WO2019187277A1 - 情報探索システム及びプログラム - Google Patents
情報探索システム及びプログラム Download PDFInfo
- Publication number
- WO2019187277A1 WO2019187277A1 PCT/JP2018/039469 JP2018039469W WO2019187277A1 WO 2019187277 A1 WO2019187277 A1 WO 2019187277A1 JP 2018039469 W JP2018039469 W JP 2018039469W WO 2019187277 A1 WO2019187277 A1 WO 2019187277A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- information
- subject
- target event
- association
- detection algorithm
- Prior art date
Links
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 212
- 238000001514 detection method Methods 0.000 claims abstract description 204
- 238000001228 spectrum Methods 0.000 claims description 150
- 238000013473 artificial intelligence Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 230000003595 spectral effect Effects 0.000 abstract description 54
- 238000003384 imaging method Methods 0.000 description 151
- 238000000034 method Methods 0.000 description 38
- 230000029553 photosynthesis Effects 0.000 description 38
- 238000010672 photosynthesis Methods 0.000 description 38
- 238000012545 processing Methods 0.000 description 31
- 238000004364 calculation method Methods 0.000 description 25
- 238000000701 chemical imaging Methods 0.000 description 22
- 230000008569 process Effects 0.000 description 17
- 238000005286 illumination Methods 0.000 description 16
- 238000004458 analytical method Methods 0.000 description 13
- 235000013399 edible fruits Nutrition 0.000 description 12
- 238000004891 communication Methods 0.000 description 11
- 230000033001 locomotion Effects 0.000 description 11
- 230000003287 optical effect Effects 0.000 description 10
- 208000005718 Stomach Neoplasms Diseases 0.000 description 9
- 206010017758 gastric cancer Diseases 0.000 description 9
- 239000000203 mixture Substances 0.000 description 9
- 201000011549 stomach cancer Diseases 0.000 description 9
- 206010028980 Neoplasm Diseases 0.000 description 8
- 239000002131 composite material Substances 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 7
- 239000000463 material Substances 0.000 description 7
- 239000002184 metal Substances 0.000 description 7
- 150000003839 salts Chemical class 0.000 description 7
- 238000006243 chemical reaction Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 210000004204 blood vessel Anatomy 0.000 description 4
- 201000011510 cancer Diseases 0.000 description 4
- 239000006185 dispersion Substances 0.000 description 4
- 239000011521 glass Substances 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000007621 cluster analysis Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 2
- 239000000919 ceramic Substances 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000010287 polarization Effects 0.000 description 2
- 210000002784 stomach Anatomy 0.000 description 2
- 238000002834 transmittance Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- KSFOVUSSGSKXFI-GAQDCDSVSA-N CC1=C/2NC(\C=C3/N=C(/C=C4\N\C(=C/C5=N/C(=C\2)/C(C=C)=C5C)C(C=C)=C4C)C(C)=C3CCC(O)=O)=C1CCC(O)=O Chemical compound CC1=C/2NC(\C=C3/N=C(/C=C4\N\C(=C/C5=N/C(=C\2)/C(C=C)=C5C)C(C=C)=C4C)C(C)=C3CCC(O)=O)=C1CCC(O)=O KSFOVUSSGSKXFI-GAQDCDSVSA-N 0.000 description 1
- 206010062129 Tongue neoplasm Diseases 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000011153 ceramic matrix composite Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 229910052729 chemical element Inorganic materials 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 229910000734 martensite Inorganic materials 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 229950003776 protoporphyrin Drugs 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000011347 resin Substances 0.000 description 1
- 229920005989 resin Polymers 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 201000006134 tongue cancer Diseases 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5838—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using colour
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/28—Investigating the spectrum
- G01J3/2823—Imaging spectrometer
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/46—Measurement of colour; Colour measuring devices, e.g. colorimeters
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/27—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands using photo-electric detection ; circuits for computing concentration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/28—Investigating the spectrum
- G01J2003/2866—Markers; Calibrating of scan
- G01J2003/2873—Storing reference spectrum
Definitions
- the present invention provides an information search system suitable for automatically searching for detection algorithm information for acquiring spectrum data necessary for discriminating a target event from a subject and various shooting conditions of a shooting apparatus for shooting the spectrum data. And programs.
- a spectral imaging apparatus that discriminates a desired event with respect to a subject by spectrally analyzing a captured image of the subject for each wavelength.
- a spectral imaging device is a high-wavelength-resolved spectroscopic information (hereinafter referred to as hyperspectral data) that can be spectrally separated over several tens of bands with a wavelength resolution of 0.1 nm to 100 nm in the wavelength range from ultraviolet to visible and further to infrared. ) Can be obtained.
- hyperspectral data high-wavelength-resolved spectroscopic information
- it is possible to analyze, for example, food freshness, building structure defects, plant photosynthesis, chemical elements contained in minerals, moisture and stains on the skin with high accuracy. It becomes. That is, according to the spectrum imaging apparatus, it is possible to detect not only the subject but also the target event in the subject, instead of simply capturing only the subject.
- Patent Documents 1 and 2 Examples of spectral imaging devices that can acquire such hyperspectral data are disclosed in Patent Documents 1 and 2, for example.
- Patent Document 1 discloses a spectrum imaging apparatus that uses a tumor site in a human body as a target event. According to the technology disclosed in Patent Document 1, a tumor site and a non-tumor site are identified by performing detection while focusing on the fluorescence wavelength corresponding to the component accumulated in the cancer cell.
- Patent Document 2 discloses an information processing apparatus for determining whether or not a subject is a fruit.
- a reference feature amount of the fruit is acquired in advance, and it is determined whether or not the subject is a fruit based on a comparison with the feature amount of the spectral image of the actually captured subject. These reference feature amounts are all based on spectrum data.
- the detection algorithm for obtaining the spectrum data of the tumor site in the human body mentioned above pays attention to the fact that protoporphyrin IX accumulated in cancer cells emits fluorescence of 635 nm, and photoprotoporphyrin emits fluorescence of 675 nm. Therefore, a detection algorithm capable of detecting these fluorescences is assembled. In constructing such a detection algorithm, it is necessary to have technical knowledge such as what is the component accumulated in these cancer cells and what wavelength of fluorescence is emitted, and furthermore, only these fluorescence is accurately detected. A lot of time and effort are required for various studies to extract well and identify accurately.
- Patent Documents 1 to 3 do not specifically disclose a technique for acquiring an optimal detection algorithm according to a target event of a subject.
- Patent Documents 1 to 3 do not particularly disclose a technique for updating the detection algorithm based on the latest acquired external environment.
- the detection accuracy can be improved by referring to the spectrum data of the subject imaged under the conventional detection algorithm.
- no technique has been proposed for referring to the spectral data of the subject imaged under such a conventional detection algorithm, including the techniques disclosed in Patent Documents 1 to 3 described above.
- the present invention has been devised in view of the above-described problems, and the object of the present invention is to automatically detect detection algorithm information for obtaining spectrum data necessary for discriminating a target event from a subject. It is an object of the present invention to provide an information search system and program that can be updated based on the latest external environment from which the detection algorithm has been acquired.
- An information search system to which the present invention is applied is an information search system that searches for detection algorithm information of spectrum data necessary for discriminating a target event from a photographed subject.
- a first association database in which three or more levels of first associations are stored in advance, target event input means for inputting information relating to a target event of a subject to be newly determined, and stored in the first association database.
- a search means for searching for one or more detection algorithm information based on information about the target event input through the target event input means with reference to the first association degree, and a photographing terminal that has photographed the subject
- Receiving means for receiving information relating to the target event of the photographed subject, wherein the first association database comprises: And updates the first association degree based on the information received by the serial receiver.
- An information search system to which the present invention is applied is an information search system that searches for detection algorithm information of spectrum data necessary for discriminating a target event from a photographed subject.
- a first association database in which three or more levels of first associations are stored in advance, target event input means for inputting information relating to a target event of a subject to be newly determined, and stored in the first association database.
- a search means for searching for one or more detection algorithm information based on the information related to the target event input through the target event input means with reference to the first association degree, and the detection searched by the search means From the imaging terminal that captured the subject based on the algorithm information, information on the target event of the captured subject
- Receiving means for receiving all the multispectral data, and data restoring means for restoring the hyperspectral data based on the multispectral data received by the receiving means, wherein the first association database is obtained by the data restoring means.
- the first association degree is updated based on the restored spectrum data.
- An information search system to which the present invention is applied is an information search system for searching for shooting conditions of a shooting device for discriminating a target event from a shot subject, and includes three or more stages of each target event of a subject and each shooting condition.
- a search unit that searches for one or more shooting conditions, and a shooting terminal that has shot the subject
- Receiving means for receiving information related to the target event, and the second linkage database is configured to receive the second linkage database based on the information received by the receiving means. And updates the degree.
- An information search system to which the present invention is applied is an information search system for searching for shooting conditions of a shooting device for discriminating a target event from a shot subject, and includes three or more stages of each target event of a subject and each shooting condition.
- a search means for searching for one or more shooting conditions based on information relating to the target event input through the target event input means with reference to the association degree, and a subject based on the shooting conditions searched by the search means
- Receiving means for receiving multispectral data as information relating to a target event of the photographed subject from the photographing terminal that photographed the image, and the receiving means
- Data restoration means for restoring hyperspectral data based on the received multispectral data
- the second association database updates the second association degree based on the spectrum data restored by the data restoration means. It is characterized by doing.
- An information search program to which the present invention is applied is an information search program for searching detection algorithm information of spectrum data necessary for discriminating a target event from a photographed subject.
- An association degree acquisition step for acquiring three or more stages of first association degrees in advance, an objective event input step for inputting information on a target event of a subject to be newly determined, and the first association degree acquired in the association degree acquisition step
- a search step for searching for one or more detection algorithm information based on the information related to the target event input in the target event input step with reference to the association degree, and from the imaging terminal that captured the subject,
- a receiving step for receiving information on the target event, and The flop, characterized in that to perform the updating of the first linkage degree based on the information received in the receiving step to the computer.
- An information search program to which the present invention is applied is an information search program for searching detection algorithm information of spectrum data necessary for discriminating a target event from a photographed subject.
- An association degree acquisition step for acquiring three or more stages of first association degrees in advance, an objective event input step for inputting information on a target event of a subject to be newly determined, and the first association degree acquired in the association degree acquisition step
- a search step for searching for one or more detection algorithm information based on information related to the target event input in the target event input step with reference to the association degree, and a subject based on the detection algorithm information searched in the search step From the shooting terminal that took the picture, Receiving a multi-spectral data as information to be received, and a data restoring step for restoring hyperspectral data based on the multi-spectral data received in the receiving step.
- the data restoring The computer is caused to update the first association degree based on the spectrum data restored in the step.
- An information search program to which the present invention is applied is an information search program for searching for shooting conditions of a shooting device for discriminating a target event from a shot subject, and includes three or more stages of each target event of a subject and each shooting condition.
- the association degree acquisition step for acquiring the second association degree in advance
- the objective event input step for inputting information related to the objective event of the subject to be newly determined
- the second association degree acquired in the association degree acquisition step .
- a reception step for receiving, and in the step of acquiring the association degree, the information is received based on the information received in the reception step. There is characterized by causing the computer to execute updating the second linkage degree.
- An information search program to which the present invention is applied is an information search program for searching for shooting conditions of a shooting device for discriminating a target event from a shot subject, and includes three or more stages of each target event of a subject and each shooting condition.
- the association degree acquisition step for acquiring the second association degree in advance
- the objective event input step for inputting information related to the objective event of the subject to be newly determined, and the second association degree acquired in the association degree acquisition step.
- a search step for searching for one or more shooting conditions based on the information about the target event input in the target event input step, and a shooting terminal that has shot the subject based on the shooting conditions searched in the search step
- a receiving step for receiving multispectral data as information relating to a target event of the photographed subject
- a data restoration step for restoring hyperspectral data based on the multispectral data received in the reception step
- the association degree obtaining step includes the second association degree based on the spectrum data restored in the data restoration step. It is characterized by having a computer perform updating.
- the present invention having the above-described configuration, it is possible to easily acquire optimal detection algorithm information of spectrum data corresponding to a target event of a subject to be determined. For this reason, each time new target events of a subject are born one after another, it is possible to reduce the burden of labor for studying an optimal detection algorithm, and to shorten the time.
- the detection algorithm in the process of updating the detection algorithm, can be optimized by referring to the spectral data of the subject imaged under the conventional detection algorithm. As a result, the detection accuracy can be improved.
- FIG. 6 is a data flow diagram from input of a target event of a subject to acquisition of imaging conditions of the imaging apparatus. It is a figure which shows the example of the information search system which provided the feedback loop.
- FIG. 1 is a block diagram showing the overall configuration of an information search system 1 to which the present invention is applied.
- the information search system 1 searches for detection algorithm information to be provided to the spectrum imaging device 4.
- the information search system 1 is connected to the algorithm database 3, the search device 2 connected to the algorithm database, and the search device 2.
- a spectral imaging device 4 and an imaging device 5 are provided.
- the algorithm database 3 is a database related to detection algorithm information to be provided to the spectral imaging device 4.
- the algorithm database 3 is a database related to shooting conditions of the shooting device 5.
- the algorithm database 3 stores information sent via a public communication network or information input by a user of this system.
- the algorithm database 3 transmits the accumulated information to the search device 2 based on a request from the search device 2.
- the search device 2 is composed of electronic devices such as a personal computer (PC), for example.
- the search device 2 is embodied in any other electronic device such as a mobile phone, a smartphone, a tablet terminal, and a wearable terminal. It may be made.
- FIG. 2 shows a specific configuration example of the search device 2.
- the search device 2 performs wired communication or wireless communication with a control unit 24 for controlling the entire search device 2 and an operation unit 25 for inputting various control commands via operation buttons, a keyboard, and the like.
- the internal bus 21 is connected to a display unit 23 as a monitor for actually displaying information.
- the control unit 24 is a so-called central control unit for controlling each component mounted in the search device 2 by transmitting a control signal via the internal bus 21. Further, the control unit 24 transmits various control commands via the internal bus 21 in accordance with an operation via the operation unit 25.
- the operation unit 25 is embodied by a keyboard or a touch panel, and an execution command for executing a program is input from the user.
- the operation unit 25 notifies the control unit 24 of this.
- the control unit 24 executes a desired processing operation in cooperation with each component including the search unit 27.
- the search unit 27 searches for detection algorithm information of spectrum data necessary for discriminating a target event from a subject imaged by the spectrum imaging device 4.
- the search unit 27 reads various information stored in the storage unit 28 as necessary information and various information stored in the algorithm database when executing the search operation.
- the search unit 27 may be controlled by artificial intelligence. This artificial intelligence may be based on any known artificial intelligence technology.
- the display unit 23 includes a graphic controller that creates a display image based on control by the control unit 24.
- the display unit 23 is realized by, for example, a liquid crystal display (LCD).
- the storage unit 28 When the storage unit 28 is composed of a hard disk, based on the control by the control unit 24, predetermined information is written to each address and is read out as necessary.
- the storage unit 28 stores a program for executing the present invention. This program is read by the control unit 24 and executed.
- FIG. 3 shows a configuration example of the spectrum imaging device 4.
- the spectrum imaging device 4 is configured by a so-called multispectral camera, a camera with a color filter exchange method, or a camera with a prism.
- the spectrum imaging device 4 captures a subject and further acquires a spectral image therefrom.
- the spectral imaging device 4 generates a spectral image based on three-dimensional spectral data having two-dimensional spatial information and one-dimensional wavelength information from the two-dimensional spectral data at each photographing position.
- the spectral image generated by the spectral imaging device 4 is composed of a plurality of two-dimensional images indicating the reflectance or transmittance of the subject for each wavelength.
- a wavelength resolution of 0.1 nm to 100 nm may be set in a wavelength range of a predetermined wavelength range of 200 nm to 13 ⁇ m, and a spectral image for each band.
- the wavelength range in the spectral image captured by the spectral imaging device 4 includes not only the visible light region but also light in the infrared region, near infrared region, and ultraviolet region.
- the spectral imaging apparatus 4 includes an objective lens 41 that takes in light emitted by the imaging target itself, light reflected or transmitted by the subject 10, that is, an imaging lens L from the subject 10, and a Y-axis direction in a three-axis orthogonal coordinate system including XYZ.
- a precision linear motion stage 42 that moves in the direction of Z, a slit plate 43 for disposing a slit opening 43a provided in the Z-axis direction on the image plane of the objective lens 41, and a light beam that has passed through the slit opening 43a as parallel light.
- the image sensor 47, the precision linear motion stage 42 and the image sensor 47 are controlled, and image data received via the image sensor 47 And a control unit 48 for performing a seed treatment.
- the spectrum imaging device 4 may use the technology disclosed in JP-A-2015-166682.
- the precision linear motion stage 42 moves the slit plate 43, the collimating lens 44, the dispersion optical element 45, the imaging lens 46, and the imaging element 47 integrally in the Y-axis direction under the control of the control unit 48.
- the dispersion optical element 45 is embodied by, for example, a diffraction grating, a prism, or the like.
- the dispersion optical element has a function of dispersing the light beam passing through the collimating lens 44 into components for each wavelength.
- the image sensor 47 is constituted by, for example, a CCD image sensor, a CMOS image sensor, or the like.
- the imaging element 47 converts light imaged on the imaging surface into an electric signal by photoelectric conversion. Then, the electrical signal converted by the image sensor 47 is transmitted to the control unit 48. If light in the infrared region, near-infrared region, and ultraviolet region is received, an image sensor 47 suitable for the light is provided.
- FIG. 4 shows a further detailed configuration of the control unit 48.
- the control unit 48 includes an imaging control unit 481 that controls the timing at which an electric signal is acquired by the image sensor 47, and a movement control unit 482 that controls the movement direction, movement amount, and movement timing of the precision linear motion stage 42 in the Y-axis direction.
- a spectral data creation unit 483 that creates spectral data based on the electrical signal from the image sensor 47, and an image processing unit 484 that performs various image processing, calibration, and the like based on the spectral data created by the spectral data creation unit 483.
- some or all of the components of the control unit 48 may be mounted in an independent personal computer (PC).
- PC personal computer
- the spectral data creation unit 483 creates two-dimensional spectral data having one-dimensional spatial information and one-dimensional wavelength information based on the electrical signal transmitted from the image sensor 47, and stores this.
- the spectral data generation unit 483 repeatedly executes these processes, and when imaging at all imaging positions is completed, obtains a hyperspectral image made up of three-dimensional spectral data having two-dimensional spatial information and one-dimensional wavelength information. It becomes possible.
- the image processing unit 484 converts the spectral image for each wavelength created by the spectral data creation unit 483 into a predetermined color system, performs color calculation processing, and generates a color analysis image. Further, the image processing unit 484 performs processing for displaying the generated color analysis image by a predetermined display method.
- the image processing unit 484 includes a calibration processing unit 484-1, a calculation unit 484-2, and a color analysis image acquisition unit 484-3.
- This calibration processing unit 484-1 performs noise removal due to dark current, sensitivity deviation correction processing between pixels, luminance calibration processing, correction of illumination unevenness of light source light in the space, and the like.
- the calculating unit 484-2 calculates each spectral radiance, each spectral luminance, and the like in the spectral image for each wavelength processed by the calibration processing unit 484-1.
- the color analysis image acquisition unit 484-3 is a standard set using various parameters calibrated by the calibration processing unit 484-1, each spectral radiance, each spectral luminance, etc. calculated by the calculation unit 484-2. A color space conversion process is performed for conversion to the color system.
- the color analysis image subjected to the color space conversion processing in the color analysis image acquisition unit 484-3 is sent to a PC or the like (not shown) and drawn on a display or the like.
- FIG. 5 shows a block configuration example of the photographing apparatus 5.
- the imaging device 5 includes a general digital camera, a multispectral camera, and any digital camera mounted on a mobile phone, a smartphone, a tablet terminal, or a wearable terminal.
- the spectral imaging device 4 can detect spectral data in all bands, while the imaging device 5 is limited to a predetermined wavelength region in addition to normal visible light imaging. It is intended to detect.
- the photographing device 5 includes an imaging optical system 51, a filter 52, an image sensor 53, and a signal processing unit 54.
- the imaging optical system 51 has at least one imaging lens 56 and collects light from the subject 10 to form an image on the imaging surface of the imaging element 53.
- the filter 52 is disposed between the subject 10 and the imaging lens 56.
- the filter 52 is disposed on the path of light that reaches the image sensor 53.
- the filter 52 is an element having a predetermined spectral transmittance. That is, the filter 52 functions to transmit only light in a preset wavelength region and reflect light in other wavelength regions.
- the type of the filter 52 is selected according to the wavelength and wavelength width of light that is actually desired to be transmitted.
- the filter 52 will be described as an example in which the filter 52 is fixedly arranged in advance in the photographing apparatus 5, but is not limited thereto. That is, the filter 52 may be configured to be able to sequentially switch a plurality of filters 52 having different wavelength regions that transmit each other.
- the image sensor 53 is configured by a CCD image sensor, a CMOS image sensor, or the like.
- the imaging element 53 converts light imaged on the imaging surface into an electrical signal by photoelectric conversion. Then, the electrical signal converted by the image sensor 53 is transmitted to the signal processing unit 54.
- the signal processing unit 54 is a circuit that processes an electrical signal sent from the image sensor 53.
- the signal processing unit 54 generates a spectrally separated image separated for each wavelength range of light from the subject 10 based on the image acquired by the image sensor 53.
- the signal processing unit 54 may perform various focus controls based on the acquired electrical signal.
- the search device 2 searches for detection algorithm information to be provided to the spectral imaging device 4 and the imaging device 5 or detection algorithm information to be provided to the imaging device 5.
- This search process starts when the user himself / herself inputs a target event of a subject to be newly photographed by the spectral imaging device 4 or the photographing device 5.
- the subject here is a generic term for objects actually photographed by the spectrum imaging device 4 or the imaging device 5, and the target event is an object to be discriminated through the spectrum imaging device 4 or the imaging device 5. Or mean things. For example, when it is desired to distinguish only salt from a mixture of salt and sugar, the subject is a mixture and the target event is salt.
- the subject when it is desired to discriminate only oil from a mixture of water and oil, the subject is a mixture and the target event is oil.
- the target event is oil.
- the subject when it is desired to determine the freshness of sushi, the subject is sushi and the target event is freshness.
- the subject when it is desired to discriminate a spot on the face, the subject is a face and the target event is a spot.
- stomach cancer when it is desired to discriminate stomach cancer from the stomach, the subject is the stomach and the target event is stomach cancer.
- the user manually inputs the target event of the subject through the operation unit 25.
- text data of a target event of a subject created in an electronic device such as another portable terminal or a PC may be input via the Internet.
- the target event of the subject transmitted or input in this way is stored in the storage unit 28.
- the information search program performs wording analysis on the target event of the subject input in step S11 and stored in the storage unit 28. (Step S12).
- wording analysis any existing text mining technology, data mining technology, language analysis processing technology, or the like may be used.
- this information search program extracts the character string of the target event of the subject to be analyzed from any grammatical structural unit such as word, morpheme, phrase, clause, etc., in any one or more units.
- any grammatical structural unit such as word, morpheme, phrase, clause, etc.
- text data “foot blood vessels” is input as the target event of the subject
- character strings such as “foot” and “blood vessels” are extracted
- text data “face moisture” is input.
- a character string such as “face” or “moisture” is extracted.
- the information search program specifies the subject and the target event from the extracted character string.
- the subject is “foot” and “face”
- the target events are “blood vessel” and “water”. Normally, since the character string constituting the subject is often before the character string constituting the target event, the subject and the target event are specified from the beginning of the extracted character string.
- the user classifies and inputs “foot” as the subject and “blood vessel” as the target event.
- the input subject and the character string of the target event are accepted as they are.
- the information search program moves to step S13 and searches for detection algorithm information having a high degree of association with the character string extracted in step S12.
- the algorithm database 3 stores a reference target event (hereinafter referred to as a reference target event) and three or more levels of relevance (hereinafter referred to as a first target event) classified into two or more types of detection algorithm information.
- the degree of association is acquired in advance.
- the detection algorithm information is an algorithm for detecting spectrum data necessary for judging a target event even if the subject is actually imaged by the spectrum imaging device 4 or the imaging device 5.
- the spectrum intensity (reflectance) changes greatly when the spectrum intensity is in the wavelength range of 500 nm to 700 nm.
- it is possible to determine the freshness of the fruit by creating a spectral image in the wavelength range of 500 nm to 700 nm.
- Specified as a characteristic wavelength is any wavelength range in which such a target event can be identified.
- any one of the wavelength ranges from 500 nm to 700 nm is specified as the characteristic wavelength.
- the characteristic wavelength may be specified by one point or a plurality of characteristic wavelengths.
- 600 nm which is the central wavelength in the above wavelength range (500 nm to 700 nm) may be used, or a wavelength at which the difference value of the spectrum intensity between the spectra is the largest.
- a convex peak is formed in each spectrum data at a wavelength of about 650 nm.
- such a singular point may be specified as a characteristic wavelength. This characteristic wavelength may be different for each target event of the subject.
- a characteristic wavelength range centered on this characteristic wavelength is set.
- the characteristic wavelength range is configured with a predetermined wavelength range set in advance, such as ⁇ 10 nm. Therefore, if the characteristic wavelength is 500 nm and the characteristic wavelength range is ⁇ 10 nm, the actual spectrum data detection range is 495 to 505 nm. This characteristic wavelength range may be different for each target event of the subject.
- the detection algorithm information may include various calculation methods.
- the characteristic wavelength and characteristic wavelength range as the individual explanatory variables x1, x2,.
- the algorithm database 3 stores such characteristic wavelengths, characteristic wavelength ranges, calculation methods, and calculation formulas that define the calculation wavelengths in association with each other for each target event for reference.
- the algorithm database 3 may be defined based on three or more levels of first associations between the target event for reference of the subject and the detection algorithm information.
- FIG. 8 shows a network in which the target event for reference of the subject and the detection algorithm information are related to each other by three or more levels of first association.
- the freshness of the fruit is the first relation when the characteristic wavelength and the characteristic wavelength range as the detection algorithm information are 970 ⁇ 10 nm, and the first relation is 80%, 1170 ⁇ 10 nm, and 880 ⁇ 15 nm.
- the first association degree is 40%, and when the calculation method is 455 ⁇ 12 nm, the first association degree is 20%. It is shown that there is.
- the calculation method is linear at three wavelengths of 630 ⁇ 5 nm, 750 ⁇ 10 nm, and 1250 ⁇ 5 nm as the characteristic wavelength and characteristic wavelength range as detection algorithm information, the moisture content of the hair is 80% first relevance, 970 ⁇ It is shown that the first relevance is 20% when the thickness is 10 nm.
- the characteristic wavelength and characteristic wavelength range as detection algorithm information is 970 ⁇ 10 nm
- the first association is 20%, 230 ⁇ 12 nm, 400 ⁇ 5 nm
- the calculation method is K-means.
- the first association degree is 40%
- the first association degree is 80% when the calculation method is cluster analysis at three wavelengths of 547 ⁇ 4 nm, 588 ⁇ 10 nm, and 939 ⁇ 5 nm.
- This first association degree may be configured by a so-called neural network.
- the first relevance is selected for determining the target event of the subject, in other words, the compatibility of the detection algorithm information selected for determining the target event of the subject via the spectrum imaging device 4 or the imaging device 5. It shows the accuracy of the detection algorithm information.
- the detection algorithm for detecting the freshness of the fruit is most compatible when the detection algorithm is 970 ⁇ 10 nm, and that the discrimination can be performed most effectively and with high accuracy. ing.
- the compatibility is 2 wavelengths of 1170 ⁇ 10 nm and 880 ⁇ 15 nm
- the calculation method is cluster analysis with 3 wavelengths of 547 ⁇ 4 nm, 588 ⁇ 10 nm and 939 ⁇ 5 nm. In some cases, this is shown to follow in the order of 455 ⁇ 12 nm.
- the notation method of the target event of the subject is not limited to the above.
- an object is a composite in which a plastic material is sandwiched between glass plates as an object, and a scratch in the plastic material is used as a reference event.
- This composite may be composed of, for example, a laminate in which a metal and a resin are laminated in a plurality of layers, or may be composed of a mixture mixed with each other such as sugar and salt. Good.
- it may be a complex body such as a ceramic matrix composite material in which ceramics is used as a base material and whiskers are added as a second layer.
- a foreign object in a composite composed of a metal and a foreign object may be used as a reference event.
- one of the complexes is the reference event for reference.
- the subject may be composed of a composite made of three or more of glass, plastic material, and ceramics, for example. Reference purpose events will be defined for each of these complexes.
- the detection algorithm information is linked to this via three or more levels of association. If the subject is composed of a composite of metal and foreign matter, in addition to the characteristic wavelength of the metal, the feature wavelength of the foreign matter and the characteristic wavelength that constitutes the detection algorithm information are taken into account, and the target event for reference is extracted from them. In this case, a suitable condition is examined in advance, and this is linked as the degree of association.
- the metal may be a mixed crystal state before and after the martensitic transformation, for example, and each phase may be a target event for reference.
- the subject itself is composed of a single-phase material instead of a composite, but the phase after the change when the single-phase material changes in time series is regarded as a reference event. You may do it.
- the information search program After shifting to step S13, the information search program performs an operation of selecting one or more detection algorithm information from the character string constituting the target event of the subject extracted in step S12.
- the detection algorithm information from the character string constituting the target event of the subject extracted in step S12 refer to the first association degree between the target reference event of the subject and the detection algorithm information shown in FIG. To do.
- the target event of the subject extracted in step S12 is “leaf photosynthesis”, 1357 ⁇ 10 nm having a high first association degree with the “leaf photosynthesis” when the first association degree is referred to.
- detection algorithm information that is not connected with an arrow may be selected.
- the calculation method is linear at 1357 ⁇ 10 nm, which is the first association with “photosynthesis of leaves” as a reference event for the subject, and at 630 ⁇ 5 nm, 750 ⁇ 10 nm, and 1250 ⁇ 5 nm.
- the detection algorithm information that is most suitable for the case where there is a case where the calculation method is linear at 630 ⁇ 5 nm, 750 ⁇ 10 nm, 1250 ⁇ 5 nm, and 970 ⁇ 10 nm, which have the first degree of association with “moisture of hair” May be estimated.
- 630 ⁇ 5 nm which is the first degree of association common to each other, may be estimated as detection algorithm information of “moisture of leaf”, or “photosynthesis of leaf”, “moisture of hair” Among them, all those having a first association degree of 40% or more may be estimated as detection algorithm information. In addition, for all detection algorithms having “photosynthesis of leaves”, “moisture of hair” and the first association degree exceeding 0%, the wavelengths weighted and averaged by the respective first association degrees are estimated as detection algorithm information. It may be.
- the target event of the subject extracted in step S12 is “tongue cancer”, such an item does not exist in the reference target event of the subject.
- “cancer” “stomach cancer” exists as a target event of the past subject, but “tongue” does not exist as a reference event for the subject.
- estimation may be made based on the past detection algorithm information of “stomach cancer”, or if there is past detection algorithm information about the “lips” or the like of the part close to the “tongue” May be estimated with reference to it.
- the detection algorithm information when the first association shown in FIG. The case where the wavelength is 230 ⁇ 12 nm and 400 ⁇ 5 nm and the calculation method is K-mens is preferentially selected.
- the subject extracted in step S12 is, for example, “paper” and the extracted target event is “foreign matter”, the subject that matches this even in the light of the first association shown in FIGS.
- the “foreign matter” as the reference event is present when the subject is a mixture of “metal” and “foreign matter”.
- the subject may be a mixture of “metal” and “foreign matter”, and detection algorithm information having a low first association degree when the reference target event is “foreign matter” may be selected. .
- the selection of the detection algorithm information is not limited to the case where the first association degree is selected in descending order, but is selected in the order of the first association degree from the lowest according to the case. Alternatively, any other priority order may be selected.
- the selection method of the detection algorithm information for the target event of the subject extracted in step S12 is not limited to the above-described method, and may be executed based on any method that refers to the first association degree. It may be. Further, the search operation in step S13 may be performed using artificial intelligence. In such a case, the first association degree may be regarded as a neural network.
- step S14 the process proceeds to step S14, and the selected detection algorithm information is displayed via the display unit 23.
- the user can immediately grasp the detection algorithm information corresponding to the target event of the subject to be determined by visually recognizing the display unit 23.
- the user sets the detection algorithm of the image processing unit 484 in the spectrum imaging device 4 or sets the detection algorithm of the imaging device 5 based on the output detection algorithm information.
- the detection algorithm is set by performing color calculation processing (hereinafter, characteristic wavelength calculation) based on the characteristic wavelength in addition to the characteristic wavelength and characteristic wavelength range. For example, when the target event of the subject is “leaf photosynthesis” and 1357 ⁇ 10 nm is selected as the detection algorithm, the characteristic wavelength calculation for displaying red is performed for pixels included in the wavelength range, For the pixels not included in the wavelength, the spectral imaging device 4 and the imaging device 5 are set so as to perform a characteristic wavelength calculation for displaying white.
- the spectral image capturing device 4 or the image capturing device 5 captures the “leaf” as the subject, thereby detecting the spectral data necessary for determining “photosynthesis” as the target event, and using this as the color analysis image. Can be displayed.
- the present invention it is possible to easily obtain optimal detection algorithm information of spectrum data according to the target event of the subject to be determined by the spectrum imaging device 4 or the imaging device 5.
- the burden of labor for studying the optimal detection algorithm can be reduced, and the time can be shortened.
- the information search system 1 to which the present invention is applied is characterized in that the optimum detection algorithm information is searched through the first association degree set in three or more stages.
- the first degree of association can be described by a numerical value of, for example, 0 to 100%, but is not limited to this, and it can be described by any level as long as it can be described by a numerical value of three or more levels. Good.
- the present invention it is possible to make a determination without overlooking detection algorithm information that is extremely low such as 1% of the first association degree. Even if the detection algorithm information has a very low degree of first association, it is connected as a small sign, and it may be useful as detection algorithm information tens or hundreds of times. You can call attention.
- a search policy can be determined in a manner of setting a threshold by performing a search based on the first degree of association of three or more stages. If the threshold value is lowered, even if the first relevance is 1%, it can be picked up without omission, but a lot of detection algorithm information that has a low possibility of detecting the target event of the subject is picked up. In some cases. On the other hand, if the threshold value is increased, only detection algorithm information that is highly likely to be able to suitably detect the target event of the subject can be narrowed down, but detection algorithm information that displays a suitable solution once every tens or hundreds of times. May be overlooked. It is possible to decide which to place importance on the basis of the idea on the user side and the system side, but it is possible to increase the degree of freedom in selecting points to place such emphasis.
- the first association degree described above may be updated. That is, the target event for reference of the subject as shown in FIG. 8 and the detection algorithm information are updated as needed.
- This update may reflect information provided via a public communication network such as the Internet.
- the first degree of association is determined according to the knowledge. Is raised or lowered. For example, when a detection algorithm having a certain degree of first association with a target event for a certain subject can be detected with high accuracy through many sites on the public communication network, it is set between them. The first degree of association is further increased.
- the update of the first degree of association is performed by the system side or the user side based on the contents of research data, papers, conference presentations, newspaper articles, books, etc. by experts other than based on information obtainable from the public communication network. It may be updated artificially or automatically. Artificial intelligence may be used in these update processes.
- the search device 2 may search for shooting conditions to be provided to the shooting device 5.
- the information search program searches for an imaging condition having a high degree of association with the acquired target event.
- the algorithm database 3 obtains in advance the second association degree between the reference purpose event and the imaging condition as shown in FIG.
- the imaging conditions referred to here are information on illumination light including the wavelength, illumination angle, brightness, conditions of a polarization filter provided in the illumination light, and spectrum imaging at the time of imaging by the spectrum imaging device 4 or the imaging device 5.
- Various imaging system parameters such as black balance and gain, and hardware parameters are also included.
- the detection algorithm information in addition to the above-described characteristic wavelength, characteristic wavelength range, and calculation method, the above-described parameters may be added. Each parameter described above may be defined as one condition for obtaining the above-described characteristic wavelength or characteristic wavelength range.
- the ranking is classified into several types such as 96-120 dpi, 120-144 dpi, 144-192 dpi.
- the first association degree may be associated with each rank.
- such a photographing condition includes “white balance OO”, a combination of “lens arrangement P” and “filter W”, “filter Q”, “filter R” and “illumination light”.
- a combination of “angle OO °”, a combination of “filter S” and “spatial resolution 133-140 dpi”, “exposure time XX ns or more”, “exposure time XX ns”, or the like is set.
- This imaging condition may be composed of a combination of a plurality of factors, or may be composed of a single factor. Also, the same shooting condition of “exposure time” is classified into “exposure time of OOns or more” and “exposure time of less than OOns”, and the second association degree is associated with each. Good.
- the “filter Q” has, for example, a transmission wavelength of the filter of 600 to 650 nm
- the “filter R” has, for example, a transmission wavelength of the filter of 340 to 400 nm
- the “filter S” has, for example, the transmission wavelength of the filter.
- Detailed conditions such as 1000 to 1100 nm are assigned.
- the object reference event of the subject and the shooting conditions are associated with each other through the first degree of association.
- filter Q is associated with “fruit freshness” at a second association degree of 80%
- white balance OO is associated with a second association degree of 20%.
- the “hair moisture” is associated with the combination of “filter S” and “spatial resolution 133-140 dpi” at the second association degree of 100%, and for the combination of “lens arrangement P” and “filter W”.
- the second association is 40%.
- “Leaf photosynthesis” is associated with “white balance OO” with a second association degree of 60%, and “less than exposure time OOns” with a second association degree of 20%.
- “Stomach cancer” is associated with the combination of “lens arrangement P” and “filter W” with a second association degree of 80%, and the second association with “filter S” and “spatial resolution 133-140 dpi”. Associated with a degree of 40%.
- “Glass defect” is associated with “filter R, illumination light angle OO °” with a second association degree of 100%, and with “exposure time of OOns or more” with a second association degree of 20%. Associated.
- This second relevance is the compatibility of the photographing conditions in the photographing apparatus 5 in making a determination based on each reference purpose event.
- the reference purpose event and thus the object purpose to be determined by this
- the combination of “lens arrangement P” and “filter W” is the most compatible as the imaging condition for “stomach cancer”, so that the most effective and highly accurate discrimination is possible. It has been shown to be. It is shown that the imaging condition for “stomach cancer” is followed by “filter S” and “spatial resolution 133-140 dpi”.
- the above-described reference event is stored in association with each other through the second association degree for each of these imaging conditions.
- the information search program may refer to the second relevance shown in FIG. 10 when searching for a shooting condition highly compatible with the newly input target event. For example, when the newly input target event is “hair moisture”, when the second association degree described above is referred to, the reference detection algorithm information corresponding to the second association degree and the “filter” having the second association degree are high. “S” and “Spatial resolution 133-140 dpi” are selected as shooting conditions. A combination of the “lens arrangement P” and the “filter W”, which has a low second association degree but is slightly recognized, may be selected as an imaging condition. Similarly, when the newly input target event is “stomach cancer”, a combination of “lens arrangement P” and “filter W” is selected.
- the selection of the shooting condition is not limited to the case where the second association degree is selected in order from the highest second association degree, and the second association degree is low depending on the case.
- the items may be selected in order, or may be selected in any other priority order.
- FIG. 11 shows an example in which shooting conditions for a combination of a reference event and a shooting condition for reference are associated with each other by three or more second relevance levels.
- the reference shooting conditions are configured by the same items as the above-described shooting conditions.
- the second degree of association shown in FIG. 11 is an example in the case where a part of the imaging condition is input as known information in addition to the target event via the operation unit 25. In other words, the target event and part of the shooting conditions have already been determined, but the remaining shooting conditions cannot be determined.
- reference event and reference shooting conditions are arranged on the left side through the second association degree, and shooting conditions to be actually searched are arranged on the right side through the second association degree. Yes.
- the node of these combinations has “white balance OO” having a first association degree of 70% and “exposure” “Time xx ns or more” is assumed to be the second association 40%.
- the reference shooting condition is “filter S”, “shutter speed XX seconds”, and the reference target event is “leaf photosynthesis”, the node of these combinations is “white balance XX”. Is the second association degree 60%, and “less than the exposure time OOns” is the second association degree 40%.
- the information search program displays the selected shooting conditions via the display unit 23 of the search device 2.
- the user can immediately grasp the imaging conditions corresponding to the detection algorithm information by visually recognizing the display unit 23.
- Artificial intelligence may also be used for the shooting condition search operation. That is, the second association degree may be configured by a neural network.
- the user designs the imaging optical system 51, the filter 52, the imaging element 53, the signal processing unit 54, and the like in the imaging device 5 based on the output imaging conditions, or sets the illumination light conditions. Alternatively, various conditions regarding photographing are determined. In addition, the user designs each component of the spectrum imaging device 4 or determines each condition based on the output photographing condition.
- the known photographing conditions may be automatically extracted.
- the means for extracting the photographing conditions is, for example, a device that reads and analyzes electronic data of manuals related to the spectrum imaging device 4 and the photographing device 5 to be used and information posted on the Internet via a text mining technique. It may be implemented by a PC or the like. Information relating to the shooting conditions may be extracted from the analyzed information and input as the above-described known shooting conditions.
- a device for measuring the actual exposure time by the spectral imaging device 4 or the imaging device 5 may be used, or the spectral imaging device 4 or The photographing apparatus 5 may be directly connected to the PC and the set exposure time may be read.
- FIG. 12 shows a data flow from input of a target event of a subject to acquisition of shooting conditions of the shooting device 5.
- the input information includes the target event of the subject, the wavelength of the illumination light that is the illumination system parameter, the illumination angle of the illumination light that illuminates the subject, the brightness of the illumination light, and the spectrum that is the imaging system parameter. These are the wavelength range, wavelength resolution, spatial resolution, spectral wavelength sensitivity, polarization filter, and the like of the imaging device 4 and the imaging device 5.
- hardware parameters may also be input.
- the detection algorithm information including the characteristic wavelength and the characteristic wavelength range and the photographing condition are searched by referring to the first association degree described above.
- the detection algorithm information obtained in this way refers to past data stored in the algorithm database 3 based on illumination system parameters and imaging system parameters in addition to the input object event of interest. Then, algorithm information and photographing conditions that are most suitable for photographing the subject by the spectral imaging device 4 and the photographing device 5 are selected.
- step S11 instead of receiving an input of a target event of a subject, an input of detection algorithm information may be received.
- step S13 by referring to the first association degree based on the detection algorithm information that has received this input, the object event of the subject is searched reversely. That is, the relationship between the input and the output in FIGS. 8 and 9 is replaced with the above-described mode, the input is the detection algorithm information, and the output is the target event of the subject.
- the salt in the mixture is the target event of the subject. It is possible to determine that there is a high possibility.
- the information search system 1 to which the present invention is applied may be provided with a feedback loop as shown in FIG.
- the algorithm database 3 receives information from the spectral imaging device 4 and the imaging device 5 that photographed the subject 9.
- the spectrum imaging device 4 and the imaging device 5 acquire detection algorithm information searched by the search device 2 based on the method described above.
- the spectrum imaging device 4 and the imaging device 5 actually capture the subject 9 based on the detection algorithm information.
- the subject 9 as the shooting target corresponds to the subject 9 input in S11, and the shooting purpose corresponds to detecting the target event input in step S11. That is, it is assumed that the detection of the photosynthesis of the leaf is input as the target event of the subject in step S11 described above, and the detection algorithm searched in step S13 is suitable for detecting the photosynthesis of the leaf. In such a case, the detected detection algorithm is input to the spectral imaging device 4 and the imaging device 5. Then, the leaves as the subject 9 are photographed through the spectrum imaging device 4 and the photographing device 5 by the input detection algorithm, and an attempt is made to detect photosynthesis.
- the spectrum imaging device 4 and the imaging device 5 transmit the spectrum data obtained by imaging the subject 9 to the algorithm database 3.
- the algorithm database 3 updates the first association degree based on such spectrum data.
- a detection algorithm having a characteristic wavelength of 1357 ⁇ 10 nm is searched for the photosynthesis of the object 9 as a target event, and based on this, a leaf as the object 9 is actually photographed. If the spectral data is suitable for expressing leaf photosynthesis, the detection algorithm is appropriate, and it is determined that the accuracy of the first relevance for searching for this is high, and updating is particularly important. do not do. On the other hand, if the obtained spectrum data does not properly represent the photosynthesis of the leaves, the detection algorithm is inappropriate, and it is determined that there is room for improvement in the accuracy of the first association for searching for this. And this will be updated.
- the formation of the first relevance shown in FIG. 8 is not limited to so-called supervised learning in which learning is performed by inputting a known object reference event and a detection algorithm. You may make it form this based on nothing learning. In such a case, the obtained data may be classified by clustering, and the first association degree may be formed based on the classification.
- It may be determined, for example, based on spectrum data whether or not the photosynthesis of leaves as the target event can be suitably expressed. In such a case, the determination may be made based on the intensity of the spectrum in each wavelength region.
- FIG. 14 (a) a case where the presence or absence of photosynthesis is determined based on the intensity of the spectrum in the wavelength range of 1347 to 1367 nm as an example of the detection algorithm will be described.
- FIG. 14B a result of imaging the leaves as the subject through the imaging device 5
- FIG. 14C a relatively large amount of photosynthesis is detected, and the peak of the spectrum is high in the wavelength range of 1347 to 1367 nm.
- FIG. 14C photosynthesis is not detected so much and the peak of the spectrum is not high in the wavelength range of 1347 to 1367 nm.
- the association degree described above may be updated. As a rule for updating the association degree, it may be performed based on the detected spectrum intensity. The higher the intensity of the spectrum in the wavelength range set as the detection algorithm, the higher the first relevance leading to that detection algorithm if it is interpreted that photosynthesis as the target event can be detected more suitably. Update as follows.
- the detection algorithm for detecting leaf photosynthesis in the first association shown in FIG. 8 has the highest association in the first association in the wavelength range of 1347 to 1367 nm (1357 ⁇ 10 nm).
- this detection algorithm 1347 to 1367 nm
- a specific threshold value is exceeded within the wavelength range of the detection algorithm as shown in FIG. It is determined that the event has been accurately detected, and the first relevance (80%) that leads to the detection algorithm is updated to be higher.
- the detection algorithm is below a specific threshold within the wavelength range of the detection algorithm, the detection algorithm Therefore, it is determined that the target event cannot be accurately detected, and the first relevance (80%) leading to the detection algorithm is updated to be lower.
- the intensity of the detected spectrum largely depends on whether or not the target event of the subject has occurred. In the above example, whether or not the leaf as the subject has developed photosynthesis as the target event. Certainly it relies heavily on crab. However, the intensity of this spectrum is not dependent solely on this, and is affected by the suitability of the detection algorithm in detecting the target event of the subject.
- the detection algorithm for detecting the target event of the subject is not suitable, and if the fitness is low, it is quite possible that the intensity of the spectrum will be lower than when the fitness is high. . If the leaves as a subject express the same degree of photosynthesis but have different spectral intensities, it is clear that they are affected by the suitability of the detection algorithm.
- the intensity of the spectrum within the wavelength range of such a detection algorithm decreases, the intensity of the spectrum outside the wavelength range may also increase.
- the intensity of this spectrum is suitable for a detection algorithm for detecting the target event of the subject, and when the suitability is high, the intensity of the spectrum within the wavelength range of the detection algorithm may be low. If the compatibility is low, the intensity of the spectrum within the wavelength range may be high.
- the spectral intensity within the wavelength region of the detection algorithm and the spectral intensity outside that wavelength region are affected.
- the first association degree is updated in accordance with the spectrum intensity expressed by receiving the influence. Accordingly, the compatibility between the target event of the subject and the detection algorithm is gradually improved through the first relevance that is sequentially updated.
- this spectral data also includes the characteristics of the image of the photographed subject itself (that is, the shape, texture, contrast, position, etc. of the subject drawn on the image). It is reflected in the same way.
- each pixel P ⁇ b> 1 and P ⁇ b> 2 in an image obtained by photographing the subject 11 via the spectral imaging device 4 or the photographing device 5 is configured by spectral data.
- Each of these spectrum data reflects the object event of the subject, the compatibility with the detection algorithm, and the image feature amount of the photographed subject itself.
- P1 is an edge portion of the image
- the influence of the image feature amount is greatly reflected in the spectrum data, while the effect of the photosynthesis of the leaf as a target event of the subject does not occur. It will not be reflected much in the data.
- P2 is not an edge portion of the image, the influence of the image feature amount is not reflected in the spectrum data so much.
- the photosynthesis of the leaf as the target event of the subject occurs, the influence is the spectrum. It will be greatly reflected in the data.
- both spectrum data in P1 and P2 are affected by compatibility with the detection algorithm.
- the spectrum data is affected by such image features, and by spatial information such as the shape, texture, contrast, and position of the subject drawn on the image. For this reason, in executing the above-described processing operation based on the spectrum data, the spectrum data is governed by the object event of the subject, the compatibility with the detection algorithm, and the spatial information.
- the compatibility between the target event of the subject and the detection algorithm may be determined. You may make it update 2 association degree.
- the hyperspectrum In FIG. 14, what is indicated by the dotted line is the hyperspectrum.
- the first association degree may be updated based on the spectrum intensity.
- a hyper spectrum imaged by the spectrum imaging device 4 may be acquired.
- the multispectrum acquired by the imaging device 5 may be restored to a hyperspectrum, and the suitability may be determined based on this hyperspectrum.
- the restoration of the multispectrum to the hyperspectrum may be executed via the search device 2 or the like, for example, based on a preset restoration algorithm.
- This restoration algorithm may be configured by a template having a correspondence relationship between the hyperspectrum and the multispectrum, and may be read out and referenced as necessary for restoration.
- Such an update process may be performed based on, for example, an update association as shown in FIG.
- the relevancy for update is set on the left side by the detection algorithm set this time and spectrum data obtained by imaging the subject, and the above-described suitability is drawn on the right side across the node 70.
- This node 70 is composed of a combination of the detection algorithm set this time and the obtained spectrum data (spectrum intensity). Each node 70 leads to suitability as an output solution. Similarly, the relevance for updating is composed of three or more levels of relevance.
- the node 70b when the spectral intensity obtained by imaging the subject based on the detection algorithm 1375 ⁇ 10 nm set this time is 10, the node 70b applies, and this node 70b has the highest relevance for updating. Compatibility is “low”. In such a case, the detection algorithm 1375 ⁇ 10 nm is determined to be less compatible from the viewpoint of the acquired spectrum intensity. Then, the first association shown in FIG. 8 is reset to be lower.
- the compatibility with the highest relevance for renewal is “high”.
- the detection algorithm 1375 ⁇ 10 nm is determined to be highly compatible from the viewpoint of the acquired spectrum intensity. Then, the first association shown in FIG. 8 is reset to be higher. Incidentally, it is not always necessary to select a higher one as long as it is based on the relevance for update, and a lower one may be selected.
- the example of searching for a search solution based on the detected spectrum intensity has been described.
- the search may be performed based on any factor as long as it is based on spectrum data. Good.
- a detection algorithm other than the detection algorithm set this time shown in FIG. 16 may be input. Then, another detection algorithm and spectrum data picked up based on the detection algorithm may be input to obtain a search solution.
- the feedback loop described above may be applied to the second association degree.
- the spectrum imaging device 4 and the imaging device 5 acquire the imaging conditions searched by the search device 2 based on the method described above.
- the spectrum imaging device 4 and the imaging device 5 actually shoot the subject 9 based on the imaging conditions.
- the subject 9 as the shooting target corresponds to the subject 9 input in S11, and the shooting purpose corresponds to detecting the target event input in step S11. That is, it is assumed that the detection of the photosynthesis of the leaf is input as the target event of the subject in step S11 described above, and the detection algorithm searched in step S13 is suitable for detecting the photosynthesis of the leaf.
- the searched imaging condition is input to the spectrum imaging device 4 or the imaging device 5.
- the leaves as the subject 9 are photographed through the spectral imaging device 4 and the photographing device 5 according to the inputted photographing conditions, and detection of photosynthesis is attempted.
- the spectrum imaging device 4 and the imaging device 5 transmit the spectrum data obtained by imaging the subject 9 to the algorithm database 3.
- the algorithm database 3 updates the second association degree based on such spectrum data.
- the subject 9 was also obtained by actually photographing the leaves of the subject 9 as a subject 9 based on the search conditions for the white balance OO for the photosynthesis of the leaf as the target event. If the spectral data can properly express the photosynthesis of leaves, the shooting conditions are appropriate, and it is judged that the accuracy of the second relevance for searching for this is high, and it is not particularly updated. . On the other hand, if the obtained spectral data cannot express the photosynthesis of the leaves in a suitable manner, it is determined that the shooting conditions are inappropriate, and there is room for improvement in the accuracy of the second relevance for searching for this. And this will be updated.
- the determination may be made, for example, based on spectrum data whether or not the photosynthesis of leaves as the target event can be suitably expressed. In such a case, the determination may be made on the basis of the intensity of the spectrum in each wavelength region as in the first association.
- the second association degree may be updated, and the update rule may be performed based on the detected spectrum intensity. If it is interpreted that the photosynthesis as the target event can be detected more appropriately as the intensity of the spectrum is higher, the second relevance that leads to the imaging condition is updated. For example, in the second association degree shown in FIG. 10, the imaging condition for detecting the photosynthesis of the leaf has the highest second association degree when the white balance OO is the highest. If the spectral intensity exceeds a specific threshold as a result of detecting the subject based on this shooting condition (white balance XX), it is determined that the target event has been accurately detected based on the shooting condition. The second relevance (80%) leading to is updated so as to be higher.
- the intensity of the detected spectrum largely depends on whether or not the target event of the subject has occurred. In the above example, whether or not the leaf as the subject has developed photosynthesis as the target event. Certainly it relies heavily on crab. However, the intensity of this spectrum is not dependent solely on this, and is affected by the suitability of the imaging conditions in detecting the target event of the subject. If the shooting conditions for detecting the target event of the subject are not appropriate and the suitability is low, it is quite possible that the intensity of the spectrum will be lower than when the suitability is high. .
- the intensity of this spectrum is appropriate for the shooting conditions for detecting the target event of the subject, and when the compatibility is high, the spectrum intensity may be low, and when the compatibility is low It is possible that the intensity of the spectrum is high.
- the spectral intensity is affected according to the compatibility between the target event of the subject and the shooting conditions.
- the second association degree is updated according to the spectrum intensity expressed by receiving the influence.
- the hyperspectrum In FIG. 14, what is indicated by the dotted line is the hyperspectrum.
- the second association degree may be updated based on the spectrum intensity.
- a hyper spectrum imaged by the spectrum imaging device 4 may be acquired.
- the multispectrum acquired by the imaging device 5 may be restored to a hyperspectrum, and the suitability may be determined based on this hyperspectrum.
- the restoration of the multispectrum to the hyperspectrum may be executed via the search device 2 or the like, for example, based on a preset restoration algorithm.
- This restoration algorithm may be configured by a template having a correspondence relationship between the hyperspectrum and the multispectrum, and may be read out and referenced as necessary for restoration.
- Such an update process may be performed based on the update association as shown in FIG.
- the imaging condition set this time and the spectrum data obtained by imaging the subject are set on the left side, and the adaptability described above is drawn on the right side with the node 70 in between.
- This node 70 is configured by a combination of the imaging conditions set this time and the obtained spectrum data (spectrum intensity). Each node 70 leads to suitability as an output solution. Similarly, the relevance for updating is composed of three or more levels of relevance.
- Such an association for update is acquired in advance. Then, the relevance for update is referred to, and the suitability as an output solution is searched based on the imaging condition set this time and the actually obtained spectrum intensity.
- This specific search method is the same as the above-described second association degree.
- the node 70b is applicable, and this node 70b is the most updated association. High degree of compatibility is “low”. In such a case, it is determined that the imaging condition (white balance OO) has low compatibility from the viewpoint of the acquired spectrum intensity. Then, the second association shown in FIG. 10 is reset so as to be lower.
- the formation of the second association is not limited to so-called supervised learning in which learning is performed by inputting a reference object event for a known subject and a detection algorithm, and this is based on unsupervised learning. You may make it form. In such a case, the obtained data may be classified by clustering, and the second association degree may be formed based on the classification.
- the example of searching for a search solution based on the detected spectrum intensity has been described.
- the search may be performed based on any factor as long as it is based on spectrum data. Good.
- other shooting conditions other than the currently set shooting conditions shown in FIG. 17 may be input. Then, other image capturing conditions and spectrum data captured based on the image capturing conditions may be input to obtain a search solution.
- the information search system 1 to which the present invention is applied may be provided with a feedback loop as shown in FIG.
- a position detection unit 81 connected to the spectrum imaging device 4 and a map information acquisition unit 82 connected to the position detection unit 81 are newly provided.
- the position detection unit 81 acquires the current position information of the spectrum imaging device 4 in real time based on the satellite positioning signal sent from the artificial satellite. If the spectrum imaging device 4 is mounted on a traveling vehicle, the position detecting unit 81 receives a satellite positioning signal as needed during the traveling of the vehicle on the road, so that the traveling road Position information at each of the above points can be acquired. The position information detected by the position detector 81 is sent to the algorithm database 3.
- the map information acquisition unit 82 stores map information including maps in Japan and maps of countries around the world.
- the map information here is embodied as a two-dimensional map describing the map in a plane, a three-dimensional map describing the map three-dimensionally, and a street view image consisting of a panoramic image taken around the road. Configured as electronic data. Based on such map information, a map can be displayed on the screen via a PC, a smartphone, a tablet terminal, etc., and various operations can be performed on the displayed map through an application. It becomes.
- the map information holding unit 82 may acquire electronic data of a map posted on the Internet as the initial map information, or electronic data of a map that is distributed free of charge or that is commercially available. May be obtained.
- the map information detected by the map information holding unit 82 is sent to the algorithm database 3.
- the update process may be performed based on the update association as shown in FIG. 19, for example.
- the relevance for update is detected by the position information detected by the position detection unit 81 and the map information holding unit 82 in addition to the detection algorithm set this time on the left side and the spectrum data obtained by imaging the subject. Map information is set, and the compatibility described above is drawn on the right side of the node 70.
- This node 70 is composed of a combination of position information and map information in addition to the detection algorithm set this time and the obtained spectrum data (spectrum intensity). Each node 70 leads to suitability as an output solution. Similarly, the relevance for updating is composed of three or more levels of relevance.
- the detection algorithm 1375 ⁇ 10 nm set this time and the spectrum intensity obtained by imaging the subject based on this is 15, and the position information XX and the map information XX, the node 70b is applied,
- the compatibility of the node 70b having the highest update association is “low”.
- the detection algorithm 1375 ⁇ 10 nm is determined to be less compatible from the viewpoint of the acquired spectrum intensity.
- the first association shown in FIG. 8 is reset to be lower.
- the information search system 1 to which the present invention is applied may be provided with a feedback loop as shown in FIG. In this feedback loop, a form detection unit 83 connected to the spectrum imaging device 4 is newly provided.
- the same components as those in the feedback loop shown in FIGS. 13 and 18 are denoted by the same reference numerals, and the following description is omitted.
- the form detection unit 83 includes a camera that captures an image of the subject 9.
- the form (shape, pattern, color, texture, etc.) of the subject 9 can be identified from the image of the subject 9 captured by the form detection unit 83.
- the form information of the subject 9 detected by the form detection unit 83 is sent to the algorithm database 3.
- the form information of the subject 9 is not limited to an image acquired by capturing an image from only a certain direction, but may be an image captured by varying the shooting range and shooting direction.
- the update process may be performed based on the update association as shown in FIG.
- the relevancy for update is set on the left side by the detection algorithm set this time and the spectrum data obtained by imaging the subject, as well as the configuration information of the subject 9 detected by the configuration detection unit 83, On the right side, the compatibility described above is drawn.
- This node 70 is composed of a combination of the detection algorithm set this time and the form information in addition to the obtained spectrum data (spectrum intensity). Each node 70 leads to suitability as an output solution. Similarly, the relevance for updating is composed of three or more levels of relevance.
- the node 70b when the detection algorithm 1375 ⁇ 10 nm set this time and the spectrum intensity obtained by imaging the subject based on this is 15, and the combination of the form information XX and the form information XX, the node 70b is applicable.
- This node 70b has “low” adaptability with the highest degree of association for update.
- the detection algorithm 1375 ⁇ 10 nm is determined to be less compatible from the viewpoint of the acquired spectrum intensity. Then, the first association shown in FIG. 8 is reset to be lower.
- the node 70d in the case of configuration information OO with the detection algorithm 970 ⁇ 10 nm, the node 70d is not suitable, and the compatibility of the node 70d having the highest relevance for relevance is “normal”. In such a case, the detection algorithm 970 ⁇ 10 nm is determined to have low adaptability from the viewpoint of the acquired spectrum intensity.
- the determination accuracy can be improved.
- the form information may include so-called spatial feature information.
- the spatial feature information here includes spatial position (arrangement) and form (shape, size, pattern, texture, color, texture, etc.).
- This spatial feature information is a concept including a feature amount on an image used in so-called deep learning technology, and is information for identifying a spatial position (arrangement) and form by extracting the feature amount.
- This spatial feature information may include a spectral feature amount extracted for each spectrum in addition to a general spatial feature amount.
- the spatial feature information may be configured by a fusion of the spatial feature amount and the spectral feature amount. Since this spectral feature value is used to extract a feature value based on a spectral image, only a desired subject can be easily separated from the background movement, etc., and the feature value can be extracted. It can be easily performed.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Library & Information Science (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
ゴリズム情報となりえる。また、これを構成する個々の説明変数x1、x2、・・・xkとしての特徴波長や特徴波長範囲についても同様に検出アルゴリズム情報となりえる。
2 探索装置
3 アルゴリズムデータベース
4 スペクトル撮像装置
5 撮影装置
9、10、11 被写体
16 画像
21 内部バス
23 表示部
24 制御部
25 操作部
26 通信部
27 探索部
28 記憶部
41 対物レンズ
42 精密直動ステージ
43 スリット板
43a スリット開口部
44 コリメートレンズ
45 分散光学素子
46 結像レンズ
47 撮像素子
48 制御部
51 結像光学系
52 フィルタ
53 撮像素子
54 信号処理部
56 撮像レンズ
70 ノード
81 位置検出部
82 地図情報保持部
83 形態検出部
481 撮影制御部
482 移動制御部
483 分光データ作成部
484 画像処理部
484-1 校正処理部
484-2 算出部
484-3 色解析画像取得部
Claims (15)
- 撮影した被写体から目的事象を判別する上で必要なスペクトルデータの検出アルゴリズム情報を探索する情報探索システムにおいて、
被写体の各目的事象と、上記検出アルゴリズム情報との3段階以上の第1連関度が予め記憶されている第1連関データベースと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力手段と、
上記第1連関データベースに記憶されている上記第1連関度を参照し、上記目的事象入力手段を介して入力された上記目的事象に関する情報に基づき、1以上の検出アルゴリズム情報を探索する探索手段と、
上記被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報を受信する受信手段とを備え、
上記第1連関データベースは、上記受信手段により受信した情報に基づいて上記第1連関度を更新すること
を特徴とする情報探索システム。 - 上記受信手段は、上記探索手段により探索された検出アルゴリズム情報に基づいて被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報を受信すること
を特徴とする請求項1記載の情報探索システム。 - 上記受信手段は、当該撮影した被写体の目的事象に関する情報として、その目的事象を撮影することにより得られたスペクトルデータを受信し、
上記第1連関データベースは、上記受信手段により受信したスペクトルデータに基づいて上記第1連関度を更新すること
を特徴とする請求項1又は2記載の情報探索システム。 - 撮影した被写体から目的事象を判別する上で必要なスペクトルデータの検出アルゴリズム情報を探索する情報探索システムにおいて、
被写体の各目的事象と、上記検出アルゴリズム情報との3段階以上の第1連関度が予め記憶されている第1連関データベースと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力手段と、
上記第1連関データベースに記憶されている上記第1連関度を参照し、上記目的事象入力手段を介して入力された上記目的事象に関する情報に基づき、1以上の検出アルゴリズム情報を探索する探索手段と、
上記探索手段により探索された検出アルゴリズム情報に基づいて被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報としてのマルチスペクトルデータを受信する受信手段と、
上記受信手段により受信されたマルチスペクトルデータに基づいてハイパースペクトルデータを復元するデータ復元手段とを備え、
上記第1連関データベースは、上記データ復元手段により復元されたスペクトルデータに基づいて上記第1連関度を更新すること
を特徴とする情報探索システム。 - 撮影した被写体から目的事象を判別するための撮影装置の撮影条件を探索する情報探索システムにおいて、
被写体の各目的事象と、各撮影条件との3段階以上の第2連関度が予め記憶されている第2連関データベースと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力手段と、
上記第2連関データベースに記憶されている上記第2連関度を参照し、上記目的事象入力手段を介して入力された上記目的事象に関する情報に基づき、1以上の撮影条件を探索する探索手段と、
上記被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報を受信する受信手段とを備え、
上記第2連関データベースは、上記受信手段により受信した情報に基づいて上記第2連関度を更新すること
を特徴とする情報探索システム。 - 上記受信手段は、上記探索手段により探索された撮影条件に基づいて被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報を受信すること
を特徴とする請求項5記載の情報探索システム。 - 上記受信手段は、当該撮影した被写体の目的事象に関する情報として、その目的事象を撮影することにより得られたスペクトルデータを受信し、
上記第2連関データベースは、上記受信手段により受信したスペクトルデータに基づいて上記第2連関度を更新すること
を特徴とする請求項5又は6記載の情報探索システム。 - 撮影した被写体から目的事象を判別するための撮影装置の撮影条件を探索する情報探索システムにおいて、
被写体の各目的事象と、各撮影条件との3段階以上の第2連関度が予め記憶されている第2連関データベースと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力手段と、
上記第2連関データベースに記憶されている上記第2連関度を参照し、上記目的事象入力手段を介して入力された上記目的事象に関する情報に基づき、1以上の撮影条件を探索する探索手段と、
上記探索手段により探索された撮影条件に基づいて被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報としてのマルチスペクトルデータを受信する受信手段と、
上記受信手段により受信されたマルチスペクトルデータに基づいてハイパースペクトルデータを復元するデータ復元手段とを備え、
上記第2連関データベースは、上記データ復元手段により復元されたスペクトルデータに基づいて上記第2連関度を更新すること
を特徴とする情報探索システム。 - 上記データベースは、更に被写体の位置情報及び/又は形態情報に基づいて各連関度を更新すること
を特徴とする請求項1~7のうち何れか1項記載の情報探索システム。 - 上記第1連関データベースは、上記3段階以上の第1連関度をニューラルネットワークで構成し、人工知能を活用することにより、上記第1連関度を更新すること
を特徴とする請求項1~4のうち何れか1項記載の情報探索システム。 - 上記第2連関データベースは、上記3段階以上の第2連関度をニューラルネットワークで構成し、人工知能を活用することにより、上記第2連関度を更新すること
を特徴とする請求項5~8のうち何れか1項記載の情報探索システム。 - 撮影した被写体から目的事象を判別する上で必要なスペクトルデータの検出アルゴリズム情報を探索する情報探索プログラムにおいて、
被写体の各目的事象と、上記検出アルゴリズム情報との3段階以上の第1連関度を予め取得する連関度取得ステップと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力ステップと、
上記連関度取得ステップにおいて取得した上記第1連関度を参照し、上記目的事象入力ステップにおいて入力された上記目的事象に関する情報に基づき、1以上の検出アルゴリズム情報を探索する探索ステップと、
上記被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報を受信する受信ステップとを有し、
上記連関度取得ステップでは、上記受信ステップにおいて受信した情報に基づいて上記第1連関度を更新すること
をコンピュータに実行させることを特徴とする情報探索プログラム。 - 撮影した被写体から目的事象を判別する上で必要なスペクトルデータの検出アルゴリズム情報を探索する情報探索プログラムにおいて、
被写体の各目的事象と、上記検出アルゴリズム情報との3段階以上の第1連関度を予め取得する連関度取得ステップと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力ステップと、
上記連関度取得ステップにおいて取得した上記第1連関度を参照し、上記目的事象入力ステップにおいて入力された上記目的事象に関する情報に基づき、1以上の検出アルゴリズム情報を探索する探索ステップと、
上記探索ステップにおいて探索した検出アルゴリズム情報に基づいて被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報としてのマルチスペクトルデータを受信する受信ステップと、
上記受信ステップにおいて受信したマルチスペクトルデータに基づいてハイパースペクトルデータを復元するデータ復元ステップとを有し、
上記連関度取得ステップでは、上記データ復元ステップにおいて復元したスペクトルデータに基づいて上記第1連関度を更新すること
をコンピュータに実行させることを特徴とする情報探索プログラム。 - 撮影した被写体から目的事象を判別するための撮影装置の撮影条件を探索する情報探索プログラムにおいて、
被写体の各目的事象と、各撮影条件との3段階以上の第2連関度を予め取得する連関度取得ステップと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力ステップと、
上記連関度取得ステップにおいて取得した上記第2連関度を参照し、上記目的事象入力ステップにおいて入力された上記目的事象に関する情報に基づき、1以上の撮影条件を探索する探索ステップと、
上記被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報を受信する受信ステップとを有し、
上記連関度取得ステップでは、上記受信ステップにおいて受信した情報に基づいて上記第2連関度を更新すること
をコンピュータに実行させることを特徴とする情報探索プログラム。 - 撮影した被写体から目的事象を判別するための撮影装置の撮影条件を探索する情報探索プログラムにおいて、
被写体の各目的事象と、各撮影条件との3段階以上の第2連関度を予め取得する連関度取得ステップと、
新たに判別すべき被写体の目的事象に関する情報が入力される目的事象入力ステップと、
上記連関度取得ステップにおいて取得した上記第2連関度を参照し、上記目的事象入力ステップにおいて入力された上記目的事象に関する情報に基づき、1以上の撮影条件を探索する探索ステップと、
上記探索ステップにおいて探索した撮影条件に基づいて被写体を撮影した撮影端末から、当該撮影した被写体の目的事象に関する情報としてのマルチスペクトルデータを受信する受信ステップと、
上記受信ステップにおいて受信したマルチスペクトルデータに基づいてハイパースペクトルデータを復元するデータ復元ステップとを有し、
上記連関度取得ステップでは、上記データ復元ステップにおいて復元したスペクトルデータに基づいて上記第2連関度を更新すること
をコンピュータに実行させることを特徴とする情報探索プログラム。
Priority Applications (9)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020187034501A KR102222913B1 (ko) | 2018-03-27 | 2018-10-24 | 정보 탐색 시스템 및 프로그램 |
AU2018415972A AU2018415972A1 (en) | 2018-03-27 | 2018-10-24 | Information retrieval system and program |
US16/309,921 US10846325B2 (en) | 2018-03-27 | 2018-10-24 | Information search system and information search program |
CN201880002236.5A CN110637224A (zh) | 2018-03-27 | 2018-10-24 | 信息搜索系统及程序 |
SG11202008391VA SG11202008391VA (en) | 2018-03-27 | 2018-10-24 | Information search system and information search program |
CA3026545A CA3026545C (en) | 2018-03-27 | 2018-10-24 | Information search system and information search program |
EP18811128.0A EP3572793A4 (en) | 2018-03-27 | 2018-10-24 | INFORMATION SEARCH SYSTEM AND INFORMATION SEARCH PROGRAM |
RU2018143650A RU2734870C2 (ru) | 2018-03-27 | 2018-10-24 | Система поиска информации и программа поиска информации |
IL277615A IL277615B1 (en) | 2018-03-27 | 2018-10-24 | Information search system and program |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2018-060572 | 2018-03-27 | ||
JP2018060572A JP6951753B2 (ja) | 2018-03-27 | 2018-03-27 | 情報探索システム及びプログラム |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2019187277A1 true WO2019187277A1 (ja) | 2019-10-03 |
Family
ID=67436969
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2018/039469 WO2019187277A1 (ja) | 2018-03-27 | 2018-10-24 | 情報探索システム及びプログラム |
Country Status (12)
Country | Link |
---|---|
US (1) | US10846325B2 (ja) |
EP (1) | EP3572793A4 (ja) |
JP (1) | JP6951753B2 (ja) |
KR (1) | KR102222913B1 (ja) |
CN (1) | CN110637224A (ja) |
AU (1) | AU2018415972A1 (ja) |
CA (1) | CA3026545C (ja) |
IL (1) | IL277615B1 (ja) |
RU (1) | RU2734870C2 (ja) |
SG (1) | SG11202008391VA (ja) |
TW (1) | TWI696924B (ja) |
WO (1) | WO2019187277A1 (ja) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6843439B2 (ja) * | 2017-05-17 | 2021-03-17 | エバ・ジャパン 株式会社 | 情報探索システム及び方法、情報探索プログラム |
JP6951753B2 (ja) * | 2018-03-27 | 2021-10-20 | エバ・ジャパン 株式会社 | 情報探索システム及びプログラム |
US20220346631A1 (en) | 2019-09-25 | 2022-11-03 | Sony Group Corporation | Imaging optical system and imaging apparatus |
JP2022039717A (ja) * | 2020-08-28 | 2022-03-10 | キヤノン株式会社 | 撮像装置およびその制御方法ならびにプログラム |
CN112905823B (zh) * | 2021-02-22 | 2023-10-31 | 深圳市国科光谱技术有限公司 | 一种基于大数据平台的高光谱物质检测识别系统及方法 |
CN113570574B (zh) * | 2021-07-28 | 2023-12-01 | 北京精英系统科技有限公司 | 一种场景特征检测的装置、搜索的装置及搜索的方法 |
CN114245021B (zh) * | 2022-02-14 | 2023-08-08 | 北京火山引擎科技有限公司 | 互动拍摄方法、电子设备、存储介质及计算机程序产品 |
CN117197269A (zh) * | 2023-08-31 | 2023-12-08 | 北华航天工业学院 | 一种基于深度学习模型的高光谱影像模拟方法 |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001343336A (ja) * | 2000-05-31 | 2001-12-14 | Nidek Co Ltd | 欠陥検査方法及び欠陥検査装置 |
JP2009211294A (ja) * | 2008-03-03 | 2009-09-17 | Nippon Hoso Kyokai <Nhk> | ニューラルネットワーク装置及びそれを用いたロボットカメラ制御装置並びにニューラルネットワークプログラム |
JP2010185719A (ja) * | 2009-02-10 | 2010-08-26 | National Agriculture & Food Research Organization | 穀粉の判別方法及び装置 |
WO2013002350A1 (ja) | 2011-06-29 | 2013-01-03 | 京都府公立大学法人 | 腫瘍部位の識別装置及び識別方法 |
JP2013535014A (ja) * | 2010-06-25 | 2013-09-09 | セルマーク セラノスティクス,リミテッド ライアビリティー カンパニー | 生物学的試片をスペクトル画像により分析する方法 |
WO2013145201A1 (ja) * | 2012-03-29 | 2013-10-03 | 株式会社日立製作所 | 画像生成システム及び画像生成方法 |
JP2015166682A (ja) | 2014-03-03 | 2015-09-24 | エバ・ジャパン 株式会社 | 分光放射輝度計 |
US20150277438A1 (en) * | 2013-05-07 | 2015-10-01 | Halliburton Energy Services Inc. | Optical sensor optimization and system implementation with simplified layer structure |
JP2017003495A (ja) | 2015-06-12 | 2017-01-05 | 株式会社リコー | 情報処理装置、情報処理プログラム、および情報処理システム |
JP2017138808A (ja) * | 2016-02-03 | 2017-08-10 | キヤノン株式会社 | 情報処理装置および情報処理方法 |
US20180053057A1 (en) * | 2016-08-18 | 2018-02-22 | Xerox Corporation | System and method for video classification using a hybrid unsupervised and supervised multi-layer architecture |
JP2018194541A (ja) * | 2017-05-17 | 2018-12-06 | エバ・ジャパン 株式会社 | 情報探索システム及び方法、情報探索プログラム |
Family Cites Families (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6919959B2 (en) * | 1999-06-30 | 2005-07-19 | Masten Opto-Diagnostics Co. | Digital spectral identifier-controller and related methods |
SE0402576D0 (sv) * | 2004-10-25 | 2004-10-25 | Forskarpatent I Uppsala Ab | Multispectral and hyperspectral imaging |
US7450761B2 (en) * | 2004-11-02 | 2008-11-11 | The Boeing Company | Spectral geographic information system |
JP4539432B2 (ja) * | 2005-05-16 | 2010-09-08 | ソニー株式会社 | 画像処理装置および撮像装置 |
DE602006009191D1 (de) * | 2005-07-26 | 2009-10-29 | Canon Kk | Bildaufnahmegerät und -verfahren |
US8154612B2 (en) | 2005-08-18 | 2012-04-10 | Qualcomm Incorporated | Systems, methods, and apparatus for image processing, for color classification, and for skin color detection |
JP4595750B2 (ja) * | 2005-08-29 | 2010-12-08 | ソニー株式会社 | 画像処理装置および方法、並びにプログラム |
US7420678B2 (en) | 2005-09-12 | 2008-09-02 | The Boeing Company | Multi-spectral imaging |
JP4642681B2 (ja) | 2005-09-30 | 2011-03-02 | 富士フイルム株式会社 | 光断層画像化装置 |
JP2009053064A (ja) | 2007-08-28 | 2009-03-12 | Sekonic Corp | カラーメータ |
KR100932843B1 (ko) * | 2008-01-29 | 2009-12-21 | 엔에이치엔(주) | 검색결과간의 연관도에 기초하여 클러스터링된 검색결과를제공하는 방법 및 시스템 그리고 검색결과를클러스터링하는 방법 및 시스템 |
JP2009206458A (ja) * | 2008-02-29 | 2009-09-10 | Canon Inc | 検出装置、露光装置およびデバイス製造方法 |
JP4508257B2 (ja) * | 2008-03-19 | 2010-07-21 | ソニー株式会社 | 構図判定装置、構図判定方法、プログラム |
AU2010203357B2 (en) | 2009-01-10 | 2014-07-10 | Carne Tender, Llc | System and method for analyzing properties of meat using multispectral imaging |
JP2011095225A (ja) | 2009-11-02 | 2011-05-12 | Olympus Corp | 画像処理装置、画像処理方法および顕微鏡システム |
US20100201800A1 (en) | 2009-02-09 | 2010-08-12 | Olympus Corporation | Microscopy system |
JP5698489B2 (ja) | 2010-09-30 | 2015-04-08 | オリンパス株式会社 | 検査装置 |
JPWO2012090416A1 (ja) | 2010-12-28 | 2014-06-05 | オリンパス株式会社 | 検査装置 |
US9019358B2 (en) * | 2011-02-08 | 2015-04-28 | Xerox Corporation | Method for classifying a pixel of a hyperspectral image in a remote sensing application |
US8600227B2 (en) | 2011-09-22 | 2013-12-03 | Xerox Corporation | Multi-filter array for a multi-resolution, multi-spectral camera |
JP6462361B2 (ja) * | 2011-11-17 | 2019-01-30 | バイエル・ヘルスケア・エルエルシーBayer HealthCare LLC | 医療診断手順についての情報を収集し、報告し、管理する方法及び技法 |
CN104080400B (zh) | 2011-12-21 | 2018-08-14 | 凯瑟琳·M·莎查夫 | 使排列于组织表面的病变成像的系统 |
NL2009015C2 (en) * | 2012-04-10 | 2013-10-15 | Biosparq B V | Method for classification of a sample on the basis of spectral data, method for creating a database and method for using this database, and corresponding computer program, data storage medium and system. |
EP3221679A4 (en) * | 2014-11-21 | 2018-11-14 | Mutti, Christopher M. | Imaging system for object recognition and assessment |
KR102251960B1 (ko) * | 2015-01-15 | 2021-05-14 | 삼성전자주식회사 | 이미지 분석 방법과 이를 수행하는 전자 장치 및 서버 |
WO2017079938A1 (en) * | 2015-11-12 | 2017-05-18 | Coyote Bioscience Co., Ltd. | Methods and systems for disease monitoring and assessment |
JP6951753B2 (ja) * | 2018-03-27 | 2021-10-20 | エバ・ジャパン 株式会社 | 情報探索システム及びプログラム |
-
2018
- 2018-03-27 JP JP2018060572A patent/JP6951753B2/ja active Active
- 2018-10-24 US US16/309,921 patent/US10846325B2/en active Active
- 2018-10-24 RU RU2018143650A patent/RU2734870C2/ru active
- 2018-10-24 EP EP18811128.0A patent/EP3572793A4/en not_active Withdrawn
- 2018-10-24 CN CN201880002236.5A patent/CN110637224A/zh active Pending
- 2018-10-24 SG SG11202008391VA patent/SG11202008391VA/en unknown
- 2018-10-24 IL IL277615A patent/IL277615B1/en unknown
- 2018-10-24 AU AU2018415972A patent/AU2018415972A1/en active Pending
- 2018-10-24 CA CA3026545A patent/CA3026545C/en active Active
- 2018-10-24 KR KR1020187034501A patent/KR102222913B1/ko active IP Right Grant
- 2018-10-24 WO PCT/JP2018/039469 patent/WO2019187277A1/ja unknown
- 2018-11-26 TW TW107142136A patent/TWI696924B/zh active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001343336A (ja) * | 2000-05-31 | 2001-12-14 | Nidek Co Ltd | 欠陥検査方法及び欠陥検査装置 |
JP2009211294A (ja) * | 2008-03-03 | 2009-09-17 | Nippon Hoso Kyokai <Nhk> | ニューラルネットワーク装置及びそれを用いたロボットカメラ制御装置並びにニューラルネットワークプログラム |
JP2010185719A (ja) * | 2009-02-10 | 2010-08-26 | National Agriculture & Food Research Organization | 穀粉の判別方法及び装置 |
JP2013535014A (ja) * | 2010-06-25 | 2013-09-09 | セルマーク セラノスティクス,リミテッド ライアビリティー カンパニー | 生物学的試片をスペクトル画像により分析する方法 |
WO2013002350A1 (ja) | 2011-06-29 | 2013-01-03 | 京都府公立大学法人 | 腫瘍部位の識別装置及び識別方法 |
WO2013145201A1 (ja) * | 2012-03-29 | 2013-10-03 | 株式会社日立製作所 | 画像生成システム及び画像生成方法 |
US20150277438A1 (en) * | 2013-05-07 | 2015-10-01 | Halliburton Energy Services Inc. | Optical sensor optimization and system implementation with simplified layer structure |
JP2015166682A (ja) | 2014-03-03 | 2015-09-24 | エバ・ジャパン 株式会社 | 分光放射輝度計 |
JP2017003495A (ja) | 2015-06-12 | 2017-01-05 | 株式会社リコー | 情報処理装置、情報処理プログラム、および情報処理システム |
JP2017138808A (ja) * | 2016-02-03 | 2017-08-10 | キヤノン株式会社 | 情報処理装置および情報処理方法 |
US20180053057A1 (en) * | 2016-08-18 | 2018-02-22 | Xerox Corporation | System and method for video classification using a hybrid unsupervised and supervised multi-layer architecture |
JP2018194541A (ja) * | 2017-05-17 | 2018-12-06 | エバ・ジャパン 株式会社 | 情報探索システム及び方法、情報探索プログラム |
Non-Patent Citations (1)
Title |
---|
See also references of EP3572793A4 |
Also Published As
Publication number | Publication date |
---|---|
IL277615A (en) | 2020-11-30 |
EP3572793A4 (en) | 2020-04-08 |
CA3026545C (en) | 2024-02-20 |
KR102222913B1 (ko) | 2021-03-04 |
IL277615B1 (en) | 2024-04-01 |
US20200065326A1 (en) | 2020-02-27 |
RU2018143650A3 (ja) | 2020-06-10 |
TWI696924B (zh) | 2020-06-21 |
KR20190113523A (ko) | 2019-10-08 |
TW201942765A (zh) | 2019-11-01 |
JP6951753B2 (ja) | 2021-10-20 |
CN110637224A (zh) | 2019-12-31 |
JP2019174200A (ja) | 2019-10-10 |
US10846325B2 (en) | 2020-11-24 |
SG11202008391VA (en) | 2020-10-29 |
AU2018415972A1 (en) | 2020-10-22 |
RU2018143650A (ru) | 2020-06-10 |
RU2734870C2 (ru) | 2020-10-23 |
CA3026545A1 (en) | 2019-09-27 |
EP3572793A1 (en) | 2019-11-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2019187277A1 (ja) | 情報探索システム及びプログラム | |
Li et al. | AOTF based molecular hyperspectral imaging system and its applications on nerve morphometry | |
CN105378538B (zh) | 用于多频谱成像的自动聚焦方法和系统 | |
JP6843439B2 (ja) | 情報探索システム及び方法、情報探索プログラム | |
Li et al. | Tongue fissure extraction and classification using hyperspectral imaging technology | |
Farooque et al. | A dual attention driven multiscale-multilevel feature fusion approach for hyperspectral image classification | |
Koirala et al. | Highlight detection and removal from spectral image | |
WO2023096971A1 (en) | Artificial intelligence-based hyperspectrally resolved detection of anomalous cells | |
WO2018212078A1 (ja) | 情報探索システム及び方法、情報探索プログラム | |
JP6925046B2 (ja) | 撮像装置の設計システム、撮像装置の設計プログラム | |
Deng et al. | Automatic estimation of rice grain number based on a convolutional neural network | |
JP6730629B2 (ja) | バンドパスフィルタの設計システム、バンドパスフィルタの設計プログラム | |
Chatoux et al. | Gradient in spectral and color images: from the di zenzo initial construction to a generic proposition | |
CN109754003A (zh) | 基于深度学习的智能机器人视觉技术应用检测系统和方法 | |
Rivas et al. | Copper concentrate dual-band joint classification using reflectance hyperspectral images in the VIS-NIR and SWIR bands | |
Cho et al. | Hyperspectral face databases for facial recognition research | |
Park et al. | An Acquisition Method for Visible and Near Infrared Images from Single CMYG Color Filter Array-Based Sensor | |
WO2021166089A1 (ja) | 評価支援装置、評価支援システム、評価支援方法およびプログラム | |
Ding et al. | Method for GPU-based spectral data cube reconstruction of integral field snapshot imaging spectrometers | |
Tao et al. | Concealed multiscale feature extraction network for automatic four-bar target detection in infrared imaging | |
Valdiviezo-N et al. | Lattice algebra approach to multispectral analysis of ancient documents | |
Chu et al. | A color correction method in uniform color space | |
Gerken et al. | Contrast enhancement of SWIR images based on four filter discrimination and principle components analysis | |
JP2023043625A (ja) | 製品品質検査システム |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 20187034501 Country of ref document: KR Kind code of ref document: A |
|
ENP | Entry into the national phase |
Ref document number: 3026545 Country of ref document: CA Kind code of ref document: A |
|
ENP | Entry into the national phase |
Ref document number: 2018811128 Country of ref document: EP Effective date: 20181211 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2018415972 Country of ref document: AU Date of ref document: 20181024 Kind code of ref document: A |