JP3995832B2 - Image sensor - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image sensor installed in a monitoring area, in particular, a non-detection target moving body such as an insect stuck to the lens of the monitoring camera, and a detection target moving body such as an intruder located away from the monitoring camera. It is related with the image sensor which can distinguish.
[0002]
[Prior art]
As a conventional image sensor, images of building entrances and other monitoring areas are continuously captured, the captured image is compared with a reference image, and the presence / absence of an intruder is determined based on the size of the area that has changed. There is an image monitoring device.
In these conventional image sensors, an abrupt change in the image in the case of an intruder in a changed area, and a gradual change in the image such as a slow curtain swing by an air conditioner or sunlight entering the monitoring area, for example. In order to make a distinction, the reference image is sequentially updated and stored with the latest captured image at regular time intervals.
[0003]
FIG. 7 is a diagram illustrating a state of an insect stuck to the lens of the surveillance camera.
As shown in FIG. 7, the insect 1 stuck to the lens of the surveillance camera installed so as to photograph a desired surveillance area is a moving body of the same size as the intruder who is far away on the photographed image. It is reflected. Furthermore, when the insect moves in the direction of the arrow 2, it appears as if the intruder has moved.
[0004]
[Problems to be solved by the invention]
As described above, in the conventional image monitoring apparatus, the insect stuck to the lens of the monitoring camera and the intruder at a position away from the monitoring camera have substantially the same size image. Further, when the insect moves, the state is similar to the movement of the intruder. Accordingly, there is a problem that the insect stuck to the lens of the surveillance camera may be mistakenly determined as an intruder, and the reliability of the surveillance device is impaired due to false alarms.
[0005]
Therefore, the object of the present invention is to solve the conventional problems, and to distinguish non-detection target moving bodies such as insects attached to the lens of the surveillance camera from detection target moving bodies such as intruders. An object of the present invention is to provide a highly reliable image sensor capable of detecting a detection target moving body without causing erroneous reporting by the target moving body.
[0006]
[Means for Solving the Problems]
Inventors etc. earnestly researched in order to solve the conventional problem mentioned above. As a result, when the image of the insect stuck to the lens of the surveillance camera is closely observed, it has a non-focus area that is out of focus. Therefore, the non-focus area on the image is determined as a non-detection target moving body. I found out that I can do it. This has an automatic focus adjustment function so that the surveillance camera can focus on the door or desk in the surveillance area, but cannot focus on the closest object like an insect stuck to the lens. It is due to that.
[0007]
Further, when the brightness of the changed area of the captured image is closely observed, the difference between the brightness of the changed area and the background decreases as the distance from the center of gravity of the changed area approaches the boundary between the changed area and the background. The insect image approaches the background brightness value as it approaches the boundary.Therefore, when the distance from the center of gravity and the brightness difference between the change region and the background decreases as it approaches the boundary, the change region is determined as a non-detection target moving body. I found out that I can do it.
[0008]
Furthermore, when the luminance distribution in the change region is obtained, it is found that the luminance distribution approximates to a function such as a quadratic function, a trigonometric function, and a Gaussian function. Therefore, the luminance distribution in the change region is obtained and the luminance distribution is It has been found that the change area can be determined as a non-detection target moving body when approximating a predetermined function.
[0009]
Further, by obtaining a difference binarized image between the current image and the reference image, and obtaining a luminance distribution of the current image corresponding to the difference binarized region in a line segment passing through the center of gravity point, the non-detection target mobile body It was found that the judgment can be made with high accuracy.
[0010]
The present invention has been made based on the above-described knowledge, and an image processing unit that captures a monitoring region, and an image process that determines the presence or absence of a detection target moving body based on a change region of an image captured by the image capturing unit. And the imaging means is focused on an area away from the imaging means where the detection target moving body may exist before and after the generation of the change area .
[0011]
According to a first aspect of the image sensor of the present invention, there is provided an imaging unit that captures a monitoring region, and an image processing unit that determines the presence or absence of a detection target moving body based on a change region of an image captured by the imaging unit. The imaging unit has a focus on a region away from the imaging unit where the detection target moving body may exist before and after the generation of the change region,
The image processing means moves the non-detection target moving body when the difference in luminance between the change area and the background decreases away from the center of gravity of the change area and approaches the boundary between the change area and the background. It is characterized by determining.
[0012]
According to a second aspect of the image sensor of the present invention, a luminance distribution in the change area is obtained, and when the luminance distribution can be approximated to a predetermined function, the change area is determined as a non-detection target moving body. Is.
[0013]
According to a third aspect of the image sensor of the present invention, the predetermined function is a quadratic function, a trigonometric function, or a Gaussian function.
[0014]
According to a fourth aspect of the image sensor of the present invention, the luminance distribution is obtained in each of a vertical direction, a horizontal direction and an oblique direction passing through the center of gravity of the image.
[0015]
According to a fifth aspect of the image sensor of the present invention, a difference binarized image between the current image and the reference image is obtained, a barycentric point of the obtained difference binarized image is obtained, and a horizontal line segment and a vertical line passing through the barycentric point are obtained. The luminance distribution of the current image corresponding to the binarized difference area is obtained for the minute and diagonal line segments.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
First, the aspect of the image sensor of this invention is demonstrated in detail.
An image sensor according to the present invention includes an imaging unit such as a monitoring camera that images a monitoring area, and an image processing unit that determines the presence or absence of a detection target moving body based on a change area of an image captured by the imaging unit. .
The image sensor according to the present invention is installed in each of a plurality of rooms in a building, for example, and images a monitoring area including a monitoring target (monitoring point), and an unfocused non-focus area that appears on the obtained image. Is determined as a non-detection target moving body, and false alarms due to sticking insects and the like are eliminated, and the presence or absence of a detection target moving body such as an intruder is accurately determined.
[0017]
Further, in the image sensor of the present invention, when the center of gravity of the change area of the captured image is obtained, the difference between the brightness of the change area and the background decreases as the distance from the center of gravity and approaches the boundary between the change area and the background, The change area is determined as a non-detection target moving body.
Furthermore, in the image sensor of the present invention, the luminance distribution in the changed region of the photographed image is obtained, and when the luminance distribution can be approximated to a predetermined function, the changed region is determined as a non-detection target moving body.
[0018]
Furthermore, in the image sensor of the present invention, the predetermined function described above is a quadratic function, a trigonometric function, or a Gaussian function. From the empirical values, the non-detection target moving body has the obtained luminance distribution very close to the above-described function.
Furthermore, in the image sensor of the present invention, the accuracy is further improved when the above-described luminance distribution is obtained not only in the horizontal direction passing through the center of gravity of the image but also in each of the vertical, horizontal and diagonal directions passing through the center of gravity. To do.
[0019]
Further, in the image sensor of the present invention, a difference binarized image between the current image and the reference image is obtained, a barycentric point of the obtained difference binarized image is obtained, a horizontal line segment passing through the barycentric point, a vertical line segment, and In the diagonal line segment, the luminance distribution of the current image corresponding to the difference binarized area is obtained.
[0020]
Further, the image processing method in the present invention will be described in detail.
Insects and the like attached to the lens of the surveillance camera are not in focus, and thus can be obtained as a defocused image. On the other hand, the intruder is obtained as an image having a texture to some extent. Therefore, by using the characteristics of both, the degree of out-of-focus of the difference extraction image (hereinafter referred to as “out-of-focus”) is calculated and processed with a predetermined threshold value, and the non-focus area is not detected as an insect. It is determined as a target moving body.
[0021]
As a method of calculating the degree of defocus, there is a method of matching the luminance value data (two-dimensional data) of the difference extracted image with a function (two-variable function) representing a defocused state, and setting the degree of matching to the degree of defocus.
[0022]
Furthermore, as another method of calculating the degree of defocusing, the luminance value data (one-dimensional data) in a straight line (line) passing through the center of gravity of the difference extracted image is extracted and matched (matched) with a function (single variable function) representing the defocused state. Thus, there is a method of setting the matching degree to the degree of defocusing in the line. In this calculation method, lines in several directions are further adopted as lines, and the average of the degree of defocusing of these lines is set as the degree of defocusing of the original difference extraction image.
[0023]
In this case, as a method of taking a line, there are, for example, two directions in the X direction and the Y direction passing through the center of gravity of the difference extraction image, and four directions including a 45 ° oblique line in addition thereto. As the function representing the out-of-focus state, the above-described quadratic function, cosine function (cosine), Gaussian function, or the like can be used approximately. As a matching method, a general matching method such as a least square method can be used.
[0024]
When the least square method is used as the matching degree, the average of the least square error is normalized by the width of the luminance value distribution and used as the matching degree. In that case, the smaller the error, the more the matching is considered. Even when other matching methods are used, the error is similarly used as the matching degree.
[0025]
Hereinafter, a more specific description will be given with reference to the drawings.
FIG. 2 is a view showing an image of an insect stuck to the lens of the surveillance camera. A non-focused non-focus area exists within a rectangular frame in the image shown in FIG. FIG. 2B shows an enlarged non-focus area. The background is blackened to make the non-focus area clear. As is clear from FIG. 2B, in the change area of the photographed image, the brightness increases as it approaches the boundary from the center (center of gravity). That is, in relation to the background, the difference in luminance from the background decreases as the distance from the center of gravity approaches the boundary.
[0026]
FIG. 6 is a diagram illustrating a relationship between the barycentric point, the luminance value, and the position of the non-focus area. As is clear from FIG. 6, the luminance value is high for the left and right objects with the center of gravity of the non-focal region as the center and away from the center toward the boundary.
[0027]
FIG. 3 is a diagram illustrating how to obtain the luminance distribution in the change region of the captured image. Image data of the non-focal region is prepared, and the center of gravity (Jx, Jy) of the non-focal region is obtained as shown in FIG. Next, luminance value data of one line passing through the emphasis (in the X direction in the example of FIG. 3A) is obtained. In order to avoid the risk of misjudgment, it is not only determined based on the luminance value data of one line, but in order to further improve accuracy, a plurality of lines passing through the emphasis (in the example of FIG. ) Brightness value data.
[0028]
FIG. 4 is a diagram illustrating a state where the luminance value data is developed. As shown in FIG. 4, when the luminance value data of one line in the X direction passing through the emphasis is developed, it is very close to a parabola (secondary curve). As described above, the luminance value distribution is empirically approximated to a quadratic function, a trigonometric function, and a Gaussian function with high probability.
[0029]
Next, the overall configuration of a monitoring system to which the image sensor of the present invention is applied will be described.
A controller is installed in the building to be monitored, and a plurality of image sensors are connected to the controller. Note that a fire sensor, an emergency button, a mode setting device, and the like may be connected to the controller. The controller is connected to a security center device provided in a remote monitoring center via a telephone line as a communication line.
[0030]
Connection between the image sensor and the controller is performed by an analog signal line and a digital signal line. The analog signal line is constituted by a coaxial cable and transmits a video signal (NTSC system), and the digital signal line is constituted by a twisted pair line and transmits a control signal by the digital signal line.
[0031]
Next, the overall operation of the monitoring system will be briefly described.
The mode setting device sets the monitoring system to a warning release mode or a warning set mode. When the image sensor detects an abnormality in the security set mode, the controller transmits an abnormality signal indicating that an abnormality has occurred in the security center device via the telephone line.
[0032]
When an abnormal signal is output from the fire sensor or the emergency button, the controller transmits data indicating the type of abnormality and the location where the abnormality occurred regardless of the security mode to the security center device via the telephone line.
In addition, the detection method of these abnormal signals, the transmission method, etc. can use what is known well in the said technical field.
[0033]
FIG. 1 is a diagram showing a general configuration of an image sensor of the present invention. As shown in FIG. 1, the image sensor is provided with a control unit 101 and a power source 102 configured by a CPU or the like. Each part inside the apparatus is controlled by the control means and receives power supply from the power source.
[0034]
The control means 101 is provided with a timing means (not shown) for controlling the timing for taking in the current image. This time measuring means is activated every 0.5 seconds, for example.
[0035]
The power source 102 includes a conversion device that converts AC power supplied from the outside into a DC voltage. Instead of providing the power source 102 in the image sensor, power may be supplied from the controller.
In addition, the storage unit 107 is connected to the control unit 101. The storage means 107 includes a program area for storing a program for causing the image sensor to execute a predetermined operation, a parameter area for storing parameters such as an installation height / angle of depression of the image sensor, a monitoring point, sensitivity setting, a work area, A status storage area that stores the status of the monitoring area, that is, whether the current state is normal or abnormal, and a mode storage area that stores either the warning set mode or the warning release mode are provided. ing.
[0036]
The storage means 107 further includes a reference image storage area, a current image storage area, and an abnormal image storage area as areas for storing images taken by the image pickup means 104. In the current image storage area, an image with a normal light amount, an image with a controlled light amount, and an image when no illumination is used, which are captured by the imaging unit 104, are stored.
[0037]
The imaging means 104 that images the monitoring area is constituted by a CCD camera and has sensitivity from the visible area to the infrared area. Infrared light projecting means 103 is provided, and infrared light is projected onto the monitoring area when it becomes dark, such as at night.
The image processing unit 106 compares and collates the current image, which is an image of the normal light amount stored in the storage unit 107, an image of the controlled light amount, and an image when the illumination is not used, with a reference image, and obtains a difference to obtain a binary value. Turn into.
[0038]
The display means 108 is configured by an LED, and is turned on when an abnormality is detected, and is turned off when no abnormality is detected, and displays the presence or absence of abnormality detection outside the image sensor.
The communication means 109 is an interface that transmits and receives signals to and from the controller, and is connected to the controller through a digital signal line. The image output means 110 is an interface for outputting an image when an abnormality occurs, and is connected to an analog signal line via the switching means 111.
[0039]
The address setting unit 112 includes a dip switch, and an address for the controller to specify the image sensor is set.
The operation unit 105 is a unit for turning on and off the power supply 102. Further, when the operation unit 105 is turned off, the switching unit 111 connects the input side and the output side of the analog signal line to bypass the image sensor from the analog signal line.
The current image is stored in the current image storage area, for example, every 0.5 seconds when the time measuring unit is activated.
[0040]
FIG. 5 is a flowchart for removing sticking insects of the image sensor of the present invention. The operation of the image sensor of the present invention will be described using the flowchart of FIG.
[0041]
An image taken with the normal light quantity captured by the imaging unit 104 is captured every time the timing unit is activated (0.5 seconds) (step S1), and then a difference binarization with a reference image captured in advance is obtained (step S2). ). At this time, when the difference binarization area does not exist, the original image is stored as a reference image in the image storage area and is determined to be normal.
[0042]
When the difference binarized area described above exists, the barycentric point of the difference binarized image is calculated (step S3). Next, at this stage, it is determined whether or not the person is an intruder (step S4). When it can be determined that there is a suspicion of the intruder, the following processing is performed.
[0043]
First, the luminance value distribution of the current image corresponding to the difference binarized area is obtained from the horizontal line segment passing through the barycentric point (step S5).
Next, the horizontal luminance value distribution is approximated by a quadratic function (step S6). The approximate calculation is performed by, for example, the least square method.
[0044]
Further, a mean square error between the approximated quadratic function and the luminance value is obtained (step S7). At this time, other than the mean square error (for example, the sum of absolute values of differences between points) may be used.
Next, the above-mentioned mean square error is divided by the brightness value width of the brightness distribution, and the mean square error is normalized (step S8). That is, by dividing by the luminance value width, the error can be evaluated without depending on the luminance value level.
[0045]
Next, if the normalized mean square error is less than or equal to a predetermined value, it can be determined that the horizontal luminance distribution is likely to be an insect (step S9). That is, the approximation is correct when the normalized mean square error is less than or equal to a predetermined value.
Next, the approximated quadratic function obtained by calculating the luminance value distribution of the current image corresponding to the difference binarized area described above with respect to the horizontal line segment passing through the barycentric point, approximating the horizontal luminance value distribution with a quadratic function A series of processes consisting of calculating the mean square error between the brightness value and dividing the mean square error by the brightness value width of the brightness distribution and normalizing the mean square error are further executed in the vertical and diagonal directions, respectively. Then, it is determined whether the luminance distribution in the vertical direction and the oblique direction is likely to be insects.
[0046]
As a result, it is determined whether or not it seems to be a worm in all the horizontal, vertical, and diagonal directions (step S10). As a result of the determination, if it is determined that it is not an insect in all directions, it is determined that it is an intruder (step S11), and if it is determined that it seems to be an insect in all directions, it is determined that it is an insect (step S12).
As described above, by processing the captured image to obtain the luminance value distribution and approximating it with a predetermined function, the non-focal region is automatically detected without the human being continuously confirming the captured image. It can be determined as a target moving body.
[0047]
【The invention's effect】
As described above, according to the present invention, a non-detection target moving body such as an insect stuck to the lens of the surveillance camera can be distinguished from a detection target moving body such as an intruder, and a false alarm is generated by the non-detection target moving body. Therefore, it is possible to provide a highly reliable image sensor that can detect the detection target moving body.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration of an image sensor according to the present invention.
FIG. 2 is a diagram showing an image of an insect stuck to a lens of a surveillance camera.
FIG. 3 is a diagram illustrating how to obtain a luminance distribution in a change area of a captured image;
FIG. 4 is a diagram illustrating a state in which luminance value data is developed.
FIG. 5 is a flowchart for removing sticking insects of the image sensor of the present invention.
FIG. 6 is a diagram illustrating a relationship between a barycentric point, a luminance value, and a position of a non-focus area.
FIG. 7 is a diagram illustrating a relationship between a barycentric point, a luminance value, and a position of a non-focus area.
[Explanation of symbols]
1. Insect 101. Control means 102. Power supply 103. Infrared light projecting means 104. Imaging means 105. Operating means 106. Image processing means 107. Storage means 108. Display means 109. Communication means 110. Image output means 111. Switching means 112. Address setting means

Claims (5)

An image sensor comprising: an imaging unit that captures a monitoring area; and an image processing unit that determines presence or absence of a detection target moving body based on a change area of an image captured by the imaging unit.
The imaging means is focused on an area away from the imaging means where the detection target moving body may exist before and after the occurrence of the change area,
The image processing means moves the non-detection target moving body when the difference in luminance between the change area and the background decreases away from the center of gravity of the change area and approaches the boundary between the change area and the background. An image sensor characterized by being determined . Obtains the luminance distribution in the change area, when the luminance distribution can be approximated to a predetermined function, and judging the change region and the non-detection target mobile, image sensor according to claim 1. The image sensor according to claim 2 , wherein the predetermined function is a quadratic function, a trigonometric function, or a Gaussian function. The image sensor according to claim 2 , wherein the luminance distribution is obtained in each of a vertical direction, a horizontal direction, and an oblique direction passing through the center of gravity of the image. A difference binarized image between the current image and the reference image is obtained, a barycentric point of the obtained difference binarized image is obtained, and a difference binary is obtained in a horizontal line segment, a vertical line segment, and a diagonal line segment passing through the barycentric point. image sensor according to claim 2 or 3, characterized in that to obtain a luminance distribution of the current image corresponding to the region.

Images