JP2014206392A - Program for plant kind discrimination and information processing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve accuracy of plant kind discrimination.SOLUTION: A method includes the steps of: extracting a first pixel of which the strength in spectrum data is not saturated, from a first image which is obtained by imaging an investigation target area for a plant kind and includes spectrum data for each pixel; extracting a second pixel having the spectrum data equal to or similar to a pixel corresponding to a position of a known plant kind from the first pixel; generating first reference spectrum data from the spectrum data of the second pixel; and determining the plant kind on the basis of the first reference spectrum data regarding the first pixel. On the other hand, the method includes processes for: extracting any other third pixel than the first pixel from a second image obtained by imaging the investigation target area for the plant kind after extinguishing an infrared region; implementing processing for restoring the infrared region on spectrum data of the third pixel; generating second reference spectrum data by performing similar processing; and determining the plant kind on the basis of the second reference spectrum data regarding the third pixel.

Description

  The present technology relates to a technology for discriminating plant types.

  In the conventional vegetation survey method, there are mainly two methods. The first investigation method is a method in which an identifier goes through the site and visually discriminates the local situation. The second investigation method is a method in which a discriminator discriminates using a photograph or an image taken from a satellite or an aircraft (that is, remote sensing). Each of these methods is used alone or in combination.

  The sensor used in remote sensing in the second survey method was previously panchromatic (black and white), but has recently changed to multispectral (color). The vegetation is identified by reading the photos and images.

  Recently, vegetation map production by GIS (Geographic Information System) has become mainstream. In GIS, by using a normal vegetation index (NDVI: Normalized Difference Vegetation Index) as reference information indicating the crown shape and color of each plant species prepared in advance, pattern matching is performed on images captured by a camera or sensor. Identify plant species.

  Furthermore, in recent years, a satellite (for example, satellite name: EO-1 (sensor name: Hyperion), satellite name: PROBA (sensor name: sensor name: CHRIS)) has been launched as a global environmental satellite and the like, and measurement by a hyperspectral sensor is also performed.

  The amount of information obtained by using hyperspectral sensors has increased dramatically over multispectral. Airborne hyperspectral sensors (also called hyperspectral cameras) have also been developed and are beginning to be used in various fields including the environment and agriculture.

  Attempts have been made to discriminate tree species using hyperspectral data, which is the output of such a hyperspectral sensor, but this does not deal with various situations.

  As a conventional technique for discriminating tree species, there is known a method of discriminating the tree type of each subcompartment section in the image data by dividing image data indicating the current state of the forest into subcompartment sections.

  Also, a plurality of band data blocks are acquired based on the appropriate analysis period of the tree species, and an object extraction map for each tree species in which upper and lower limits are set for the luminance value of each band data block is generated, and the NDVI of each tree species is masked. A technique for processing and extracting tree species distribution is known.

  In addition, the brightness value of the forest image data taken from the sky is flattened at the peaks and valleys, and the canopy shape and its texture feature amount are obtained by dividing the area with respect to the spatial change of the brightness value of the flattened image data, There is known a technique for determining a tree type based on a texture characteristic amount of a known tree crown.

  Spectral data obtained by a multi-band sensor or hyperspectral sensor has spectral data including wavelength information and light intensity information for each image coordinate (that is, for each pixel or pixel). It can be said that the data has a three-dimensional structure having elements as data.

  On the other hand, since the reflection spectrum has characteristics according to the tree species, it has become possible to discriminate the tree species by using hyperspectral data from which high-accuracy spectrum data can be obtained.

  That is, the reference spectrum of the tree to be discriminated is compared with the spectrum data to be identified, the similarity is obtained by a spectral angle mapper, etc., a pixel or pixel having a high similarity is extracted, and the plant is discriminated. .

  However, in the reflection spectrum of plants, features appear not only in the visible light region but also in the infrared region, and when focusing on the wavelength intensity in the visible light region, the spectral intensity of the wavelength in the infrared region is Saturates. In order to avoid spectral intensity saturation in the infrared region, if the incident light intensity is lowered in the entire wavelength region using a dimming means, the spectral resolution in the visible light region may be distorted or the wavelength resolution may be reduced due to a decrease in the intensity ratio. As a result, there is a problem that correct discrimination is not performed.

JP 2010-086276 A JP 2006-085517 A JP 2006-285310 A JP 2003-344048 A JP 2011-169896 A

  Therefore, the objective of this technique is to provide the technique for improving the discrimination | determination precision of a plant classification according to one side.

  An information processing method according to an aspect of the present technology is (A) stored in a data storage unit, obtained by photographing a plant-type survey target area, and from a first image including spectrum data for each pixel. The first pixel whose spectral intensity is not saturated in the spectral data is extracted, and (B) the second pixel having the same or similar spectral data as the pixel corresponding to the position whose plant type is known from the first pixel. Extracting a pixel, (C) generating first reference spectrum data from spectrum data of the second pixel, (D) determining the plant type for the first pixel based on the first reference spectrum data, (E) stored in the data storage unit, obtained by photographing the investigation target area of the plant type after dimming the infrared region, and including spectral data for each pixel The third pixel other than the first pixel is extracted from the image of (3), (F) the process of restoring the infrared region is performed on the spectral data of the third pixel, and (G) the third pixel is The fourth pixel, which is the same or similar to the pixel corresponding to the position corresponding to the known plant type and has the spectrum data after the restoration process, is extracted, and the spectrum after the restoration process of the fourth pixel is performed. The method includes generating second reference spectrum data from the data, and (H) determining a plant type for the third pixel based on the second reference spectrum data.

  In one aspect, plant type discrimination accuracy can be improved.

FIG. 1 is a functional block diagram of the information processing apparatus according to the embodiment. FIG. 2 is a diagram illustrating an example of spectrum data without saturation. FIG. 3 is a diagram illustrating an example of spectral data with saturation. FIG. 4 is a diagram showing an example of spectrum data when the infrared region is dimmed. FIG. 5 is a diagram illustrating an example of spectrum data obtained by restoring the infrared region. FIG. 6 is a diagram illustrating a processing flow according to the embodiment. FIG. 7 is a diagram illustrating a processing flow according to the embodiment. FIG. 8 is a diagram illustrating an example of a hyperspectral image. FIG. 9 is a diagram illustrating an example of the first pixel group. FIG. 10 is a diagram illustrating a determination result for the first pixel group. FIG. 11 is a diagram illustrating an example of the second pixel group. FIG. 12 is a diagram illustrating a determination result for the second pixel group. FIG. 13 is a diagram illustrating the merged determination result. FIG. 14 is a diagram illustrating an example of the similarity between pine and cocoon based on cedar. FIG. 15 is a diagram illustrating an example of the degree of similarity for cedar and cocoon based on pine. FIG. 16 is a diagram illustrating an example of the degree of similarity for cedar and pine based on firewood. FIG. 17 is a diagram illustrating a discrimination result of the comparative example. FIG. 18 is a functional block diagram of a computer.

  FIG. 1 shows a functional block diagram of information processing apparatus 100 according to the present embodiment. The information processing apparatus 100 according to the present embodiment includes a first data storage unit 101, a first determination unit 110, a second determination unit 120, a second data storage unit 130, a synthesis unit 140, and a determination result storage. Part 150 and output part 160.

  In the first data storage unit 101, a first image and a second image taken by a hyperspectral camera for a certain survey target area, and a position (image that the plant type is known for each plant type in the target survey area) Data) are stored.

  Note that the first image is an image in which at least one line is confirmed to have a spectral intensity that is not saturated in the entire wavelength region (for example, 350 nm to 1050 nm). That is, for example, one line of spectrum data as shown in FIG. 2 is obtained. In the example of FIG. 2, the horizontal axis represents wavelength (nm) and the vertical axis represents spectral intensity. Here, it is assumed that the spectrum intensity is saturated at 0.2. In the example of FIG. 2, the spectral intensity is less than 0.2 in the entire wavelength region, and there is no saturated wavelength region. However, some pixels include pixels from which spectral data in which the spectral intensity in the infrared region is saturated as shown in FIG.

  The second image is an image in which the infrared region is dimmed by photographing a hyperspectral camera with a filter for dimming the infrared region for the same investigation target area as the first image. . At this time, a filter having a flattening attenuation rate in the infrared region is used. Further, after confirming that the spectral intensity after dimming is not saturated in the entire wavelength region, a filter having an appropriate dimming rate is adopted. Further, for the filter, the light attenuation rate for each wavelength is measured. For example, FIG. 4 shows spectral data obtained as a result of imaging the same position as in FIG. 3 with a filter having a light attenuation rate of 30%. As shown in FIG. 4, the spectral intensity is not saturated in the entire wavelength region. In this way, the measurement is performed within the measurement range while keeping the characteristics of the plant reflection spectrum in the infrared region. In addition, if {1 / (1-light attenuation rate)} × spectral intensity is calculated for each wavelength, the original spectral intensity can be restored. In the example of FIG. 4, the spectrum data as shown in FIG. 5 is restored.

  The first determination unit 110 includes a first pixel extraction unit 111, a second pixel extraction unit 112, a first reference spectrum generation unit 113, and a first matching processing unit 114. The first pixel extraction unit 111 extracts, from the first image, a first pixel group whose spectral intensity is not saturated in the entire wavelength region. The first pixel extraction unit 111 stores the extracted coordinates of the first pixel group in the second data storage unit 130. The second pixel extraction unit 112 extracts, for each plant type, the spectrum data of the pixel having the known plant type from the first pixel group extracted by the first pixel extraction unit 111, and extracts each plant type. First candidate pixels having spectral data identical or similar to the spectral data of the selected pixels are extracted. The first reference spectrum generation unit 113 generates first reference spectrum data from the spectrum data of the first candidate pixel for each plant type. The first matching processing unit 114 performs matching of the first pixel group based on the first reference spectrum data of each plant type.

  The second determination unit 120 includes a third pixel extraction unit 121, a restoration processing unit 122, a third data storage unit 123, a fourth pixel extraction unit 124, a second reference spectrum generation unit 125, and a second matching process. Part 126.

  The third pixel extraction unit 121 extracts a second pixel group other than the first pixel group from the second image stored in the first data storage unit 101. The restoration processing unit 122 performs restoration processing of the infrared region on the spectral data of the second pixel group based on the light attenuation rate, generates restored spectral data, and stores the restored spectral data in the third data storage unit 123. . The fourth pixel extraction unit 124 extracts, for each plant type, a pixel having a known plant type from the second pixel group, and is the same as or similar to the spectrum data after restoration of the pixel extracted for each plant type. A second candidate pixel having the restored spectrum data is extracted. The second reference spectrum generation unit 125 generates second reference spectrum data from the spectrum data after the restoration of the second candidate pixels for each plant type. The second matching processing unit 126 performs matching of the second pixel group based on the second reference spectrum data of each plant type.

  The combining unit 140 combines the determination result of the first matching processing unit 114 and the determination result of the second matching processing unit 126 and stores the result in the determination result storage unit 150. The output unit 160 outputs the determination result stored in the determination result storage unit 150 to a display device, a printing device, another computer, or the like.

  Next, processing in the present embodiment will be described with reference to FIGS. First, the first pixel extraction unit 111 of the first determination unit 110 determines a first pixel group whose spectral intensity is not saturated in the infrared region from the first image stored in the first data storage unit 101. Extract (step S1). In addition, the first pixel extraction unit 111 stores the extracted coordinates of each pixel of the first pixel group in the second data storage unit 130 (step S3).

  Further, the second pixel extraction unit 111 reads out data of coordinates with known plant types from the first data storage unit 101, and extracts pixels of coordinates with known plant types from the first pixel group for each plant type. At the same time, a first candidate pixel having spectral data that is the same as or similar to the spectral data of the pixel extracted for each plant type is extracted from the first pixel group (step S5). Generally, since the number of pixels with known plant types is small, it is a problem in accuracy to generate reference spectrum data from only the spectrum data of these pixels. Therefore, for each plant type, the average spectral intensity and standard deviation are calculated for each wavelength from the spectral data of pixels that are known to be the plant type. Then, for all wavelengths, a pixel whose spectrum intensity falls within the range of average spectrum intensity−standard deviation to average spectrum intensity + standard deviation is extracted from the first pixel group as the first candidate pixel.

  Then, the first reference spectrum generation unit 113 generates a first reference spectrum from the spectrum data of the first candidate pixel for each plant type, and outputs the first reference spectrum to the first matching processing unit 114 (step S7). Specifically, the average spectral intensity of each wavelength is calculated from the spectral data of the first candidate pixel and adopted as the first reference spectrum.

  The first verification processing unit 114 uses the first reference spectrum of each plant type, and each pixel included in the first pixel group corresponds to any plant type or any plant type. A collation process for determining whether or not to perform is executed, and the processing result is output to the synthesis unit 140 (step S9). The similarity between the spectrum data of each pixel included in the first pixel group and the first reference spectrum of each plant type is calculated with a predetermined algorithm, and whether the calculated similarity is equal to or higher than a threshold value, It is judged whether it corresponds to each plant classification. For example, when discriminating between cedar, camellia and pine, whether or not the degree of similarity between the cedar and the first reference spectrum is greater than or equal to the threshold, and the degree of similarity between the cedar and the first reference spectrum is greater than or equal to the threshold. It is determined whether or not there is a similarity between the pine and the first reference spectrum equal to or greater than a threshold value.

  The processing shifts to the processing of FIG. 7 via the terminal A, and the third pixel extraction unit 121 of the second determination unit 120 extracts the second data from the second image stored in the first data storage unit 101. Based on the coordinates of the first pixel group stored in the storage unit 130, a second pixel group that is a pixel group other than the first pixel group is extracted (step S11). In addition, the process after step S11 can be performed asynchronously with the process of the 1st determination part 110, if step S3 is completed.

  Then, the restoration processing unit 122 reads the spectral data of the pixels belonging to the second pixel group from the first data storage unit 101, restores the infrared region in the spectral data of the second pixel group, and the restored spectral data Is stored in the third data storage unit 123 (step S13). As described above, since the spectrum data as shown in FIG. 4 is obtained, {1 / (1−attenuation rate)} × spectral intensity is calculated for each wavelength, and after restoration as shown in FIG. Spectral data is generated.

  Further, the fourth pixel extraction unit 124 reads out data of coordinates with known plant types from the first data storage unit 101, and extracts pixels of coordinates with known plant types from the second pixel group for each plant type. At the same time, a second candidate pixel having spectral data that is the same as or similar to the spectral data of the pixel extracted for each plant type is extracted from the second pixel group (step S15). As in step S5, since the number of pixels with known plant types is generally small, it is a problem in accuracy to generate reference spectrum data from only the spectrum data of these pixels. Therefore, for each plant type, the average spectral intensity and standard deviation are calculated for each wavelength from the restored spectral data of the pixels that are known to be the plant type. Then, for all wavelengths, a pixel whose spectrum intensity falls within the range of average spectrum intensity−standard deviation to average spectrum intensity + standard deviation is extracted as the second candidate pixel from the second pixel group.

  Then, the second reference spectrum generation unit 125 generates a second reference spectrum from the spectrum data of the second candidate pixel for each plant type, and outputs the second reference spectrum to the second matching processing unit 126 (step S17). Specifically, the average spectral intensity of each wavelength is calculated from the restored spectral data of the second candidate pixel and adopted as the second reference spectrum.

  Then, the second matching processing unit 126 uses the second reference spectrum of each plant type, and each pixel included in the second pixel group corresponds to which plant type or any plant type. A collation process for determining whether or not to perform is executed, and the processing result is output to the synthesis unit 140 (step S19). The similarity between the restored spectrum data of each pixel included in the second pixel group and the second reference spectrum of each plant type is calculated by a predetermined algorithm, and whether the calculated similarity is equal to or greater than a threshold value Thus, it is determined whether or not each plant type is applicable. For example, when discriminating cedar, camellia, and pine, whether or not the degree of similarity between the cedar and the second reference spectrum is greater than or equal to the threshold, and the degree of similarity between the cedar and the second reference spectrum is greater than or equal to the threshold. It is determined whether or not there is a similarity between the pine and the second reference spectrum equal to or greater than a threshold value.

  Thereafter, the synthesizing unit 140 merges the determination result by the first matching processing unit 114 and the determination result by the second matching processing unit 126 and stores the result in the determination result storage unit 150 (step S21). Since the first pixel group and the second pixel group are different pixel groups, they may be simply merged.

  Then, the output unit 160 outputs the determination result stored in the determination result storage unit 150 to a display device, a printing device, another computer connected to the network, or the like (step S23). For example, an image colored by plant type is generated and output.

[Example]
For example, it is assumed that a hyperspectral image (a first image and a second image) as shown in FIG. 8 is obtained. However, since the first image and the second image cannot be distinguished from humans, only one is shown.

  Among these, as the first pixel group, a pixel group as shown in FIG. 9 is extracted. Here, it is assumed that there are only 10 pixels that are known to be cedar in the first pixel group, and that 60 first candidate pixels are extracted by performing the processing described above. Then, a first reference spectrum for cedar is generated from 60 pixels as described above. In addition, a first candidate pixel for the pine is extracted based on a pixel that is known to be a pine in the first pixel group, and a first reference spectrum for the pine is generated from the first candidate pixel for the pine. Further, a first candidate pixel for 檜 is extracted based on a pixel known as 檜 in the first pixel group, and a first reference spectrum for 檜 is generated from the first candidate pixel for 檜.

  By calculating the similarity between the first reference spectrum of cedar, pine and pine thus obtained and the spectrum data of each pixel included in the first pixel group, and comparing with the threshold value set for each. A discrimination result as shown in FIG. 10 is obtained. In FIG. 10, there are other areas such as forest roads where the white portion is in addition to the area determined to be a pine, the area determined to be a cedar, and the area determined to be a cocoon.

  Further, since the second pixel group is a pixel group other than the first pixel group, a pixel group as shown in FIG. 11 is extracted. Here, it is assumed that there are only 7 pixels that are known as cedar in the first pixel group, and there are 70 second candidate pixels extracted by performing the processing described above. Then, the second reference spectrum for the cedar is generated from the 70 pixels as described above. Further, a second candidate pixel for the pine is extracted based on a pixel known as a pine in the second pixel group, and a second reference spectrum for the pine is generated from the second candidate pixel for the pine. Further, a second candidate pixel for 檜 is extracted based on a pixel known as 檜 in the second pixel group, and a second reference spectrum for 檜 is generated from the second candidate pixel for 檜.

  The similarity between the second reference spectrum of the cedar, pine and pine thus obtained and the restored spectrum data of each pixel included in the second pixel group is calculated, and the threshold value set for each When compared, a discrimination result as shown in FIG. 12 is obtained. In FIG. 12, in addition to the area determined to be a pine, the area determined to be a cedar, and the area determined to be a cocoon, the white portion is another area such as a forest road.

  The synthesizer 140 merges the determination results as shown in FIGS. 10 and 12 and stores the data as shown in FIG. 13 in the determination result storage 150.

  In such an embodiment, the similarity between the spectrum data for pixels known as cedar (pixels included in the first pixel group) and the first reference spectrum for cedar is 0.998. became. Similarly, the similarity between the post-restoration spectrum data for pixels known as cedars (pixels included in the second pixel group) and the second reference spectrum for cedars was 0.998. When such first and second reference spectra are used to calculate the degree of similarity with pixels that are known to be pine or moth, a result as shown in FIG. 14 is obtained.

  Looking at the result of FIG. 14, since the similarity between pine and cocoon is low, cedar can be distinguished from pine and cocoon by appropriately setting a threshold value.

  Similarly, the similarity between the spectrum data for a pixel known as a pine (a pixel included in the first pixel group) and the first reference spectrum for the pine was 0.981. Similarly, the similarity between the restored spectral data for a pixel known as a pine (a pixel included in the second pixel group) and the second reference spectrum for the pine was 0.983. When such first and second reference spectra are used to calculate the degree of similarity with pixels that are known as cedars and camellias, the results shown in FIG. 15 are obtained.

  Looking at the results in FIG. 15, since the degree of similarity is low in both cedar and cocoons, pine can be distinguished from cedar and cocoons by appropriately setting a threshold value.

  Furthermore, the similarity between the spectrum data for a pixel known as 檜 (a pixel included in the first pixel group) and the first reference spectrum for 檜 was 0.921. Similarly, the similarity between the restored spectral data for the pixel known as 檜 (pixels included in the second pixel group) and the second reference spectrum for 檜 was 0.930. Although the value is smaller than that of cedar and pine, a sufficiently high similarity is calculated. When such first and second reference spectra are used to calculate the degree of similarity with pixels known to be cedar or pine, a result as shown in FIG. 16 is obtained.

  Looking at the result of FIG. 16, since the degree of similarity of cedar and pine is lower than the above degree of similarity, cocoons can be distinguished from cedar and pine by appropriately setting a threshold value.

[Comparative example]
On the other hand, when cedar, pine and cocoon reference spectra were generated from spectrum data in which the infrared region was not saturated, and the discrimination process was performed for the same investigation target region, the results shown in FIG. 17 were obtained. Comparing FIG. 17 with FIG. 13, it can be seen that there are many areas determined as cedar and cypress in FIG.

  In this case, the similarity between the reference spectrum for cedar and the spectrum data for the pixel known as cedar is 0.958, and the similarity with the spectrum data for the pixel known as pine is 0.875. Thus, the degree of similarity with the spectral data for a pixel known to be 檜 is 0.940. The difference in the degree of similarity between cedar and camellia is getting smaller.

  Further, the similarity between the reference spectrum for pine and the spectrum data for the pixel known as pine is 0.981, and the similarity with the spectrum data for the pixel known as cedar is 0.925. The similarity with the spectral data for the pixel known as 檜 is 0.930. Pines are relatively easy to distinguish.

  Further, when the similarity between the reference spectrum for 檜 and the spectrum data for the pixel known as 檜 is calculated, it becomes 0.921, and the similarity with the spectrum data for the pixel known as cedar becomes 0.910. The similarity to the spectral data for the pixel known as pine is 0.880. The difference in the degree of similarity between cedar and camellia is very small.

  Accordingly, it is difficult to set the threshold value, and erroneous determination occurs, and a result as shown in FIG. 17 is obtained.

  Thus, unless the characteristics of the infrared region are correctly reflected to generate a reference spectrum, it is impossible to distinguish between cedars and straw.

  As described above, according to the present embodiment, it is possible to generate a highly accurate reference spectrum even when there are few pixels with known plant types or when there is a large variation in spectrum data, and it is possible to improve discrimination accuracy. .

  Although the embodiment of the present technology has been described above, the present technology is not limited to this. For example, the functional block diagram shown in FIG. 1 is an example, and may not match the program module configuration or the data storage unit configuration. As for the processing flow, as long as the processing result does not change, the processing order may be changed or parallel processing may be performed.

  Note that the information processing apparatus 100 described above is a computer apparatus, and as shown in FIG. 18, a memory 2501, a CPU (Central Processing Unit) 2503, a hard disk drive (HDD: Hard Disk Drive) 2505, and a display device. A display control unit 2507 connected to 2509, a drive device 2513 for the removable disk 2511, an input device 2515, and a communication control unit 2517 for connecting to a network are connected by a bus 2519. An operating system (OS) and an application program for executing the processing in this embodiment are stored in the HDD 2505, and are read from the HDD 2505 to the memory 2501 when executed by the CPU 2503. The CPU 2503 controls the display control unit 2507, the communication control unit 2517, and the drive device 2513 according to the processing content of the application program, and performs a predetermined operation. Further, data in the middle of processing is mainly stored in the memory 2501, but may be stored in the HDD 2505. In an embodiment of the present technology, an application program for performing the above-described processing is stored in a computer-readable removable disk 2511 and distributed, and installed from the drive device 2513 to the HDD 2505. In some cases, the HDD 2505 may be installed via a network such as the Internet and the communication control unit 2517. Such a computer apparatus realizes various functions as described above by organically cooperating hardware such as the CPU 2503 and the memory 2501 described above and programs such as the OS and application programs. .

  The above-described embodiment can be summarized as follows.

  The information processing method according to the present embodiment is (A) stored in the data storage unit, and is obtained by photographing the survey target area of the plant type, and from the first image including the spectrum data for each pixel, A first pixel in which spectral intensity is not saturated in the spectral data is extracted, and (B) a second pixel having spectral data that is the same as or similar to a pixel corresponding to a position whose plant type is known from the first pixel. (C) generating first reference spectrum data from the spectrum data of the second pixel, (D) determining the plant type for the first pixel based on the first reference spectrum data, ( E) a second data stored in the data storage unit, obtained by taking an image of an investigation target area of a plant type after dimming an infrared region, and including spectral data for each pixel A third pixel other than the first pixel is extracted from the image, (F) a process of restoring the infrared region is performed on the spectral data of the third pixel, and (G) from the third pixel, A fourth pixel that is the same or similar to a pixel corresponding to a known position of the plant type and that has the processed spectral data to be restored is extracted, and the restored spectral data of the fourth pixel is restored. The second reference spectrum data is generated from (H), and (H) includes a process of determining the plant type for the third pixel based on the second reference spectrum data.

  Thus, even in a situation where the spectrum intensity is saturated in the infrared region, the reference spectrum is generated by appropriately evaluating the characteristics of the infrared region so that the plant type can be accurately identified. become.

  In addition, the process which extracts the 2nd pixel described above extracts the pixel which has the spectral data contained in the width | variety of a standard deviation from the average of the spectral data of the pixel (b1) corresponding to the position where a plant classification is known. Processing may be included. In this way, it is possible to appropriately extract the second pixel used to generate the first reference spectrum even when the number of pixels corresponding to the position where the plant type is known is small.

  In addition, the process of extracting the fourth pixel described above includes (g1) the standard deviation width from the average of the restored spectrum data of the pixel corresponding to the position where the plant type is known. You may make it include the process which extracts the pixel which has the spectrum data after the process to decompress | restore. In this way, similarly to the second pixel, the fourth pixel used to generate the second reference spectrum is appropriately selected even when the number of pixels corresponding to the position where the plant type is known is small. Can be extracted.

  Further, the restoration process described above may include a process of multiplying the reciprocal of the value obtained by subtracting the dimming rate from (f1) 1 and multiplying the intensity of each wavelength in the infrared region. This makes it possible to properly restore the spectrum data.

  Note that a program for causing a computer to perform the information processing method can be created. The program can be a computer-readable storage medium such as a flexible disk, a CD-ROM, a magneto-optical disk, a semiconductor memory, or a hard disk. It is stored in a storage device. The intermediate processing result is temporarily stored in a storage device such as a main memory.

  The following supplementary notes are further disclosed with respect to the embodiments including the above examples.

(Appendix 1)
A first pixel that is stored in the data storage unit and is obtained by photographing the survey target area of the plant type and includes spectral data for each pixel, and the spectral intensity in the spectral data is not saturated. Extract
From the first pixel, a second pixel having spectral data that is the same as or similar to a pixel corresponding to a position where the plant type is known is extracted;
Generating first reference spectral data from spectral data of the second pixel;
For the first pixel, a plant type is determined based on the first reference spectrum data,
From the second image, which is stored in the data storage unit, obtained by photographing the investigation target area of the plant type after dimming the infrared region, and including spectral data for each pixel, Extract a third pixel other than one pixel,
Performing a process of restoring the infrared region on the spectral data of the third pixel;
From the third pixel, extract a fourth pixel that is the same as or similar to a pixel corresponding to a position where the plant type is known and that has the processed spectrum data to be restored,
Generating second reference spectrum data from the restored spectrum data of the fourth pixel,
A program for causing a computer to execute a process of determining a plant type for the third pixel based on the second reference spectrum data.

(Appendix 2)
The process of extracting the second pixel includes
The program according to appendix 1, including a process of extracting a pixel having spectral data included in a standard deviation range from an average of spectral data of pixels corresponding to positions where the plant type is known.

(Appendix 3)
The process of extracting the fourth pixel includes
Supplementary note 1 including a process of extracting a pixel having the spectrum data after the restoration that is included in a standard deviation width from an average of the spectrum data after the restoration of the pixel corresponding to the position where the plant type is known Or the program according to 2.

(Appendix 4)
The process to restore is
The program according to any one of appendices 1 to 3, including a process of multiplying a reciprocal of a value obtained by subtracting a dimming ratio from 1 to an intensity of each wavelength in the infrared region.

(Appendix 5)
A first pixel that is stored in the data storage unit and is obtained by photographing the survey target area of the plant type and includes spectral data for each pixel, and the spectral intensity in the spectral data is not saturated. And extracting a second pixel having spectral data that is the same as or similar to a pixel corresponding to a position whose plant type is known from the first pixel, and extracting the first pixel from the spectral data of the second pixel. A first determination processing unit that generates reference spectrum data and determines a plant type for the first pixel based on the first reference spectrum data;
From the second image, which is stored in the data storage unit, obtained by photographing the investigation target area of the plant type after dimming the infrared region, and including spectral data for each pixel, A third pixel other than the first pixel is extracted, the infrared data is restored to the spectral data of the third pixel, and the plant type is located at a known position from the third pixel. A fourth pixel that is the same as or similar to the corresponding pixel and has the processed spectral data to be restored is extracted, and a second reference spectrum is extracted from the restored spectral data of the fourth pixel. A second determination processing unit that generates data and determines a plant type for the third pixel based on the second reference spectrum data;
An information processing apparatus.

101 1st data storage part 110 1st determination part 111 1st pixel extraction part 112 2nd pixel extraction part 113 1st reference spectrum generation part 114 1st collation process part 120 2nd determination part 121 3rd pixel extraction part 122 Restoration process Unit 123 third data storage unit 124 fourth pixel extraction unit 125 second reference spectrum generation unit 126 second collation processing unit 130 second data storage unit 140 synthesis unit 150 discrimination result storage unit 160 output unit

Claims (5)

From the first image that is stored in the data storage unit and obtained by photographing the survey target area of the plant type and includes the spectrum data for each pixel, the first pixel whose intensity in the spectrum data is not saturated is obtained. Extract and
From the first pixel, a second pixel having spectral data that is the same as or similar to a pixel corresponding to a position where the plant type is known is extracted;
Generating first reference spectral data from spectral data of the second pixel;
For the first pixel, a plant type is determined based on the first reference spectrum data,
From the second image, which is stored in the data storage unit, obtained by photographing the investigation target area of the plant type after dimming the infrared region, and including spectral data for each pixel, Extract a third pixel other than one pixel,
Performing a process of restoring the infrared region on the spectral data of the third pixel;
From the third pixel, extract a fourth pixel that is the same as or similar to a pixel corresponding to a position where the plant type is known and that has the processed spectrum data to be restored,
Generating second reference spectrum data from the restored spectrum data of the fourth pixel,
A program for causing a computer to execute a process of determining a plant type for the third pixel based on the second reference spectrum data. The process of extracting the second pixel includes
The program according to claim 1, further comprising: extracting pixels having spectral data included in a standard deviation range from an average of spectral data of pixels corresponding to positions where the plant type is known. The process of extracting the fourth pixel includes
The process of extracting the pixel which has the spectrum data after the said process which is contained in the width | variety of a standard deviation from the average of the spectrum data after the said process which restore | restores the pixel corresponding to the said plant type known position, The process is included. The program according to 1 or 2. The process to restore is
The program according to any one of claims 1 to 3, including a process of multiplying an inverse of a value obtained by subtracting a dimming ratio from 1 to an intensity of each wavelength in the infrared region. From the first image that is stored in the data storage unit and obtained by photographing the survey target area of the plant type and includes the spectrum data for each pixel, the first pixel whose intensity in the spectrum data is not saturated is obtained. Extracting, extracting from the first pixel a second pixel having spectral data that is the same as or similar to a pixel corresponding to a position whose plant type is known, and from the spectral data of the second pixel, a first reference A first determination processing unit configured to generate spectrum data and determine a plant type based on the first reference spectrum data for the first pixel;
From the second image, which is stored in the data storage unit, obtained by photographing the investigation target area of the plant type after dimming the infrared region, and including spectral data for each pixel, A third pixel other than the first pixel is extracted, the infrared data is restored to the spectral data of the third pixel, and the plant type is located at a known position from the third pixel. A fourth pixel that is the same as or similar to the corresponding pixel and has the processed spectral data to be restored is extracted, and a second reference spectrum is extracted from the restored spectral data of the fourth pixel. A second determination processing unit that generates data and determines a plant type for the third pixel based on the second reference spectrum data;
An information processing apparatus.