JP7805836B2 - River flow estimation system - Google Patents

Description

This invention relates to technology for estimating river flow rates, and more specifically to a river flow rate estimation system that can estimate the flow rate of a river based on images obtained by photographing the river.

Knowing the amount of water flowing in a river (flow rate) is useful, and is extremely important when flooding due to heavy rain is predicted. For over 100 years, the basic method for measuring flow rate during floods has been float observation. This float observation method involves dropping a float into the river and observing the time it takes for it to flow down a certain distance to determine the flow rate, and then multiplying this flow rate by the cross-sectional area of the river (the part through which water flows) to determine the flow rate. However, float observation required a minimum of five observers, which meant it was very labor-intensive.

On the other hand, in cases where it is difficult to dispatch many observers, such as rivers in mountainous regions, river flow is generally observed using the cross-section measurement method. With this method, the river cross section is measured at intervals of several centimeters, which requires a minimum of two to three observers, and the time required for observation at one location is estimated to be 15 to 30 minutes. In other words, determining river flow using the cross-section measurement method requires a great deal of effort, just like float observations.

It is possible to measure flow directly using measuring equipment without relying on float observations or cross-sectional measurement methods, but using measuring equipment during floods is dangerous, and even if the measuring equipment were permanently installed, there is a risk of it being damaged or readings being missed, so this is not a realistic method.

In recent years, therefore, attempts have been made to install cameras or video cameras in noteworthy locations on rivers and use the images and videos to determine river flow rates. For example, Patent Document 1 proposes technology for determining river flow rates by capturing images of a river at two or more time periods and using these images to perform PIV (Particle Image Velocimetry) analysis.

Japanese Patent Application Laid-Open No. 2008-216010

Until now, the mainstream method for determining river flow rate has been to first determine the flow velocity and then calculate the flow rate. For example, in float observations, the speed of the float is measured, and in Patent Document 1, PIV analysis is performed to determine the river flow velocity, and the river flow rate is then calculated by multiplying the flow velocity by the cross-sectional area of the river.

However, measuring river flow velocity using float observations requires a great deal of effort and does not always result in an accurate flow velocity. Furthermore, even if PIV analysis is performed as in Patent Document 1, accurate flow velocity may not always be obtained depending on the image conditions. And if accurate flow velocity cannot be obtained, it is naturally impossible to accurately determine the river flow rate.

The objective of the present invention is to solve the problems of the past, namely, to provide a river flow estimation system that uses images obtained by photographing a river and can determine river flow rate without calculating flow velocity.

The present invention focuses on the fact that river flow is estimated based on the number of pixels related to the river in an image obtained by photographing the river, and is an invention based on an unprecedented concept.

The river flow estimation system of the present invention is a system that estimates river flow based on observed images of a river, and includes a river section setting means, a river pixel counting means, and a calculated flow output means. The river section setting means is a means for setting a river width side line, and the river pixel counting means is a means for counting the total number of pixels in the observed image related to the river width side line as the river pixel count. The calculated flow output means is a means for outputting the calculated flow by inputting the river pixel count into a flow calculation formula. The river width side line is set to be the river width of the river included in the observed image. The flow calculation formula is a function that represents the relationship between the river pixel count and the calculated flow, and is set based on the actually counted river pixel count and the actual observed river flow rate. The calculated flow rate output by the calculated flow output means can then be estimated as the flow rate of the river included in the observed image.

The river flow estimation system of the present invention can also set river width lateral lines by inputting observed images into a trained model. In this case, the trained model is generated by machine learning observed images on which river width lateral lines have been set.

The river flow estimation system of the present invention can also be configured so that the river area setting means sets the river area. In this case, the river pixel counting means counts the total number of pixels in the observed image that belong to the river area as the river pixel count.

The river flow estimation system of the present invention can also set river width laterals by inputting observed images into a trained model. In this case, the trained model is generated by machine learning observed images in which river areas are set.

The river flow estimation system of the present invention can also be configured so that the flow calculation formula is set as a quadratic or cubic formula.

The river flow rate estimation system of the present invention has the following effects.
(1) The flow rate can be determined simply by acquiring observation images. For example, by permanently installing a camera on-site, unmanned operation and labor saving can be achieved. As a result, observation costs can be reduced and observers are not placed in dangerous situations.
(2) Furthermore, since the flow rate can be grasped simply by installing a camera on-site, it can be applied in a variety of locations.
(3) Since there is no need to calculate the flow velocity, it is possible to avoid the inconvenience of calculating an inaccurate river flow rate due to an inaccurate flow velocity.

1 is a block diagram showing the main configuration of a river flow rate estimation system according to the present invention; (a) is a model diagram showing a schematic representation of the river width side line set by the trained model of the river section setting means, and (b) is a model diagram showing a schematic representation of the river area set by the trained model of the river section setting means. A mathematical formula explaining that flow rate is proportional to the 8/3 power of water depth. FIG. 10 is a graph illustrating a flow rate calculation formula set as a quadratic expression of the number of river pixels. FIG. 2 is a flowchart showing an example of a main processing flow of the river flow rate estimation system.

An example of an embodiment of the river flow estimation system of the present invention will be explained with reference to the drawings.

Figure 1 is a block diagram showing the main components of the river flow estimation system 100 of the present invention. As shown in this figure, the river flow estimation system 100 of the present invention is configured to include a river section setting means 101, a river pixel counting means 102, and a calculated flow output means 103, and can also be configured to include a flow calculation formula setting means 104, a model generation means 105, an image acquisition means 106, an output means 107, an observed image storage means 108, a trained model storage means 109, and a flow calculation formula storage means 110.

The river section setting means 101, river pixel counting means 102, calculated flow rate output means 103, flow rate calculation formula setting means 104, and model generation means 105 that constitute the river flow rate estimation system 100 of the present invention can be manufactured as dedicated devices, or a general-purpose computer device can be used. This computer device is equipped with a processor such as a CPU, memory such as ROM and RAM, and may further include input means such as a mouse and keyboard, and display means such as a display, and can be configured, for example, as a personal computer (PC) or server.

Furthermore, the observed image storage means 108, trained model storage means 109, and flow rate calculation formula storage means 110 can use the storage device of a general-purpose computer (e.g., a PC), or can be built on a database server. If built on a database server, it can be placed on a local network (LAN: Local Area Network), or it can be a cloud server that stores data via the Internet.

Below, we will explain in detail each of the main elements that make up the river flow estimation system 100.

(Image Acquisition Means)
The image acquisition means 106 acquires images (hereinafter referred to as "observation images") obtained by photographing the river as still images or videos, and can be a digital camera, a digital video camera, a smartphone, a tablet PC, or the like. Observation images can be acquired using the image acquisition means 106 carried by the observer, or using the image acquisition means 106 of a so-called fixed camera permanently installed on-site, or satellite images can also be used. The still images and videos acquired by the image acquisition means 106 are stored in the observation image storage means 108 (Figure 1). Note that the observation images taken by the image acquisition means 106 of the fixed camera may be transmitted via communication means and then stored in the observation image storage means 108.

(River section setting means)
The river section setting means 101 is a means for setting a portion of the observed image corresponding to a river (hereinafter simply referred to as a "river section"). The river flow rate estimation system 100 of the present invention can be broadly divided into a form (hereinafter referred to as a "first form") in which a lateral line (hereinafter referred to as a "river width lateral line") indicating the width of a river (hereinafter simply referred to as a "river width") is set as a river section, as shown in Figure 2(a) , and a form (hereinafter referred to as a "second form") in which an area indicating a river (hereinafter referred to as a "river area") is set as a river section, as shown in Figure 2(b) . That is, the river section setting means 101 in the first form sets the river width lateral line as the river section, and the river section setting means 101 in the second form sets the river area as the river section.

The river section setting means 101 can be configured so that the operator can set the river width side line and river area while checking the observation image displayed on the display (output means 107). In this case, it is recommended to use a pointing device (mouse, touch panel, pen tablet, touchpad, trackpad, trackball, etc.) or keyboard as the river section setting means 101.

The river section setting means 101 can also be configured to automatically (mechanically) set river width side lines and river areas from observed images. In this case, the river area can be automatically extracted using conventional image recognition technology, and then the river width side lines and river areas can be set. Alternatively, the river width side lines and river areas can be set using a "trained model."

The trained model is generated by the model generation means 105. In the first form, the model generation means 105 generates the trained model by performing machine learning (e.g., deep learning) using observation images in which river width side lines are set as training data. On the other hand, in the second form, the model generation means 105 generates the trained model by performing machine learning using observation images in which river areas are set as training data. The trained model generated by the model generation means 105 is stored in the trained model storage means 109 (Figure 1).

When the river section setting means 101 uses a trained model, the river width side line and river area are set by inputting an observed image into the trained model. For example, in Figure 2(a), the trained model to which the observed image is input sets the river width side line, and in Figure 2(b), the trained model to which the observed image is input sets the river area.

(Means for counting river pixels)
The river pixel counting means 102 is a means for counting the number of pixels (pixels) constituting the observation image that correspond to the river portion (hereinafter referred to as the "river pixel count"). That is, the river pixel counting means 102 in the first form reads out the observation image, reads out the river width side line set by the river portion setting means 101, extracts pixels related to the river width side line (constituting the river width side line), and counts the total number as the river pixel count. On the other hand, the river pixel counting means 102 in the second form reads out the observation image, reads out the river area set by the river portion setting means 101, extracts pixels related to the river area (constituting the river area), and counts the total number as the river pixel count.

In the first embodiment, the river pixel counting means 102 can also be configured to count the number of river pixels of the river width side line based on the number of pixels that make up the river area. For example, a first axis (e.g., a vertical axis) and a second axis (e.g., a horizontal axis) perpendicular to the first axis can be set for the observed image, and the river area can be considered to be composed of pixels arranged along the second axis aligned in the first axis direction, with the average number of pixels aligned along the second axis being counted as the number of river pixels of the river width side line. Therefore, in this case, the river portion setting means 101 sets the river area as a river portion.

(Flow rate calculation formula setting means)
The flow rate calculation formula setting means 104 is a means for setting the "flow rate calculation formula." This flow rate calculation formula is a function that represents the relationship between the flow rate obtained by calculation (hereinafter referred to as the "calculated flow rate") and the number of river pixels, and is set based on the number of river pixels that have actually been recorded (hereinafter referred to as the "actual river pixel number") and the flow rate of the river that has actually been observed (hereinafter referred to as the "actual river flow rate"). The procedure for setting the flow rate calculation formula will be explained below.

First, the actual river flow rate is obtained by measuring the river flow rate, and the actual river pixel count is obtained by acquiring observed images at that time. It is advisable to prepare multiple combinations of actual river pixel count and actual river flow rate by measuring the flow rate and acquiring observed images under conditions with different flow rate values. Once multiple combinations of actual river pixel count and actual river flow rate have been prepared, a flow rate calculation formula is set using the least squares method or the like. The flow rate calculation formula set by the flow rate calculation formula setting means 104 is stored in the flow rate calculation formula storage means 110 (Figure 1).

Generally, a river's flow rate is calculated by multiplying its cross-sectional area by its flow velocity, as shown in equation (1) in Figure 3, and the flow velocity is often calculated using equation (2) in Figure 3. When the cross-sectional shape is roughly trapezoidal, as shown in Figure 3, the flow rate can be expressed as equation (8) using equations (3) to (7). If the riverbed B in equation (8) approaches zero (i.e., approaches a triangle), the flow rate Q becomes proportional to the 8/3 power of the water depth h (equation (9)). In other words, flow rate Q can be calculated as a function with water depth h as a variable, and it is expected to be highly correlated with quadratic and cubic expressions for water depth h. Therefore, the inventors of the present invention designed a flow rate calculation formula (a function expressing the relationship with the number of river pixels) as a quadratic or cubic expression with the number of river pixels as a variable, and found that, as expected, a high correlation was observed. For example, the flow rate calculation formula shown in Figure 4 is set as a quadratic expression of the number of river pixels, but its correlation coefficient R2 is a high value of 0.88. In other words, it is more preferable to set the flow rate calculation formula as a quadratic or cubic expression of the number of river pixels.

(Calculated flow rate output means)
The calculated flow rate output means 103 is a means for outputting the flow rate calculated by the flow rate calculation formula (hereinafter referred to as the "calculated flow rate") to an output means 107 such as a display or printer (Fig. 1). That is, the calculated flow rate output means 103 reads out the number of river pixels counted by the river pixel number counting means 102, reads out the flow rate calculation formula set by the flow rate calculation formula setting means 104, and outputs the calculated flow rate by inputting the number of river pixels into the flow rate calculation formula. The calculated flow rate output here can then be estimated as the flow rate of the river when it was captured in the observed image.

(Processing flow)
The main processing of the river flow rate estimation system 100 will be described in detail below with reference to Fig. 5. Fig. 5 is a flow diagram showing an example of the flow of the main processing of the river flow rate estimation system 100, in which the processing to be executed is shown in the center column, what is necessary for that processing is shown in the left column, and what results from that processing is shown in the right column.

To estimate river flow using the river flow estimation system 100, an observation image is first acquired, as shown in Figure 5 (Step 201 in Figure 5). As mentioned above, the observation image can be acquired using an image acquisition means 106 carried by an observer, or it can be acquired using an image acquisition means 106 of a so-called fixed camera permanently installed on-site.

When an observation image is acquired, the river section setting means 101 sets the river section (river width side line or river area) from the read observation image (Step 202 in Figure 5). The river pixel counting means 102 then reads the observation image and river width side line and counts the number of river pixels (Step 203 in Figure 5), and the calculated flow rate output means 103 reads the number of river pixels and the set flow rate calculation formula and outputs the calculated flow rate (Step 204 in Figure 5).

The river flow estimation system of the present invention is particularly useful for river administrators, including national and local governments. Considering that the present invention can quickly predict floods and other river-related water damage, thereby protecting many residents from disasters, the present invention is not only applicable to industry but is also expected to make a significant contribution to society.

100 River flow rate estimation system of the present invention 101 River section setting means (of river flow rate estimation system) 102 River pixel number counting means (of river flow rate estimation system) 103 Calculated flow rate output means (of river flow rate estimation system) 104 Flow rate calculation formula setting means (of river flow rate estimation system) 105 Model generation means (of river flow rate estimation system) 106 Image acquisition means (of river flow rate estimation system) 107 Output means (of river flow rate estimation system) 108 Observed image storage means (of river flow rate estimation system) 109 Learned model storage means (of river flow rate estimation system) 110 Flow rate calculation formula storage means (of river flow rate estimation system)

Claims (3)

A system for estimating the flow rate of a river based on observation images of the river, comprising:
a river area setting means for setting a river area;
a river pixel counting means for counting the total number of pixels relating to the river region in the observation image as the number of river pixels;
a calculated flow rate output means for outputting a calculated flow rate by inputting the number of river pixels into a flow rate calculation formula;
the river region is set to be a river region included in the observed image,
The flow rate calculation formula is a function that represents the relationship between the number of river pixels and the calculated flow rate, and is set based on the number of river pixels that is actually recorded and the actual value of the flow rate of the river that is actually observed,
The calculated flow rate output by the calculated flow rate output means can be estimated as the flow rate of the river included in the observation image.
A river flow rate estimation system characterized by: the river area setting means sets the river area by inputting the observed image into a trained model;
The trained model is generated by machine learning the observed image in which the river area is set.
2. The river flow rate estimation system according to claim 1 . The flow rate calculation formula is set as a quadratic or cubic formula,
3. The river flow rate estimation system according to claim 1 or 2 .