Water level measurement device and shoreline extraction method

1. A water level measurement device comprising:

processing circuitry to

successively select pixels of interest from an image area designated from a captured image captured by a monitoring camera;

upon selection of each pixel of interest,

extract, as identification images, a plurality of image areas that come in contact with the currently-selected pixel of interest,

calculate an identification strength indicating a degree to which an area corresponding to each of the plurality of identification images corresponding to the currently-selected pixel is a water area, on a basis of a result of machine learning related to identification between the water area and a non-water area, and

calculate an evaluation value of the currently-selected pixel on the basis of the identification strengths of the corresponding plurality of identification images;

extract a subset of the pixels of interest as a shoreline in the captured image on a basis of comparison of the evaluation values calculated for the respective pixels of interest; and

calculate a water level within an image capturing range of the monitoring camera on a basis of the shoreline extracted.

2. The water level measurement device according to claim 1 , wherein the processing circuitry

extracts a learning image from the captured image; and

executes the machine learning related to the identification between the water area and the non-water area by using the learning image extracted.

3. The water level measurement device according to claim 1 , wherein

the monitoring camera is a camera having a function of three-dimensional measurement within the image capturing range, and

the processing circuitry calculates the water level within the image capturing range of the monitoring camera, on a basis of three-dimensional measured data by the monitoring camera.

4. A water level measurement device comprising:

processing circuitry to

successively extract sets of identification images from an image area designated from a captured image captured by a monitoring camera, each set comprising a pair of identification images one of which is directly above the other in the captured image;

upon extraction of each set of identification images,

calculate an identification strength indicating a degree to which an area corresponding to each identification image in the currently-extracted set of identification images is an water's edge, on a basis of a result of machine learning related to identification between a water area, the water's edge and a non-water area, and

calculate an evaluation value for the currently-extracted set of identification images on a basis of the respective identification strengths;

calculate a position of a shoreline in the captured image by comparing the evaluation values of the respective sets of identification images and determining, on a basis of the comparison, the sets in which both identification images come into contact with the shoreline; and

calculate a water level within an image capturing range of the monitoring camera on a basis of a position of the shoreline calculated.

5. The water level measurement device according to claim 4 , wherein the processing circuitry

extracts a learning image from the captured image; and

executes the machine learning related to the identification between the water area, the water's edge and the non-water area by using the learning image extracted.

6. A water level measurement device comprising: processing circuitry to successively extract pairs of identification images from a captured image captured by a monitoring camera, each pair including one identification image that is above the other in the captured image; identify a position of a shoreline in the captured image, on a basis of a result of machine learning related to identification of each of the identification images in the extracted pairs as one of a water area, a waters edge and a non-water area; and calculate a water level within an image capturing range of the monitoring camera on a basis of the position of the shoreline identified.

7. The water level measurement device according to claim 6 , wherein the processing circuitry

extracts a learning image from the captured image; and

executes machine learning related to identification of the water's edge by using the learning image extracted.

8. The water level measurement device according to claim 7 , wherein

the processing circuitry executes the machine learning related to the identification of the water's edge, by using, as a label of teacher data, a position of a shoreline in the learning image extracted.

9. A shoreline extraction method comprising:

selectively selecting pixels of interest from an image area designated from a captured image captured by a monitoring camera;

upon selection of each pixel of interest,

extracting, as identification images, a plurality of image areas that come in contact with the currently-selected pixel of interest;

calculating an identification strength indicating a degree to which an area corresponding to each of the plurality of identification images corresponding to the currently-selected pixel is a water area, on a basis of a result of machine learning related to identification between the water area and a non-water area, and

calculating an evaluation value of the currently-selected pixel on the basis of the identification strengths of the corresponding plurality of identification images; and

extracting a subset of the pixels of interest as a shoreline in the captured image on a basis of comparison of the evaluation values calculated for the respective pixels of interest.