Image processing device, image processing method, image processing program and image monitoring system

Abstract

To provide an image processing device, an image processing method, an image processing program, and an image monitoring system that enable stable video analysis of a water surface or a snowfall surface from a camera image without being affected by moving objects such as rain and snow.SOLUTION: An image processing device 3 of an image monitoring system that monitors a water level or a snow accumulation amount based on a captured image includes an image acquisition unit 4 that acquires an image, a noise determination unit 5 that determines whether or not moving objects such as trails of rainfall or snowfall are captured in the image, and an image processing unit 6 that, if the moving objects such as the trails of rainfall or snowfall are captured in the image, processes the image into an image suitable for improving inference accuracy by boundary inference.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an image processing device, an image processing method, an image processing program, and an image monitoring system. [Background technology]

[0002] There is an image surveillance system that monitors remote locations by installing a camera at a location to be monitored and acquiring the images captured by the camera via a network. In this image surveillance system, a noise removal processing technology is proposed to make the camera images that contain external noise such as rain and snow appear clearer to the user, thereby providing high-quality surveillance images.

[0003] For example, Patent Document 1 discloses a technology that identifies the environment from a camera image and appropriately determines whether or not to perform fog removal, brightness correction, and rain removal processes, thereby improving the visibility of the image. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2023-116151 Summary of the Invention [Problem to be solved by the invention]

[0005] When an image surveillance system using video analysis monitors the surface of a water body such as a reservoir at night, an infrared camera equipped with an infrared projector function may be used to compensate for the lack of light at night. Even in this case, it is necessary to slow down the shutter speed to obtain a reasonably clear image.

[0006] In this case, in a rainy or snowy environment, the rain and snow fall while reflecting infrared light, causing noise in the trajectory of the snow. As a result, when performing video analysis using the image to determine the water level, there is a problem in that the water surface cannot be accurately determined, which cannot be addressed by the technology disclosed in Patent Document 1.

[0007] An object of the present invention is to provide an image processing device, an image processing method, and an image monitoring system that enable stable image analysis such as water surface discrimination from camera images without being affected by moving objects such as rain and snow. [Means for solving the problem]

[0008] The above problem is solved as follows.

[0009] (1) An image processing device for an image monitoring system that monitors water levels or snow accumulation based on captured images, comprising: an image acquisition unit that acquires the images; a noise determination unit that determines whether a moving object is captured in the images; and an image processing unit that, if a moving object is captured in the images, processes the images into images suitable for improving the accuracy of boundary inference.

[0010] (2) The image processing device according to (1), further comprising a boundary inference unit that calculates a boundary position from the image.

[0011] (3) The image processing device according to (1), further comprising a transmitting unit for inputting the image to an external boundary inference unit that calculates a boundary position from the image.

[0012] (4) In the image processing device described in (1), the image acquisition unit sets a range necessary to improve inference accuracy by boundary inference from the acquired image.

[0013] (5) In the image processing device described in (1), the image processing unit detects edges of the image, extracts horizontal lines, and calculates boundary positions from the horizontal lines.

[0014] (6) In the image processing device described in (1), the image processing unit calculates an image that is optimal for calculating the boundary position from statistical information on images that are consecutive in time series over a predetermined period of time.

[0015] (7) An image monitoring system that monitors water levels or snow accumulation based on captured images, comprising: an image acquisition unit that acquires the images; a noise determination unit that determines whether a moving object is captured in the images; an image processing unit that processes the images into images suitable for improving the accuracy of boundary inference if a moving object is captured in the images; a shooting unit that captures video data with the shutter speed automatically adjusted according to the amount of ambient light; a boundary inference unit that calculates the boundary position from the images; and a notification unit that notifies the user if the calculation result by the boundary inference unit is outside a specified range.

[0016] (8) In the image monitoring system described in (7), if the boundary inference unit is not within the device constituting the image monitoring system, the image monitoring system is provided with a communication unit that transmits the image processed by the image processing unit to an external boundary inference unit.

[0017] (9) In the image monitoring system described in (7), the boundary inference unit calculates the height of the water surface or snow surface from a boundary position based on a specific point in the image.

[0018] (10) In the image monitoring system described in (7), the image monitoring system is connected to a network, and includes an alarm issuing unit that issues an alarm via the network when a measurement value falls outside a predetermined range, and a notification count limiting unit that controls the notification unit to notify only once within a predetermined period.

[0019] (11) An image processing method for an image monitoring system that monitors water levels or snow accumulation based on captured images, comprising the steps of acquiring the image, determining whether a moving object is captured in the image, and, if a moving object is captured in the image, processing the image into an image suitable for improving the accuracy of boundary inference.

[0020] (12) An image processing program for causing a computer to execute the steps of acquiring a captured image, determining whether a moving object is captured in the image, and, if a moving object is captured in the image, processing the image into an image suitable for improving the accuracy of boundary inference. [Effects of the Invention]

[0021] According to the present invention, it is possible to provide an image processing device and an image monitoring system that are capable of stably performing image analysis such as determining the water surface from camera images without being affected by moving objects such as rain and snow. [Brief explanation of the drawings]

[0022] [Figure 1] 1 is a diagram showing the configuration of an image monitoring system 1 for monitoring the water level of a reservoir according to an embodiment. [Figure 2] FIG. 10 is a flow diagram illustrating the operation of the image monitoring system to detect the water level of a reservoir as a monitoring target. [Figure 3] FIG. 10 is a diagram showing an example of a captured image of a reservoir. [Figure 4] FIG. 2 is a diagram showing an example of a measuring plate. [Figure 5] FIG. 10 is a diagram showing an estimated water surface (dashed line) obtained by an image processing unit and an area (hatched area) where pixel values ​​have been changed. [Figure 6] FIG. 10 is a flow diagram illustrating another operation of the image monitoring system for detecting the water level of a reservoir as a monitoring target. [Figure 7] 10A and 10B are diagrams illustrating other processing contents of the image processing unit. [Figure 8] FIG. 10 is a diagram illustrating a water level detection method. [Figure 9] 10A and 10B are diagrams illustrating another water level detection method. [Figure 10] 10A and 10B are diagrams illustrating another water level detection method. DETAILED DESCRIPTION OF THE INVENTION

[0023] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing the configuration of an image monitoring system 1 for monitoring the water level of a reservoir according to an embodiment.

[0024] The image monitoring system 1 comprises a photographing device 2 that photographs the reservoir to be monitored, an image processing device 3 that removes image noise caused by moving objects such as the trajectory of rainfall or snowfall in the photographed image of the reservoir, and corrects the image to make it easier to detect the water level of the reservoir, and a water level detection server 9 that provides a water level detection service that detects the water level of the reservoir based on the photographed image of the reservoir.

[0025] The water level detection server 9 of the image monitoring system 1 includes a boundary inference unit 90 that infers the boundary position from an image. That is, the boundary inference unit 90 determines the water level of a reservoir from a captured image of the reservoir.

[0026] The photographing device 2, image processing device 3, and water level detection server 9 are connected via a network. Images of the reservoir photographed by the photographing device 2 are sent to the image processing device 3 via the network, and the photographed images of the reservoir from which noise has been removed by the image processing device 3 are sent to the water level detection server 9 via the network.

[0027] The water level detection server 9 is a server computer equipped with a boundary inference unit 90, and is realized as, for example, a cloud server. Note that the image processing device 3 may be equipped with the boundary inference unit 90 and have a water level detection function.

[0028] The image capturing device 2 is a so-called network surveillance camera. It is equipped with a visible light camera, an infrared camera, and an infrared light illuminator. When the illuminance is high during the day, the image capturing device 2 outputs a color image using the visible light camera, and at night or when the illuminance falls below a certain level, the infrared camera captures an image using the infrared light illuminator and outputs a monochrome image. The switching between color and monochrome image output and the camera exposure control (shutter speed and aperture) are automatically adjusted by the image capturing device 2 according to the ambient illuminance (amount of ambient light) or are performed according to commands from the image processing device 3.

[0029] The image processing device 3 includes an image acquisition unit 4, a noise determination unit 5, an image processing unit 6, a communication unit 7, an alarm issuing unit 91, and a notification count limiting unit 92.

[0030] The image acquisition unit 4 acquires photographed images of the reservoir to be monitored from the photographing device 2 via a network such as a LAN (Local Area Network) or a connection means conforming to standards such as USB (Universal Serial Bus). The image acquisition unit 4 acquires photographed images at regular intervals (for example, every 10 minutes). Furthermore, as will be described in detail later, the image acquisition unit 4 sets the processing range to the range of the photographed image of the reservoir that is necessary to improve the inference accuracy by boundary inference.

[0031] The noise determination unit 5 determines whether or not the captured image of the monitored object contains anything that could reduce the detection accuracy of the water level detection service. For example, the noise determination unit 5 determines whether or not moving objects, such as raindrops during rainfall or snowflakes during snowfall, are present in the captured image of a reservoir. For example, the noise determination unit 5 generates an image with reduced noise by averaging multiple frame images extracted from a video. The noise determination unit 5 then evaluates the difference between the noise-reduced image and the frame images to determine whether or not moving objects, such as raindrops during rainfall or snowflakes during snowfall, are present.

[0032] The image processing unit 6 removes noise from the image due to moving objects such as raindrops during rainfall and snow during snowfall, and performs image enhancement and correction processing to improve the accuracy of water surface detection in the water level detection service. The image processing unit 6 also removes anything outside a predetermined pixel range of the water surface, making it easier to detect the water surface. In other words, the image processing unit 6 sets the processing range to the range necessary to improve the inference accuracy by boundary inference from the captured image of the reservoir. Details of the processing by the image processing unit 6 will be described later. The image processing unit 6 processes the image into an image suitable for improving the inference accuracy by boundary inference.

[0033] The image processing unit 6 detects edges in the image, extracts horizontal lines, and calculates the boundary positions from these horizontal lines. The boundary positions are water surfaces or snowy surfaces. This makes it easy to distinguish between the trajectory of a moving object and water surfaces or snowy surfaces. The image processing unit 6 then calculates an optimum image for calculating the boundary position from statistical information on consecutive images in a time series over a predetermined period, thereby eliminating moving objects that cause noise in the image.

[0034] The communication unit 7 transmits the captured image of the monitoring target processed by the image processing unit 6 to the water level detection server 9 and requests it to detect the water level. Then, the communication unit 7 acquires the water level value of the monitoring target reservoir from the water level detection server 9. The alarm issuing unit 91 issues an alarm email or an alarm message to the user when the water level of the reservoir, determined by the water level detection server 9 from a photographed image of the reservoir, exceeds a predetermined range. The notification count limiting unit 92 controls the alarm issuing unit 91 so that the notification is only issued once within a predetermined period. This prevents the same message from being issued multiple times, resulting in so-called SPAM emails. The alarm issuing unit 91 may also issue an alarm message via, for example, a social networking service (SNS).

[0035] Specifically, the image processing device 3 is an information processing device (computer) consisting of a CPU (Central Processing Unit), a storage device, a communication device, and an input / output device. The CPU executes a program stored in the storage device to realize the functions of a noise determination unit 5 and an image processing unit 6. The CPU also controls the communication device based on the program to realize the functions of an image acquisition unit 4 and a communication unit 7.

[0036] The operation of the image monitoring system 1 to detect the water level of a reservoir as a monitoring target at a predetermined interval will be described below with reference to the flow diagram in Fig. 2. The flow diagram in Fig. 2 shows the operation of detecting the water level by taking a single image.

[0037] In step S1, the photographing device 2 photographs the reservoir and acquires a photographed image.

[0038] FIG. 3 is a diagram showing an example of a photographed image of a reservoir. The image monitoring system 1 detects the water surface boundary 32 of the reservoir 31 from the captured image, recognizes the scale of the position of the water surface boundary 32 on the water measuring plate 33 (see FIG. 4), and detects the water level of the reservoir.

[0039] The image of the reservoir in Figure 3 shows multiple snow reflection trajectories 34. Moving objects such as these snow reflection trajectories 34 become disturbance noise when detecting the position of the water surface boundary 32. These snow reflection trajectories 34 are generated by slowing down the shutter speed of the image capture device 2, and are particularly common in images captured in dark environments such as at night.

[0040] Returning to FIG. 2, in step S2, the image acquisition unit 4 acquires photographed images of the reservoir from the photographing device 2 at regular time intervals (for example, every 10 minutes) via a network, USB, or other standard-compliant interface. The image acquisition unit 4 may also acquire moving images of the reservoir from the photographing device 2 instead of photographed images. The moving images are composed of a plurality of frames.

[0041] In step S3, the image acquisition unit 4 extracts an inference target range for detecting the water level from the acquired photographed image of the reservoir. For example, in the photographed image of the reservoir in Fig. 3, the user sets in advance a rectangular area including the measuring plate 33 and the water surface boundary 32 as a target range necessary for improving the inference accuracy by boundary inference.

[0042] In step S3, the image acquisition unit 4 performs range setting processing to extract a target range necessary for improving the inference accuracy by boundary inference, thereby reducing the noise judgment (to be described later) and the processing load of the image processing unit 6. In addition, since the amount of information similar to the water surface is reduced, the accuracy of water surface detection can be improved.

[0043] In step S4, the noise determination unit 5 determines whether the target range of the captured image of the monitored object contains disturbance noise equal to or greater than a threshold that may reduce the accuracy of water level detection. If disturbance noise equal to or greater than the threshold is contained (Yes in S4), the process proceeds to step S51, where the target range of the captured image of the monitored object is corrected so as not to reduce the accuracy of water surface detection. If disturbance noise equal to or greater than the threshold is not contained (No in S4), the process proceeds to step S6.

[0044] The noise determination unit 5 determines, for example, moving objects such as raindrops during rainfall and snow (snow reflection trajectories) during snowfall in a processed image of a reservoir as disturbance noise. More specifically, when it snows, the snow reflection trajectories appear white (high brightness), as shown in FIG. 3, and the brightness distribution differs from when there is no snowfall. When the change in brightness distribution from when there is no rainfall or snowfall in the processed image of the reservoir is equal to or greater than a threshold, the noise determination unit 5 determines that there is disturbance noise caused by moving objects such as rainfall or snowfall.

[0045] In step S51, the image processing unit 6 sets the pixel values ​​of pixels outside the pixel range of the water surface, which has been set in advance, to a predetermined fixed value for the target range of the captured image of the monitoring target. The image processing unit 6 fixes the pixel values ​​by setting them to 0, for example, and can eliminate unnecessary edge enhancement, which will be described later.

[0046] In step S52, the image processing unit 6 performs edge enhancement processing on the target area of ​​the captured image of the monitoring target, for which the pixel values ​​of the pixels outside the pixel area of ​​the water surface have been set to a predetermined value in step S51. The edge enhancement processing may be performed by applying a general edge enhancement filter such as a Gaussian filter, a Sobel filter, or a Laplacian filter.

[0047] In step S53, the image processing unit 6 performs line extraction processing of horizontal lines within the target range of the captured image of the monitoring target that has been edge-enhanced in step S52. The line extraction processing may be a general line extraction processing such as a line extraction method using a Hough transform.

[0048] In step S54, the image processing unit 6 finds, from among the horizontal lines extracted in step S53, the horizontal line that is most likely to represent the water surface based on its position and length within the target range of the captured image, and sets this as the estimated water surface.The image processing unit 6 then changes the pixel values ​​of the area of ​​the captured image of the monitoring target that corresponds to the area below the estimated water surface within the target range of the captured image to a preset pixel value of the water surface (a monochromatic pixel value).The trajectory of rain or snow appears as a vertical or diagonal straight line at the edge of the image, so even if a moving object is captured in the captured image, it can be distinguished from the horizontal line indicating the water surface.

[0049] The image processing unit 6 may also paint pixel values ​​of the area of ​​the captured image of the monitoring target corresponding to the area below the estimated water surface with a water surface texture in fine weather. This makes it possible to create an image that is more suitable for improving inference accuracy when the water level detection server 9 estimates the water surface using AI (artificial intelligence). As described above, in step S54, the image processing unit 6 changes the pixel values ​​of the area below the estimated water surface within the target range of the captured image of the monitoring target to preset pixel values.

[0050] Specifically, the image processing unit 6 determines the position coordinates of the measuring plate in advance within the target range of the captured image, and determines the horizontal line that crosses the range of the measuring plate as the estimated water surface. In addition, when multiple water surfaces are estimated, the slope of the water surface line in fine weather may be set in advance, and the estimated water surface with the closest slope may be selected.

[0051] Fig. 5 shows an estimated water surface 35 determined by the image processing unit 6 and a region 36 in which pixel values ​​have been changed in the photographed image of the reservoir described in Fig. 3. Here, the image processing unit 6 defines the entire region below the estimated water surface 35 as the pixel value change region 36. The image processing unit 6 corrects pixel values ​​in an area 36 below the estimated water surface 35 in the captured image of the reservoir so that the snow reflection trajectory 34 is not captured, thereby preventing a decrease in the accuracy of water surface detection using the snow reflection trajectory 34. In other words, the image processing unit 6 processes the captured image into one suitable for improving the accuracy of boundary inference.

[0052] Returning to Fig. 2, in step S6, the communication unit 7 notifies the water level detection server 9 of the captured image of the monitored object, which has been corrected in step S54 by changing the pixel values ​​of the area 36 below the estimated water surface 35, and requests that the water level detection server 9 detect the water level. The water level detection server 9 performs the water level detection process, and the operation of the image monitoring system 1 ends.

[0053] Among moving objects, disturbance noise from raindrops during rainfall and snow (snow reflection trajectories) during snowfall occurs in the vertical direction of the captured image of the monitored object due to the influence of gravity. In the above, the image processing unit 6 focuses on this, extracts a horizontal line as an estimated water surface, and performs correction processing on the captured image of the monitored object. Next, other correction processing in the image processing unit 6 will be described.

[0054] FIG. 6 is a flow diagram illustrating another operation of the image monitoring system 1 for detecting the water level of a reservoir as a monitoring target. Steps S1 to S4 and step S6 in FIG. 6 are the same as those in FIG. 2, and therefore will not be described here.

[0055] The image monitoring system 1, which operates according to the flow chart in Fig. 6, focuses on the fact that disturbance noise caused by raindrops during rainfall or snow (snow reflection trajectories) during snowfall in a photographed image of a monitored object has high brightness and large pixel values. Based on this, the image processing unit 6 performs correction processing to remove the disturbance noise from the photographed image of the monitored object.

[0056] In step S55, the image processing unit 6 calculates the minimum value of each pixel in consecutive images in time series for each pixel of multiple captured images of the monitoring target taken over a certain period of time.The image processing unit 6 then generates an image with the calculated minimum value as its pixel value, and uses this as a corrected image of the captured images of the monitoring target.This makes it possible to remove moving objects from the image and obtain a noise-free image.Note that the method is not limited to the minimum value of each pixel in consecutive images in time series, and any statistical information such as the median or average value of each pixel may be used to calculate an optimal image for calculating the boundary position.

[0057] 7 is a diagram illustrating the processing content of step S55 by the image processing unit 6. Frames 71 to 73 are images of multiple monitored objects captured over a certain period of time. The image processing unit 6 calculates the minimum pixel value for pixels at the same position in frames 71 to 73, and sets this as the pixel value of the pixel at the same position in a corrected image 74 of the processed monitored image. This generates a corrected image 74 from which pixels with large pixel values ​​due to disturbance noise have been removed.

[0058] In the above, it has been explained that the image processing unit 6 generates a corrected image from multiple captured images of the monitored object taken over a certain period of time, but it is also possible to obtain multiple frames of captured images of the monitored object in step S55 and generate a corrected image in the same manner.

[0059] 6, in step S55, the corrected image of the photographed image of the monitoring target is subjected to edge enhancement processing so that the luminance difference of the water surface boundary appears. The edge enhancement at this time may be a general level correction processing.

[0060] Next, a water level detection method in the water level detection server 9 will be described with reference to Figures 8 to 10. The following method is different from the method of recognizing the water level numerical value from the scale on the water measuring plate in Figure 4.

[0061] 8 is a diagram illustrating a method for determining the water level from the positions of the upper and lower ends of the measuring plate 33 in the captured image of the monitored object and the position of the detected water surface boundary 32 of the reservoir 31. As shown in FIG. 8, the upper end of the measuring plate 33 indicates the maximum water level Hmax, and the lower end of the measuring plate 33 indicates the minimum water level Hmin. The user measures the positions of the upper and lower ends of the measuring plate 33 in the captured image of the monitored object in advance and sets them in the water level detection server 9.

[0062] The water level detection server 9 calculates the difference between the positions of the upper and lower ends of the measuring plate 33 in the captured image of the monitored object, and the difference between the position of the upper end of the measuring plate 33 and the position of the water surface boundary 32. The ratio of these differences is equal to the ratio of the difference between the actual water level at the upper end of the measuring plate 33 and the water level at the lower end, and the difference between the actual water level at the upper end of the measuring plate 33 and the water level at the water surface boundary 32. From this, the water level of the water surface boundary 32 is calculated.

[0063] 9 is a diagram illustrating a method for determining the water level from the actual water levels at the top and bottom of the photographed image of the monitoring target and the position of the water surface boundary 32 of the reservoir 31 detected in the photographed image of the monitoring target. As shown in Fig. 9, the user sets in the water level detection server 9 the water level at the top end of the photographed image of the monitoring target as the maximum value Hmax and the water level at the bottom end as the minimum value Hmin.

[0064] The water level detection server 9 detects the water surface boundary 32 and calculates the length from the top to the bottom of the captured image of the monitoring target, i.e., the height of the captured image of the monitoring target, and the difference between the position of the top of the captured image of the monitoring target and the position of the water surface boundary 32. The ratio of these differences is equal to the ratio of the difference between the maximum value Hmax and the minimum value Hmin to the difference between the actual water level at the top of the captured image of the monitoring target and the water level of the water surface boundary 32. In this way, the water level of the water surface boundary 32 is calculated.

[0065] 10 is a diagram explaining a method for determining the water level from the position of the background block pattern in the captured image of the monitored object, the position of the water surface boundary 32, the water level corresponding to the size of one block pattern, and the water level corresponding to the reference block pattern. The user measures the water level B corresponding to the size of the block pattern in advance and sets it in the water level detection server 9. The user also measures the water level Href of the reference block in advance by measuring the scale corresponding to the reference block pattern on the water measuring plate 33, and sets this information in the water level detection server 9.

[0066] The water level detection server 9 calculates the distance from a reference block pattern in a captured image of the monitored object to the water surface boundary 32, calculates how many block patterns the calculated distance corresponds to, and converts the calculated number of block patterns into a water level.The water level of the water surface boundary 32 is calculated by subtracting the converted water level from the water level corresponding to the reference block pattern.

[0067] 8 to 10, the water level detection server 9 calculates the water level of the water surface boundary 32 from the relative position of the water surface boundary 32 when using the upper and lower ends of the measuring plate 33, the frame of the photographed image of the monitoring target, or the background pattern as a reference, and the reference water level. This makes it possible to detect the water level of the water surface boundary 32 in ways other than by reading the scale on the measuring plate 33 pointed to by the water surface boundary 32, thereby improving the reliability of detection.

[0068] The above describes an image monitoring system 1 that monitors the water level of a reservoir, but since it can remove disturbance noise caused by moving objects such as snowfall, hail, insects, birds, and animals in the same way as rainfall, it can also be implemented in an image monitoring system that monitors the amount of snowfall. Furthermore, the image processing device of the embodiment is not limited to an image monitoring system for monitoring the water level of a reservoir or the amount of snowfall, but can also be applied to an image monitoring system in which moving objects such as raindrops or snow are treated as disturbance noise in processed images of a monitored object.

[0069] The present invention is not limited to the above-described examples, and includes various modifications. The above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those having all of the described configurations. The image processing device 3 does not need to include the image acquisition unit 3, the noise determination unit 4, and the image processing unit 5, and each functional unit may be provided in a separate device. [Explanation of symbols]

[0070] 1. Image surveillance system 2. Imaging equipment 3. Image processing device 4 Image acquisition unit 5 Noise detection section 6 Image processing section 7. Communication unit (transmitter) 9 Water Level Detection Server 90 Boundary reasoning part 91 Alarm Dispatch 92 Notification count limit section

Claims

1. An image processing device of an image monitoring system that monitors a water level or an amount of snow based on a captured image, an image acquisition unit that acquires the image; a noise determination unit that determines whether a moving object is captured in the image; an image processing unit that processes the image into an image suitable for improving inference accuracy by boundary inference when a moving object is captured in the image; An image processing device comprising:

2. further comprising a boundary inference unit that calculates a boundary position from the image; The image processing device according to claim 1 .

3. a transmitting unit for inputting the image to an external boundary inference unit that calculates a boundary position from the image; The image processing device according to claim 1 .

4. 2. The image processing device according to claim 1, the image acquisition unit sets a range necessary for improving inference accuracy by boundary inference from the acquired image. Image processing device.

5. 2. The image processing device according to claim 1, the image processing unit detects edges of the image, extracts horizontal lines, and calculates boundary positions from the horizontal lines; Image processing device.

6. 2. The image processing device according to claim 1, the image processing unit calculates an optimum image for calculating the boundary position from statistical information of images that are consecutive in time series over a predetermined period of time; Image processing device.

7. An image monitoring system for monitoring water levels or snow accumulation based on captured images, an image acquisition unit that acquires the image; a noise determination unit that determines whether a moving object is captured in the image; an image processing unit that processes the image into an image suitable for improving inference accuracy by boundary inference when a moving object is captured in the image; a photographing unit that photographs video data with a shutter speed automatically adjusted according to the amount of ambient light; a boundary inference unit that calculates a boundary position from the image; a notification unit that notifies a user when a calculation result by the boundary inference unit is outside a predetermined range; An image surveillance system equipped with

8. 8. The image monitoring system according to claim 7, If the boundary inference unit is not included in a device that constitutes the image monitoring system, a communication unit that transmits the image processed by the image processing unit to an external boundary inference unit; An image monitoring system comprising:

9. 8. The image monitoring system according to claim 7, the boundary inference unit calculates the height of the water surface or the snow surface from a boundary position based on a specific point in the image; Image surveillance system.

10. 8. The image monitoring system according to claim 7, The image monitoring system is connected to a network and includes an alarm issuing unit that issues an alarm via the network when a measurement value falls outside a predetermined range, and a notification count limiting unit that controls the notification unit to issue a notification only once within a predetermined period. Video surveillance system

11. An image processing method for an image monitoring system that monitors water levels or snow accumulations based on captured images, comprising: acquiring the image; determining whether a moving object is captured in the image; If a moving object is captured in the image, processing the image into an image suitable for improving inference accuracy by boundary inference; An image processing method comprising:

12. On the computer, Steps to retrieve the captured images, determining whether a moving object is captured in the image; a step of processing the image into an image suitable for improving inference accuracy by boundary inference when a moving object is captured in the image; An image processing program for executing the above.

Images