WO2020188692A1

WO2020188692A1 - Water level measuring device, water level measuring method and water level measuring program

Info

Publication number: WO2020188692A1
Application number: PCT/JP2019/011134
Authority: WO
Inventors: 祐貴大西; 茂樹桑原
Original assignee: 三菱電機株式会社
Priority date: 2019-03-18
Filing date: 2019-03-18
Publication date: 2020-09-24

Abstract

In conventional water level measuring devices, there were instances where high precision water level measurement could not be performed when the surrounding environment changed due to season, time, weather, etc.　A water level measuring device (100) of the present invention is provided with an image identification unit (12) and a water level calculation unit (13). The image identification unit (12) identifies whether a region corresponding to an image for identification, which is an image of a region including a predetermined reference point for an acquired photographed image, is a water region or a non-water region, said identification being made on the basis of a plurality of water region images, which correspond to the surrounding environment affecting image identification, and a plurality of non-water region images corresponding to the surrounding environment. The water level calculation unit (13) calculates the water level in the photographed image on the basis of the identification results by the image identification unit (12) and a reference point water level value corresponding to the reference point.

Description

Water level measuring device, water level measuring method and water level measuring program

The present invention relates to a water level measuring device for measuring a water level from an image, a water level measuring method, and a water level measuring program.

Conventionally, a technology for measuring the water level of a river or the like has been developed using an image or a video image taken by a surveillance camera installed in the river or the like.

For example, the water level measuring device of Patent Document 1 measures the water level of a river by using an image including something other than the water of the river (for example, a specific structure) taken by a surveillance camera or the like. Is. Specifically, the water level measuring device of Patent Document 1 distinguishes between a water region and a non-water region by machine learning from an image of a river taken by a surveillance camera. Determine the boundary coordinates. The water level is measured using the boundary coordinates and the base point water level value of the coordinates.

WO2018-092238

However, with the above-mentioned conventional water level measuring device, high-precision water level measurement may not be possible when the surrounding environment changes due to seasons, time, weather, etc.

The present invention has been made to solve the above-mentioned problems, and provides a water level measuring device, a water level measuring method, and a water level measuring program that stably measure a water level even in a changing surrounding environment. The purpose is.

The water level measuring device according to the present invention is determined in advance with respect to acquired captured images based on images of a plurality of water areas according to the surrounding environment affecting image identification and images of a plurality of non-water areas according to the surrounding environment. An image identification unit that identifies whether the region corresponding to the identification image, which is an image of the region including the base point, is a water region or a non-water region, the identification result by the image identification unit, and the base point water level corresponding to the base point. It is equipped with a water level calculation unit that calculates the water level in the captured image based on the value.

According to the present invention, stable water level measurement can be realized in consideration of changes in the surrounding environment.

It is a functional block diagram which shows the main part of the water level measuring apparatus which concerns on Embodiment 1 of this invention. It is a hardware block diagram of the water level measuring apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows the operation of the base point setting part of the water level measuring apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows the operation of the image identification part and the water level calculation part of the water level measuring apparatus which concerns on Embodiment 1 of this invention. This is an example of an image taken by a surveillance camera. It is a flowchart which shows the operation of the learning image generation part of the water level measuring apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows the operation of the image learning part of the water level measuring apparatus which concerns on Embodiment 1 of this invention.

Embodiment 1.
In the first embodiment, as an example, a case where the target of water level measurement is a river and the river is photographed by a surveillance camera will be described below.

FIG. 1 is a functional block diagram showing a main part of the water level measuring device 100 according to the first embodiment of the present invention.

In FIG. 1, the water level measuring device 100 is connected to the surveillance camera 1, the operation input device 2, and the display device 5. The connection method may be wired or wireless.

The surveillance camera 1 captures a river to be measured by the water level measuring device 100. The range to be photographed by the surveillance camera 1 (hereinafter referred to as “photographing range”) includes river water and substances other than river water (for example, buildings and land around the river). Hereinafter, the image captured by the surveillance camera 1 is referred to as a “captured image”. The surveillance camera 1 outputs image data indicating a captured image to the water level measuring device 100.

The operation input device 2 receives an input of an operation by a worker who uses the water level measuring device 100 (hereinafter, simply referred to as a "worker"). The operation input device 2 is composed of, for example, a keyboard 3 and a mouse 4. Further, the display device 5 is composed of, for example, a display 6 such as a liquid crystal display or an organic EL (Electro Luminescence) display.

Hereinafter, the main parts of the water level measuring device 100 that measures the water level by performing learning in consideration of changes in the surrounding environment will be described.

The water level measuring device 100 includes a base point setting unit 11 that sets a base point that is a reference point for water level measurement, a region in which river water exists in an image (hereinafter referred to as "water region"), and other than river water. The boundary line between the water region and the non-water region of the image is identified from the learning results of the image learning unit 15 that performs machine learning related to the identification from the region (hereinafter referred to as “non-water region”). It includes an image identification unit 12, a water level calculation unit 13 for calculating the water level, and a learning image generation unit 14 for generating a learning image in consideration of changes in the surrounding environment.

The base point setting unit 11 acquires the image data output by the surveillance camera 1. The base point setting unit 11 responds to the operation input to the operation input device 2, and indicates the coordinate value of the base point, which is an arbitrary point that serves as a reference for measuring the water level in the captured image, and the value indicating the water level corresponding to the base point (hereinafter,). "Base point water level value") and is set. More specifically, the base point setting unit 11 sets the coordinate values of one or more base points in the captured image and the base point water level values corresponding to each of the one or more base points. The base point setting unit 11 outputs the coordinate value of each base point and the base point water level value corresponding to each base point to the water level calculation unit 13.

The base water level value is not limited to the value of the height measured from the sea level, that is, the value above sea level, but is a value representing a relative height with respect to an arbitrary place in a certain range of places. Is also good.

The image identification unit 12 acquires the image data output by the surveillance camera 1. The image identification unit 12 sets a specific area which is a whole area or a part of the whole area of the photographed image among the photographed images indicated by the acquired image data as a target area (hereinafter referred to as “target area”). The image identification unit 12 identifies the boundary line between the water region and the non-water region of the image to be identified (hereinafter referred to as “identification image”) corresponding to the target region.

When the target area is a specific area, the image identification unit 12 cuts out an area including the coordinate values of each base point set by the base point setting unit 11.

The water level calculation unit 13 calculates the water level value in the identification image of the target area of the surveillance camera 1 by using the identification result by the image identification unit 12 and the base point water level value at each coordinate output by the base point setting unit 11. The water level calculation unit 13 outputs the calculation result and the identification result of the image identification unit 12 to the learning image generation unit 14.

The learning image generation unit 14 acquires the image data output by the surveillance camera 1. The learning image generation unit 14 is a water region of at least one of the identification result and the identification image corresponding to the calculation result output from the water level calculation unit 13, and the image of the past time and the image of the future time. Using the identification result and the calculation result of the boundary line between the non-water area and the non-water area, from the captured image shown by the image data acquired from the surveillance camera 1, a plurality of images of the water area according to the surrounding environment affecting the image identification. Alternatively, images of a plurality of non-water areas according to the surrounding environment are determined. The learning image generation unit 14 uses image data of the image determined to be an image of a plurality of water regions according to the surrounding environment or an image of a plurality of non-water regions according to the surrounding environment as image data for learning. Output to the learning unit 15. Specifically, the image data of the image of a plurality of water areas according to the surrounding environment is the image data in which there is a high possibility that the water of the river is captured in the target area of the photographed image, and is according to the surrounding environment. The image of the plurality of non-water regions is image data of an region in the target region of the captured image in which there is a high possibility that something other than river water is captured. Hereinafter, an image determined to be an image of a plurality of water regions according to the surrounding environment or an image of a plurality of non-water regions according to the surrounding environment is referred to as a "learning image". The learning image generation unit 14 outputs the learning image to the image learning unit 15. The learning image generation unit 14 is an example of a reference image generation unit.

The image learning unit 15 executes machine learning related to the discrimination between the water region and the non-water region by using the learning image.

The combination of the image identification unit 12 and the image learning unit 15 uses, for example, a "neural network". A neural network is a computer that inputs a plurality of image data together with correct answers in advance and trains them to determine whether or not what is copied in the newly input image data is a specific target. It is one of the mechanisms to operate so as to output the result. For example, the model includes a convolutional neural network and a model using a support vector machine (Support Vector Machine, SVM), and machine learning related to discrimination between a water region and a non-water region can be executed for a training image. Any model may be used. Since the structure of the neural network and the machine learning by the neural network are conventional techniques and various known methods can be used, the description thereof will be omitted.

Next, the hardware configuration of the water level measuring device 100 according to the first embodiment will be described.

FIG. 2 is a hardware configuration diagram of the water level measuring device 100 according to the first embodiment of the present invention. The configuration of the water level measuring device 100 according to the first embodiment of the present invention will be described with reference to FIG.

As shown in FIG. 2, the water level measuring device 100 is composed of a computer, and is composed of a processor 21, a memory 22, and a bus 23.

The bus 23 is a signal path that electrically connects each device and exchanges data.

The processor 21 is composed of, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a microprocessor, a microcontroller, a DSP (Digital Signal Processor), and the like. Alternatively, the processor 21 is a combination of the above.

The memory 22 includes, for example, a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), an EEPROM (Electrically Memory Disk), an EEPROM (Electrically Memory), and an EEPROM (Electrically Memory). It is composed of a non-volatile or volatile semiconductor memory such as a disk, a DVD (Digital Versaille Disc) or an HDD (Hard Disk Drive), a magnetic disk, an optical disk, a magneto-optical disk, or the like. Alternatively, the memory 22 is a combination of the above-mentioned ones.

The memory 22 stores a program for causing the computer to function as the base point setting unit 11, the image identification unit 12, the water level calculation unit 13, the learning image generation unit 14, and the image learning unit 15 shown in FIG. Further, the memory 22 stores the image data generated by the learning image generation unit 14. When the processor 21 reads out and executes the program stored in the memory 22, the base point setting unit 11, the image identification unit 12, the water level calculation unit 13, the learning image generation unit 14, and the image learning unit 15 shown in FIG. 1 The function of is realized.

The processor 21 or the memory 22 realizes the functions of the base point setting unit 11, the image identification unit 12, the water level calculation unit 13, the learning image generation unit 14, and the image learning unit 15 by different processors or different memories. You may.

Further, each function of the water level measuring device 100 can be realized by hardware, software, firmware, or a combination thereof.

Note that the devices of the water level measuring device 100 and the devices are connected via communication. The communication to be used may be wired communication or wireless communication.

Next, the operation of the water level measuring device 100 will be described.

FIG. 3 is a flowchart showing the operation of the base point setting unit 11 of the water level measuring device 100 according to the first embodiment of the present invention. The operation of the base point setting unit 11 will be described below with reference to FIG.

In step ST1, the base point setting unit 11 acquires the image data output by the surveillance camera 1. The base point setting unit 11 sets the coordinate values of one or more base points in the captured image and the base point water level values corresponding to each base point according to the operation input to the operation input device 2. The base point setting unit 11 outputs the coordinate value of each base point and the base point water level value corresponding to each base point to the water level calculation unit 13.

Specifically, for example, the base point setting unit 11 acquires the image data output by the surveillance camera 1 and displays the captured image indicated by the image data on the display 6. The base point setting unit 11 sets a base point for one or more or each coordinate designated by the operator using the mouse 4 with respect to the captured image displayed on the display 6. The base point setting unit 11 calculates the coordinate values of each designated base point. The base point setting unit 11 sets the base point water level value corresponding to each base point input by the operator using the keyboard 3 for each designated base point. Alternatively, the memory 22 may store the base point water level value corresponding to each coordinate, and the base point setting unit 11 may set the base point water level value by acquiring the base point water level value from the memory 22.

In addition, in order to set the base point water level value corresponding to each base point, it is necessary to measure the size and uneven shape of the buildings and land around the river in advance. Since various known methods can be used for the survey, the description thereof will be omitted.

Step ST1 is executed only once, for example, when the surveillance camera 1 is installed and the water level measuring device 100 is started to be used. Although it is assumed that step ST1 is executed only once, it may be executed repeatedly. For example, when an operation for instructing the change of the coordinates of the base point or the change of the base point water level value of the base point is performed, step ST1 may be performed and the base point setting unit 11 may update the coordinates of the base point or the base point water level value of the base point. ..

After the step ST1 is completed, the surveillance camera 1 continuously executes the process of shooting the shooting range and outputting the image data indicating the shot image. The process of outputting image data indicating the captured image of the surveillance camera 1 may not be continuously executed or may be fragmentary. Further, the timing of the process of outputting the image data is not particularly limited. The water level measuring device 100 automatically executes the process shown in the flowchart of FIG. 4 and the process shown in the flowchart of FIG. 6 for each of the image data sequentially output by the surveillance camera 1.

FIG. 4 is a flowchart showing the operation of the image identification unit 12 and the water level calculation unit 13 of the water level measuring device 100 according to the first embodiment of the present invention. The operation of the image identification unit 12 and the water level calculation unit 13 of the water level measuring device 100 will be described below with reference to FIG.

In step ST11, the image identification unit 12 acquires the image data output by the surveillance camera 1. The image data indicates, for example, one or a plurality of captured images captured by the surveillance camera 1 at the latest timing with respect to step ST11. The image data may be past data or the like, and there is no time limit.

In step ST12, the image identification unit 12 cuts out an identification image from the captured image indicated by the image data acquired in step ST11. At this time, the area including the coordinate values of the base points set in step ST1 of FIG. 3 becomes the target area.

FIG. 5 is an example of the image 30 captured by the surveillance camera 1.

FIG. 5 is a photographed image 30 in which the sky 31, the mountain 32, the sandy area 33, the river 34, and the grassland 35 are photographed by the surveillance camera 1. As shown in the identification images D1 to D3, the identification image to be identified in step ST12 may be the identification image D1 in which the target area is the entire area of the captured image, or the specific area in which the target area is a part of the entire area. The identification image D2 and the identification image D3 may be used.

Returning to step ST12 of FIG. 4, the image identification unit 12 divides the target area into a predetermined area below the target area in the identification image, and is the center of the predetermined image which is the image of the predetermined area the water region? , Identify whether it is a non-water area. At this time, the image identification unit 12 executes the identification process of whether it is a water region or a non-water region based on the result of machine learning by the image learning unit 15. Machine learning by the image learning unit 15 will be described later. The predetermined area is an example of the area corresponding to the identification image, and may be the same area as the target area or a smaller area than the target area.

In step ST13, the image identification unit 12 identifies the boundary line between the water region and the non-water region of the identification image by repeatedly executing the process of determining the boundary between the water region and the non-water region in step ST12. To do. The image identification unit 12 shifts the center of the predetermined image by one pixel or a plurality of pixels in a certain direction, and repeatedly identifies whether the center of the predetermined image is a water region or a non-water region until it corresponds to the identification image. To do. The image identification unit 12 shifts the center of the predetermined image by one pixel or a plurality of pixels in a certain direction, but the center may be changed at random. Further, the image identification unit 12 divides the target area into one or more areas of various sizes, and until the divided image corresponds to the identification image, whether the center of the divided image is the water area or the non-water area. You may identify whether it is.

The image identification unit 12 determines the boundary between the coordinates identified as the water region and the coordinates identified as the non-water region, for example, the identification of the water region and the identification of the non-water region in two adjacent pixels or the adjacent specific region. Is determined by exchanging. The image identification unit 12 determines the boundary between the water region and the non-water region by switching the identification of the water region and the identification of the non-water region until it corresponds to the identification image, and the water region and the non-water region of the identification image are determined. Identify the border with the area. If the boundary line between the water region and the non-water region of the identification image cannot be identified, step ST13: NO is set, and the process returns to step ST12. When the image identification unit 12 can identify the boundary line between the water region and the non-water region of the identification image, step ST13: YES is set, the identification result is output to the water level calculation unit 13, and the process proceeds to the next step. The identification result output by the image identification unit 12 in step ST13 is stored in the memory 22 until a predetermined condition is satisfied (for example, an operation for instructing deletion is performed after a certain period of time has passed). To.

In step ST14, the water level calculation unit 13 uses the identification result output by the image identification unit 12 in step ST13 and the base point water level value corresponding to each coordinate output by the base point setting unit 11 in step ST1 to monitor the camera. Calculate the water level value in the shooting range of 1. Further, the water level calculation result output by the water level calculation unit 13 in step ST14 is stored in the memory 22 and stored until a predetermined condition is satisfied (for example, an operation for instructing deletion is performed after a certain period of time has passed). Will be done.

Hereinafter, the processing of steps ST11 to ST14 is referred to as "water level measurement processing". After the step ST1 is completed, the water level measuring device 100 satisfies a predetermined condition (for example, when an operation for instructing the end of the water level measuring process is performed, when the power of the water level measuring device 100 is turned off, or monitoring. The water level measurement process is repeatedly executed until the communication connection between the camera 1 and the water level measuring device 100 is disconnected). Although it is assumed that the water level measurement process is repeatedly executed, it may be executed only once without repeating the process.

FIG. 6 is a flowchart showing the operation of the learning image generation unit 14 of the water level measuring device 100 according to the first embodiment of the present invention. The operation of the learning image generation unit 14 of the water level measuring device 100 will be described below with reference to FIG.

In step ST21, the learning image generation unit 14 has the identification result output by the image identification unit 12 in step ST13 and the water level calculation result output by the water level calculation unit 13 in step ST14 stored in the memory 22 in the water level measurement process. To get.

In step ST22, the learning image generation unit 14 determines the water region and the non-water region from the identification result output by the image identification unit 12 in step ST13 and the water level calculation result output by the water level calculation unit 13 in step ST14. Obtain the coordinates of the boundary line and the water level value corresponding to the coordinates. The learning image generation unit 14 stores the coordinates of the acquired boundary line between the water region and the non-water region and the water level value of the coordinates in the memory 22.

In step ST23, the learning image generation unit 14 selects an area in which the learning image should be generated. Normally, in a device that improves the identification accuracy by learning, the identification accuracy does not change so much no matter how much the same learning image is learned. On the other hand, for example, the color of the leaves changes from green to brown as the seasons change in the grassland, the color of water changes from blue to dark blue as time passes from day to night, and water as the weather changes from sunny to cloudy. When there are images with various patterns due to changes in the color of the surrounding environment even in the same area, such as when the color of is changed from blue to dark blue, learning the images with various patterns improves the identification accuracy by learning. The identification accuracy of the device to be used is dramatically improved.

In the water level measuring device 100, in order to improve the accuracy of discrimination between the water region and the non-water region by the image identification unit 12 in the region of the boundary line between the water region and the non-water region, the image learning unit 15 learns the image. Do. Therefore, the learning image generation unit 14 needs to generate learning images of various patterns due to changes in the surrounding environment in the region of the boundary line between the water region and the non-water region. The learning image generation unit 14 detects the region or coordinates where the water region and the non-water region are interchanged in the region of the boundary line between the water region and the non-water region, and detects the exchange between the water region and the non-water region. By using the identification image for the time during which the water region and the non-water region can be clearly distinguished in the region or coordinates as the learning image, various patterns of learning images due to changes in the surrounding environment are generated.

The learning image generation unit 14 calculates the change in the temporal or spatial boundary line by using one or more of the boundary line coordinates and the water level value acquired in step ST22. Specifically, in the coordinates for calculating the change of the boundary line, the change of the boundary line is calculated from a plurality of identification results and water level calculation results at different times. Here, the change of the boundary line means, for example, the temporal fluctuation of the boundary line, the temporal fluctuation of the water level value in the target area, the color fluctuation, etc., of the water level value of the river and the surrounding environment. It's a change. The water level of a river usually changes continuously over time. In addition, the change in the boundary between the water region and the non-water region occurs in a part or a plurality of regions in the image, and it does not occur all at once in the entire region such as the river water suddenly drying up. There is.

In step ST24, the learning image generation unit 14 detects a region or coordinates in which the water region and the non-water region are interchanged in the region of the boundary line between the water region and the non-water region. The area or coordinates in which the water area and the non-water area detected by the learning image generation unit 14 are interchanged are areas or coordinates in which images having various patterns are present due to changes in the surrounding environment. The learning image generation unit 14 determines the region or coordinates in which the change in the boundary line calculated in step ST23 satisfies the first specific condition as the learning image extraction region or the learning image extraction coordinates.

The first specific condition is a condition when images of various patterns exist due to changes in the surrounding environment in a certain area or a certain coordinate. The first specific condition is, for example, that the identification result of step ST13 is always constant in a specific time zone for several hours to several days, but the identification result of step ST13 is periodically or irregularly in another time zone. Or when the coordinates of the boundary line between the water region and the non-water region change continuously within a continuous fixed time when the identification of step ST13 is performed continuously for several hours to several days. is there. In addition, although several hours and several days are given as an example, there is no time limit such as several years.

In step ST25, the learning image generation unit 14 uses an image at a time during which the water region and the non-water region can be clearly distinguished in the region or coordinates where the replacement of the water region and the non-water region is detected as the learning image. .. Specifically, the learning image generation unit 14 generates an image as a learning image at a time in which the water region and the non-water region can be clearly distinguished by the learning image extraction region or the learning image extraction coordinates. .. When the image corresponding to the learning image extraction area or the learning image extraction coordinates determined in step ST24 satisfies the second specific condition, the learning image generation unit 14 immediately before the time when the image was taken. From the coordinates of the boundary line between the water region and the non-water region for a certain period of time before, immediately after, after a certain period of time, or for a plurality of periods, it is identified whether the center coordinate is a water region or a non-water region. It is stored in the memory 22 as a learning image together with the identification information.

The second specific condition is that in the learning image extraction region or the image corresponding to the learning image extraction coordinates, there is little change between the water region and the non-water region in a certain period, and the training image extraction region or the training image extraction region or This is a condition when the image corresponding to the learning image extraction coordinates is likely to be an image corresponding to a water region or an image corresponding to a non-water region. By using an image that is likely to correspond to a water region and an image that is likely to correspond to a non-water region as a learning image, the identification accuracy of the water level measuring device 100 can be improved. The second specific condition is, for example, when the change in the water level value in the captured image in the same time zone for several hours to several days is below the threshold value, or for several hours to several days in a row. For example, the fluctuation of the coordinates of the boundary line between the water region and the non-water region is less than the threshold value. In addition, although several hours and several days are given as an example, there is no time limit such as several years.

Regarding the identification of whether the center coordinates are the water region or the non-water region, for example, the learning image generation unit 14 determines immediately before, before, immediately after, after a certain period, or after the time when the image is taken. If the learning image extraction region or the learning image extraction coordinates of a plurality of periods is a water region, it is identified as a water region, and if it is a non-water region, it is identified as a non-water region.

Hereinafter, the processing of steps ST21 to ST25 is referred to as "learning image generation processing". The water level measuring device 100 instructs to generate a learning image when a predetermined condition is satisfied (for example, immediately after the water level measuring device 100 is installed, when the variation of the water level measurement result is equal to or greater than the threshold value, or when a certain period of time has elapsed. Executed when an operation is performed). Further, the water level measuring device 100 satisfies a predetermined condition (for example, when an operation for instructing the end of the water level measuring process is performed, when the power of the water level measuring device 100 is turned off, or when the monitoring camera 1 and the water level are turned off. The learning image generation process is repeatedly executed until the communication connection between the measuring devices 100 is disconnected (for example). Although it is assumed that the learning image generation process is repeatedly executed, it may be executed only once without repeating the process.

FIG. 7 is a flowchart showing the operation of the image learning unit 15 of the water level measuring device 100 according to the first embodiment of the present invention. The operation of the image learning unit 15 of the water level measuring device 100 will be described below with reference to FIG. 7.

In step ST31, the image learning unit 15 acquires the image data corresponding to the learning image stored in step ST25 stored in the memory 22 and the identification information.

In step ST32, the image learning unit 15 executes machine learning related to the discrimination between the water region and the non-water region. In the machine learning by the image learning unit 15, when an identification image having the same characteristics as the learning image of the image that is likely to correspond to the water region is input in the learning image generation process, the identification image is used. When the corresponding area is identified as the water area and an identification image having the same characteristics as the learning image of the image which is likely to correspond to the non-water area is input in the learning image generation process. , The learning is aimed at identifying that the region corresponding to the identification image is a non-water region.

Hereinafter, the processes of steps ST31 to 32 are collectively referred to as "machine learning process". After the water level measuring device 100 is installed, the water level measuring device 100 satisfies a predetermined condition (for example, when an operation instructing the start of machine learning processing is input to the operation input device 2, the learning image is stored in the memory. When stored in 22, a predetermined condition is satisfied (machine when the water level measurement process such as maintenance of the monitoring camera 1 is stopped in order to reduce the processing load of the water level measurement device 100 during the execution of the water level measurement process). When an operation instructing the end of the learning process is input to the operation input device 2, the power of the water level measuring device 100 is turned off, or the communication connection between the monitoring camera 1 and the water level measuring device 100 is disconnected) Until, the machine learning process is automatically and repeatedly executed. Although it is assumed that the machine learning process is repeatedly executed, it may be executed only once without repeating.

That is, as time elapses from the start of water level measurement by the water level measuring device 100, the number of machine learning processes executed increases, and the number of learning images used by the image learning unit 15 for learning increases. .. In general, machine learning has the property that the accuracy of output improves as the amount of image data corresponding to the input learning image increases. Therefore, by repeating the machine learning process, the accuracy of identification by the image identification unit 12 can be gradually improved. As a result, the accuracy of the water level calculation by the water level calculation unit 13 can be improved, and the measurement by the water level measuring device 100 can be stabilized.

As described above, in the water level measuring device 100 of the first embodiment, the learning image generation unit 14 generates a learning image in consideration of the change in the surrounding environment, and the image learning unit 15 considers the change in the surrounding environment. Since the learning is performed, the image identification unit 12 can discriminate between the water region and the non-water region with high accuracy. As a result, the accuracy of the water level calculation by the water level calculation unit 13 can be improved, and the measurement by the water level measuring device 100 can be stabilized.

Further, since the water level measuring device 100 of the first embodiment sets the coordinate value of the base point according to the operation input to the operation input device 2, it is also used for water level measurement of a river or the like in which a water level plate is not installed. be able to.

Since the water level measuring device 100 of the first embodiment automatically and repeatedly executes the machine learning process, the accuracy of identification by the image identification unit 12 can be gradually improved.

Since the water level measuring device 100 of the first embodiment automatically and repeatedly executes the learning image generation process, the labor and burden are significantly reduced as compared with the case where the operator manually identifies the image and generates the learning image. can do.

Since the water level measuring device 100 of the first embodiment automatically and repeatedly executes the learning image generation process, the water level can be measured with high accuracy even if the worker does not manually collect the learning images in advance.

The base point setting unit 11 of the first embodiment sets the coordinate values of the plurality of base points in the captured image and the base point water level values corresponding to each of the plurality of base points. By increasing the number of base points in this way, the water level calculation unit 13 can calculate a finer water level.

In the first embodiment, the target of the water level measurement by the water level measuring device 100 is a river, but the target is not limited to the river. The water level measuring device 100 can also be used for measuring the water level of, for example, lakes, marshes, oceans, dams, irrigation canals or reservoirs.

In the first embodiment, the surveillance camera 1 is used to photograph the river to be measured by the water level measuring device 100, but the river to be measured by the water level measuring device 100 even if it is not the surveillance camera 1. Anything that can take a picture of.

In the first embodiment, in the water level measuring device 100, the learning image generation unit 14 automatically generates a learning image, and the image learning unit 15 performs machine learning of the images of the water region and the non-water region. The operator may do it manually. Specifically, the worker extracts the image of the water region and the image of the non-water region according to the most probable surrounding environment from the plurality of captured images, so that the learning image generation unit 14 and the image learning unit 15 It is an alternative. That is, in the first embodiment, the water level measuring device 100 automatically and repeatedly executes the learning image generation process, but the operator manually identifies the image and the water level measuring device 100 inputs the image. It may be.

In the first embodiment, the image identification unit 12 discriminates whether the center of the predetermined image is the water region or the non-water region in the identification image, but even if it is not the center, the water is about the predetermined position. It may be possible to identify whether it is a region or a non-water region.

In the first embodiment, the base point setting unit 11 sets the coordinate values of one or more base points in the captured image and the base point water level values corresponding to each of the one or more base points. However, the coordinate values of one base point are set. And the base point water level value corresponding to the one base point may be set. However, from the viewpoint of enabling the water level calculation unit 13 to calculate a finer water level, the base point setting unit 11 sets the coordinate values of the plurality of base points and the base point water level values of the plurality of base points. Is preferable.

In the first embodiment, the shape of the target area by the image identification unit 12 is a region such as the identification images D1 to D3, but the shape is not limited to the quadrangle and may be any shape. Further, the target area may include the coordinate values of the base points, and the position where the base points are arranged in the target area is not limited to the center thereof. For example, the target area may have a base point arranged at a corner thereof.

By the way, the water level measuring device, the water level measuring method, and the water level measuring program shown in the above-described embodiment are merely examples, and within the scope of the invention, modifications of any component of the embodiment such as combination with other devices. Alternatively, any component of the embodiment can be omitted.

1 Surveillance camera, 2 Operation input device, 3 Keyboard,
4 mouse, 5 display device, 6 display,
11 Basis point setting unit, 12 Image identification unit, 13 Water level calculation unit,
14 Image generation unit for learning, 15 Image learning unit,
21 processors, 22 memories, 23 buses,
30 captured images, 31 sky, 32 mountains, 33 sands,
34 rivers, 35 grasslands, D1 to D3 identification image examples,
100 Water level measuring device.

Claims

Based on the images of a plurality of water areas according to the surrounding environment that affect the image identification and the images of a plurality of non-water areas according to the surrounding environment, the acquired image is an image of a region including a predetermined base point. An image identification unit that identifies whether the region corresponding to a certain identification image is the water region or the non-water region,

A water level calculation unit that calculates the water level in the captured image based on the identification result by the image identification unit and the base point water level value corresponding to the base point.

A water level measuring device equipped with.
The surrounding environment that affects the image identification is the surrounding environment in which the color changes.

The water level measuring device according to claim 1.
The color change is a color change due to at least one of season, time, and weather.

The water level measuring device according to claim 2.
The image identification unit determines a boundary line between the water region and the non-water region by identifying whether the region corresponding to the identification image is the water region or the non-water region.

The water level measuring device according to any one of claims 1 to 3.
A reference image generation unit that generates at least one of a plurality of water region images according to the surrounding environment and a plurality of non-water region images according to the surrounding environment from the identified identification image based on the identification result.

The water level measuring device according to any one of claims 1 to 4.
An image learning unit that performs machine learning related to discrimination between the water region and the non-water region based on the image generated by the reference image generation unit.

The water level measuring device according to claim 5.
In addition to the identification result, the reference image generation unit calculates a change in the boundary line between the water region and the non-water region from the water level calculation result of the water level calculation unit, and a plurality of waters according to the surrounding environment. For at least one of the image of the region and the image of a plurality of non-water regions according to the surrounding environment, the region or coordinates for which the change of the boundary line satisfies the first specific condition is the region or coordinates for learning image extraction or for learning. Generate from the image corresponding to the image extraction coordinates

The water level measuring device according to claim 6.
The first specific condition is continuous when the identification result is always constant in a specific time zone for several hours to several days, but the identification result is replaced regularly or irregularly in another time zone. At least one of the cases where the coordinates of the boundary line between the water region and the non-water region change continuously within a certain period of time.

The water level measuring device according to claim 7.
When the image for learning image extraction region or the image corresponding to the coordinates for image extraction for learning satisfies the second specific condition, the reference image generation unit may generate a plurality of water regions according to the surrounding environment. An image or an image of a plurality of non-water areas according to the surrounding environment.

The water level measuring device according to any one of claims 7 to 8.
The reference image generation unit is constant immediately before, immediately before, immediately after, and constant before the time when the image of the plurality of water regions corresponding to the surrounding environment or the image of the plurality of non-water regions corresponding to the surrounding environment is taken. After a period of time, or based on the coordinates of the boundary between the water area and the non-water area for a plurality of periods, it is possible to identify whether the center coordinate is the water area or the non-water area.

The image learning unit performs machine learning related to the discrimination between the water region and the non-water region based on the identification of whether the center coordinates are the water region or the non-water region.

The water level measuring device according to claim 9.
The second specific condition is that when the change in the water level value in the captured image in the same time zone for several hours to several days is equal to or less than the threshold value, the water is taken for several hours to several days in a row. At least one of the cases where the fluctuation of the coordinates of the boundary line between the region and the non-water region is less than or equal to the threshold value.

The water level measuring device according to any one of claims 9 to 10.
Operation input unit A base point setting unit that sets the coordinate value of the base point and the base point water level value corresponding to the base point according to the operation input to the input device.

The water level measuring device according to any one of claims 1 to 11.
Based on the images of a plurality of water areas according to the surrounding environment that affect the image identification and the images of a plurality of non-water areas according to the surrounding environment, the acquired images are images of the area including a predetermined base point. A step of identifying whether the region corresponding to a certain identification image is the water region or the non-water region, and

A step of calculating the water level in the captured image based on the identification result of the identification step and the base point water level value corresponding to the base point.

Water level measurement method having.
Based on the images of a plurality of water areas according to the surrounding environment that affect the image identification and the images of a plurality of non-water areas according to the surrounding environment, the acquired images are images of the area including a predetermined base point. A process of identifying whether the region corresponding to a certain identification image is the water region or the non-water region,

A process of calculating the water level in the captured image based on the identification result of the identification process and the base point water level value corresponding to the base point.

Water level measurement program to execute.