JP2967483B1 - Detection target extraction processing device - Google Patents

Abstract

[Summary] [Object] In an outdoor environment where direct external light is incident, detection is performed without being affected by illuminance fluctuation or shadow, and accurately extracting a detection target without losing the characteristics of the detection target, and detecting without erroneous detection. An object extraction processing device is provided. [Constitution] Auto iris lens 1, double speed CCD color camera 2, RGB converter 3, A / D converter 41, 4
2, 43, RGB image memories 51, 52, 53, a hue operation unit 6, a reference image memory 7, a sample image memory 8, and a difference processing unit 9, and the difference due to the hue of the video output from the 2x speed CCD color camera. By performing the processing, the detection target is accurately extracted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an illuminance fluctuation, particularly in a portion where direct external light is applied, such as outdoors, and a shadow portion, in an abnormality detection device, a parked vehicle detection device, and a person detection device using an image processing device. The present invention relates to a detection target extraction processing apparatus when the apparatus is used in a general environment having a large illuminance difference as described above.

[0002]

2. Description of the Related Art Conventionally, image processing is performed based on the image of a television camera installed in a place where illuminance changes constantly due to changes in sunlight, such as outdoors or indoors having windows and openings, and detection is performed. As a device for extracting an object, there are an abnormality detection device, a parked vehicle detection device, a number of people detection device, and the like. In any of these, the time during which the sunlight changes is shorter than the period of the video signal sent from the television camera so that the change in the image due to the change in the illuminance is not abnormal and is not erroneously detected as a detection target such as a car or a person. By taking advantage of the long time, perform the difference processing between the reference image and the sample image between two consecutive screens or within the shortest possible time, or use the latest image that has detected that there is no target object as the reference image A method of updating and performing a difference process between the reference image and the sample image is used.

[0003] In addition, it has been practiced to use a normal auto iris lens to keep the average density of the entire image constant and to reduce the influence of illuminance fluctuation.

[0004]

However, in the above-mentioned conventional method, the temporal change of the entire image including the detection target is detected in the difference processing between two screens in a continuous or short time. However, there arises a problem that the characteristics of the detection target are lost or that the stationary detection target is not extracted.
In addition, when the illuminance changes while the detection target is on the processed image, the background cannot be updated, which causes a problem that it is difficult to extract only the detection target. In addition, when a normal auto iris lens is used, if there is a sudden change in illuminance that the auto iris lens cannot follow, or if there is a large detection target in the detection field of view, the auto iris lens is the average density of the entire image, or the maximum or minimum density. Is controlled so as to maintain a constant value, which causes a problem that erroneous auto iris lens control is performed. Further, a shadow portion caused by direct external light is erroneously extracted as a detection target regardless of the background.

The present invention has been made in consideration of the above-described circumstances, and in a general environment in which the illuminance constantly fluctuates, the characteristics of the detection target are lost or the shadow of the detection target is erroneously extracted without being affected by the environmental fluctuation. It is an object of the present invention to provide a detection target extraction processing apparatus that does not erroneously detect and can perform a detailed characteristic analysis after the extraction of the target by accurately extracting the detection target without causing the detection.

[0006]

According to the present invention, there is provided a double-speed CCD color camera equipped with an auto iris lens, an RGB converter, an A / D converter, an RGB image memory, a hue calculator, a reference image memory, and a sampler. The image processing apparatus includes an image memory and a difference processing unit, and performs difference processing based on hue image data.

[0007]

When the illuminance is changed, the aperture value is controlled to an appropriate value by the auto iris lens. At the same time, the saturation of the video signal is avoided by the double speed CCD color camera for the extremely bright part and the extremely dark part. Accordingly, it is possible to always stably extract only the detection target without losing the hue.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a detection object extraction processing apparatus according to the present invention. In FIG. This is an auto iris lens for controlling the aperture of the lens by 2b, and is usually controlled so that the average value of all the incident light amounts to the CCD element becomes constant. 2x speed CCD
The camera 2 has images with different exposure amounts (for example, 1/60 second and 1
/ 2000 sec) twice the frequency (1 Hz) of normal (60 Hz)
20Hz), image processing LSI using two images
The image is edited into an optimal image that does not saturate the brightness gradation of both the brighter and darker parts of the image, and the normal NTSC (Nation
al TelevisionSystem Commi
tte) is output as a color composite video signal 2a (the 2 × speed CCD camera is commercially available from Matsushita Communication Industrial Co., Ltd. as a super dynamic CCD camera).

Reference numeral 3 denotes an RGB converter for converting the NTSC color composite video signal 2a into RGB analog video signals 3a, 3b, and 3c. Reference numerals 41, 42, and 43 denote RGB analog video signals 3a, 3b, and 3c as n bits. A / D converters 51, 52, and 53 convert digital image data 41a, 42a, and 43a that express a black level as 0 and a white level as ²ⁿ in shades of gray, and A / D converted digital image data 41a. , 42a, and 43a are stored in the RGB image memories 51, 52, and 53, respectively. RGB digital image data 51a, 52a,
53a, each pixel of each of the RGB digital image data 51a, 52a, 53a is calculated based on the following equation. The hue calculation unit 7 for calculating the hue H ° accordingly has two consecutive hue image data 6a output from the hue calculation unit 6.
The difference between the screens is obtained, and the difference amount, which is the number of pixels having a difference equal to or greater than a certain threshold value, is monitored for a certain period of time. It is determined that there is no detection target (for example, an intruder) inside, and the hue image data obtained last is referred to as the reference image data 6b.
The reference image memory 8 stores the hue image data obtained from the next of the reference image among the hue image data 8a sequentially transferred from the hue calculation unit 6 as sample image data 6C.
The sample image memory 9 stores the difference between the reference image data 7a transferred from the reference image memory 7 and the sample image data 8a transferred from the sample image memory 8, and the result of the calculation is equal to or more than a certain value. That is, the sample image portion having a hue change is detected object object image data 9a.
A difference processing unit that outputs as

The operation of the detection object extraction processing apparatus of the present embodiment having the above configuration will be specifically described below using an actual processed image.

The images (a) to (f) shown in FIG. 2 show the double-speed CCD color camera 2 in the vicinity of the entrance where external light is directly incident, using the detection object extraction processing apparatus described with reference to FIG. It displays the processed image inside the device when installed and used, and the processed image when using a general auto iris lens and CCD color camera.
(1) No intruder, indirect external light incident, (2) No intruder,
The operation will be described for each case of direct external light incidence, (3) presence of an intruder, and direct external light incidence.

(1) In case of no intruder and indirect external light incident FIG. 2 (a) is a color image picked up by the 2 × -speed CCD color camera 2, and in FIG. (Medium light blue), a2 part is a high luminance wall (high lightness yellow), a3 is a low luminance door (low lightness brown), and a4 part is a medium luminance plant (medium lightness green). The aperture control of the auto iris lens 1 works by the change of the indirect external light,
The average luminance in the field of view of the 2 × -speed CCD color camera 2 is kept constant.

Here, in order to clarify the relationship between the control of the auto iris lens 1 and the brightness of each part in FIG. 2A, a description will be given with reference to FIG. In FIGS. 3A to 3E, the vertical axis represents the luminance gradation (8
2), the horizontal axis represents the area of each part, and the auto iris lens 1 multiplies the average luminance gradation of FIG. 2A, that is, the area of each part of FIG.
Control is performed so as to keep the averaged luminance gradation value (indicated by a broken line in FIG. 3A) constant. In FIG. 3A, the luminance gradation of each part of a1 to a4 is not saturated (0 or 255).

The color image shown in FIG. 2A is output from the double-speed CCD color camera 2 as an NTSC color composite video signal 2a, and is converted into an RGB converter 3 and an A / D converter 4.
The hue is calculated by the hue calculation unit 6 via the RGB image memories 51, 52, and 53, and the hue is sequentially output as hue image data 6a. The content of the hue image data 6a is as follows: a1 in FIG.
Part is yellow H = 99 °, a3 part is brown H = 55 °, a
Four parts have green H = 167 °.

The hue images obtained after monitoring the sequentially output hue image data for a certain period of time are stored in the reference image memory 7 as reference images. After the reference image data 6a is transferred from the hue calculation unit 6 to the sample image memory 8, an operation of sequentially storing the next output hue image as a sample image in the sample image memory 8 is performed. When the amount of incident indirect external light changes in this state, the hue calculated by the hue calculator 6 is not obtained because the brightness gradation of the image portions a1 to a4 is not saturated by the aperture control of the auto iris lens 1 described above. The image does not change, and the same hue data as before the indirect external light change is stored in the sample image memory 8. The difference processing unit 9 calculates the difference in hue between the reference image stored in the reference image memory 7 and the sample image stored in the sample image memory 8. In this case, since the images have the same hue, the difference is constant. No difference image larger than the value is generated, and an image without the detection target is output as the detection target image data 9a. The same operation is performed in this case also when a normal auto iris lens and a CCD color camera are used.

(2) No Intruder and Direct External Light Incident FIG. 2C shows a state in which the hue image obtained from the color image of FIG. 2A is stored in the reference image memory 7 as reference image data. As shown, when external light is directly incident on a part c5 of the floor portion, the double-speed CCD color camera uses FIG.
Brightness gradation of each part of the image by 1/60 second exposure shown in
Using two images of the luminance gradation of each part of the image by the exposure of 1/2000 second shown in FIG. 3C, the image is edited into an image in which the luminance gradation of each part of the image is not saturated. In this case, b5 in FIG.
Since the brightness gradations of the portion and the b1, b2, b4, and b3 portions of FIG. 3C are saturated, the b5 portion is the optimum using FIG. 3C, and the other portions are the optimum using FIG. 3B. As an image, FIG.
The image shown in FIG.
Output as a TSC color composite video signal 2a, an RGB conversion unit 3, A / D conversion units 41, 42, 43,
The hue is calculated by the hue calculator 6 via the RGB image memories 51, 52 and 53, sequentially output as hue image data 6a and transferred to the sample image memory 8.
In each of the parts c1, c2, c3, c4, and c5 in (c), since the luminance gradation is not saturated, the hue is the reference image shown in FIG.
(A), the difference processing unit 9 does not generate a difference image with a certain threshold value or more, and the detection target image data 9a is an image having no detection target. Here, in order to clarify the operation of the present invention, a case where a normal CCD color camera is used as the configuration of the present invention will be described.

FIG. 2B shows a color image using the auto iris lens 1 and a normal CCD color camera.
As in (c), a case where external light is directly incident on a part of the floor as shown by b5 in FIG. 2 (b) is shown.

As described above, the aperture control of the auto iris lens 1 is performed so as to keep the average luminance in the visual field constant, and therefore, as shown in FIG. Is saturated to the white side (luminance gradation 255), and the door portion b3 is saturated to the black side (luminance gradation 0). therefore,
When the hue of each part of the image shown in FIG. 2B is calculated by the hue calculation unit 6, the floor b1 and the wall b where no external light is directly incident are obtained.
2. The same hue as that before direct light is incident on the plant b4 is obtained, but the floor b5 on which direct external light is incident is achromatic white (no hue), and the door b3 is achromatic black (no hue). 2 as a difference image having a certain threshold value or more from the difference processing unit 9.
The b5 part and the b3 part of (b) are output.

(3) When there is an intruder and external light is incident directly FIG. 2E shows a state in which the hue image obtained from the color image of FIG. 2A is stored in the reference image memory 7 as reference image data. As shown, external light is directly incident on part e5 of the floor,
In the case where there is an intruder e6, the same operation as in the case of (2) above, where there is no intruder and direct external light is applied, except for the intruder e6 and the shadow e7 of the intruder. Is not saturated with the output from the double-speed CCD color camera like the door part e3, the hue can be calculated, and the hue becomes the same as that of the floor part e5. As a result, the detection target image enters as shown in FIG. 2 (f). Only those who are obtained. FIG. 2D shows an image obtained when a normal CCD color camera is used.
3, the hue of the floor d5 and the shadow d7 of the intruder cannot be calculated due to saturation of the luminance gradation, and an erroneous difference image is output.

As described above, as an embodiment of the detection object extraction processing apparatus of the present invention, a case where the detection object extraction processing apparatus is used in an abnormality detection apparatus has been described. The present invention is also applicable to a person detection device, a parked vehicle detection device, and the like, in which a camera is installed outdoors where external light is directly incident.

[0021]

As described above, a normal auto iris lens detects an object to be detected in hue by using a double-speed CCD color camera in an abnormality detection device or the like installed outdoors where luminance gradation is saturated. Since the extraction is performed, erroneous detection due to a low-brightness color or high-brightness color object or shadow is prevented, and a highly reliable detection target extraction processing apparatus can be provided.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a detection target object extraction processing device according to the present invention.

FIG. 2 is a processed image for explaining the operation of the embodiment.

FIG. 3 is a luminance gradation graph for explaining the operation of the embodiment.

[Explanation of symbols]

 1. Auto iris lens 2.2x speed CCD color camera 3. RGB converter 41.42.43. A / D converter 51.52.53. RGB image memory 6. Hue calculator 7. 7. Reference image memory Sample image memory 9. Difference processing unit

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-8-241412 (JP, A) JP-A-10-289393 (JP, A) JP-A 8-182786 (JP, A) (58) Field (Int.Cl. ⁶ , DB name) G06T 1/00 H04N 7/18 H04N 9/04-9/11

Claims (1)

(57) [Claims] An auto iris lens for controlling an aperture amount of a lens according to an amount of light incident on a CCD element, and outputting two images having different exposure amounts by a double speed CCD element;
A 2 × -speed CCD color camera capable of obtaining an optimum image in which the luminance gradation of both the bright and dark portions is not saturated by an image processing LSI using the 2 × images;
RGB conversion means for converting an NTSC color composite video signal output from a D color camera into each of RGB analog video signals, and converting each of the RGB analog video signals to A /
A / D conversion means for D-conversion, an RGB image memory for storing each of the RGB digital images obtained by the A / D conversion means, and calculation of the hue of each pixel from each of the RGB digital images stored in the RGB image memory Hue calculating means,
A reference image memory that stores a hue image obtained as a reference image after monitoring a hue image obtained by the hue calculation means for a certain period of time, and a sample image memory that stores a hue image obtained continuously from the reference image as a sample image And a difference means for calculating a difference between the reference image stored in the reference image memory and the sample image stored in the sample image memory.