JP2021043485A - Image analysis program, image analysis method and image analysis device - Google Patents

Abstract

To provide an image analysis program, an image analysis method and an image analysis device, capable of identifying an object based on an image captured from the sky and generating information related to the identified object.SOLUTION: In an image analysis device 1, an image analysis program applies object-based image analysis to a near-infrared band image 31 in which a region on the earth including an object is captured, to generate a first identification image in which the region of the object is identified in the near-infrared band image 31, and to execute a process of generating related information 35 related to the object based on the first identification image.SELECTED DRAWING: Figure 2

Description

The present invention relates to image analysis programs, methods and devices.

The aquaculture industry on the surface of the sea is thriving in the coastal areas of Japan, and the aquaculture industry plays an important role as a source of marine products to respond to the increase in the world's population. It has been pointed out that the aquaculture industry is built on the blessings of nature, and that excessive aquaculture may adversely affect the ecosystem. For this reason, in recent years, fishery rights holders are required to grasp the number of aquaculture facilities actually installed on the sea surface and the spatial arrangement such as latitude and longitude.

For example, according to Article 90 of the revised Fisheries Law, which is scheduled to come into effect in Japan, fishery rights holders are obliged to report the status of utilization of fishing grounds to the prefectural governors. Maritime Domain Awareness (MDA) is also positioned as one of the basic measures in Japan's Ocean Policy, the Third Basic Ocean Plan, and it is required to collect, aggregate and share information in the ocean. It's on the way.

In recent years, the use of remote sensing technology from the sky by artificial satellites and aircraft has become widespread. For example, Patent Document 1 discloses a technique for detecting a target on the sea surface by comparing and collating a satellite image with a digital map.

Japanese Unexamined Patent Publication No. 2000-353234

The aquaculture industry is carried out based on fishery rights, but the number and location information of aquaculture facilities actually installed on the surface of the sea are voluntarily managed by the fishery cooperatives, etc., who are the owners of the fishery rights in each region. The exact number and location information are not known. Many aquaculture facilities are installed in coastal areas, and it is inefficient and impractical to quantitatively grasp aquaculture facilities through regular field surveys.

Although the technique of Patent Document 1 is a technique for detecting a target on the sea surface, it is a technique for generating information related to the target such as the size (area), number, and position information of the detected target. is not it.

The present invention provides an image analysis program, method and apparatus for identifying an object based on an image captured from the sky and generating information related to the identified object.

The present invention for achieving the above object includes, for example, the following aspects.
(Item 1)
On the computer
Object-based image analysis is applied to a near-infrared band image in which a region on the earth including an object is captured to generate a first identification image in which the region of the object is identified in the near-infrared band image. And
Generates relevant information related to the object based on the first identification image.
An image analysis program for executing processing.
(Item 2)
On the computer
The buffer region including the region of the identified object based on the first identification image has a higher resolution than the near-infrared band image and is the same as the earth in the near-infrared band image. Set the upper area to the captured high-definition image,
Object-based image analysis is applied to the high-definition image within the buffer region to generate a second identification image in which the region of the object is identified in the high-definition image.
Based on the second identification image instead of the first identification image, related information related to the object is generated.
Item 4. The image analysis program according to Item 1, for executing the process.
(Item 3)
On the computer
The reference area surrounding the buffer area is set in the high-definition image, and the reference area is set.
For the high-definition image, the object-based image analysis is applied to the high-definition image within the range of the buffer region by using the brightness value of the reference region as a threshold value.
Item 2. The image analysis program according to Item 2, for executing the process.
(Item 4)
Item 2. The image analysis program according to Item 2 or 3, wherein the high-definition image is a panchromatic band image.
(Item 5)
On the computer
Item 4. The image analysis program according to any one of Items 1 to 4, which executes a process of generating information regarding at least one of the size and number of the objects.
(Item 6)
On the computer
Item 2. The image according to any one of Items 1 to 5, wherein a process for generating information regarding the position of the object is executed based on the latitude and longitude information of the imaging range associated with the near-infrared band image. Analysis program.
(Item 7)
On the computer
Item 6. The image analysis program according to any one of Items 1 to 6, wherein a process of applying an image mask to a land region in the near-infrared band image is executed.
(Item 8)
Item 6. The image analysis program according to any one of Items 1 to 7, wherein the object is an object floating in a water area.
(Item 9)
Object-based image analysis is applied to a near-infrared band image in which a region on the earth including an object is captured to generate a first identification image in which the region of the object is identified in the near-infrared band image. First identification image generation step and
A related information generation step of generating related information related to the object based on the first identification image, and a related information generation step.
Image analysis methods, including.
(Item 10)
Object-based image analysis is applied to a near-infrared band image in which a region on the earth including an object is captured to generate a first identification image in which the region of the object is identified in the near-infrared band image. First identification image generation unit and
A related information generation unit that generates related information related to the object based on the first identification image, and a related information generation unit.
An image analysis device.

According to the present invention, it is possible to provide an image analysis device, a method and a program for identifying an object based on an image captured from the sky and generating information related to the identified object.

It is a figure for demonstrating the usage mode of the image analysis apparatus which concerns on one Embodiment of this invention. It is a block diagram for demonstrating the function of the image analysis apparatus which concerns on one Embodiment of this invention. It is a flowchart for demonstrating the procedure of the image analysis method which concerns on one Embodiment of this invention. This is an example of a near-infrared band image of the analysis target region. This is an example of a near-infrared band image and a high-definition image enlarged and displayed for the unit region U. It is a schematic diagram for demonstrating the buffer area and the reference area set in an image when applying an object-based image analysis. This is an example of a first identification image generated for a unit region of a near-infrared band image. This is an example of a second identification image generated for a unit area of a high-definition image. This is an example of an identification image generated for a unit area different from the unit area shown in FIG. 4 in the analysis target area. It is a block diagram for demonstrating the function of the image analysis apparatus which concerns on other embodiment of this invention. It is a flowchart for demonstrating the procedure of the image analysis method which concerns on other embodiment of this invention. It is a figure for demonstrating the result of the field survey which carried out in 2nd Example. This is an example of the result of image analysis performed in the second embodiment. It is a figure for demonstrating the comparison result with the measured value in the field survey performed in the 2nd Example.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and drawings, the same reference numerals indicate the same or similar components, and thus duplicate description of the same or similar components will be omitted.
[Outline of Invention]

FIG. 1 is a diagram for explaining a usage mode of the image analysis device according to the embodiment of the present invention.

The image analysis device 1 according to one embodiment is an device that identifies an object 9 based on an image captured from the sky and generates information 35 related to the identified object 9. In the present embodiment, the object 9 for which information is generated by image analysis is, for example, a farming raft installed in a state of floating on the sea surface in a coastal area. The information 35 generated by the image analysis is information related to the aquaculture raft, such as the size (area), number, and position information (latitude and longitude) of the aquaculture raft.

For image analysis, for example, an optical satellite image taken from the sky of the object 9 by the satellite 8 is used. In this embodiment, the optical satellite images are, for example, a near-infrared band image 31 and a high-definition image 32. The high-definition image 32 has a higher resolution than the near-infrared band image 31. For the high-definition image 32, for example, a panchromatic band image can be used. The near-infrared band image 31 and the high-definition image 32 are paired, and these are images obtained by capturing the same region on the earth including the object 9.

The optical satellite image received from the satellite 8 is stored in, for example, an external server 2, and the image analysis device 1 acquires the optical satellite image via the network N. The image analysis device 1 performs image analysis based on the acquired optical satellite images (near-infrared band image 31 and high-definition image 32), and generates information 35 related to the object 9.
[Device configuration]

FIG. 2 is a block diagram for explaining the function of the image analysis apparatus according to the embodiment of the present invention.

The image analysis device 1 (1A) according to one embodiment includes a data processing unit 12, an auxiliary storage device 13, an input unit 14, a display unit 15, and a communication interface unit (communication I / F unit) 16. ing. The image analysis device 1 can be configured by using a tablet terminal, a smartphone or the like (hereinafter, referred to as a tablet terminal or the like).

In the present embodiment, the image analysis device 1 includes an auxiliary storage device 13, an input unit 14, a display unit 15, and a communication I / F unit 16 as hardware configurations. Although not shown, the image analysis apparatus 1 further includes a processor such as a CPU that performs data processing and a memory that the processor uses for a work area of data processing as a hardware configuration.

The auxiliary storage device 13 is a non-volatile storage device that stores an operating system (OS), various control programs, data generated by the programs, and the like. For example, a flash memory, an eMMC (embedded Multi Media Card), or an SSD. (Solid State Drive) etc. In the present embodiment, the auxiliary storage device 13 stores a near-infrared band image 31, a high-definition image 32, a first identification image 33, a second identification image 34, related information 35, and an image analysis program P. To.

The near-infrared band image 31 is an image obtained by capturing a region on the earth including the object 9. The high-definition image 32 has a higher resolution than the near-infrared band image 31, and is an image obtained by capturing the same region on the earth as the near-infrared band image 31. The near-infrared band image 31 and the high-definition image 32 are paired, and these are images obtained by capturing the same region on the earth including the object 9. In the present embodiment, the near-infrared band image 31 and the high-definition image 32 use an optical satellite image captured from above the object 9 by the satellite 8. In the near-infrared band image 31 and the high-definition image 32 handled in the present embodiment, metadata including information on the scale and latitude and longitude information of the imaging range is recorded. Illustratively, the image data of the optical satellite image of the near-infrared band image 31 and the high-definition image 32 is the data of the luminance value (digital number) represented by 11 bits to 12 bits.

In the present embodiment, the near-infrared band image 31 uses an image in the near-infrared wavelength band created from the multispectral band image. A multispectral band image is an image in which electromagnetic waves of a plurality of wavelength bands such as near infrared, mid-infrared, and thermal infrared are recorded among electromagnetic waves reflected or radiated from an object 9 observed by satellite 8. means. In this embodiment, a panchromatic band image is used for the high-definition image 32. The panchromatic band image means an image in which only the electromagnetic wave of one specific wavelength band among the electromagnetic waves reflected or radiated from the object 9 observed by the satellite 8 is recorded.

The first identification image 33 is an image generated by applying object-based image analysis to the near-infrared band image 31. In the first identification image 33, the region 41 of the object 9 is identified by object-based image analysis.

The second identification image 34 is an image generated by applying object-based image analysis to the high-definition image 32 in which the buffer area 42 is set. In the second identification image 34, the region 44 of the object 9 is identified by object-based image analysis.

The related information 35 is information related to the object 9. In this embodiment, the relevant information 35 includes at least one of the size (area) and number of objects 9. When the near-infrared band image 31 is associated with latitude and longitude information in the imaging range, the related information 35 can include information about the position of the object 9.

The image analysis program P is a computer program for realizing each unit 21 to 28 in the data processing unit 12 described later, which is a functional block by software. The image analysis program P may be installed in the image analysis device 1 via a network such as the Internet connected by the communication I / F unit 16. Alternatively, the image analysis program P is installed in the image analysis device 1 by causing the image analysis device 1 to read a computer-readable non-temporary tangible recording medium such as a memory card on which the image analysis program P is recorded. May be good. The image analysis program P can be an application such as a tablet terminal.

The input unit 14 can be composed of, for example, a mouse or a keyboard, and the display unit 15 can be composed of, for example, a liquid crystal display, an organic EL display, or the like. In the present embodiment, the input unit 14 and the display unit 15 are integrated as a touch panel.

The communication I / F unit 16 transmits / receives data to / from an external device such as an external server 2 via a wired or wireless network. The communication I / F unit 16 may be various wireless or wired connections such as Bluetooth®, Wi-Fi®, and ZigBee®.

In the present embodiment, the image analysis device 1 includes a data processing unit 12 as a software configuration. The data processing unit 12 is a functional block realized by the processor executing the image analysis program P.

The near-infrared band image acquisition unit 21 acquires a near-infrared band image 31 in which a region on the earth including the object 9 is captured.

The high-definition image acquisition unit 22 has a higher resolution than the near-infrared band image 31, and acquires a high-definition image 32 in which the same region on the earth as the near-infrared band image 31 is captured.

The mask application unit 23 applies the image mask to the land region in the near-infrared band image 31.

The first identification image generation unit 24 applies object-based image analysis to the near-infrared band image 31 to generate a first identification image 33 in which the region 41 of the object 9 is identified in the near-infrared band image 31. To do.

Based on the first identification image 33, the buffer area setting unit 25 sets the buffer area 42 including the area 41 of the identified object 9 in the high-definition image 32.

The reference area setting unit 26 sets the reference area 43 surrounding the buffer area 42 in the high-definition image 32.

The second identification image generation unit 27 applies object-based image analysis to the high-definition image 32 within the range of the buffer region 42, and the second identification image in which the region 44 of the object 9 is identified in the high-definition image 32. Generate 34. The second identification image generation unit 27 can apply object-based image analysis to the high-definition image 32 within the range of the buffer region 42 by using the brightness value of the reference region 43 as a threshold value for the high-definition image 32.

The related information generation unit 28 generates related information 35 related to the object 9 based on the second identification image 34. The related information generation unit 28 can generate information regarding at least one of the size and the number of the objects 9. The related information generation unit 28 can generate information regarding the position of the object 9 based on the latitude and longitude information of the imaging range associated with the near-infrared band image 31.
[Processing procedure]

FIG. 3 is a flowchart for explaining the procedure of the image analysis method according to the embodiment of the present invention.

In step S1 (near-infrared band image acquisition step), the near-infrared band image 31 of the analysis target region in which the region on the earth including the object 9 is captured is acquired. In the present embodiment, the near-infrared band image 31 and the high-definition image 32 of the analysis target region including the region on the earth to be analyzed are captured by the satellite 8 in advance and stored in the external server 2, and the images are stored in the external server 2. The analysis device 1 acquires the near-infrared band image 31 from the external server 2.

FIG. 4 is an example of a near-infrared band image of the analysis target region. In the illustrated near-infrared band image 31, the black region in the center is the sea area (water area), and the gray or white region around the sea area is the land area. In the sea area in the center of the drawing, it can be confirmed that a plurality of aquaculture rafts 9 which are objects to be identified by image analysis processing and generate related information 35 are imaged.

In step S2 (image mask application step), the image mask is applied to the land region in the near infrared band image 31. In the present embodiment, since the aquaculture cage 9 floating in the sea area is identified by the image analysis processing, the land area such as the ground, forest, field, paddy field, residential area, etc., which is not the target of the image analysis processing, is used. Apply an image mask. In the image analysis process described below, the image analysis process is not applied to the region in the near-infrared band image 31 to which the image mask is applied.

In step S3 (high-definition image acquisition step), a high-definition image 32 of the analysis target region in which the same region on the earth as the near-infrared band image 31 is captured is acquired. In the present embodiment, the image analysis device 1 acquires a high-definition image 32 paired with the near-infrared band image 31 from the external server 2.

In step S4, the area to be analyzed is divided into a plurality of unit areas U. The unit area U is an area to be used as a processing unit when the image analysis apparatus 1 performs image analysis processing in the following steps S5 to S8.

FIG. 5 is an example of a near-infrared band image and a high-definition image enlarged and displayed for the unit region U. (A) is a near-infrared band image 31, and (B) is a high-definition image 32.

The near-infrared band image 31 and the high-definition image 32 to be analyzed are each divided into a plurality of unit regions U. The near-infrared band image 31 and the high-definition image 32 are paired, and the unit region U is also paired between the near-infrared band image 31 and the high-definition image 32. That is, when performing the image analysis processing of the following steps S5 to S8 on the unit region U at the position illustrated in FIG. 4, the near-infrared band image of the unit region U in which the same region shown in FIG. 5 is imaged. Image analysis processing is performed using 31 and the high-definition image 32.

In step S5 (first identification image generation step), object-based image analysis is applied to the near-infrared band image 31, and the first identification image in which the region 41 of the object 9 is identified in the near-infrared band image 31. Generate 33. In the present embodiment, the threshold value of the brightness value (also called the image value or the digital number in the optical satellite image) used in the object-based image analysis is, for example, the brightness value of the near-infrared band image 31 displayed on the display unit 15. Based on the histogram of, for example, it is arbitrarily determined by the operator. The determined threshold value of the brightness value is input to the image analysis device 1 by the operator, for example, via the input unit 14.

FIG. 6 is a schematic diagram for explaining a buffer area and a reference area set in an image when applying object-based image analysis. FIG. 7 is an example of a first identification image generated for a unit region of a near-infrared band image.

The region 41 is a region of the object 9 identified by applying object-based image analysis to the near-infrared band image 31. In the present embodiment, the region 41 is a set of pixels determined to be the aquaculture raft 9 installed on the sea surface by object-based image analysis. Object-based image analysis is a method of image analysis in which neighboring pixels having similar properties are combined as one object and the combined objects are used as a unit. According to object-based image analysis, it is possible to obtain results close to visual judgment.

By identifying the region 41 in the near-infrared band image 31, the buffer region 42 and the reference region 43 are set in the high-definition image 32 paired with the near-infrared band image 31. The buffer area 42 and the reference area 43 are virtual areas set in the high-definition image 32 when the second identification image 34 is generated. In other words, the buffer region 42 and the reference region 43 are virtual regions for performing a two-step object-based image analysis using the near-infrared band image 31 and the high-definition image 32. That is, the outline of the region of the object 9 is identified by the first-stage object-based image analysis using the near-infrared band image 31, and the two stages within the range of the buffer region 42 using the high-definition image 32. Object-based image analysis of the eye identifies the region of object 9 in detail.

The buffer area 42 is an area including the area 41 of the object 9. In the present embodiment, the buffer area 42 is an area including the area 41 and an area within a range of a predetermined number of pixels (for example, about 2 pixels to 5 pixels) from the outer periphery of the area 41. The reference area 43 is an area surrounding the buffer area 42 and does not include the area 41 and the buffer area 42. In the present embodiment, the reference area 43 is an area within a predetermined number of pixels (for example, about 1 pixel to 5 pixels) from the outer circumference of the buffer area 42.

In step S6 (buffer area setting step), the buffer area 42 including the area 41 of the identified object 9 is set in the high-definition image 32 based on the first identification image 33. Preferably, following this step, the reference area 43 surrounding the buffer area 42 is set in the high-definition image 32.

As described with reference to FIGS. 6 and 7, for each of the plurality of regions 41 identified as the object 9 in the first identification image 33, the buffer region 42 in the paired high-definition image 32. And the reference area 43 is set.

In step S7 (second identification image generation step), the object-based image analysis is applied to the high-definition image 32 within the range of the buffer region 42, and the region 44 of the object 9 is identified in the high-definition image 32. Identification image 34 is generated. In this step, for the high-definition image 32, the object-based image analysis can be applied to the high-definition image 32 within the range of the buffer region 42 by using the brightness value of the reference region 43 as a threshold value.

FIG. 8 is an example of a second identification image generated for a unit region of a high-definition image. Regarding the region of the object 9 to be identified, when the region 41 in the first identification image 33 shown in FIG. 7 and the region 44 in the second identification image 34 shown in FIG. 8 are compared, the region shown in FIG. 8 is compared. It is understood that 44 more accurately identifies the contour of the object 9.

FIG. 9 is an example of an identification image generated for a unit area different from the unit area shown in FIG. 4 in the analysis target area. (A) is the first identification image 33, and (B) is the second identification image 34. Regarding another unit region U shown in FIG. 9, it is understood that the region 44 in the second identification image 34 identifies the contour of the object 9 more accurately.

In step S8 (related information generation step), the related information 35 related to the object 9 is generated based on the second identification image 34. In the present embodiment, the object 9 identified as the region 44 in the second identification image 34 is based on a known image processing technique that counts the number of regions and calculates the size of the region. Information about at least one of the size and number of is generated as related information 35. Referring to FIG. 8, the area of the object 9 (cultured raft 9) calculated based on the size of the region 44 is shown as an example near the region 44 in the second identification image 34. ..

In this step, further, information on the position of the object 9 can be generated as related information 35 based on the latitude and longitude information of the imaging range associated with the near-infrared band image 31. Normally, the near-infrared band image 31 imaged by the satellite 8 records metadata including information on the scale and latitude and longitude information on the imaging range. In the present embodiment, based on this metadata and the position of the area 44 in the second identification image 34 of the unit area U, information regarding the position of the area 44 of the object 9 is generated as the related information 35. ..

In step S9, the processing is repeated from step S5 until all the divided unit areas U are processed (Yes in step S9), and related information 35 related to the object 9 included in the unit area U is generated one after another. ..
[effect]

As described above, according to the image analysis apparatus, method and program according to the embodiment of the present invention, it is possible to identify an object based on an image captured from the sky and generate information related to the identified object. The generated related information can be used as information related to the identified object. For example, when the object is aquaculture raft, the fishery right holder can quantitatively grasp the number of aquaculture facilities actually installed on the sea surface and the spatial arrangement such as latitude and longitude with high accuracy.

In addition, since the identification of the aquaculture raft, which is the object, and the generation of related information are performed based on the image taken from the sky of the object by satellite, etc., the fishery right holder needs to conduct an inefficient field survey. There is no. When an optical satellite image image taken by a satellite is used for image analysis, the optical satellite image is periodically captured by the satellite orbiting the earth's satellite orbit, so that related information can also be generated regularly. Is.

Further, in the near-infrared band image 31, the difference in the brightness value between the aquaculture raft 9 which is the object 9 and the sea surface is greatly recorded. As a result, the range of the threshold value at which a good identification result can be obtained by using the object-based image analysis is expanded, and the threshold value can be easily set.
[Other Embodiments]

Unless otherwise specified, the image analysis apparatus 1 (1B) and the image analysis method according to the other embodiments described below are the same as the image analysis apparatus and the image analysis method according to the above-described embodiment, and thus overlap. The description is omitted.

In the above embodiment, the object-based image analysis is applied to the high-definition image 32 to generate the second identification image 34, and the related information 35 is generated based on the second identification image 34. The procedure for generating is not limited to this. In another embodiment, the image analysis is performed using the near-infrared band image 31 without using the high-definition image 32. That is, in another embodiment, the related information 35 is generated based on the first identification image 33 generated by applying the object-based image analysis to the near-infrared band image 31.

FIG. 10 is a block diagram for explaining the function of the image analysis apparatus according to another embodiment of the present invention. In the image analysis device 1B according to another embodiment, in the image analysis device 1A according to the above embodiment described with reference to FIG. 2, the high-definition image acquisition unit 22, the buffer area setting unit 25, the reference area setting unit 26, The second identification image generation unit 27 is omitted. Further, the related information generation unit 28 generates related information 35 related to the object 9 based on the first identification image 33 instead of the second identification image 34.

FIG. 11 is a flowchart for explaining the procedure of the image analysis method according to another embodiment of the present invention. In the image analysis method according to the other embodiment, steps S3, S6, and S7 are omitted in the image analysis method according to the above embodiment described with reference to FIG. Further, in step S8, the related information 35 related to the object 9 is generated based on the first identification image 33 instead of the second identification image 34. Referring to FIG. 7, the area of the object 9 (cultured raft 9) calculated based on the size of the region 41 is shown as an example near the region 41 in the first identification image 33. ..
[Other forms]

Although the present invention has been described above by the specific embodiment, the present invention is not limited to the above-described embodiment.

In the above embodiment, the image mask is applied to the land area in the near-infrared band image 31, but for example, when it is possible to grasp in advance that the entire area to be processed is the sea area, the image mask is applied. The process can be omitted.

In the above embodiment, the area to be analyzed is divided into a plurality of unit areas U, and image analysis processing is performed for each of the divided unit areas U. However, the analysis target is not divided into a plurality of unit areas U. Image analysis processing may be performed on the region.

In the above embodiment, the near-infrared band image 31 and the high-definition image 32 of the analysis target region are acquired from the external server 2 and the image analysis processing is performed. However, the near-infrared band image 31 and the high-definition image 32 are acquired. Originally, it is not limited to the external server 2. The near-infrared band image 31 and the high-definition image 32 can also be acquired directly from the satellite 8. Further, for example, when the image analysis program of the present invention is installed on the satellite 8 itself, the procedure of acquiring the near-infrared band image 31 and the high-definition image 32 from the outside can be omitted.

In the above embodiment, the image analysis device 1 is realized as an integrated device, but the image analysis device 1 does not have to be an integrated device, and a processor, a memory, an auxiliary storage device 13, and the like are arranged in different places. May be networked to each other. Some or all of the functions of the data processing unit 12 and the data items in the auxiliary storage device 13 may be cloud-based on the external server 2 connected via the communication I / F unit 16.

In the above embodiment, the functional blocks 21 to 28 constituting the data processing unit 12 are realized by software, but some or all of the functional blocks 21 to 28 may be realized as hardware. The processing of the functional blocks 21 to 28 constituting the data processing unit 12 does not have to be processed by a single processor, and may be distributed and processed by a plurality of processors. Further, instead of the processor, an FPGA (Field Programmable Gate Array) may perform the processing, or for example, a GPU (Graphics Processing Unit) may be used as an accelerator to assist the parallel arithmetic processing performed by the processor. That is, the processing performed by the processor means that the processing performed by the processor or FPGA using an accelerator such as a GPU is also included.

In the above embodiment, the image analysis device 1 is configured by using a tablet terminal, a smartphone, or the like, but the image analysis device 1 may be configured by using a general-purpose computer such as a personal computer.

In the above embodiment, the optical satellite image is used for the image analysis, but the image used for the image analysis is not limited to the optical satellite image, and may be an image taken from, for example, an aircraft or a drone. That is, the image used for the image analysis may be an image captured from the sky above the object 9. The aircraft or drone may be equipped with an image pickup device capable of capturing at least two types of images, for example, a near-infrared band image 31 and a high-definition image 32.

In the above embodiment, the panchromatic band image is used for the high-definition image 32, but the high-definition image 32 is not limited to the panchromatic band image. The high-definition image 32 may have a higher resolution than the near-infrared band image 31, and may be an image in which the same region on the earth as the near-infrared band image 31 is captured.

In the above embodiment, the object 9 for which information is generated by image analysis is, for example, a farming raft installed while floating on the sea surface in a coastal area, but the object 9 is an object (floating object) floating in a water area. Not limited to. The object 9 may be, for example, an aircraft or a vehicle stored in a land area such as a desert area as a mothball, and the object 9 may be an object that can be extracted from a background image by object-based image analysis. Just do it.

Examples of the present invention are shown below to further clarify the features of the present invention.
[First Example]

In the first embodiment, the number and area of the object 9 obtained by the image analysis method according to the other embodiment of the present invention (that is, the number and area of the region 41 identified in the near-infrared band image 31). ) Was compared with the measured value in the field survey. Further, the same comparison can be obtained by image analysis of satellite images (RGB images) of other formats and by image analysis of only panchromatic band images without using near infrared band images. I also went to the results.

The target area was Hirota Bay, which is located on the prefectural border between Iwate and Miyagi prefectures. Hirota Bay is a rias-type overseas with a bay opening length of about 4.8 km, a depth of about 9 km, an area of about 37.1 km ² , and a maximum water depth of about 56 m in the bay and at the mouth of the bay. Sea surface aquaculture of oysters, wakame seaweed, scallops, etc. is carried out in the bay. In this example, we identified aquaculture rafts with a length of about 10 m and a width of about 4 m used for oyster cultivation, calculated the number and area of the identified aquaculture rafts, and compared them with the measured values in the field survey. ..

For the image analysis, a multispectral band image with a spatial resolution of 2.0 m, which was observed by the artificial satellite Pleiades1A on April 30, 2015, was used. From this multispectral band image, two types of images, an RGB image (R: band 3, G: band 2, B: band 1) and a near infrared band image (band 4) were created, and these 2 were created. Image analysis was performed for each of the types of images. In addition, image analysis was performed when only a panchromatic band image having a spatial resolution of 0.5 m was used. For image analysis, software ENVI5.5 (Harris GeoSpatial, USA) was used to perform rule-based object-based analysis. As a condition for image analysis, a spectrum excluding the spectrum region of the sea surface was set as a farming raft.

A field survey was conducted on September 29, 2015, and it was confirmed that the aquaculture rafts were regularly arranged in each section. Regarding the spatial arrangement, by visually judging using the panchromatic band image, it was confirmed that 1053 aquaculture rafts were installed in the entire sea area of the target area.

From image analysis using RGB images, 897 farmed rafts were detected. This was a detection rate of about 85% as compared with the result by visual judgment. The average area per aquaculture raft was about 48 m ² . Since the actual area of the aquaculture raft is about 40 m ² , it was detected larger than the actual area in the image analysis using the RGB image. This difference in the size of the area corresponds to ^{the area (about 8.0 m 2} ) of 2 pixels of the spatial resolution of the near-infrared band image by the artificial satellite Pleiades 1A used for the image analysis.

Image analysis using near-infrared band images detected 1030 aquaculture rafts. This was a detection rate of about 98% as compared with the result by visual judgment. In addition, the average area per aquaculture raft was about 49 m ² , and the area was detected to be about 2 pixels larger than the actual area, as in the case of using the RGB image. The reason why the number of aquaculture rafts was detected less than the actual number is that, for example, in aquaculture rafts arranged close to each other and aquaculture rafts where fishing boats were moored, each object is not distinguished. It is considered that the reason is that it is analyzed as. In addition, it is considered that the average area per aquaculture raft was also detected to be larger than the actual area.

Image analysis using only panchromatic band images detected 1319 farmed rafts. This was an increase in detection rate of about 25% compared to the result by visual judgment. The average area per aquaculture raft was about 30 m ² , which was detected smaller than the actual area. The reason why the number of aquaculture rafts was detected more than the actual number is that the spatial resolution (0.5 m) of the panchromatic band image used for image analysis shows that one aquaculture raft is a plurality of small rafts of about 2 to 4 rafts. The reason may be that it was analyzed as an object. In addition, it is considered that the average area per aquaculture raft was also detected to be smaller than the actual area.

As described above, from the first embodiment, when detecting the number of cultivated rafts by object-based image analysis, it is not always necessary to use a panchromatic band image having a high spatial resolution, and a spatial resolution of about a multispectral band image is sufficient. Was shown to be. It was also shown that the detection rate of the aquaculture raft is improved by using the near-infrared band image rather than using the RGB image.
[Second Example]

In the second embodiment, the area of the object 9 obtained by each of the image analysis methods according to the above two types of embodiments of the present invention was compared with the measured value in the field survey. That is, in the second embodiment, the area of the region 41 identified by the first-stage object-based image analysis using the near-infrared band image 31 and the range of the buffer region 42 using the high-definition image 32. Each of the areas of the region 44 identified by the second-stage object-based image analysis in was compared with the measured values in the field survey. The target area was Nakazuraura, Ishinomaki City, Miyagi Prefecture, and aquaculture rafts installed floating on the sea surface of Nakazuraura were detected as objects, and the area was calculated as related information to the objects. The actual measurement of the aquaculture raft in the field survey was conducted on May 21, 2019.

FIG. 12 is a diagram for explaining the results of the field survey conducted in the second embodiment. The RGB image (black and white display) of Nakazuraura, Ishinomaki City, Miyagi Prefecture, which was the target area, is shown in (A), and the area of the aquaculture raft actually measured in the field survey is shown in Table (B). The actual measurement at the site was performed on the aquaculture rafts installed at the points indicated by numbers 1 to 10 in the RGB image of (A).

FIG. 13 is an example of the result of image analysis performed in the second embodiment. (A) is a first identification image, and (B) is a second identification image. The identification images shown in (A) and (B) are enlarged and displayed for the sea area of the aquaculture raft shown by Nos. 6 to 8. Further, in (A), the area of each aquaculture raft generated based on the first identification image is displayed. Similar to (A), (B) also displays the area of each aquaculture raft generated based on the second identification image.

FIG. 14 is a diagram for explaining the comparison result with the measured value in the field survey conducted in the second embodiment. (A) is a graph for comparing the area of the aquaculture raft obtained by each of the above two types of image analysis methods with the measured value of the area of the aquaculture raft in the field survey. (B) is a table showing the difference between the measured value and the area of the aquaculture raft obtained by each of the above two types of image analysis methods. In the table (B), the difference from the measured value is shown for each aquaculture raft installed at the points indicated by numbers 1 to 10.

Referring to the graph of (A), for example, for the aquaculture rafts of No. 6 and No. 10, the value of the area obtained by the image analysis based on the near infrared band image without using the high-definition image is about about the measured value. It was calculated to be as wide as 20%. With reference to the table in (B), the standard deviation of the difference between the measured area and the area obtained by each of the above two types of image analysis methods was calculated. For image analysis based on near-infrared band images, the standard deviation was 11.21 m ² , and for image analysis based on high-definition images, the standard deviation was 3.60 m ² . From this, it was confirmed that the image analysis based on the high-definition image has less variation in the obtained area.

As described above, from the second embodiment, when the number of aquaculture rafts is detected by object-based image analysis and the area of the detected aquaculture rafts is calculated, the near-infrared band image and the high-definition image are used in two steps. It has been shown that it is preferable to perform object-based image analysis.

1 (1A, 1B) Image analyzer 2 External server 8 Satellite 9 Object (aquaculture raft)
12 Data processing unit 13 Auxiliary storage device 14 Input unit 15 Display unit 16 Communication interface unit 21 Near infrared band image acquisition unit 22 High-definition image acquisition unit 23 Mask application unit 24 Identification image generation unit 25 Buffer area setting unit 26 Reference area setting Part 27 Identification image generation part 28 Related information generation part 31 Near infrared band image 32 High-definition image (panchromatic band image)
33 First identification image 34 Second identification image 35 Related information 41 Object area 42 Buffer area 43 Reference area 44 Object area N Network P Image analysis program U Unit area

Claims (10)

On the computer
Object-based image analysis is applied to a near-infrared band image in which a region on the earth including an object is captured to generate a first identification image in which the region of the object is identified in the near-infrared band image. And
Generates relevant information related to the object based on the first identification image.
An image analysis program for executing processing. On the computer
The buffer region including the region of the identified object based on the first identification image has a higher resolution than the near-infrared band image and is the same as the earth in the near-infrared band image. Set the upper area to the captured high-definition image,
Object-based image analysis is applied to the high-definition image within the buffer region to generate a second identification image in which the region of the object is identified in the high-definition image.
Based on the second identification image instead of the first identification image, related information related to the object is generated.
The image analysis program according to claim 1, wherein the process is executed. On the computer
The reference area surrounding the buffer area is set in the high-definition image, and the reference area is set.
For the high-definition image, the object-based image analysis is applied to the high-definition image within the range of the buffer region by using the brightness value of the reference region as a threshold value.
The image analysis program according to claim 2, wherein the process is executed. The image analysis program according to claim 2 or 3, wherein the high-definition image is a panchromatic band image. On the computer
The image analysis program according to any one of claims 1 to 4, which executes a process of generating information regarding at least one of the size and number of the objects. On the computer
The method according to any one of claims 1 to 5, wherein a process for generating information regarding the position of the object is executed based on the latitude and longitude information of the imaging range associated with the near-infrared band image. Image analysis program. On the computer
The image analysis program according to any one of claims 1 to 6, wherein a process of applying an image mask to a land region in the near-infrared band image is executed. The image analysis program according to any one of claims 1 to 7, wherein the object is an object floating in a water area. Object-based image analysis is applied to a near-infrared band image in which a region on the earth including an object is captured to generate a first identification image in which the region of the object is identified in the near-infrared band image. First identification image generation step and
A related information generation step of generating related information related to the object based on the first identification image, and a related information generation step.
Image analysis methods, including. Object-based image analysis is applied to a near-infrared band image in which a region on the earth including an object is captured to generate a first identification image in which the region of the object is identified in the near-infrared band image. First identification image generation unit and
A related information generation unit that generates related information related to the object based on the first identification image, and a related information generation unit.
An image analysis device.

Images