Portable surface-roughness measurement method considering waviness

Abstract

[Problem] To provide a scheme that both minimizes the effect of waviness and adjusts position, posture, and scale when using only two-dimensional imaging to evaluate roughness in a measured surface having macroscopic three-dimensional waviness [Solution] A substrate is used that integrally supports a sampling surface capable of elastically following the measured surface and transferring and retaining fine structures, and an identification marker including a reference pattern of known geometry. After attaching and peeling off the sampling surface as a target, the sampling surface and the reference pattern are acquired by a single imaging, planarized by projection homography based on reference points, and subjected to absolute scale-calibration using known dimensions. Additionally, structures are then extracted from the sampling surface region, and at least the nearest neighbor distance average is calculated as a roughness index.

Description

Portable surface roughness measurement method taking waviness into account

[0001] The present invention relates to a technique for measuring the roughness of a material surface, and more particularly to a portable measurement method for suppressing the influence of macroscopic waviness of the surface to be measured by geometric correction using identification markers, and for obtaining a roughness index from a two-dimensional image of the surface to be measured transferred onto a sampling surface.

[0002] Surface roughness can be measured in one, two, or three dimensions depending on the shape of the object being measured and the purpose. Generally, in two-dimensional measurements, the object surface is treated as a reference plane, and the unevenness on that plane is recorded.

[0003] On the other hand, many object surfaces are accompanied by macroscopic irregularities (waviness) with a period longer than the reference length inherent to the roughness of the surface of the material being observed. Conventionally, atomic force microscopes and confocal laser microscopes have been known as devices for measuring such three-dimensional surface shapes, but they have issues such as high cost and large size of the device, burdensome pre-processing, and long scanning times. Here, "roughness" in this specification refers to the short wavelength component relative to the reference length inherent to the roughness of the object being observed, and "waviness" refers to the component with a longer wavelength than the roughness.

[0004] In recent years, technologies have been proposed that emphasize the speed of measurement and aim to increase the speed only in two dimensions. For example, Patent Document 1 discloses a technology for high-speed two-dimensional measurement of the roughness of a rolling roll surface using a non-contact sensor, but this technology is based on the assumption that the surface is flat, and is difficult to apply to objects that are affected by waviness.

[0005] Even in three-dimensional measurements that take undulation into account, there remain two challenges: (Problem 1) a method for recording unevenness while taking undulation into account, and (Problem 2) a method for accurately adjusting the position, orientation, and scale relationship between the measuring device and the object.

[0006] JP 2024-89852 A

[0007] Ueda, S. ;Ye, X. , arXiv:2404.19513, 2024. Barbour, J. etal. , Entomol. Exp. Appl. , 60:289-300, 1991. Hoffland, E. etal. , J. Chem. Ecol. , 26:2697-2711, 2000. Li, Q. etal. , ASABEAnnualInt. Meeting, 2014. Sobral, M. etal. , PNAS, 118(33): e2005865118, 2021. Wilkens, R. T. etal. , Oecologia, 106:181-191, 1996. Galdon-Armero, J. etal. , PlantJ. , 96(3):607-619, 2018. Escobar-Bravo, R. etal. , PlantCellPhysiol. , 59(12):2462-2475, 2018. Hoffland, E. ; Jeger, M.; J. ; vanBeusichem, M.; L. , PlantSoil, 218:239-247, 2000.

[0008] The descriptions of prior art documents in this specification are intended to aid in understanding the technical background, and are not intended to be an admission by the applicant that all or part of the matters described in the documents were well known or publicly known at the time of filing this application, or that they impair the novelty or inventive step of the present invention.

[0009] The problems to be solved are that in conventional three-dimensional measurements that take undulation into consideration, in order to solve the above (Problem 1), it is necessary to use high-precision sensors and complex signal processing algorithms, or to arrange a large number of sensors, and in order to solve the above (Problem 2), it is necessary to use precise positioning mechanisms and jigs, which makes the equipment expensive and large-scale, and limits the measurement environment to limited locations such as factories and research laboratories.

[0010] The present invention provides the following technical means to solve the above (Problem 1) and (Problem 2). For the sake of convenience in this specification, the process symbols S1 to S7 correspond one-to-one to (A) to (G) of claim 1.

[0011] According to a first aspect of the present invention, there is provided a method for measuring the roughness of a surface to be measured, the method comprising: (S1) using a measuring instrument comprising: a sampling surface that conforms to and comes into close contact with the surface to be measured by elastic deformation, and that is capable of transferring and retaining a finer roughness than a waviness component of the surface to be measured; (S2) an identification marker that includes a reference pattern of known geometry used to estimate the position, orientation, and magnification of the sampling surface; and (S3) a base that supports (S1) and (S2) together; (S4) bringing the sampling surface into contact with the surface to be measured to form a transferred pattern of the fine structure of the surface; and (S5) peeling the sampling surface from the surface to be measured, shaping the sampling surface to such an extent that it can be retained in the same plane as the reference pattern on a flat reference surface provided on the base, and measuring the roughness of the surface to be measured by an optical sensor. The method includes: (S6) calculating a projection homography based on the correspondence of the non-collinear reference points of the reference pattern or the independent geometric constraints, and flattening the sampling surface image using the homography, while performing absolute scale calibration based on the known dimensions, and further correcting for geometric distortion or response distortion specific to the sensor as needed; and (S7) extracting a region corresponding to the sampling surface from the adjusted dataset and calculating an index including at least a nearest neighbor distance average as a roughness index.

[0012] In the above-described invention, (S1) is not particularly limited as long as it is a material capable of transferring the structure and presence of the structure on the surface of the object. For example, adhesive or painted tape, film, sheet, etc. can be used. Furthermore, (S1) can transfer the uneven shape of the surface of the object by adhering to it. This allows the unevenness of the surface of even an object with three-dimensional undulations to be recorded as two-dimensional information, thereby solving (Problem 1), which is to record unevenness taking into account the undulations of the surface of the object. Furthermore, (S2) is, for example, a two-dimensional marker, barcode, specific pattern, or shape of (S3) printed on (S1), and its position information can be identified by image processing. Furthermore, by using (S2) provided in (S1), the positional relationship between the object and the optical sensor can be adjusted, thereby solving (Problem 2).

[0013] According to a second aspect of the present invention, there is provided a method for measuring the roughness of a material surface, characterized in that in the first aspect of the present invention, the step (S7) includes a step of calculating, based on the spatial distribution state of the structures transferred to the sampling surface, an index that indicates a spatial distribution in which the influence of the defects is suppressed, even if the structures on the surface to be measured are partially defective.

[0014] A method for measuring the roughness of a surface of a material according to a second aspect of the present invention divides the entire two-dimensional spatial distribution of the structure transferred in (S1) into arbitrary partial regions. For each divided region, an index representing the spatial distribution of the structure is calculated and recorded, and this process is repeated recursively the number of times equal to the number of divided regions. A representative value is calculated from the finally obtained distribution of indices to evaluate the roughness of the surface to be measured.

[0015] The spatial distribution index is not particularly limited as long as it reflects the spatial characteristics of the structures, such as the number of structures, the area of ​​the structures, or the distance between structures.

[0016] In this way, by statistically analyzing the distribution of structures in a partial region rather than the entire sampling surface, the influence of partial loss of structures on the surface to be measured can be suppressed, enabling more accurate roughness measurement.

[0017] According to a third aspect of the present invention, in the first or second aspect of the present invention, the measurement target surface is the surface of a plant, and prior research has reported (prior findings 1) that the surface structure of a plant can change in response to stress conditions such as nutritional or environmental stress. Utilizing prior findings 1, there is provided a method for diagnosing stress in a plant, comprising: evaluating the stress state of the plant based on the roughness of the measurement target surface.

[0018] The above-mentioned prior finding 1 has been reported in (Non-Patent Document 2) to (Non-Patent Document 8).

[0019] According to a fourth aspect of the present invention, there is provided a nutritional diagnosis system according to the third aspect of the present invention, wherein the measurement target surface is the surface of a plant under stress, and which utilizes the prior knowledge 1 and comprises an optical sensor, an arithmetic processing device, and a computer program executable by the arithmetic processing device, wherein the plant stress diagnosis method is executed by the computer program, and the computer program acquires data output from the optical sensor, evaluates the nutritional state of the plant, and outputs the evaluation result.

[0020] A nutritional diagnostic system according to a fourth aspect of the present invention measures the surface roughness of a plant according to the plant stress diagnostic method according to the third aspect of the present invention. The measured roughness data is then output from the optical sensor and input to a processing device, and a computer program evaluates the nutritional state of the plant based on the input roughness data and outputs the evaluation result.

[0021] According to a fifth aspect of the present invention, in the third aspect of the present invention, the measurement target surface is the surface of a plant under stress, and prior research has reported that a plant's attractiveness to pests can change depending on the stress state (prior findings 2). Utilizing prior findings 2, there is provided a pest risk diagnosis system comprising an optical sensor, a processing device, and a computer program executable by the processing device, wherein the plant stress diagnosis method is executed by the computer program, and the computer program acquires data output from the optical sensor, evaluates the pest risk of the plant, and outputs the evaluation results.

[0022] In the pest risk diagnosis system according to the fifth aspect of the present invention, the surface roughness of the plant is measured according to the plant stress diagnosis method according to the third aspect of the present invention.

[0023] Next, the measured roughness data is output from the optical sensor and input to a processing device, and the computer program evaluates the pest risk of the plant based on the input roughness data and outputs the evaluation results.

[0024] The above-mentioned prior finding 2 has been reported in (Non-Patent Document 3), (Non-Patent Document 6), and (Non-Patent Document 9).

[0025] The method for measuring the roughness of a material surface according to the first aspect of the present invention can realize portable surface roughness measurement that takes waviness into account.

[0026] By using simple measuring tools and image processing technology, it is now possible to accurately measure roughness taking into account the waviness of the surface being measured in a variety of locations and environments, without using the expensive, large three-dimensional measuring machines that were used in the past.

[0027] This can be said to be a dimensional compression technique that enables processing of a three-dimensional uneven shape as two-dimensional information by transferring it to (S1).

[0028] This eliminates the need for the vertical scanning mechanism previously required for three-dimensional measuring devices, and the complex calculations that previously required taking into account information in the vertical direction relative to the reference plane can now be performed using only information in the planar direction, making it possible to make the device smaller and simpler and to reduce processing time.

[0029] In the case of roughness measurement by the method of the second aspect of the present invention, by comparing the rate of change of the roughness measurement value by the method of the second aspect of the present invention with the method of simply counting the structures on (S1) when a structure is partially damaged, it can be shown that the method of the second aspect of the present invention is a more robust measurement method that is less susceptible to the effects of damage.

[0030] According to the plant stress diagnosis method relating to the third aspect of the present invention, by applying the roughness measurement technology of material surfaces to plants, it becomes possible to easily evaluate the stress state of plants, which has previously been difficult to do.

[0031] This was achieved by using a new index of surface roughness according to the present invention, based on the knowledge that the surface structure of plants changes depending on the stress state.

[0032] According to the third aspect of the present invention, in plants with developed leaf surface structures, stress diagnosis can be performed more quickly, at lower cost, and accurately than conventional methods such as visual observation, which requires skilled techniques, or biochemical analysis, which requires expensive reagents, equipment, and complicated processes.

[0033] According to the nutritional diagnostic system of the fourth aspect of the present invention, by using the plant stress diagnostic method of the third aspect of the present invention, the nutritional state of a plant can be evaluated based on an objective indicator, namely, surface roughness.

[0034] This makes it possible to evaluate nutritional status based on more objective data, rather than the vague judgments that previously relied on subjective factors such as experience and intuition, and more appropriate fertilizer management can be carried out based on the evaluation results.

[0035] According to the pest risk diagnosis system of the fifth aspect of the present invention, by using the plant stress diagnosis method of the third aspect of the present invention, the pest risk of a plant can be evaluated based on an objective indicator, namely surface roughness.

[0036] This will enable more flexible pest management that reflects the actual condition of the plants, rather than the conventional pest risk assessment based on control schedules that do not take into account the situation on site.

[0037] The present invention can be widely used for on-site measurements using portable devices, such as nutritional diagnosis and pest management in the agricultural field, surface inspection of industrial materials, and on-site maintenance inspection.

[0038] 3B. (Example 1) A plan view of a measuring instrument according to an embodiment of the present invention. (Example 1) A bottom view of a measuring instrument according to an embodiment of the present invention. (Example 1) (a) The surface of a tomato leaf, (b) The sampling surface after the unevenness of the tomato leaf surface has been transferred, (c) An enlarged view of FIG. 3B. (Example 1) A conceptual diagram of step (S4) of attaching and transferring the sampling surface 11 of the measuring instrument 10 to the measurement target surface 21 of the measurement target 20. (Example 2) (a) Digital data of unevenness record, (b) Unevenness record of the specified sampling surface area, (c) Transferred structures, (d) Number of structures. (Example 2) A diagram comparing the roughness index (NND) of the present invention with the conventional method (Point Count). (Example 3) ROC curve showing the performance of a binary classification model. (Example 4)

[0039] Hereinafter, with reference to the drawings, a detailed description will be given of an embodiment of the present invention by giving a specific example. The specific example is one of the embodiments, and if there are other embodiments, they will be described in the following modified examples. In this specification and drawings, parts having substantially the same functional configuration are assigned the same reference numerals, and their description will be omitted. Definition of Terms

[0040] In this specification, roughness refers to the component of the three-dimensional shape of the surface being measured that has a shorter wavelength relative to the waviness of the object. The spatial wavelength considered to be roughness is typically 0.05 to 1.0 mm, and waviness refers to components exceeding 1 mm, but these can be appropriately determined by those skilled in the art depending on the object. Waviness refers to a macroscopic shape component with a longer wavelength than the roughness. "Transferable" refers to a state in which, after a sampling surface is in close contact with the surface being measured and then peeled off, the pixel contrast or edge response corresponding to the microstructure of the surface being measured is reproduced in an image of the sampling surface captured by an optical sensor, and a roughness index can be calculated based on this reproduced information. A reference pattern of known geometry refers to an information element that can estimate the position, orientation, and magnification through image processing using a figure, symbol, or tool shape with predefined known dimensions and known layout. An identification marker refers to a planar pattern or substrate shape that is located on the same plane as the sampling surface and is used to estimate the position, orientation, and magnification based on known dimensions and layout. Examples include two-dimensional coded markers, checkerboards, dot arrays, hatching patterns, holes, notches, corners, and edges formed in a substrate. A reference pattern refers to a pattern element included in the identification marker that provides four or more non-collinear reference points or independent geometric constraints that can uniquely determine a projection homography. Examples of independent geometric constraints include (i) four or more non-collinear reference points, (ii) mutually orthogonal edge pairs and known ratios, (iii) known lattice cell dimensions and node correspondences, and (iv) a rectangular frame containing a known length scale plus diagonal intersections. Geometry or response distortion refers to geometric distortions such as perspective distortion and signal response distortions such as nonlinear luminance response. A structure refers to an image element (e.g., a point-like, line-like, or area-like feature) that is caused by the micromorphology of the surface to be measured and can be identified as a local intensity change on an image of the sampling surface. When calculating the nearest neighbor distance average, the center of gravity of each connected component is extracted as a representative point, thereby obtaining the structure arrangement as a point cloud. The roughness index is a quantitative value calculated from the structure extraction results, and includes the density of the structures, nearest neighbor distance, co-occurrence matrix feature amount, etc. The existence density refers to the surface density of the structures, expressed by the number of structures per unit area or the area ratio, etc.An optical sensor refers to an imaging device capable of acquiring images (e.g., a digital camera, a smartphone camera, a flatbed scanner, etc.). "Maintaining in the same plane" refers to a state in which the sampling surface and reference pattern are held by being attached or pressed against a flat reference surface of the substrate or by similar shaping, and are contained within the depth of field of the imaging system. In other words, it refers to a state in which the camera is focused and has a degree of planarity that allows image shaping by a single projective transformation. Specifically, it refers to a planarity in which at least the root mean square of the reprojection error of the reference point is below a threshold appropriately set depending on the imaging conditions. A flat reference surface refers to a surface of a planar view object provided on the substrate that supports the sampling surface and reference pattern, and serves as a reference surface for maintaining both in the same plane. "Single imaging" refers to capturing the sampling surface area and the reference pattern through the same imaging operation (same frame). Note that, as needed in this specification, "optional process" refers to the presence or absence of such process being reasonably selectable by a person skilled in the art depending on the imaging conditions and device characteristics.

[0041] FIG. 1 is a plan view of a measuring device according to an embodiment of the present invention, and FIG. 2 is a bottom view thereof.

[0042] The measuring instrument 10 includes a sampling surface 11 and an identification marker 12. In this embodiment, the measuring instrument 10 is made of cardboard.

[0043] The sampling surface 11 is a flexible sheet-like member that can be brought into close contact with the surface of the object to be measured, and is capable of transferring the irregularities of the surface. In this embodiment, the sampling surface 11 is made of cellophane tape, which is a specific example of a measuring tool 2, and is attached to a measuring tool 10 as shown in FIG.

[0044] The identification marker 12 has information for identifying the sampling surface 11 from the uneven record, and is used to estimate the position, orientation, and magnification based on known geometry. Specifically, it is used to identify the sampling surface area (A) and to estimate the orientation and magnification based on known geometry (B). In this example, as a specific example 3 of a measuring tool, the identification marker 12 is printed with high contrast in an adjacent area on a substrate that is coplanar with the sampling surface 11.

[0045] The measuring instrument 10 is provided with the sampling surface 11 and the identification marker 12, and is therefore capable of transferring unevenness taking into account the undulations of the surface to be measured.

[0046] As shown in FIG. 3, in this example, the irregularities present on the surface of a tomato leaf shown in FIG. 3(a) are transferred to the measuring device shown in FIG. 3(b) and its enlarged view, FIG. 3(c).

[0047] Next, an embodiment of the method for measuring the roughness of a material surface according to the first aspect of the present invention will be described with reference to the conceptual diagram of the embodiment shown in FIG.

[0048] (Step S4) In FIG. 4, the sampling surface 11 of the measuring instrument 10 is attached to the measurement target surface 21 of the measurement target 20, thereby transferring the structure on the target surface to the sampling surface 11.

[0049] (Step S5) The sampling surface 11 is peeled off from the measurement object 20 and held substantially flush on the flat reference surface of the substrate, utilizing the property of the sampling surface 11 to return to a substantially flat shape due to elastic recovery. Next, an optical sensor images the sampling surface 11 and the identification markers 12 on the flat reference surface to obtain the reference pattern and the sampling surface area in a single image.

[0050] (Step S6) A reference pattern of known dimensions contained in the identification marker 12 is detected from the captured image, and a projective homography is calculated based on the correspondence of at least four non-collinear reference points or geometric constraints sufficiently independent to uniquely define the projective homography, thereby flattening the sampled surface image. Subsequently, absolute scale calibration is performed based on the known dimensions of the reference pattern. Specifically, the number of pixels corresponding to a known actual length in the flattened image is measured, and a conversion factor (millimeters per pixel) is calculated from the ratio, and various distance quantities are converted to actual lengths. If necessary, geometric distortion or brightness response distortion due to the lens is corrected.

[0051] (Step S7) A region of the sampling surface 11 is identified from the adjusted data set based on the position information of the identification markers 12, and the surface roughness of the material in that region is measured. As a specific example, structures transferred to the sampling surface 11 are detected by image processing, and the coordinates of the center of gravity of each structure are obtained. In that region, at least the average of the nearest neighbor distances is calculated as a roughness index. If necessary, auxiliary indexes such as the number of structures and area ratio may also be calculated. If the flatness criterion is not met, the image is re-imaged, and the process returns to step S5.

[0052] The method for measuring the roughness of a material surface according to the first aspect of the present invention includes the steps S4 to S7 described above, thereby adjusting the positional relationship between the object to be measured 20 and the optical sensor, and realizing roughness measurement that takes into account the waviness of the surface of the object to be measured.

[0053] As shown in Figure 5, when this embodiment is used, digital data of the concave-convex record of a specified area of ​​the sampling surface shown as an example in Figure 5(b) can be obtained from the digital data of the concave-convex record shown as an example in Figure 5(a), the transferred structures shown as an example in Figure 5(c) can be detected, and the number of structures shown as an example in Figure 5(d) can be calculated.

[0054] The pass / fail of the coplanar retention is determined by setting a threshold value for the root mean square (in pixels) of the reprojection error relative to the reference point, and if the error exceeds the threshold value, re-imaging is required. In one embodiment, a root mean square of the reprojection error of 0.5 pixels or less is passed, a value between 0.5 and 1.0 is conditionally passed, and a value exceeding 1.0 is re-imaging. However, this threshold value can be changed as appropriate by a person skilled in the art depending on the imaging resolution, the placement of the reference point, the depth of field, etc.

[0055] In this example, a method for measuring the roughness of a material surface when a structure on the measurement target surface is partially missing will be described according to the second aspect of the present invention.

[0056] This example is an improvement on the method described above (step S7). It detects structures transferred to the sampling surface 11 through image processing and evaluates the roughness of the surface to be measured. Here, only the region corresponding to the sampling surface is masked from the geometrically corrected image, and the coordinates of the center of gravity of each structure are obtained by performing smoothing, adaptive binarization, small object removal, and connected component labeling in that order. For each structure, the Euclidean distance to the nearest other structure is calculated, and the average value of these distances for all structures is defined as the nearest neighbor distance (NND). Compared to simple number density, NND is smoothed by the overall distance distribution even when localized structural defects occur, thereby statistically suppressing the impact of defects. Therefore, it is suitable as the "index that suppresses the impact of defects" described in claim 2. Texture indices based on number density, area ratio, co-occurrence matrix, etc. may also be used in combination as needed.

[0057] (Step S7-1) As a specific example 1 of the defect roughness measurement method, the entire two-dimensional spatial distribution of structures is divided into partial regions including only a certain structure and its nearest neighboring structures.

[0058] (Step S7-2) In each divided region, as a specific example 2 of the defect roughness measurement method, the Euclidean distance between a certain structure and its nearest neighboring structure is calculated and recorded.

[0059] (Step S7-3) The processes of (Step S7-1) and (Step S7-2) are repeated recursively as many times as the number of detected structures as a specific example 3 of the defect roughness measuring method.

[0060] (Step S7-4) As a specific example 4 of the defect roughness measurement method, the average value is calculated from the obtained distance distribution and used as the roughness index of the surface to be measured.

[0061] 6 shows a comparison of the roughness index (NND) of the measurement surface in this example with the roughness index (PointCount) based on the number of structures in step S7, where the roughness index (NND) in this example is a value calculated in steps S7-1 to S7-4. Using simulation data simulating the spatial distribution of various surface structures, a comparison of the absolute rates of change in the roughness index between cases where half of the surface of the material is missing structures and cases where it is not, shows that this example is less susceptible to the effects of missing structures and is a more robust measurement method.

[0062] Next, an example of the plant stress diagnosis method according to the third aspect of the present invention will be described. In this example, a tomato will be used as an example of a specific example 1 of the stress diagnosis method.

[0063] As shown in FIG. 4 , in this embodiment, the sampling surface 11 of the measuring instrument 10 described in Example 2 is first attached to the surface of a target tomato leaf. Then, the surface roughness of the tomato leaf is measured using the same procedures as in Example 2 (steps S4) to S7). Next, the measured surface roughness of the tomato leaf is compared with a pre-prepared surface roughness model of tomato leaves created under various stress doses. In the comparison with the model, the measured surface roughness is input into the model, and the corresponding fertilizer stress dose is evaluated. This diagnostic process does not rely on manual observation or simple mental steps. Instead, the image data acquired from the optical sensor is automatically subjected to the following steps by a computing device: geometric correction and absolute scale calibration based on (S6), mechanical calculation of the nearest neighbor distance average based on (S7), and mechanical judgment processing based on a predetermined trained model or threshold. As a result, the diagnostic results are output from the device as specific control or instructions, such as fertilizer amount adjustment, alarm generation, and automatic generation of a work plan, thereby achieving useful technical effects directly linked to work in real environments.

[0064] As a specific example of stress diagnosis method 2, it can be used to diagnose fertilizer stress among other plant stresses. For fertilizer stress diagnosis, a surface roughness model of tomato leaves created under various fertilizer stress doses is used.

[0065] The method for creating a surface roughness model is as follows. First, tomatoes are grown with different fertilizer concentrations, and the fertilizer stress index and surface roughness of tomato leaves under the various fertilizer concentrations are measured. In this example, the nitrate ion concentration in the leaf sap is used as the fertilizer stress index, as in specific example 3 of the stress diagnosis method. Next, the relationship between the measured surface roughness and the fertilizer stress index is analyzed, and a model representing the relationship between the fertilizer stress dose and surface roughness is created. As in specific example 4 of the stress diagnosis method, the model is formulated as a binary classification model based on whether or not the nitrate ion concentration in the leaf sap exceeds a threshold value.

[0066] FIG. 7 shows an ROC (Receiver Operating Characteristic) curve evaluating the performance of the binary classification model in this example. The horizontal axis represents the false positive rate (FPR), and the vertical axis represents the true positive rate (TPR). AUC (Area Under the Curve) indicates the area under the ROC curve, taking values ​​from 0 to 1, with values ​​closer to 1 indicating higher discrimination ability of the model. FIG. 7 shows the ROC curve of the binary classification model. The horizontal axis represents FPR, and the vertical axis represents TPR. The AUC is 0.81.

[0067] 7 shows the performance evaluation results of a binary classification model created using data acquired by the present inventor based on this example. For a related academic report, please refer to (Non-Patent Document 1).

[0068] 1 shows the configuration of a nutritional diagnostic system according to one embodiment of the present invention. As described in claim 4, the nutritional diagnostic system includes an optical sensor, a processing unit, and a computer program executable by the processing unit. The computer program has a function of causing the processing unit to execute the method according to claim 1 or 2.

[0069] The optical sensor is similar to that described in Example 2 and outputs data regarding the roughness of the plant surface obtained using the measuring device 10.

[0070] The arithmetic processing unit is a data processing unit that acquires the data output from the optical sensor and executes the computer program.

[0071] The computer program causes the arithmetic processing device to execute a process for evaluating the nutritional state of a plant using the plant stress diagnosis method according to the third aspect of the present invention and outputting the evaluation results. As a specific example of a nutritional diagnosis system, the computer program can display the evaluation results on a display or print them from a printer. Furthermore, if necessary, it can generate and output information recommending fertilization based on the evaluation results. The computer program includes a function for inputting a roughness index such as NND, calculating the nutritional state or pest risk using threshold judgment or a trained classifier or regression, and automatically generating and outputting recommended amounts, timing, and alerts for fertilization and pest control based on the results. The computer program of the present invention may be provided by being recorded on a non-transitory recording medium (e.g., semiconductor memory, magnetic disk, optical disk, etc.), or may be distributed from a server to a terminal via a communication line. A client / server configuration may be used in which all or part of the processing is performed on the server side and the remaining processing is performed on the terminal side.

[0072] [Modifications] The present invention is not limited to the above-described first to fifth embodiments, and various modifications are possible within the scope of the present invention. In the present invention, it is desirable that, during imaging, the sampling surface 11 and the identification marker 12 (including a reference pattern of known geometry) are arranged in substantially the same plane on a flat reference surface of the substrate, and that the images are acquired by a single imaging. The following modifications can be made within the scope of satisfying this requirement.

[0073] In the measuring instrument specific example 1, the material of the measuring instrument 10 is not particularly limited. For example, resin, metal, wood, glass, etc. may be used.

[0074] In the specific example of the measuring instrument 2, the material of the sampling surface 11 is not particularly limited as long as it is a material that can transfer the structure and the presence of the structure on the surface of the object. For example, a tape, film, sheet, or the like coated with an adhesive or paint can be used. Furthermore, although the arrangement of the sampling surface 11 is shown as being located on a portion of the plane of the measuring instrument 10, the present invention is not limited to this. The sampling surface 11 may be located over the entire surface of the measuring instrument 10, the measuring instrument 10 may have two or more sampling surfaces 11, or the sampling surface 11 may be located on any surface of the measuring instrument 10. In either case, the sampling surface 11 is kept substantially flush with the flat reference surface during imaging.

[0075] In the specific example of the measuring tool 3, the type of the identification marker 12 is not particularly limited as long as it contains information for identifying the sampling surface 11 from the uneven record. For example, a barcode, a specific pattern or shape, or even the shape of the measuring tool 10 or the sampling surface 11 itself can be used as the identification marker 12. Furthermore, in FIG. 1 , two identification markers 12 are provided on a flat surface, in an area adjacent to the sampling surface 11 on the substrate. However, the present invention is not limited to this arrangement. The identification markers 12 may be configured integrally with the sampling surface 11, configured separately, provided as a single marker, provided as a single marker, or provided as a separate marker, or provided on any surface of the measuring tool 10. However, in any of these arrangements, the identification marker 12 is held together with the sampling surface 11 on the flat reference surface so that the reference pattern portion of the identification marker 12 is substantially in the same plane as the sampling surface 11 during imaging, and both are acquired by a single imaging. The identification marker 12 can be formed by printing (inkjet, laser printing, silkscreen, etc.), etching, laser processing, forming an opening by punching or milling, attaching a label, etc. In addition, the method of printing the identification marker 12 does not necessarily require printing directly on the sampling surface 11, but may involve attaching a separately prepared identification marker 12 to the sampling surface 11 and the above (S3). In this case, too, the identification marker 12 is co-attached or co-supported with the sampling surface 11 so that the reference pattern portion of the identification marker 12 is substantially in the same plane as the sampling surface 11 on the flat reference surface during imaging.

[0076] In the second example of the roughness measurement method, a monocular image obtained by a single image capture is subjected to flattening transformation by calculating a projective homography based on a reference pattern. If necessary, correction of lens geometric distortion and luminance response distortion may also be performed.

[0077] In the specific example 3 of the roughness measurement method, there are no particular limitations on the index as long as it reflects the spatial arrangement of the structures transferred to the sampling surface 11. For example, in feature conversion of image data, a method of detecting a specific structure from an image and calculating its density, a method of processing the image directly and extracting features based on a gray level co-occurrence matrix, texture spectrum analysis, automatic feature extraction using a neural network, etc. can be used.

[0078] In specific example 1 of the defect roughness measurement method, the entire two-dimensional spatial distribution of structures is divided into partial regions that include a certain structure and one of its nearest neighboring structures around it, but the present invention is not limited to this, and the size of the partial region and the number of nearest neighboring structures can also be changed as desired.

[0079] In a second specific example of the defect roughness measurement method, another distance measure between a structure and its nearest neighboring structure may be calculated.

[0080] In a specific example 3 of the defect roughness measurement method, in consideration of the case where the number of structures becomes enormous, the number of structures may be sampled to an arbitrary number and calculation may be performed.

[0081] In a fourth specific example of the defect roughness measuring method, statistical values ​​other than the average value, such as the median, variance, skewness, and kurtosis, may be calculated from the distribution of the obtained distances.

[0082] Although tomato was shown in Example 1 of the stress diagnosis method, the plant stress diagnosis method according to the third aspect of the present invention can also be applied to other plant species, such as watermelon, eggplant, tobacco, hemp, potato, soybean, and other plants that have developed surface structures and are important for agricultural use.

[0083] Although the specific example 2 of the stress diagnosis method shows the diagnosis of fertilizer stress, the plant stress diagnosis method according to the third aspect of the present invention can also be applied to other stress states.

[0084] In the specific example 3 of the stress diagnosis method, the nitrate ion concentration in leaf sap was used as an index of stress, but the plant stress diagnosis method according to the third aspect of the present invention can also be applied to other stress indexes. For example, the index is preferably selected depending on the type of plant, environmental conditions, etc., and examples include total nitrogen content, ammonium ion concentration, protein content, amino acid profile, total phosphorus content, soluble phosphate concentration, total potassium content, polyphenol content, sugar content, acidity, and total lipid content.

[0085] In the specific example 4 of the stress diagnosis method, the nitrate ion concentration in the leaf sap is formulated as a binary classification model of whether or not it exceeds a threshold value, but the present invention is not limited to this. For example, various machine learning models such as a support vector machine, a random forest, or a neural network can be used as a model representing the relationship between the fertilizer stress dose and surface roughness. Furthermore, in addition to the binary classification model, a model that classifies the degree of fertilizer stress into multiple stages, a regression model that predicts the fertilizer stress dose as a continuous value, or the like can also be used.

[0086] In the nutritional diagnostic system specific example 1, the evaluation results can be displayed on a display or printed out from a printer, but the output destination is not particularly limited. For example, the results can be sent by email, saved in cloud storage, or sent to a dedicated smartphone app.

[0087] In the pest risk diagnosis system specific example 1, the evaluation results can be displayed on a display or printed out from a printer, but the output destination is not particularly limited. For example, the results can be sent by email or plotted on a map so that the distribution of pest risks in the field can be visually grasped. Enabling conditions

[0088] As an example of the best mode for carrying out the present invention, the following conditions are suitable, but not limited to: The sampling surface is a cellophane-based adhesive tape. The thickness is approximately 0.05 mm (e.g., 0.04 to 0.06 mm). The adhesive strength is approximately 3 to 5 Newtons per 10 mm. A two-dimensional coded marker is used, with each side of the marker approximately 10 to 15 mm (e.g., 12 mm) and a clear margin around it, the width of which is at least 10% of the side length. Printing should be performed with high contrast, and a resolution of 600 dots per inch (dpi) or higher is desirable. After printing, the side length is measured, and anything exceeding the tolerance (e.g., + / - 0.2 mm) is rejected. At least two markers should be placed on the same plane as the sampling surface, spaced apart, so that at least four non-collinear reference points can be obtained in the captured image. The substrate is white cardboard, with a standard thickness of approximately 0.2 mm (e.g., 0.18 to 0.22 mm). It is desirable that the surface be matte, with minimal ink bleeding and minimal dimensional change after printing. After printing, the image is allowed to dry for several minutes, and if waviness occurs, lightly press it on a flat plate to ensure flatness. After capturing the image, a projective transformation is calculated based on the reference points of the markers, and the image is flattened. In the flattened image, the number of pixels corresponding to a known actual length (e.g., 12 mm) is measured, and a conversion coefficient is obtained from this value to convert the distance amount into an actual length. For reference, the imaging example described in (Non-Patent Document 1) may be followed, but the conditions necessary for implementing the present invention are in accordance with the enabling conditions described in this specification.

[0089] 10 Measuring tool 11 Sampling surface 12 Identification marker 20 Measurement object 21 Measurement object surface

Claims

1. A method for measuring the roughness of a surface to be measured, comprising: (a) a sampling surface that conforms to and adheres to the surface to be measured by elastic deformation, and is capable of transferring and retaining a finer roughness than the waviness component of the surface to be measured; (b) an identification marker that includes a reference pattern of known geometry used to estimate the position, orientation, and magnification of the sampling surface; and (c) a base that co-supports (a) and (b), using a measuring tool comprising: (d) a step of bringing the sampling surface into contact with the surface to be measured to form a transferred pattern of the fine structure of the surface; and (e) a step of peeling the sampling surface from the surface to be measured, shaping the sampling surface to a degree that it can be retained in the same plane as the reference pattern on a flat reference surface provided on the base, and acquiring the sampling surface area and the reference pattern in a single image using an optical sensor, wherein the reference pattern includes at least four non-collinear reference points with known dimensions, or has a configuration that imparts independent geometric constraints sufficient to uniquely determine a projection homography. (f) an adjustment step of calculating a projection homography based on the correspondence of the non-collinear reference points of the reference pattern or the independent geometric constraints, flattening the sampling surface image using the homography, performing absolute scale calibration based on the known dimensions, and further correcting geometric distortion or response distortion specific to the sensor as necessary; and (g) a step of extracting a region corresponding to the sampling surface from the adjusted dataset, and calculating an index including at least a nearest neighbor distance average as a roughness index.

2. A method for measuring the roughness of a surface to be measured according to claim 1, wherein step (g) comprises the steps of: extracting structures within the sampling surface area from the adjusted data set, obtaining the coordinates of the center of gravity of each structure, calculating the Euclidean distance for each structure to the shortest other structure, and calculating the average value of the distances across all structures as a roughness index.

3. A method for diagnosing stress in plants using the method according to claim 1 or 2, wherein the surface to be measured is the surface of a plant, and the method includes a step of evaluating and outputting the stress state of the plant based on the roughness index.

4. A nutritional diagnosis system comprising: an optical sensor; a processing unit; and a computer program executable by the processing unit, wherein the computer program has the function of causing the processing unit to execute the method described in claim 1 or 2, and is configured to cause the processing unit to execute the following processes: (i) acquiring, from the optical sensor, image data in which a sampling surface area and a reference pattern are acquired by a single image capture on a flat reference surface; (ii) aligning a coordinate system based on detection of the reference pattern, flattening transformation using at least four non-collinear reference points or a projective homography calculated based on independent geometric constraints sufficient to uniquely define a projective homography, absolute scale calibration based on known dimensions of the reference pattern, and correcting sensor-specific geometric distortion or response distortion as necessary; (iii) extracting a sampling surface area from the adjusted dataset; (iv) calculating a nearest neighbor distance average from the sampling surface area as a roughness index; (v) evaluating the nutritional state of a plant based on the roughness index; and (vi) outputting the evaluation results.

Images