Processing apparatus, method, and non-transitory computer-readable recording medium
View Patent ↗A processing apparatus is connected to one or more terminals including a hyperspectral sensor that generates compressed image data which is hyperspectral information of a target compressed as two-dimensional image information over a network. The processing apparatus includes a storage device that stores data sets of samples and a first reconstruction table for generating hyperspectral data from the compressed image data, the data set of each sample includes hyperspectral data of the sample and data indicative of a property value of the sample; and a processing circuit that generates a statistical model for estimating the property value from the hyperspectral data on the basis of the data sets of the samples and generates a second reconstruction table by editing the first reconstruction table in accordance with the statistical model thus generated.
1. A processing apparatus that is connected to one or more apparatuses including a hyperspectral sensor that generates compressed image data which is hyperspectral information of a target compressed as two-dimensional image information, the processing apparatus comprising:
a storage device that stores data sets of samples and a first reconstruction table for generating hyperspectral data from the compressed image data, the data set of each of the samples including hyperspectral data of the sample and data indicative of a property value of the sample; and
a processing circuit that generates a model for converting the hyperspectral data into spectral data relevant with the property value on a basis of the data sets of the samples and generates a second reconstruction table for generating spectral data relevant with the property value from the compressed image data on a basis of the model and the first reconstruction table.
2. The processing apparatus according to claim 1 , wherein
the hyperspectral data includes luminance information for each of wavelength bands included in a target wavelength range;
the first reconstruction table is data for generating the luminance information for each of the wavelength bands from the compressed image data; and
the second reconstruction table is data for generating luminance information obtained by weighted-addition of the pieces of luminance information of the wavelength bands from the compressed image data.
3. The processing apparatus according to claim 1 , wherein
a data size of the second reconstruction table is smaller than a data size of the first reconstruction table.
4. The processing apparatus according to claim 1 , wherein
the processing circuit calculates weights corresponding to wavelength bands relevant with the property value by machine learning based on the hyperspectral data of the samples and generates the second reconstruction table from the first reconstruction table on a basis of the weights.
5. The processing apparatus according to claim 1 , wherein
the hyperspectral data includes luminance information for each of wavelength bands included in a target wavelength range;
the first reconstruction table is data for generating the luminance information for each of the wavelength bands from the compressed image data; and
the processing circuit decides weights corresponding to the wavelength bands by machine learning based on the hyperspectral data of the samples, decides some wavelength bands of a relatively low weight from among the wavelength bands on a basis of the weights, and generates the second reconstruction table by integrating pieces of information concerning the some wavelength bands in the first reconstruction table.
6. The processing apparatus according to claim 1 , wherein
the hyperspectral sensor includes a filter array including filters that have different transmission spectra; and
the first reconstruction table is data reflecting a spatial distribution of the transmission spectra of the filter array.
7. The processing apparatus according to claim 1 , wherein
the processing circuit transmits the second reconstruction table to the apparatus.
8. The processing apparatus according to claim 1 , wherein
the processing circuit generates the model, the first reconstruction table, and the second reconstruction table for each predetermined classification.
9. The processing apparatus according to claim 1 , wherein
the processing circuit acquires partial compressed image data generated by extracting data of a region from the compressed image data by the apparatus, generates the spectral data corresponding to the region from the partial compressed image data by using the second reconstruction table, and transmits the spectral data to the apparatus.
10. The processing apparatus according to claim 1 , wherein
the processing circuit acquires the compressed image data from the apparatus, generates spectral data relevant with the property value from the compressed image data by using the second reconstruction table, and transmits the spectral data to the apparatus.
11. The processing apparatus according to claim 1 , wherein
in a case where it is determined that it is necessary to change the model and the second reconstruction table on a basis of spectral data generated from the compressed image data on a basis of the second reconstruction table and the hyperspectral data in the data sets of the samples, the processing circuit generates the model again on a basis of the data sets of new samples and updates the second reconstruction table in accordance with the model thus generated again.
12. A method generated by a computer, the method comprising:
acquiring a first reconstruction table for generating hyperspectral data from compressed image data which is hyperspectral information of a target compressed as two-dimensional image information;
acquiring a model generated on a basis of a data set of a sample including hyperspectral data of the sample and data indicative of a property value of the sample, the model being for converting the hyperspectral data into spectral data relevant with the property value; and
generating a second reconstruction table for generating spectral data relevant with the property value from the compressed image data on a basis of the model and the first reconstruction table.
13. The method according to claim 12 , wherein
the hyperspectral data includes luminance information for each of wavelength bands included in a target wavelength range;
the first reconstruction table is data for generating the luminance information for each of the wavelength bands from the compressed image data; and
the second reconstruction table is data for generating luminance information obtained by weighted-addition of the pieces of luminance information of the wavelength bands from the compressed image data.
14. The method according to claim 12 , wherein
a data size of the second reconstruction table is smaller than a data size of the first reconstruction table.
15. The method according to claim 12 , wherein
the second reconstruction table is generated by synthesizing pieces of information of some of the wavelength bands included in the first reconstruction table.
16. A non-transitory computer-readable recording medium storing a program causing a computer to execute a process comprising:
acquiring a first reconstruction table for generating hyperspectral data from compressed image data which is hyperspectral information of a target compressed as two-dimensional image information;
acquiring a data set of a sample including hyperspectral data of the sample and data indicative of a property value of the sample;
generating the model for converting the hyperspectral data into spectral data relevant with the property value on a basis of the data set; and
generating a second reconstruction table for generating spectral data relevant with the property value from the compressed image data on a basis of the model and the first reconstruction table.