighted average calculation processing 111 in step 08 results in an image in which the virtual image area 121 to be created in the image 106 in which the background image has been updated is filled in and the area 125 with the luminance value 0 is masked. In the combining process step 224, the mask image 108 and the detected object presence area holding image 107 created in the background image holding step 221 are added by the adding process 116. As a result, the area 120 where the masked detection object exists
Does not perform the update process, obtains an image 109 in which the other areas are updated, and this image 109 becomes the background image 101, and returns to the image input step 202.

Next, the binarization processing 602 of the binarization processing step 204 and the binarization unit 11 of the binarization processing step 220
The difference between the threshold values used in No. 2 will be described with reference to FIG. This figure shows an example of a state of a change in luminance value of the intruding object and its surroundings on the difference image. In addition,
In this embodiment, an infrared camera is used as a means for obtaining an image, and FIG. 3 shows an image from the infrared camera.

In FIG. 1, the luminance distribution of a line 306 in the horizontal direction of an object (indicated by, for example, a person / area 301) of the image is shown below. A line 306 is a line indicating a position at which the change in the luminance value is measured. An area 302 is a lump in which the area 301 of the detected object has a luminance value equal to or larger than the threshold value 303. A line 305 is a line having a luminance value of 0. A threshold line 303 for determining the intruding object area 301 is a threshold in the binarizer 602 in the binarization processing step 203. A line 304 shown surrounding the area 301 above,
313 and 314 respectively connect the same luminance level,
It is represented as a contour line. This brightness level is
The same reference numerals are used at the bottom of the figure.

In the following description, an intruding object detection using an infrared camera will be described as an embodiment.

Binarization processing 1 of binarization processing step 220
The threshold at 12 is represented by line segment 304. Thus, this binarization threshold value 304 is
1 is an area 304P including the surrounding portion. Thus, the section determined by the threshold value 304 is the section 308
It is. The area indicated by this section 308 is the areas 122 and 124 holding the background image and the virtual image 12 of the intruding object.
The areas 123 and 125 mask one part.

FIG. 4 shows a relationship between a change in luminance value at an arbitrary pixel of the input image, a temporal change in luminance value at a pixel at the same position in the background image, and threshold lines 303 and 304. is there. In the figure, the horizontal axis is the time axis, and the vertical axis is an example of the case where the brightness value of the difference image is created by the absolute value. A line segment 402 indicates a change in the luminance value at an arbitrary pixel of the input image. A line segment 401 indicates a change in luminance value of a background image created by weighted averaging of input images of the same pixel point. The luminance value line 401 of this background image
Lines 30 that are parallel at equal intervals above and below
Reference numerals 3 and 304 denote lines (see FIG. 3) indicating threshold values used for object detection, respectively. That is, lines 303 and 304
Is a line indicating a threshold value of an area for determining an area not to be updated. In the section 403, an intruding object appears, and the luminance value changes rapidly in the section. At this time, the interval 307 between the threshold value 303 for detecting an object and the threshold value 304 for determining an area not to be updated is set at an interval that is not affected by random noise indicated by the luminance value of the input image, thereby obtaining a background image. In the update processing step 208, a virtual image having a high probability of being detected due to the influence of random noise is prevented from being created.

As one method of determining the threshold value 304 for determining the area where the background is not updated, the luminance value change width 404 of the input image when no intruding object exists is subtracted from the threshold value 303 for detecting the object. Select a value.

In the above embodiment, an example using an infrared camera has been described. However, an infrared camera is particularly effective because a sudden change in luminance is small. Further, by setting a wider range 304P with respect to the image range 301 of the object, even if the intruding object slightly moves, its influence can be prevented.

In the above description, these intruding object detection operations are performed by computer processing. However, these operations may be performed by hardware.

[0024]

According to the present invention, when the threshold value for detecting an object does not satisfy the above-described condition, the erroneous detection that occurs when the object to be detected is stopped for a long time can be greatly reduced by the conventional method. Becomes In particular, the present invention can provide a good detection operation when a sudden change in luminance value is small, such as a far-infrared camera image.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram illustrating an operation in a background image update process according to the present invention.

FIG. 2 is a flowchart of one embodiment of the present invention.

FIG. 3 is a diagram for explaining an area indicated by a threshold according to the embodiment of the present invention.

FIG. 4 is an explanatory diagram for explaining a relationship between a luminance value of an image and a threshold according to the embodiment of the present invention.

FIG. 5 is a flowchart of object detection image processing according to the related art.

FIG. 6 is an explanatory diagram for explaining an object detection method according to the related art.

FIG. 7 is an explanatory diagram for explaining a background image update process in the related art.

FIG. 8 is an explanatory diagram for explaining erroneous detection by a conventional method.

[Explanation of symbols]

101: background image, 102: input image, 103: difference image, 104: binary image, 105: inverted binary image, 1
06: Image with updated background image, 107: Image with intruding object area retained from background image, 108: Image with virtual image of intruding object masked from image with updated background image, 10
9: background image without virtual image, 110: subtractor, 111: weighted average calculator, 112: binarizer, 113: inversion calculator, 114: logical product calculator, 115: adder, 120:
Intruding object, 121: virtual image of intruding object, 122, 123:
Area where an intruding object is present, 124: an area holding an area where an intruding object is present in a background image, 125: area in which a virtual image of an intruding object in an image obtained by updating the background image is masked, 201: background image creating step, 202: image input step, 203: difference processing step, 204: binarization processing step, 205: intruding object determination step, 2
07: warning processing step, 208: background image update processing step, 220: binarization processing step, 221: background image holding step, 223: background image mask step, 224: synthesis processing step, 301: detected object Area, 302: curve of luminance value change, 303: threshold for detecting an object, 304: threshold for determining an unupdated area, 305: line of difference value 0, 306: measurement line of luminance value change, 307: threshold value difference, 401: luminance value line of one pixel of background image, 402: luminance value line of one pixel of input image, 403: interval where an intruding object was present, 404: random noise of input image 601: binarized image, 602: binarizer, A: change in input image, B: change in background image, C: change in difference image,: temporal change.

Continued on the front page F-term (reference) 5B057 AA19 BA08 BA29 CH08 DA08 DB02 DC32 5C054 CA05 CC03 ED17 FC01 FC04 FC05 FE28 HA18 5C084 AA02 AA07 AA08 AA13 BB04 CC16 CC19 DD15 GG42 GG43 GG52 GG56 GG57 GA048 GA51 HA01 LA04

Claims (8)

[Claims] 1. A video signal obtained by imaging with a TV camera is A / D converted into digital image data, and a background image for obtaining an image of an intruder into an imaging area of the TV camera is obtained from the image data. The invading object detection image processing method for detecting an intruding object that has entered the imaging area of the TV camera by comparing the background image with image data newly imaged by the TV camera. An intruding object detection image processing method, wherein when an object is detected, updating of the background image of at least a portion where the intruding object is detected is stopped. 2. A video signal obtained by imaging with a TV camera is A / D converted to digital image data, and a background image for obtaining an image of an intruder into an imaging area of the TV camera is obtained from the image data. Detects an intruder into the image area of the TV camera by comparing the background image created by the background image update process with the weighted average calculation and the image data newly captured by the TV camera. In the intruder detection image processing method, when the intruder is detected, the updating process of the background image of the surrounding area including the portion where the intruder is detected is stopped,
In addition, the intruder detection image processing method is characterized in that the background image update processing of the imaging area other than the area is continuously performed. 3. A video signal obtained by imaging with a TV camera is converted into digital image data by A / D conversion, and a background image for obtaining an image of an intruder into an imaging area of the TV camera is obtained from the image data. Detects an intruder into the image area of the TV camera by comparing the background image created by the background image update process with the weighted average calculation and the image data newly captured by the TV camera. In the intruder detection image processing method, when an intruder is detected, a threshold of a luminance level including a portion where the intruder is detected is set, and an intruder area is determined based on the threshold. And updating the background image of the imaging area other than the area while continuously stopping the update processing of the background image of the area. 4. The TV according to claim 1, wherein
An intruder detection image processing method, wherein the image obtained from the camera is an infrared image. 5. An image data conversion means comprising a TV camera for detecting an intruder and converting a video signal of the TV camera which has taken an image of a specific area into image data, and obtained by the image data conversion means. Means for creating a background image for obtaining the image of the intruder from the obtained image data by a weighted average calculation, means for storing the image data created by the background image creating means, and means for acquiring the background image created by the background image creating means. In the intruder detection image processing apparatus having means for obtaining a difference image from the reference image obtained from the TV camera and the input image from the TV camera to detect an intruder, when the intruder detector detects an intruder, the background is detected. An intruder detection image processing apparatus, comprising: means for stopping updating of a background image in an image creating means. 6. A means having a TV camera for detecting an intruder, converting a video signal obtained by capturing an image of a specific area such as a restricted area into image data, and an image obtained by the image data converting means. Means for creating a background image for obtaining an image of the intruder from the data by a weighted average calculation, means for storing image data obtained by the background image creating means, and a reference image obtained by the creating means; TV
In an intruder detection image processing apparatus having an intruder detecting means for detecting an intruder by obtaining a difference image from an input image from a camera, the intruder includes the intruder when the intruder detector detects an intruder. Area setting means for setting the specific area, means for stopping updating of the background image of the area set by the area setting means, and means for updating the background image update of the specific area excluding the area containing the intruder And an intruder detection image processing apparatus. 7. A TV camera for detecting an intruder, an image of a specific area such as an off-limits area is formed through a lens, and an obtained video signal is converted into image data by A / D conversion. Means, from the image data obtained by the image data conversion means, means for creating a background image for obtaining the image of the intruder by a weighted average calculation, means for storing the image data obtained by the creation means, In an intruder detection image processing apparatus having a means for detecting an intruder by obtaining a difference image from a reference image obtained by the creating means and an input image for detecting an intruder, when an intruder is detected, An intruder detection image processing apparatus, comprising: means for stopping updating of a background image. 8. The intruder detection image processing apparatus according to claim 5, wherein the TV camera is an infrared television camera.