JP5534300B2 - How to create a remote sensing calibration curve - Google Patents

Description

In the present invention, in remote sensing, there are two states where the illuminance is considered to be uniform (hereinafter referred to as cloudy weather) in a state where there is no cloud shadow in the field to be imaged (clear sky) or a cloud shadow in the entire field (cloudy sky) (hereinafter, The present invention relates to a method for creating a calibration curve that can be measured not only in a “uniform illuminance state” but also in a state in which cloud shadows are partially mixed in the field (hereinafter referred to as “cloud uneven state”).

  Conventionally, as described in Japanese Patent Application Laid-Open No. H10-228707, field photography by remote sensing is performed in a uniform illuminance state during fine weather or cloudy weather.

  In addition, as described in Patent Document 2, a calibration curve used for calculating crop information in remote sensing is a calibration curve that is less affected by differences in photographing conditions such as light characteristics of the measurement target crop and sunlight intensity. Therefore, it is considered to create a calibration curve by adding various imaging conditions to explanatory variables.

  However, since the calibration curve creation method described in Patent Document 2 seems to use imaging data in a uniform illuminance state, the calibration curve created by the creation method is used in the cloud uneven state. If the photographic data is calculated, the measurement accuracy may be reduced. For this reason, it is not desirable to perform shooting in a cloud uneven state, and thus the shooting time is limited. Therefore, the cloud unevenness state is a hindrance factor when performing photographing efficiently and systematically.

  By the way, Patent Document 3 describes that image data in a cloud uneven state is converted and used. According to Patent Document 3, when monitoring the pollution state of a water area such as a lake, a histogram is created in advance from reference image data, and the image data of the cloud unevenness state is converted using the histogram. Yes.

  In this method, it is considered that the reference image data and the image data in the cloud uneven state need to be images captured in the same imaging range. As in Patent Document 3, there is no problem in the fixed point observation of a water area such as a lake where the change in the measurement target is relatively small. However, in the field, for example, the field crop as the measurement target grows every day. It seems to be necessary to obtain image data as a reference for each growth stage. Therefore, it cannot be said that it is an efficient method in the remote sensing concerning a field.

  For this reason, it is desired that image data captured not only in a uniform illumination state but also in a cloud uneven state can be used in the same manner as image data captured in a uniform illumination state without any special processing. .

JP 2003-009664 A JP 2008-175537 A Japanese Patent Laid-Open No. 3-154851

  In view of the above problems, the present invention enables to use image data taken not only in a uniform illuminance state but also in a cloud uneven state, in the same manner as image data photographed in a uniform illuminance state, without special processing. It is an object of the present invention to make it possible to create a calibration curve corresponding to image data not only in a uniform illumination state but also in a cloud uneven state.

In order to solve the above-mentioned problems, the present invention provides a remote sensing calibration curve calculation method for calculating crop information in a field, and receives reflected light from crops growing in a plurality of areas in the field. and measuring for each of the areas by the imaging by the crop information crop in the field, a step of further determining the chemical analysis for each of the zones, the explanatory variable information of the reflected light obtained by the imaging, the crop information as an objective variable, and a step of calculating a regression coefficient of multiple regression equation determined in advance, the measurements are fine weather conditions, clouds shade cloudy conditions and cloud partially mixed in the field to be photographed Meanwhile the Hare rows in the weather of the states in uneven state, the calculation, together with the use of the reflected light information on weather in each state in the explanatory variables, the use of multiple information of the reflected light at the weather in each state Technical means Taken it was.

In addition, in the method of creating a calibration curve for remote sensing , a technical means was used in which weather in each of the clear sky state, the cloudy state, and the cloud unevenness state was used as a dummy variable.

Moreover, a correction method for improving the measurement accuracy of the calibration curve prepared by the method of creating according to 請 Motomeko 1 or 2, the reflected light from the crops are growing in the field, shooting by the camera And applying the information of the reflected light obtained by the imaging to the provisional calibration curve in which the constant term of the calibration curve or / and the coefficient of the dummy variable is 0, and the area of the constant area in the field Calculating first crop information of a crop growing in the plant, measuring second crop information of a crop growing in the area by a method of directly irradiating the leaf blades of the crop, and Including a step of calculating a difference between the first crop information and the second crop information, and using the difference as a constant term value of the provisional calibration curve as a calibration curve after correction Took appropriate measures.

And the technical means of measuring the crop by remote sensing using the calibration curve corrected by the correction method was taken.

  According to the present invention, in the multivariate analysis at the time of creating a calibration curve, not only the illuminance uniform state but also the image data of the cloud uneven state is used, so that a calibration curve considering the cloud uneven state can be created. For this reason, it has become possible to take a picture of a farm field even in a cloud uneven state, which was not able to be taken in the past, and to perform the work efficiently in a planned manner.

  In the present invention, since the value corresponding to the bias component (constant term) of the calibration curve is obtained using the measurement value obtained by directly measuring the crop so as not to be affected by sunlight or the like, the measurement accuracy can be improved. It has become possible.

It is explanatory drawing which showed the imaging | photography state. It is a figure which shows the state which divided the image | photographed image of the agricultural field into the several area of a fixed area. It is the figure which showed three types of states, "Sunny", "Cloudy", and "Cloudy unevenness". It is the figure which showed the measurement accuracy of the new calibration curve at the time of fine weather. It is the figure which showed the measurement accuracy of the new calibration curve at the time of cloudy weather. It is the figure which showed the measurement accuracy of the new calibration curve at the time of cloud irregularity. It is the figure which showed the measurement accuracy of the old calibration curve at the time of fine weather. It is the figure which showed the measurement accuracy of the old calibration curve at the time of cloudy weather. It is the figure which showed the measurement accuracy of the old calibration curve at the time of cloud nonuniformity.

  The best mode for carrying out the present invention will be described below with reference to the drawings. In creating a calibration curve, first, a method for photographing a field by remote sensing will be described. Here, the case where the crop growing in the field is rice and the measurement object is the leaf nitrogen content of the rice (rice) will be described as an example.

  The present invention is not limited to paddy rice, and can be used in the case of crops such as upland rice and wheat, and the measurement object is not limited to the nitrogen content of the crop, and the grain after harvesting. Can also be used to create a calibration curve such as the predicted protein value.

  Since remote sensing for photographing a farm field obliquely from above with a camera installed on the ground is described in Patent Document 2 and Patent Document 3, only the outline will be described here. FIG. 1 shows an example of photographing a field 1, and a camera 3 is installed at a predetermined position toward the field 1 where paddy rice is grown. Note that the field 1 is exposed to sunlight, and the imaging range of the camera 3 includes the entire field 1. The camera 3 is connected to the data processing device 20. The data processing device 20 may be any device that can process image data captured by the camera 3. For example, a general personal computer can be used.

  In FIG. 1, the data processing device 20 is built in the lifting device 2, and the lifting device 2 is loaded on the truck 4. The camera 3 is moved up and down by the lifting device 2. By the way, the farm field 1 here may be a single farm field divided by “cold” or a “farm field group” including a plurality of farm fields.

  Reference numeral 21 denotes a portable measuring device, which will be described later, that measures crop information of the crop by a method of directly irradiating the leaf blades of the crop.

  The camera 3 includes a CCD, and the CCD receives reflected light from sunlight (natural light) from the paddy rice through a lens in the camera 3. Image data (reflected light) obtained by the CCD receiving light is sent to the data processing device 20. The camera 3 can shoot RGB R (red) signals, G (green) signals, B (blue) signals, and infrared (or near infrared) signals.

  The camera used in the present invention is not particularly limited as long as the photographed information can be digitized into the same state as the image data photographed by a digital camera equipped with a CCD. In addition, using two cameras, R signal, G signal and B signal with one camera, NIR signal in near infrared region (NIR) or IR signal with infrared region (IR) with the other camera. You may make it acquire.

  What is photographed by the camera 3 is the reflected light of the rice plant in the field 1 due to sunlight. The information of the reflected light obtained by receiving the light is sent to the data processing device 20 after photographing. The incident light to the camera 3 is different between the reflected light from the field to the camera 3 and the reflected light from the field near the camera 3 and the reflected light from the field far from the camera 3. Therefore, it is preferable to correct a difference in incident light due to different incident angles.

  In the present invention, the calibration curve is created using the information (image data) of the reflected light of the rice plant sent to the data processing device 20. As shown in FIG. 2, the information on the reflected light obtained by receiving the reflected light of the rice plants in the field 1 is divided into a plurality of areas of a certain area and grown in each area of the certain area. Find the average value of the reflected light of each rice plant.

  In addition, rice leaves are collected from a plurality of rice plants growing in each area of the predetermined area, and the nitrogen content of the collected rice leaves is directly determined by chemical analysis such as Dumas method, and the average value of the nitrogen content is calculated. Calculate for each area.

  In each area of the constant area, the average value of the reflected light is an explanatory variable, the average value of the nitrogen content is an objective variable, and the factor due to clouds is a dummy variable, To do.

  A calibration curve is created by performing multivariate analysis using a plurality of sample data from each area in the field 1.

  In FIG. 2, since the field 1 is divided into 30 areas, 30 pieces of sample data can be obtained from one image obtained by photographing the field 1. In addition, when the said image has image | photographed the agricultural field group containing a several agricultural field, each agricultural field can be made into one area.

  By the way, since sample data obtained from one image basically has a common dummy variable, it is necessary to take an image for each type of factor caused by clouds. There are three possible causes of clouds: “clear weather” with no clouds, “cloudy weather” completely covered with clouds, and “cloud unevenness” that partially includes shadows of clouds. For this reason, it is necessary to shoot each of three types of states, “sunny”, “cloudy”, and “cloud unevenness”, and obtain sample data of dummy variables corresponding to the factors caused by clouds.

  Here, with reference to FIG. 3, “fine weather”, “cloudy weather”, and “cloud unevenness” in the present invention will be described. “Sunny” indicates that the field to be photographed is completely free of shadows due to clouds (FIG. 3 (a)). “Cloudy” means that the field to be photographed is completely covered with clouds and the entire field is cloudy. The state which exists is shown (FIG.3 (b)). Then, “cloud unevenness” indicates a state in which a shadow by a cloud exists in a part of the field and the shadow and a “clear sky” part are mixed as shown in FIG. Note that the shaded area in FIG. 3 represents the shaded portion of the cloud.

  Next, a method for creating a calibration curve will be specifically described. In the calibration curve creation method of the present invention, for example, a calibration curve is created by the multiple regression equation (Formula 1) shown below.

  In Equation 1, N is the nitrogen content of the measurement object, F0, F1, F2, F3, F4, E1, and E2 are multiple regression coefficients, R1, R2, R3, and R4 are the reflected light information for each wavelength, and D1 , D2 represents a dummy variable.

  The nitrogen content N is a value obtained by collecting rice leaves from a plurality of rice plants growing in a certain area in the field, and obtaining and averaging the nitrogen contents of the plurality of rice leaves by chemical analysis. Therefore, the nitrogen content is an objective variable of sample data in the area.

  The information of each reflected light represented by R1, R2, R3, and R4 is an average value for each area of the reflected light of rice grown in the area. In the present invention, as reflected light information, R signal of RGB signal, G signal, B signal, and reflected light of four wavelengths in the NIR signal, which is a wavelength in the near infrared region, are measured. It is used as an explanatory variable for sample data in the area.

  D1 and D2 are dummy variables that take “1” or “0” according to three shooting conditions. In this embodiment, a dummy variable is provided for each image obtained by photographing a farm field. There are three possible causes of clouds: “clear weather” with no clouds, “cloudy weather” completely covered with clouds, and “cloud unevenness” that partially includes shadows of clouds. Therefore, dummy variables (D1 and D2) are provided according to the weather, and combinations of dummy variables as shown in Table 1 are used.

  The information (R1, R2, R3, and R4) of each reflected light in Equation 1 is preferably logarithmic, and it is desirable to perform multivariate analysis using a multiple regression equation as shown in Equation 2.

  By the way, in this embodiment, a calibration curve is created by multivariate analysis using dummy variables. After the calibration curve is created, the field is photographed, and the nitrogen content of the crop is obtained from the image data obtained by photographing. When calculating a value such as using the calibration curve, it is necessary to input “1” or “0” to each dummy variable. If the number of images is small, this is not a big problem. However, if the number of images is large, it is necessary to input “1” or “0” for each dummy variable for each image. .

  In order to solve this problem, in the present invention, when actually using the calibration curve represented by Formula 1 or 2, it is necessary to input “1” or “0” to each dummy variable for each image. While omitted, a method for improving the measurement accuracy is implemented.

  The method will be described. If “1” or “0” is not input to the dummy variable of the formula 1 or 2, the sum of the constant term represented by the multiple regression coefficient F0 and each constant term represented by the multiple regression coefficients E1 and E2 It is difficult to calculate (hereinafter referred to as “intercept”, which is a linear function, referred to as “bias”). Therefore, an accurate bias cannot be calculated unless “1” or “0” is input to the dummy variable. Therefore, in the method, a value corresponding to the bias is separately obtained.

  The bias component is obtained by using, for example, a portable leaf component measuring device (hereinafter referred to as “measuring device 21”) as described in JP-A-10-96692. According to the measuring device 21, it is possible to easily measure the nitrogen content of the rice growing in the field, and to directly measure the rice leaves and to avoid the influence of sunlight etc. The content rate can be determined. In addition, about how to obtain | require a bias component, you may obtain | require by the method generally known other than using an apparatus like the measuring apparatus 21. FIG.

  Next, how to obtain the bias amount using the measuring device 21 will be described. In this case, Formula 3 shown below obtained by removing the term for calculating the bias from Formula 2 created by the above-described calibration curve creation method is used as the provisional calibration curve. Naturally, instead of the formula 2, a formula obtained by removing the term for calculating the bias amount from the formula 1 may be used.

  Here, it is assumed that the field 1 is captured by the camera 3 in a state where the provisional calibration curve (Equation 3) is stored in the data processing device 20. In this case, as shown in FIG. 1, the nitrogen content of the rice growing in the area Y of a certain area in the field 1 is the information of each reflected light obtained by the photographing from the crops in the area Y. The information can be calculated by substituting the reflected light information into the temporary calibration curve. It is assumed that this value is 3.0% (hereinafter referred to as “first crop information”). The shape and area of the area are not particularly limited, but considering that image data is processed by a personal computer, the shape is preferably a quadrangle, and a square is particularly preferable. When the shape is a quadrangle, the length of one side is preferably about 8 m.

  Next, it is assumed that the nitrogen content of the rice growing in the area Y is 3.8% (hereinafter referred to as “second crop information”) as measured by the measuring device 21. The measured value by the measuring device 21 is an average value obtained by measuring a plurality of rice leaves growing in the area Y. The measurement with the measuring device 21 may be performed by measuring one leaf per rice body, and the number of the leaves is preferably 20 or more. More preferably, the number is 25 or more. Further, the measurement value (average value) may be obtained by performing chemical analysis instead of the measurement by the measurement device 21.

  Since the first crop information can be stored in the data processing device 20 and the second crop information is stored in the measuring device 21, the second crop information is input to the data processing device 20 to obtain data. Store in the processing device 20. The input method may be either automatic or manual. The data processing device 20 calculates a difference between the first crop information and the second crop information. Here, since the first crop information is 3.0% and the second crop information is 3.8%, the first crop information is 0.8% lower than the second crop information. . Therefore, since the difference is + 0.8%, correction of + 0.8% is necessary for the calculation result by the provisional calibration curve. Therefore, in this embodiment, this difference (+ 0.8%) is used as the bias (constant term) of the provisional calibration curve. Therefore, the provisional calibration curve is expressed by the following Equation 4,

It becomes. The calibration curve expressed by Equation 4 can be used to determine the nitrogen content of the rice plant in the field 1 taken to determine the bias. Therefore, by substituting the information (R1, R2, R3, and R4) of the reflected light in the areas other than the area Y into the calibration curve (Formula 4), the nitrogen content of the rice bodies in each area can be easily obtained. Can do. The calibration curve can be used for image data obtained by photographing a field different from the field 1, but there is a risk that the measurement accuracy may be reduced. The field 1 is not limited to a single field, and may be a field group including a plurality of fields. In this way, it is possible to easily measure the crop information of the crops cultivated in the field (here, the nitrogen content of rice) over a wide range. In general, the average value of the nitrogen content is obtained for each field, and this average value is used as the nitrogen content of each field.

  In order to confirm the effect of this example, the above-mentioned calibration curve (hereinafter referred to as “new calibration curve”) created using image data of “clear weather”, “cloudy weather” and “cloud unevenness”, “clear weather” and Comparison was made with a conventional calibration curve created using “cloudy sky” image data (hereinafter referred to as “old calibration curve”).

  In creating the old calibration curve, two images of the same field taken with the camera 3 from the same position were used. These two images are a “sunny” image taken on September 16, 2008 and a “cloudy” image taken on September 18, 2008. The field was divided into 30 areas, 30 sample data were created from one image, and a total of 60 sample data were obtained. Then, an old calibration curve was created by multivariate analysis (multiple regression analysis) using these 60 sample data and the above mathematical formula 2. The sample data is obtained by the method described in the above embodiment.

  Further, the bias of the old calibration curve is obtained by calculating the value calculated by the temporary calibration curve for one area of the total of 60 sample data, using the formula expressed by Equation 3 above as a temporary calibration curve. It was set as the value calculated | required from the difference of the value and the chemical analysis value of the said area.

  In order to create a new calibration curve, in addition to the “clear sky” image and “cloudy sky” image used to create the old calibration curve, a “cloud” obtained by photographing the same field as those images with the camera 3 from the same position. "Mura" image was used. The shooting was performed on September 18, 2008. As in the previous calibration curve creation, the field was divided into 30 sections, 30 sample data were created from one image, and a total of 90 sample data were obtained. Then, a new calibration curve was created by multivariate analysis using these 90 sample data and Equation 2 above. The sample data is obtained by the method described in the above embodiment. Further, the total 60 sample data created from the two images of “clear sky” and “cloudy sky” are the same as those used when the old calibration curve was created.

  Similarly to the old calibration curve, the bias of the new calibration curve was calculated with the provisional calibration curve for one area of the sample data of 90 in total, using the expression expressed by Equation 3 above as the provisional calibration curve. A value was obtained, and a value obtained from the difference between this value and the chemical analysis value of the area was used.

  By the way, the new calibration curve and the old calibration curve have the objective variable as the leaf blade nitrogen content of rice plant, the explanatory variable as the reflected light of sunlight (natural light) from the rice plant, “Cloudy sky” and “cloud unevenness” are obtained as dummy variables. The cultivar (rice) in the field is Hinohikari and the growing season is the ripening season.

  For verification comparison of the old calibration curve and the new calibration curve, three images different from the image used at the time of preparing the calibration curve were used. These three images are composed of a “fine weather” image S, a “cloudy” image D, and a “cloud unevenness” image C. These three images S, D, and C are taken by the camera 3 when the image used for creating the calibration curve is taken, and the same field as the image is taken from the same position. ing.

  Then, for each of the images S, D and C, the field in the image is divided into 30 areas in the same manner as when creating the calibration curve, and the crop leaves are divided into the old calibration curve and the new calibration curve for each area. The body nitrogen content was calculated, and a comparison was made using the values obtained by calculation and the chemical analysis values of 30 areas. The results are shown in FIGS. 4 to 9 and Table 2.

  FIGS. 4 to 6 show the case where a new calibration curve is used, in which FIG. 4 shows the measurement accuracy in fine weather, FIG. 5 shows the measurement accuracy in cloudy weather, and FIG. 7 to 9 show the case where the old calibration curve is used, in which FIG. 7 shows the measurement accuracy in clear weather, FIG. 8 in cloudy weather, and FIG. 4 to 9 are graphs each having a chemical analysis value on the horizontal axis and a calculated value on the vertical axis.

  Table 2 shows the correlation coefficient and standard error between the values calculated with the old calibration curve and the new calibration curve, and the chemical analysis values for the image S in fine weather, the image D in cloudy weather, and the image C in cloudy weather. (SEP) is shown. According to Table 2, there is no significant difference between the old calibration curve and the new calibration curve at the time of fine weather and cloudy weather. However, at the time of cloud unevenness, the correlation coefficient of the old calibration curve is 0.779 and the standard error is 0.314, while the measurement accuracy is lowered, whereas the correlation coefficient of the new calibration curve is 0.942 and the standard error It was 0.084, and the measurement accuracy was equal to or better than that in fine weather and cloudy weather. For this reason, it can be said that the new calibration curve can be sufficiently used even in a cloud uneven state.

  In the method of creating a calibration curve of the present invention, as shown in Equation 5 below, the reflected light information of the R (red) signal, the reflected light information of the G (green) signal, and the reflected light of the B (blue) signal And a multiple regression equation that takes the logarithm of the ratio of the information of the reflected light of the NIR signal.

  Further, in the method of creating a calibration curve of the present invention, without using a dummy variable, multivariate is simply obtained by mixing image data in clear weather, image data in cloudy weather, and image data in cloudy unevenness. Analysis may be performed to create a calibration curve. In this case, any one of the mathematical formulas 6, 7 and 8 may be used.

  When a calibration curve is created using any one of Equations 6 to 8, since no dummy variable is used, it is not necessary to input a dummy variable when using the created calibration curve. Note that the bias component may be obtained separately by the measuring device 21 or the like.

  The method for creating a calibration curve according to the present invention can be used not only for photographing from the ground but also for remote sensing in which a camera is mounted on a balloon, a radio-controlled flying device (airplane, helicopter) or a manned airplane to photograph a field. is there. The photographing from the ground means photographing with a camera so that the field is looked down from a height of 1 m to 30 m from the ground, and it is desirable to use a telephoto lens when photographing far away. By using a telephoto lens, it becomes possible to secure resolution at the time of analysis up to an image of a far image, and a wider range can be measured at a time.

DESCRIPTION OF SYMBOLS 1 Farm 2 Lifting device 3 Camera 4 Truck 20 Data processing device 21 Measuring device

Claims (4)

A method for creating a remote sensing calibration curve for calculating crop information in a field, wherein the reflected light from a crop growing in a plurality of areas in the field is measured for each of the areas by photographing with a light receiving means When,
The crop information of the crop in said field, a step of determining more chemical analysis for each of the zones,
The explanatory variable information of the reflected light obtained by photographing, as a target variable the crop information, and a step of calculating a regression coefficient of multiple regression equation determined in advance,
The measurement is sunny conditions, whereas earthenware pots row in weather each state of cloud uneven state shade cloudy conditions and cloud partially mixed in the field to be photographed, said calculating, the reflected light at the weather in each state A method for creating a calibration curve for remote sensing , wherein a plurality of pieces of information on reflected light in the weather in each state is used . In creating a claim Remote Sensing of the calibration curve described in 1, sunny conditions, how to create a calibration curve of the remote sensing, which comprises using as the cloudy state, and the dummy variable weather of the states in clouds uneven state. A calibration curve correction method created by the creation method according to claim 1 or 2,
Measuring the reflected light from the crop growing in the field by photographing with a camera;
Applying the reflected light information obtained by the imaging to a tentative calibration curve where the constant term of the calibration curve or / and the coefficient of the dummy variable is 0, a crop growing in a certain area in the field Calculating the first crop information of
Measuring second crop information of a crop growing in the area by a method of directly irradiating the leaf blades of the crop;
Calculating a difference between the first crop information and the second crop information,
A calibration curve correction method, wherein the difference is the value of the constant term of the provisional calibration curve, which is a corrected calibration curve. A method for measuring a crop by remote sensing, wherein a calibration curve corrected by the calibration method for a calibration curve according to claim 3 is used. "

Images