Processing device, processing method, and program

Abstract

This processing device 1 comprises an extraction unit 22 for calculating a representative value for each of a plurality of segments that have similar pixel feature values and that are divided from satellite image data 11 photographed at a prescribed time of day, and extracting two segments for which the calculated representative values are different.

Description

[Technical field]

[0001] The present invention relates to a processing device, a processing method, and a program. [Background technology]

[0002] Sensors mounted on satellites measure the emission or reflection of electromagnetic waves from the atmosphere surrounding the Earth, the earth's surface, the sea surface, etc. Sensors do not actually measure the physical quantities or states of objects. A model is used to convert the electromagnetic waves measured by the sensor into a desired value. Data assimilation is being developed in the field of weather forecasting (Non-Patent Document 1). Data assimilation provides a complementary link between observations and numerical models. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Japan Meteorological Agency, "Chapter 3 Data Assimilation", [online], [Retrieved May 27, 2021], Internet〈URL:https: / / www.jma.go.jp / jma / kishou / books / nwptext / 45 / 1_chapter3.pdf〉 Summary of the Invention [Problem to be solved by the invention]

[0004] Noise caused by changes in the natural environment can make it difficult to obtain sufficient model accuracy. To solve this problem, actual measurement data is needed to calibrate the model before applying it.

[0005] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a technology that enables efficient selection of observation points for data suitable for calibration before applying a model in data assimilation. [Means for solving the problem]

[0006] A processing device according to one embodiment of the present invention includes an extraction unit that calculates a representative value for each of a plurality of segments having similar pixel feature values, which are divided from satellite image data captured at a specified time, and extracts two segments having different calculated representative values.

[0007] In one embodiment of the processing method of the present invention, a computer calculates a representative value for each of a number of segments having similar pixel feature values ​​divided from satellite image data captured at a specified time, and extracts two segments having different calculated representative values.

[0008] One aspect of the present invention is a program for causing a computer to function as the processing device. Effect of the Invention

[0009] According to the present invention, it is possible to provide a technique for efficiently selecting observation points for data suitable for calibration before applying a model in data assimilation. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram illustrating functional blocks of a processing device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating an example of satellite image data. [Diagram 3] FIG. 3 is a diagram illustrating an example of segments divided from satellite image data. [Figure 4] FIG. 4 is a diagram illustrating the change over time of the representative value. [Diagram 5] FIG. 5 is a diagram illustrating an example of divisions of the representative values. [Figure 6] FIG. 6 is a diagram illustrating an example of a platform route. [Figure 7] FIG. 7 is a flowchart illustrating a processing method performed by the processing device according to the embodiment of the present invention. [Figure 8] FIG. 8 is a diagram illustrating the hardware configuration of a computer used in the processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the description of the drawings, the same parts are given the same reference numerals and the description will be omitted.

[0012] (Processing device) A processing device 1 according to an embodiment of the present invention identifies a measurement segment from satellite image data 11 captured by a satellite to obtain measurement data to be used in calibration before applying a data assimilation model, and a measurement point in the measurement segment. The processing device 1 also identifies a measurement segment from which measurement data to be used in a simulation is obtained, and a measurement point in the measurement segment. The measurement data is measured by flying a platform such as a drone from a predetermined base to the measurement point, and measuring the measurement point with a measuring device mounted on the platform.

[0013] 1, the processing device 1 includes satellite image data 11, segment data 12, actually measured segment data 13, and actually measured point data 14, as well as functions of a division unit 21, an extraction unit 22, an identification unit 23, and an output unit 24. Each piece of data is stored in a storage device such as a memory 902 or a storage 903. Each function is implemented in a CPU 901.

[0014] The satellite image data 11 is an image of the earth taken from a satellite. The satellite image data 11 has values ​​of multiple electromagnetic wave bands for each pixel. Specifically, each pixel has values ​​of not only red (R), green (G), and blue visible light, but also values ​​of each band such as near infrared and thermal infrared. The values ​​of each band that a pixel has are also called feature values.

[0015] The satellite image data 11 may be a collection of multiple image data taken at multiple times and multiple altitudes, as shown in Fig. 2. The satellite image data 11 is taken at a predetermined time interval by the orbit of the satellite. In the example shown in Fig. 2, the satellite image data 11 has data of images taken at five altitudes at time t-1, data of images taken at five altitudes at time t, and data of images taken at five altitudes at time t+1.

[0016] The segment data 12 specifies a number of segments divided from satellite image data 11 captured at a given time and altitude. The segment data 12 associates, for example, the capture time and altitude as well as a segment identifier and an identifier of a pixel belonging to the segment. The segment data 12 is generated by a dividing unit 21. The satellite image data 11 is divided into a number of segments as shown in Fig. 3. In the example shown in Fig. 3, the satellite image data 11 is divided into five segments, but the number of segments is not limited to this.

[0017] The measured segment data 13 identifies a segment to be measured for calibration or simulation among a plurality of segments identified by the dividing unit 21. The measured segment data 13 includes, for example, a segment identifier and an altitude. The measured segment data 13 is generated by the extracting unit 22.

[0018] The actual measurement point data 14 identifies a point of the actual measurement target among the segments of the actual measurement target. The actual measurement point data 14 is generated by the identification unit 23. The actual measurement point data 14 includes an identifier of a pixel of the actual measurement target, or a latitude and longitude corresponding to the pixel. The actual measurement point data 14 may also include position information such as altitude.

[0019] The division unit 21 divides the satellite image data 11 captured at a predetermined time into a plurality of segments to generate segment data 12. For example, the division unit 21 specifies pixels having similar pixel feature values ​​and adjacent to each other as a segment (for example, see “Object-oriented forest stand type classification using texture features and spectrum features of high-resolution satellite images” (URL: https: / / www.jstage.jst.go.jp / article / jsprs1975 / 46 / 2 / 46_2_27 / _pdf / -char / ja)).

[0020] The extraction unit 22 calculates a representative value for each of a plurality of segments divided from the satellite image data 11 captured at a predetermined time, and extracts two segments with different calculated representative values. The extracted segments are targets for obtaining actual measurement data used in calibration, which removes electromagnetic waves from objects other than the measurement target, or simulations. The actual measurement data may be obtained at actual measurement points within the extracted segments. Here, "different representative values" may include not only cases where the representative values ​​are not the same, but also cases where the representative values ​​are different by a predetermined value or more. For example, the extraction unit 22 arranges the calculated representative values ​​in a predetermined order, and extracts two segments with different orders. The extraction unit 22 outputs actual measurement segment data 13 including identifiers of the extracted segments.

[0021] First, the process of extracting a plurality of segments to be used for calibration will be described.

[0022] The extraction unit 22 calculates a representative value for each of a plurality of segments from the feature values ​​of the pixels belonging to that segment. The representative value is a value that characterizes the attribute of the segment, such as the average or median of the feature values ​​of the pixels belonging to the segment, or the feature value of the pixel at the center of gravity of the segment. When the feature value of each pixel includes multiple elements, such as values ​​for each band, the representative value is the magnitude of a vector specified by the multiple elements.

[0023] After calculating the representative value for each segment, the extraction unit 22 arranges the representative values ​​in a predetermined order. The extraction unit 22 extracts two segments whose representative values ​​are different from each other, such as a segment with the largest representative value and a segment with the smallest representative value. The extraction unit 22 may extract a plurality of segments to be used for calibration from among the plurality of segments included in the satellite image data 11, the segments in which the drone can move.

[0024] For example, the extraction unit 22 divides the representative value into a plurality of sections and extracts a segment from each section having a maximum value and a minimum value. If there is one segment that belongs to the section having the maximum value, the extraction unit 22 extracts that segment. Similarly, if there is one segment that belongs to the section having the minimum value, the extraction unit 22 extracts that segment. The processing device 1 can support stable calibration by selecting a point to be actually measured for calibration from two segments whose representative values ​​diverge.

[0025] When there are multiple segments that belong to the section having the maximum or minimum value, the extraction unit 22 selects one actually measured segment from the multiple segments. The extraction unit 22 selects one actually measured segment by focusing on the change over time or the distance.

[0026] First, a process of selecting an actual measurement segment with attention to the change over time will be described. When there are multiple segments belonging to a section having a maximum value or a minimum value, the extraction unit 22 may extract a segment whose representative value has a small difference from the representative value in satellite image data captured at a time different from the predetermined time. When there are multiple segments belonging to a section having a maximum value, the extraction unit 22 extracts a segment whose representative value has a small change over time as an actual measurement segment from among the multiple segments. Similarly, when there are multiple segments belonging to a section having a maximum value, the extraction unit 22 extracts a segment whose representative value has a small change over time as an actual measurement segment from among the multiple segments. When there are multiple candidates for a segment whose representative value deviates from an already selected segment, the processing device 1 extracts a segment whose representative value has a smaller change over time. The processing device 1 can support stable calibration by selecting a point to be measured for calibration from a segment whose representative value has a smaller change over time. The extraction unit 22 may identify the change over time of the representative value from two satellite image data captured at different times, or may identify the change over time of the representative value from three or more satellite image data captured at different times.

[0027] Here, the change over time will be described with reference to FIG. 4. In the embodiment of the present invention, the satellite image data 11 has satellite image data of the same latitude, longitude, and altitude for each hour. The feature value of each segment may change over time depending on its conditions. FIG. 4 shows the change in the representative value over time for two segments. In FIG. 4, the solid line indicates that the change over time is small. A segment with a small change in the representative value over time is suitable for measuring actual measurement data for calibration, since stable correction is possible. On the other hand, the dashed line indicates that the change in the representative value over time is large. A segment with a small change in the representative value over time is suitable for measuring actual measurement data for simulation, since a wide variety of data can be reflected in the simulation.

[0028] For example, as shown in Fig. 5, the extraction unit 22 divides the range of the representative value into a plurality of sections, such as 10, and associates with each section an identifier of a segment having a representative value belonging to that section. In the example shown in Fig. 5, segments are extracted from each of section 1 having the maximum value and section 10 having the minimum value. Since the number of segments belonging to section 1 having the maximum value is 1, the extraction unit 22 extracts segment A from section 1 as the actual measurement segment.

[0029] On the other hand, the number of segments belonging to category 10, which has the minimum value, is 2. Therefore, extraction unit 22 refers to satellite image data captured at another time and calculates the representative value for each of segment B and segment C. Extraction unit 22 extracts, from segment B or segment C, the segment with the smaller difference between the representative value in satellite image data 11 and the representative value in satellite image data captured at the other time.

[0030] Next, a process of selecting an actual measurement segment with a focus on distance will be described. When there are multiple segments belonging to a division having a maximum or minimum value, a segment close in distance to the segment selected from the other division may be extracted. For example, when there is one divisional segment having a maximum value and multiple segments belonging to a division having a minimum value, the extraction unit 22 extracts a segment close to the segment belonging to the division having the maximum value from among the multiple segments belonging to the division having the minimum value as an actual measurement segment. This makes it possible to suppress the flight distance of the drone in the actual measurement.

[0031] Next, a process for extracting a plurality of segments to be used in a simulation will be described.

[0032] The extraction unit 22 extracts segments whose representative values ​​have a large difference from the representative values ​​in the satellite image data 11 captured at a time different from the predetermined time as actual measurement segments for the simulation. The extraction unit 22 may extract a plurality of segments to be used in the simulation from among the plurality of segments included in the satellite image data 11, which are segments in which the drone can move. The extraction unit 22 outputs actual measurement segment data 13 including identifiers of the extracted segments. By setting segments with large time changes as actual measurement segments for the simulation, the processing device 1 can reflect a variety of actual measurement values ​​in the simulation.

[0033] The extraction unit 22 may also select one of the segments selected for calibration as a segment for simulation, thereby making it possible to reduce the travel distance of the drone and efficiently collect actual measurement values.

[0034] The identification unit 23 identifies an actual measurement point by a measuring instrument among the segments identified in the actual measurement segment data 13. The identification unit 23 outputs the identified point as actual measurement point data 14. The identification unit 23 identifies a pixel in the satellite image data 11 and may output the pixel itself as the actual measurement point, or may output the latitude and longitude corresponding to the pixel as the actual measurement point.

[0035] First, the process of identifying the actual measurement points used for calibration will be described.

[0036] The identification unit 23 identifies an actual measurement point from each of the two segments extracted for calibration. The identification unit 23 identifies pixels in the extracted segment that are a predetermined number of pixels away from an adjacent segment, and that have a small difference from a feature value in satellite image data captured at a time different from the predetermined time. The identification unit 23 identifies the position of the identified pixel as an actual measurement point for calibration.

[0037] In identifying actual measurement points for calibration, the identifying unit 23 first excludes pixels within a few pixels from the boundary of a segment from candidates for actual measurement points. Pixels close to the boundary of a segment may be affected by the feature values ​​of adjacent segments. By excluding such pixels from the actual measurement points, the identifying unit 23 can select actual measurement points from pixels that more closely reflect the features of the segment.

[0038] The identifying unit 23 identifies pixels with small time changes among the pixels excluding pixels close to the boundary of the segment as points to be measured for calibration. By using pixels with small time changes as actual measurement points for calibration, the processing device 1 can support stable calibration.

[0039] Next, a process for identifying actual measurement points to be used in the simulation will be described.

[0040] The identification unit 23 identifies actual measurement points for the simulation from the segments extracted for calibration or the segments extracted for the simulation. The identification unit 23 identifies pixels in the extracted segments that are a predetermined number of pixels away from adjacent segments, and that have a large difference from the feature value in the satellite image data 11 captured at a time different from the predetermined time. The identification unit 23 identifies the position of the identified pixel as an actual measurement point for the simulation.

[0041] In identifying actual measurement points for simulation, the identifying unit 23 first excludes pixels close to the boundary of a segment from candidates for actual measurement points. Pixels close to the boundary of a segment may be influenced by the feature values ​​of adjacent segments. By excluding such pixels from the actual measurement points, the identifying unit 23 can select actual measurement points from pixels that more closely reflect the features of the segment.

[0042] The identifying unit 23 identifies pixels with large time changes among the pixels excluding pixels close to the boundary of the segment as points to be measured for the simulation. By setting the pixels with large time changes as actual measurement points for the simulation, the processing device 1 can support the simulation using a variety of actual measurement values.

[0043] The output unit 24 outputs the actually measured segment data 13 and the actually measured point data 14 .

[0044] The output unit 24 outputs the two segments extracted for calibration by the extraction unit 22 as actual measurement segments for calibration that removes electromagnetic waves from sources other than the measurement target. The output unit 24 further outputs the actual measurement points identified by the identification unit 23 from each of the two segments extracted for calibration.

[0045] The output unit 24 outputs the segments extracted for the simulation by the extraction unit 22 as actual measurement segments for the simulation. The output unit 24 further outputs the actual measurement points identified by the identification unit 23 from each of the two segments extracted for the simulation.

[0046] The output unit 24 outputs at least two measured segments for calibration and one measured segment for simulation as segments included in the measured segment data 13. The measured segment for simulation may be the same as or different from one of the two measured segments for calibration.

[0047] The output unit 24 outputs two actual measurement points identified from each of the two actual measurement segments for calibration and one actual measurement point identified from the actual measurement segment for simulation as points included in the actual measurement point data 14. Note that even if the segment for calibration and the segment for simulation are the same in the embodiment of the present invention, the actual measurement points for calibration and the actual measurement points for simulation are different. The actual measurement points for calibration have small changes over time, and the actual measurement points for simulation have large changes over time.

[0048] The output unit 24 may output, for example, a route for moving to two actual measurement points for calibration and one actual measurement point for simulation. In this case, the route is preferably in the order of one actual measurement point for calibration, an actual measurement point for simulation, and another actual measurement point for calibration. By providing a time difference between the two points for calibration, it is possible to obtain more stable data. The route may also be set so that the second actual measurement point for calibration is measured last. By measuring one point each for calibration and simulation, it is possible to obtain the minimum amount of data and achieve the purpose of the measurement. Also, by measuring the second point for calibration when there is a margin in the power of a platform such as a drone, it is possible to support more accurate calibration.

[0049] As shown in Figure 6, a route is set to draw a figure eight for one measurement point. In Figure 6, the route is set in the order of (1) to (8). This reduces the mechanical distance on the platform. Specifically, in order to observe the rectangle centered on the measurement point P on an average basis, the diagonal lines are crossed within the rectangle and the circumference is also measured sequentially.

[0050] The output unit 24 sets the route, moving speed, number of platforms, etc. so that the measurement by the platform is completed within the time it takes for the satellite to synchronize with new satellite image data captured by the satellite. Here, "synchronization with satellite image data" means that the measurement by the platform is completed within the time it takes for the satellite to capture new satellite image data. Calibration, simulation, etc. become possible by linking the actual measurement data of the actual measurement point P with the new satellite image data captured by the satellite.

[0051] The processing device 1 may include a control unit (not shown) that monitors whether the platform is moving according to the route output by the output unit 24. The control unit may track the position of the platform, and when the platform deviates a predetermined distance from the route, output a warning that the platform has deviated from the route and prompt the platform to return to the route.

[0052] With reference to FIG. 7, a processing method according to an embodiment of the present invention will be described.

[0053] First, in step S1, the processing device 1 divides satellite image data at a predetermined time and a predetermined altitude into a plurality of segments.

[0054] In step S2, the processing device 1 extracts two or more segments for calibration based on the representative value of each segment. In step S3, the processing device 1 specifies actual measurement points for calibration from each segment extracted in step S2.

[0055] In step S4, the processing device 1 extracts segments for simulation based on the representative value of each segment. In step S5, the processing device 1 specifies actual measurement points for simulation from the segments extracted in step S4.

[0056] In step S6, the processing device 1 outputs data on the route that passes through the actual measurement points identified in steps S3 and S5.

[0057] When identifying actual measurement points for a plurality of altitudes, the processing device 1 repeats the processes of steps S1 to S5 for the satellite image data for each altitude.

[0058] The processing device 1 according to the embodiment of the present invention can locally and dynamically acquire data that meets a specific purpose, such as calibration, using satellite image data of a relatively wide area acquired by a permanent observation system such as a satellite, thereby reducing observation costs and enabling measurement of actual data suitable for the desired satellite image data.

[0059] The processing device 1 also divides the satellite image data into a plurality of segments with similar feature values, and extracts a segment that matches the purpose of measurement from the plurality of segments. The processing device 1 identifies an actual measurement point suitable for measurement from the extracted segments. This enables the processing device 1 to efficiently select a desired actual measurement point.

[0060] The processing device 1 according to the embodiment of the present invention can efficiently select, in data assimilation, observation points of data suitable for calibration before application of a model.

[0061] The processing device 1 of the present embodiment described above is, for example, a general-purpose computer system including a CPU (Central Processing Unit, processor) 901, a memory 902, a storage 903 (HDD: Hard Disk Drive, SSD: Solid State Drive), a communication device 904, an input device 905, and an output device 906. In this computer system, the CPU 901 executes a program loaded on the memory 902, thereby realizing each function of the processing device 1.

[0062] The processing device 1 may be implemented in one computer or in multiple computers, and may also be a virtual machine implemented in a computer.

[0063] The program of the processing device 1 can be stored in a computer-readable recording medium such as an HDD, SSD, USB (Universal Serial Bus) memory, CD (Compact Disc), or DVD (Digital Versatile Disc), or can be distributed via a network.

[0064] The present invention is not limited to the above-described embodiment, and various modifications are possible within the scope of the present invention. [Explanation of symbols]

[0065] 1 Processing equipment 11 Satellite image data 12 Segment Data 13 Actual Segment Data 14. Actual measurement point data 21 Division 22 Extraction part 23 Specific section 24 Output section 901 CPU 902 Memory 903 Storage 904 Communication equipment 905 Input Device 906 Output Device

Claims

1. An extraction unit that calculates a representative value for each of a plurality of segments having similar pixel feature values, which are divided from satellite image data taken at a specified time, and extracts two segments having different calculated representative values; The two segments are measurement segments for calibration to remove electromagnetic waves from other than the measurement target, The extraction unit is Dividing the representative value into a plurality of sections and extracting a segment from each section having a maximum value and a minimum value; When there are a plurality of segments that belong to the category having the maximum or minimum value, a segment whose representative value has a small difference from a representative value in satellite image data captured at a time different from the predetermined time is extracted. Processing unit.

2. An extraction unit that calculates a representative value for each of a plurality of segments having similar pixel feature values, which are divided from satellite image data taken at a specified time, and extracts two segments having different calculated representative values; The two segments are measurement segments for calibration to remove electromagnetic waves from other than the measurement target, The extraction unit is Dividing the representative value into a plurality of sections and extracting a segment from each section having a maximum value and a minimum value; If there are a plurality of segments that belong to the category having the maximum or minimum value, a segment that is close in distance to the selected segment is extracted from the other category. Processing unit.

3. The two segments are measurement segments for calibration to remove electromagnetic waves from other than the measurement target, A determination unit that, among pixels in the extracted segment that are a predetermined number of pixels away from adjacent segments, identifies pixels that have a small difference from a feature value in satellite image data captured at a time different from the predetermined time, and determines the position of the identified pixel as an actual measurement point. The processing device according to claim 1 or 2, further comprising:

4. an identifying section that identifies pixels in the extracted segments that are a predetermined number of pixels away from adjacent segments and have a large difference in feature value from satellite image data captured at a time different from the predetermined time, and identifies the positions of the identified pixels as actual measurement points for simulation; The processing device according to claim 1 or 2, further comprising:

5. The extraction unit further A segment in which the representative value has a large difference from a representative value in satellite image data captured at a time different from the predetermined time is extracted as an actual measurement segment for simulation. The processing device according to claim 1 or 2.

6. a computer calculates a representative value for each of a plurality of segments having similar pixel feature values, the representative value being divided from satellite image data captured at a predetermined time, and extracts two segments having different calculated representative values; The two segments are measurement segments for calibration to remove electromagnetic waves from other than the measurement target, The computer, Dividing the representative value into a plurality of sections and extracting a segment from each section having a maximum value and a minimum value; When there are a plurality of segments that belong to the category having the maximum or minimum value, a segment whose representative value has a small difference from a representative value in satellite image data captured at a time different from the predetermined time is extracted. Processing methods.

7. A computer calculates a representative value for each of a plurality of segments having similar pixel feature values ​​divided from satellite image data taken at a specified time, and extracts two segments having different calculated representative values; The two segments are measurement segments for calibration to remove electromagnetic waves from other than the measurement target, The computer, Dividing the representative value into a plurality of sections and extracting a segment from each section having a maximum value and a minimum value; If there are a plurality of segments that belong to the category having the maximum or minimum value, a segment that is close in distance to the selected segment is extracted from the other category. Processing methods.

8. A program for causing a computer to function as the processing device according to claim 1 or 2.

Images