Determination device and evaluation device

Abstract

The determination device 10 includes a tree trunk internal image acquisition unit 102 that acquires an internal image Im of the inside of the tree trunk Tt obtained by irradiating the trunk Tt of a potted tree T with electromagnetic waves W in an irradiation direction B that intersects the height direction H of the tree T, and a determination unit 108 that determines the age At of the tree T based on the number Ng of annual rings G formed inside the tree trunk Tt, which are captured in the internal image Im.

Description

Technical Field

[0001] The present disclosure relates to a determination device and an evaluation device.

Background Art

[0002] Japanese Patent Application Laid-Open No. 11-232427 discloses a method for measuring the number of annual rings of wood. The end face of the wood is imaged by imaging means such as a television camera, and the number of annual rings of the wood is measured based on the luminance change of the image of the end face.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] When an end face is provided on a tree planted in a pot like a potted plant, the tree loses its value as an ornamental. Therefore, in order to measure the number of annual rings of a tree planted in a pot using the technique disclosed in Japanese Patent Application Laid-Open No. 11-232427, an end face cannot be provided on the tree. That is, there is a problem that the tree age of a tree planted in a pot cannot be accurately determined.

[0005] The present disclosure aims to solve the above-described problems.

Means for Solving the Problems

[0006] A first aspect of the present disclosure includes a trunk internal image acquisition unit that acquires an internal image inside the trunk obtained by irradiating electromagnetic waves in an irradiation direction intersecting the height direction of the tree with respect to the trunk of a tree planted in a pot, and a determination unit that determines the tree age of the tree based on the number of annual rings formed inside the trunk imaged in the internal image.

[0007] A second aspect of this disclosure is an evaluation device comprising: a training data acquisition unit that acquires training data associated with feature quantities of the tree including the determination result by the determination device according to the first aspect, an appearance image obtained by capturing the appearance of the tree with a camera, and an evaluation value assigned in advance as an evaluation of the tree; and a learning unit that uses the training data to train a learning model that takes the feature quantities of the tree and the appearance image as input data and outputs an estimated value of the evaluation of the tree as output data. [Effects of the Invention]

[0008] According to this disclosure, the age of potted trees can be determined with high accuracy. [Brief explanation of the drawing]

[0009] [Figure 1] Figure 1 illustrates the direction and position of electromagnetic waves irradiated to acquire an internal image of the trunk of a potted tree. [Figure 2] Figure 2 illustrates the relationship between the number of tree rings captured in internal images of the tree trunk and the tree's age. [Figure 3] Figure 3 is a block diagram schematically showing the configuration of the judgment device and evaluation device. [Figure 4] Figure 4 illustrates the pixel values ​​for each pixel position in an internal image after contrast adjustment. [Figure 5] Figure 5A shows an example of tree age determination results based on the number of annual rings captured in an internal image of the tree trunk. Figure 5B shows an example of potted plant duration determination results based on changes in the spacing of annual rings captured in an internal image of the tree trunk. Figure 5C shows an example of tree age determination results when the tree was repotted, based on changes in the spacing of annual rings captured in an internal image of the tree trunk. [Figure 6] Figure 6 illustrates a learning model that generates estimated evaluation values ​​for potted trees. [Figure 7]Figure 7A is a flowchart illustrating the procedure for determining the age and duration of potted trees. Figure 7B is a flowchart illustrating the procedure during the learning phase of a learning model. Figure 7C is a flowchart illustrating the procedure during the estimation phase using the learning model. [Modes for carrying out the invention]

[0010] Figure 1 is a diagram illustrating the irradiation direction B and irradiation position Tw of electromagnetic waves W irradiated to acquire an internal image of the trunk Tt of a potted tree T. The tree T shown in Figure 1 is planted in a pot Lp containing soil Ls. The potted tree T is cultivated, for example, as a bonsai. The tree T includes a root system Tr, a trunk Tt located above the root system Tr along the height direction H of the tree T, and a canopy Tc from which branches and leaves spread out from the trunk Tt. Note that the root system Tr is not necessarily buried in the soil Ls. The root system Tr of a potted tree T cultivated as a bonsai may be exposed above the soil Ls, as shown in Figure 1.

[0011] When a potted tree T is evaluated as a bonsai, several characteristics of the tree T may be considered. These characteristics may include, for example, the age of the tree T and the period during which the tree T was in a pot. These characteristics may further include the species of the tree T, the shape of the tree T, the root system of the tree T, how well the tree T looks in its pot, the history of the tree T, and any awards or accolades the tree T has received.

[0012] Among the characteristics of a tree T, its age can be determined based on the number of annual rings formed inside its trunk Tt. However, in order to determine the number of annual rings by visual inspection or imaging with a camera, it is necessary to damage the potted tree T to expose the cross-section of the trunk Tt, thereby exposing all the annual rings. In that case, the tree T would lose its value as a bonsai or ornamental plant.

[0013] In this embodiment, the number of annual rings is determined without damaging the tree T by using an internal image of the inside of the trunk Tt obtained by irradiating the trunk Tt of a potted tree T with electromagnetic waves W. The electromagnetic waves W irradiated onto the trunk Tt are, for example, X-rays. However, the electromagnetic waves W irradiated onto the trunk Tt may also be other electromagnetic waves such as gamma rays or microwaves.

[0014] As shown in Figure 1, electromagnetic waves W are irradiated from the electromagnetic wave irradiation device E onto the trunk Tt of the tree T, which is neither the root Tr nor the canopy Tc. Preferably, the electromagnetic waves W are irradiated onto the side surface Tw of the trunk Tt, which is, for example, about 5 cm away from the root Tr. The electromagnetic waves W are irradiated from the electromagnetic wave irradiation device E onto the side surface Tw of the trunk Tt in an irradiation direction B that intersects with the height direction H of the tree T. This allows an internal image of the inside of the trunk Tt to be obtained.

[0015] Figure 2 illustrates the relationship between the number of tree rings G (Ng) captured within the internal image Im of a tree trunk Tt and the tree age At. Figure 2 shows a correspondence between the tree rings G appearing in a hypothetical cross-section of the tree trunk Tt and the images of the tree rings G captured within the internal image Im obtained by irradiating the side surface of the tree trunk Tt with electromagnetic waves W. Note that the example internal image Im shown in Figure 2 is a positive image, so the images of the tree rings G within this internal image Im are represented by dense pixel values.

[0016] Since the internal image Im is obtained by irradiating the side surface of the trunk Tt of a tree T with electromagnetic waves W, the number Ng of tree rings G captured within the internal image Im corresponds to twice the age At of the tree T. Therefore, the age At can be determined based on the number Ng of tree rings G captured within the internal image Im. That is, the age At corresponds to half the number Ng of tree rings G in the internal image Im.

[0017] The interval Ga of the tree rings G can vary according to the growth environment of the tree T in a year. The tree T used as a potted plant grows in a natural growth environment before being potted. As it grows, the tree rings G are formed. Subsequently, the tree T is replanted in a pot Lp and grows in a potted environment. As it grows, more tree rings G are formed.

[0018] The interval Ga of the tree rings G is different between the natural growth environment and the potted environment. In the natural growth environment, since the tree T can absorb abundant nutrients, the interval Ga of the tree rings G becomes relatively large. In the potted environment, the nutrients that the tree T can absorb are limited, and the thickness of the trunk Tt of the tree T is restricted by the size of the pot Lp, so the interval Ga of the tree rings G becomes relatively small.

[0019] Therefore, when the growth environment of the tree T is switched from the natural growth environment to the potted environment, the interval Ga of the tree rings G changes and becomes smaller. The value obtained by subtracting the tree age Af corresponding to the tree ring G with the changed interval Ga from the current tree age At corresponds to the potted period Qp during which the tree T was potted. That is, based on the change in the interval Ga of the tree rings G imaged in the internal image Im, the potted period Qp can be determined.

[0020] Note that after the growth environment of the tree T is switched from the natural growth environment to the potted environment, there may be a so-called pot upgrade where the tree T is further transferred to a larger pot Lp. In that case, the interval Ga of the tree rings G imaged again in the internal image Im changes and becomes larger. Therefore, based on the change in the interval Ga of the tree rings G imaged in the internal image Im, the tree age As of the tree T when the pot upgrade of the tree T was performed can be determined.

[0021] Figure 3 is a schematic block diagram showing the configuration of the determination device 10 and the evaluation device 20. The determination device 10 uses an internal image Im of the trunk Tt of a potted tree T to determine the tree age At and the potting period Qp of the trunk Tt. The evaluation device 20 uses the feature quantities and external image of the tree T, including the determination result from the determination device 10, and a learning model that has been trained in advance using training data, to generate an estimated evaluation value for the potted tree T.

[0022] The determination device 10 is a computer having an arithmetic unit 100 and a storage unit 150. The arithmetic unit 100 includes one or more processors such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). In other words, the arithmetic unit 100 includes processing circuitry.

[0023] The memory unit 150 is a non-transient storage medium that can be read by a computer. The memory unit 150 includes volatile memory such as RAM (Random Access Memory) and non-volatile memory such as ROM (Read Only Memory) or flash memory. The volatile memory is used as the processor's working memory. The non-volatile memory stores programs executed by the processor and other necessary data.

[0024] Computer programs (computer software) executed by a processor can also be called computer program products. Computer program products are not limited to computer programs stored on non-transient storage media, but also include computer programs that are transmitted, distributed, or downloaded via the internet, etc.

[0025] The calculation unit 100 includes a tree trunk internal image acquisition unit 102, a contrast adjustment unit 104, a peak detection unit 106, a determination unit 108, and a determination result output unit 110. The calculation unit 100 executes a program stored in the storage unit 150 to realize the tree trunk internal image acquisition unit 102, the contrast adjustment unit 104, the peak detection unit 106, the determination unit 108, and the determination result output unit 110.

[0026] At least a portion of the tree trunk internal image acquisition unit 102, contrast adjustment unit 104, peak detection unit 106, determination unit 108, and determination result output unit 110 may be implemented by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array), or by an electronic circuit including discrete devices.

[0027] The tree trunk internal image acquisition unit 102 acquires an internal image Im of the inside of the tree trunk Tt from an external storage device, etc., and stores it in the storage unit 150. The internal image Im is obtained by irradiating the trunk Tt of a potted tree T with electromagnetic waves W in an irradiation direction B that intersects the height direction H of the tree T, as described above, and is stored in advance in an external storage device, etc.

[0028] The contrast adjustment unit 104 performs contrast adjustment by changing the contrast of the internal image Im acquired by the internal tree trunk image acquisition unit 102 and stored in the storage unit 150. For this contrast adjustment, for example, the CLAHE (Contrast-Limited Adaptive Histogram Equalization) algorithm is used. This makes it possible to emphasize the change in contrast in areas where the change in contrast in the original image is small (e.g., flat areas) and to suppress noise amplification in areas where the change in contrast in the original image is large (e.g., edge areas).

[0029] The CLAHE algorithm uses a limiting parameter to restrict changes in contrast. The smaller the value of this limiting parameter, the more the change in contrast is suppressed. By performing contrast adjustments with different limiting parameter values, multiple internal images (Im) can be obtained depending on the values ​​of multiple limiting parameters.

[0030] The peak detection unit 106 detects peak pixel values ​​within the internal image Im. In the positive image of the internal image Im, the pixel values ​​at pixel positions corresponding to the tree rings G formed inside the tree trunk Tt are larger than the values ​​in the regions between adjacent tree rings G. Therefore, within the internal image Im, the peak pixel values ​​of pixels arranged in a pixel row along the direction intersecting the image of the tree rings G can correspond to the image of the tree rings G. The detection of peak pixel values ​​will be described later using Figure 4. If multiple internal images Im are acquired due to contrast adjustment with different limiting parameter values, peak pixel value detection is performed in each of the multiple internal images Im.

[0031] The determination unit 108 determines the age At of the tree T based on the number of tree rings G captured in the internal image Im. As described above, the peak of the pixel value can correspond to the image of the tree rings G. The determination unit 108 determines the number Ng of tree rings G captured in the internal image Im based on the number of peaks of the pixel value detected by the peak detection unit 106. The determination unit 108 determines the age At based on the number Ng of tree rings G captured in the internal image Im. The determination unit 108 determines half the value of the number Ng of tree rings G as the age At. This makes it possible to accurately determine the age At of the potted tree T.

[0032] As described above, if a peak in pixel values ​​is detected in each of the multiple internal images Im obtained by performing contrast adjustment with different limit parameter values, the determination unit 108 determines the tree age At corresponding to each internal image Im. In other words, multiple tree age At values ​​can be calculated.

[0033] In that case, the determination unit 108 determines the range of the multiple tree age At values ​​calculated for each of the multiple internal images Im as the tree age At determination result. For example, if three tree age At values ​​are calculated for each of the three internal images Im, the determination unit 108 determines the range that is greater than or equal to the minimum value of those three tree age At values ​​and less than or equal to the maximum value as the tree age At determination result. This provides a more realistic determination result that takes into account the uncertainty of the internal images Im.

[0034] The determination unit 108 further determines the potted planting period Qp of the tree T based on the change in the spacing Ga of the tree rings G captured in the internal image Im. As described above, the value obtained by subtracting the tree age Af corresponding to the tree ring G with changed spacing Ga from the current tree age At corresponds to the potted planting period Qp of the tree T. The spacing Ga of the tree rings G corresponds to the distance D between the peak of the pixel value detected by the peak detection unit 106 and the adjacent peak. This distance D is expressed by the number of pixels corresponding to the difference in pixel positions between one of the adjacent peaks and the other.

[0035] The determination unit 108 identifies the pixel position Pf of the peak where the amount or rate of change of the distance D exceeds a predetermined value determined in advance by experiment. The peak at the pixel position Pf thus identified corresponds to the tree ring G with a changed spacing Ga. The tree ring G with a changed spacing Ga corresponds to the tree age Af when the growing environment of the tree T is switched from a natural growing environment to a potted plant environment.

[0036] The determination unit 108 determines the potted planting period Qp of the tree T by subtracting the tree age Af corresponding to the pixel position Pf of the peak where the amount or rate of change of distance D exceeds a predetermined value from the current tree age At. This allows for accurate determination of the potted planting period Qp of the tree T. If the tree T was actively cared for during the potted planting period Qp, such care may affect the evaluation of the potted tree T. Therefore, it is preferable to be able to accurately determine the potted planting period Qp in this manner.

[0037] As described above, if a peak in pixel values ​​is detected in each of the multiple internal images Im obtained by performing contrast adjustment with different limit parameter values, the determination unit 108 determines the potted planting period Qp corresponding to each internal image Im. In other words, multiple values ​​for the potted planting period Qp can be calculated.

[0038] In that case, the determination unit 108 determines the determination result for the potted planting period Qp as a range of values ​​for the potted planting period Qp calculated in relation to the multiple internal images Im. For example, if three values ​​for the potted planting period Qp are calculated in relation to three internal images Im, the determination unit 108 determines the determination result for the potted planting period Qp as a range that is greater than or equal to the minimum value and less than or equal to the maximum value among those three potted planting period Qp values. This provides a more realistic determination result that takes into account the uncertainty of the internal images Im.

[0039] The determination unit 108 may further identify pixel positions Ps of peaks where the amount or rate of change in the distance D between adjacent peaks exceeds a predetermined value determined by experiment, outside of pixel position Pf (closer to the surface of the tree trunk Tt). The peaks at the pixel positions Ps thus identified correspond to annual rings G with changed spacing Ga, and correspond to a later period than the tree age Af described above. The annual rings G with changed spacing Ga correspond to the tree age As when the potted tree T was repotted. Based on the pixel positions Ps of peaks where the amount or rate of change in distance D exceeds a predetermined value, the determination unit 108 determines the tree age As of the repotting of the tree T.

[0040] As described above, if a peak in pixel values ​​is detected in each of the multiple internal images Im obtained by performing contrast adjustment with different limit parameter values, the determination unit 108 determines the age As of the tree T at the time of potting, corresponding to each internal image Im. In other words, multiple values ​​for the age As at the time of potting can be calculated.

[0041] In that case, the determination unit 108 determines the determination result for the age of the repotted tree as a range of values ​​for the repotted tree age As calculated in relation to the multiple internal images Im. For example, if three values ​​for the repotted tree age As are calculated in relation to three internal images Im, the determination unit 108 determines the determination result for the repotted tree age As to be a range that is greater than or equal to the minimum value of those three age As values ​​and less than or equal to the maximum value. This provides a more realistic determination result that takes into account the uncertainty of the internal images Im.

[0042] The judgment result output unit 110 outputs judgment data indicating the judgment result from the judgment unit 108 to the storage unit 150 and / or evaluation device 20, etc.

[0043] The evaluation device 20 is a computer having an arithmetic unit 200 and a storage unit 250. The arithmetic unit 200 includes one or more processors such as a CPU or GPU. That is, the arithmetic unit 200 includes processing circuits. The storage unit 250 is a non-transient storage medium that can be read by the computer. The storage unit 250 includes volatile memory such as RAM and non-volatile memory such as ROM or flash memory. The volatile memory is used as the working memory of the processor. The non-volatile memory stores programs executed by the processor and other necessary data.

[0044] The calculation unit 200 includes a training data acquisition unit 202, a learning unit 204, an input data acquisition unit 206, an evaluation estimation unit 208, and an evaluation estimated value output unit 210. The calculation unit 200 executes a program stored in the storage unit 250, thereby realizing the training data acquisition unit 202, the learning unit 204, the input data acquisition unit 206, the evaluation estimation unit 208, and the evaluation estimated value output unit 210.

[0045] At least a portion of the training data acquisition unit 202, the learning unit 204, the input data acquisition unit 206, the evaluation estimation unit 208, and the evaluation estimated value output unit 210 may be implemented by an integrated circuit such as an ASIC or FPGA, or by an electronic circuit including discrete devices.

[0046] The training data acquisition unit 202 acquires training data that associates the feature quantity Mic of tree T, which includes the judgment result from the judgment device 10, the appearance image Mia obtained by capturing the appearance of tree T with a camera, and evaluation values ​​that have been assigned in advance as an evaluation of tree T.

[0047] The judgment data output by the judgment result output unit 110 of the judgment device 10 is acquired by the training data acquisition unit 202 as the judgment result by the judgment device 10, which is included in the feature quantity Mic of tree T. That is, the age At and the potting period Qp of tree T are included in the feature quantity Mic of tree T acquired by the training data acquisition unit 202. In addition, the age As of tree T when it was repotted may also be included in the feature quantity Mic of tree T.

[0048] Furthermore, feature data entered by an operator using the input device 30 may be included in the feature Mic of tree T acquired by the training data acquisition unit 202. This feature data may include, for example, data showing the history of tree T and the awards bestowed upon tree T. The input device 30 may be a keyboard, mouse, touch panel, stylus pen, character recognition device, etc.

[0049] The appearance image Mia of tree T may include multiple appearance images Mia of tree T taken from different orientations. The training data acquisition unit 202 may also acquire feature quantities Miac of tree T based on the appearance image Mia of tree T. In that case, the feature quantities Miac of tree T may include, for example, the tree species of tree T, the tree shape of tree T, the root spread of tree T, and the reflection of tree T in its pot.

[0050] The learning unit 204 uses training data to train a learning model that takes the feature vector Mic of tree T and the appearance image Mia of tree T as input data, and pre-assigned evaluation values ​​for tree T as output data. The feature vector Mic of tree T used as training data for the learning model includes the feature vector Mic of tree T that includes the judgment result from the judgment device 10, and the feature vector Mic of tree T indicated by the feature vector data input by the operator using the input device 30. This results in a learning model that can accurately estimate the evaluation of tree T.

[0051] In other words, the feature variable Mic for tree T includes the age At of tree T, the period Qp during which tree T was grown in a pot, the history of tree T, and any awards awarded to tree T. The feature variable Mic for tree T may further include the age As of tree T when it was transplanted into a pot.

[0052] Alternatively, instead of the external image Mia of tree T itself, the feature quantities Miac of tree T obtained by the training data acquisition unit 202 based on the external image Mia may be used as the training data for the learning model. In other words, the tree species of tree T, the tree shape of tree T, the root spread of tree T, and the reflection of tree T in the pot may be used as the training data for the learning model. In that case, the image data of the external image Mia of tree T is used indirectly rather than directly as the training data.

[0053] The trained model, having completed training using the aforementioned training data, is used to evaluate tree T, with the feature vector Mic and the external image Mia of tree T as input data, and pre-assigned evaluation values ​​for tree T as output data. The trained model generates the output data based on the input data.

[0054] The input data acquisition unit 206 acquires the feature quantity Mic of tree T, which includes the judgment result from the judgment device 10, and the appearance image Mia of tree T, as input data for a trained model that has been trained using training data. The judgment data output by the judgment result output unit 110 of the judgment device 10 is acquired by the input data acquisition unit 206 as the judgment result from the judgment device 10 included in the feature quantity Mic of tree T. That is, the age At and the potting period Qp of tree T are included in the feature quantity Mic of tree T acquired by the input data acquisition unit 206. Furthermore, the age As of tree T when it was repotted may also be included in the feature quantity Mic of tree T.

[0055] Furthermore, feature data entered by an operator using the input device 30 may be acquired by the input data acquisition unit 206. This feature data may include, for example, data indicating the history of tree T and the awards awarded to tree T. In this case, the history of tree T and the awards awarded to tree T are further included in the feature Mic of tree T.

[0056] The external image Mia of tree T may include multiple external images Mia of tree T taken from different orientations. The input data acquisition unit 206 may also acquire feature quantities Miac of tree T based on the external image Mia of tree T. In this case, the feature quantities Miac of tree T may include, for example, the tree species of tree T, the tree shape of tree T, the root spread of tree T, and the reflection of tree T in its pot.

[0057] The evaluation estimation unit 208 uses the feature quantities Mic and the appearance image Mia of tree T as input data, along with a trained model that has been trained using training data, to generate an estimated evaluation value for the potted tree T as output data. This allows for the accurate and stable generation of an estimated evaluation value for tree T.

[0058] The feature vector Mic and the external image Mia of tree T, which serve as input data, are acquired by the input data acquisition unit 206 as described above. Therefore, the feature vector Mic of tree T includes the age At of tree T and the potted planting period Qp of tree T. The feature vector Mic of tree T may further include the history of tree T and any awards bestowed upon tree T. In addition, the feature vector Mic of tree T may further include the age As of tree T when it was repotted.

[0059] Furthermore, based on the external image Mia of tree T, the feature vector Miac of tree T can be obtained by the input data acquisition unit 206 as described above. In this case, the feature vector Miac of tree T may include, for example, the tree species of tree T, the tree shape of tree T, the root spread of tree T, and the reflection of tree T in its pot, which are used as input data. In other words, the image data of the external image Mia of tree T is used indirectly rather than directly as input data.

[0060] The evaluation estimate output unit 210 stores the estimated value data, which represents the estimated evaluation value for tree T generated by the evaluation estimation unit 208, in the storage unit 250 or an external device, and also outputs it to an output device 40 such as a display.

[0061] Figure 4 is a diagram illustrating the pixel values ​​for each pixel position in the contrast-adjusted internal image Im. As described above, the peak detection unit 106 of the determination device 10 detects the peak pixel values ​​of pixels arranged in a row of pixels aligned in a direction intersecting the image of the tree rings G within the contrast-adjusted internal image Im of the trunk Tt of the potted tree T.

[0062] Figure 4 illustrates the change in pixel values ​​of pixels in a pixel sequence when there are Pm+1 pixels in the pixel sequence from pixel position 0 to pixel position Pm within the internal image Im. In this embodiment, the internal image Im is a positive image. In this case, the pixel value at pixel position Pc near the center of the tree trunk Tt is relatively large because the thickness of the tree trunk Tt along the direction of electromagnetic wave W irradiation B is greater than that of the outermost pixel positions Pe(Pei, Pej) of the tree trunk Tt.

[0063] Furthermore, in the internal image Im after contrast adjustment, the peaks in pixel values ​​within the internal image Im may correspond not only to the image of the tree rings G but also to noise. Therefore, the peak detection unit 106 may detect peaks that satisfy predetermined conditions as peaks corresponding to the image of the tree rings G. This allows for accurate detection of peaks corresponding to the image of the tree rings G.

[0064] The predetermined conditions include, for example, the following two conditions (i) and (ii). The predetermined pixel value used in condition (i) and the predetermined distance used in condition (ii) are both determined in advance by experiment. (i) The difference between the minimum pixel value among the pixels in the local range that includes the pixel position where the pixel value shows a peak in the pixel row, and the pixel value of that peak, is greater than or equal to a predetermined pixel value. (ii) The distance D (as described above) between adjacent peaks is greater than or equal to a predetermined distance.

[0065] The peaks detected by satisfying conditions (i) and (ii) correspond to the image of the tree rings G. Therefore, the number of peaks Np detected by the peak detection unit 106 is equal to the number of tree rings G Ng mentioned above. In other words, the number of detected peaks Np corresponds to twice the age At of the tree T. The age At corresponds to half the number of peaks Np.

[0066] Assume that tree T initially grew in a natural growing environment, and was still growing in a natural growing environment even when the first annual ring G corresponding to the pixel position Po(Poi, Poj) of the peak closest to the pixel position Pc near the center of the tree trunk Tt was formed inside the tree trunk Tt. Figure 4 shows the value Da of the distance D between the peak at pixel position Po(Poi, Poj) and the adjacent peak in the pixel row along the radial direction of the tree trunk Tt, moving away from pixel positions Pc and Po(Poi, Poj).

[0067] Figure 4 shows the peak pixel positions Pf(Pfi, Pfj) in the direction further away from pixel positions Pc, Po(Poi, Poj). Between pixel positions Po(Poi, Poj) and Pf(Pfi, Pfj), the value of the distance D between adjacent peaks is maintained at a value close to Da. While the tree rings G corresponding to these peaks are being formed, the tree T is growing in a natural growing environment.

[0068] Figure 4 shows the value Db of the distance D between the peak at pixel position Pf(Pfi, Pfj) and the adjacent peak in the pixel row along the radial direction of the tree trunk Tt, moving away from pixel positions Pc and Pf(Pfi, Pfj). Db is smaller than Da. Figure 4 also shows the pixel position Ps(Psi, Psj) of the peak, moving further away from pixel positions Pc and Pf(Pfi, Pfj). Between pixel positions Pf(Pfi, Pfj) and Ps(Psi, Psj), the value D of the distance between adjacent peaks is maintained close to Db.

[0069] In other words, at pixel positions Pf(Pfi, Pfj), the distance D between adjacent peaks changes from Da to Db, which is smaller than Da. Suppose that the amount or rate of change of this distance D from Da to Db exceeds a predetermined value. In that case, as described above, the determination unit 108 of the determination device 10 identifies the pixel positions Pf(Pfi, Pfj) of the peaks where the amount or rate of change of distance D exceeds the predetermined value.

[0070] The peaks at the identified pixel positions Pf(Pfi, Pfj) correspond to the tree rings G with changed spacing Ga. At tree age Af corresponding to the identified pixel positions Pf(Pfi, Pfj), it is thought that the growth environment of tree T switched from a natural growth environment to a potted environment. Therefore, the changes in pixel values ​​in the range Rn from pixel position Po(Poi, Poj) to pixel position Pf(Pfi, Pfj) correspond to the natural growth period during which tree T was growing naturally.

[0071] Figure 4 shows the value of distance Dc between the peak at pixel position Ps(Psi, Psj) and the adjacent peak in the pixel row along the radial direction of the tree trunk Tt, moving away from pixel positions Pc and Ps(Psi, Psj). Dc is smaller than Db.

[0072] Figure 4 shows the peak pixel positions Pt(Pti, Ptj) in the direction further away from the pixel positions Pc and Ps(Psi, Psj). Pixel positions Pt(Pti, Ptj) correspond to the most recent tree ring G of tree T. Between pixel positions Ps(Psi, Psj) and Pt(Pti, Ptj), the distance D between adjacent peaks is maintained at a value close to Dc.

[0073] In other words, at pixel positions Ps(Psi, Psj), the distance D between adjacent peaks changes from Db to Dc, which is greater than Db. Suppose that the amount or rate of change of this distance D from Db to Dc exceeds a predetermined value. In that case, as described above, the determination unit 108 of the determination device 10 identifies the pixel positions Ps(Psi, Psj) of the peaks where the amount or rate of change of distance D exceeds the predetermined value.

[0074] The peaks at the identified pixel positions Ps(Psi, Psj) correspond to tree rings G with altered spacing Ga. It is thought that the potted tree T was repotted at the tree age As corresponding to the identified pixel positions Ps(Psi, Psj).

[0075] Therefore, the change in pixel values ​​in the range Rpf from pixel position Pf(Pfi, Pfj) to pixel position Ps(Psi, Psj) corresponds to the period when tree T was initially potted in the first pot Lp after being switched from a natural growing environment to a potted environment. The change in pixel values ​​in the range Rps from pixel position Ps(Psi, Psj) to pixel position Pt(Pti, Ptj) corresponds to the period when tree T was potted in the larger second pot Lp after being repotted. The change in pixel values ​​in the range Rp, which combines ranges Rpf and Rps, corresponds to the potted period Qp during which tree T was potted.

[0076] Furthermore, even after the latest annual ring G of tree T was formed, it can be assumed that the tree T has continued to be planted in the second pot Lp up to the present. In that case, the range Rps described above corresponds to the range from pixel position Ps(Psi, Psj) to the outermost pixel position Pe(Pei, Pej) of the tree trunk Tt.

[0077] Figure 5A shows an example of the determination result of the tree age At based on the number Ng of annual rings G captured in the internal image Im of the tree trunk Tt. As described above, the determination unit 108 of the determination device 10 determines the current tree age At of the tree T based on the number Np of pixel value peaks detected by the peak detection unit 106 of the determination device 10. At that time, pixel value peaks are detected in each of the multiple internal images Im obtained by performing contrast adjustment with different limiting parameter values ​​that limit the change in contrast.

[0078] Let the constraint parameters be x, y, and z, and the relationship x ≤ y ≤ z hold. For constraint parameter values ​​x, y, and z, let the number of detected peaks Np be Npx, Npy, and Npz, respectively. In this case, the relationship Npx ≤ Npy ≤ Npz holds. This determines the equation Npx ≤ Np ≤ Npz, which represents the range of the number of peaks Np.

[0079] The tree age At is half the number of peaks Np. For the limit parameter values ​​x, y, and z, the tree age At values ​​are calculated as Atx, Aty, and Atz, respectively. Of these three tree age At values, the minimum is Atx and the maximum is Atz. Therefore, the formula Atx ≤ At ≤ Atz, which indicates the range of tree age At values, is determined as the result of the determination by the determination unit 108.

[0080] Figure 5B shows an example of the determination result of the potted planting period Qp based on the change in the spacing Ga of the annual rings G captured in the internal image Im of the tree trunk Tt. As described above, the determination unit 108 identifies the pixel position Pf of a peak where the amount of change or rate of change of the distance D between adjacent peaks exceeds a predetermined value. The determination unit 108 determines the potted planting period Qp of the tree T by subtracting the age Af of the tree T at the time when its growing environment was switched from a natural growing environment to a potted planting environment, which corresponds to the pixel position Pf, from the current age At.

[0081] For constraint parameter values ​​x, y, and z, the pixel positions Pf at which the distance D between adjacent peaks changes from Da to a smaller value Db are Pfx, Pfy, and Pfz, respectively. The tree age Af values ​​corresponding to the pixel position Pf values ​​Pfx, Pfy, and Pfz, when the growing environment of tree T is switched from a natural growing environment to a potted plant environment, are Afx, Afy, and Afz, respectively. In this case, the relationship Afx ≤ Afy ≤ Afz holds. That is, the minimum value of these three tree age Af values ​​is Afx, and the maximum value is Afz. Therefore, the equation Afx ≤ Af ≤ Afz, which shows the range of tree age Af values, is determined.

[0082] The potting period Qp is obtained by subtracting the determined tree age Af from the current tree age At, as shown in Figure 5A. For limit parameter values ​​x, y, and z, the values ​​of the potting period Qp are calculated as Qpx, Qpy, and Qpz, respectively. Of these three values ​​of the potting period Qp, the minimum is Qpx and the maximum is Qpz. Therefore, the formula Qpx ≤ Qp ≤ Qpz, which indicates the range of the potting period Qp, is determined as the result of the determination by the determination unit 108.

[0083] Figure 5C shows an example of the determination result of the age As of a potted tree T based on the change in the spacing Ga of the annual rings G captured in the internal image Im of the tree trunk Tt. As described above, the determination unit 108 identifies the pixel position Ps of a peak outside of the pixel position Pf where the amount of change or rate of change of the distance D between adjacent peaks exceeds a predetermined value. The determination unit 108 determines the age As of the potted tree T at the time of repotting, corresponding to the pixel position Ps.

[0084] For constraint parameter values ​​x, y, and z, the pixel positions Ps where the distance D between adjacent peaks changes from Db to a value greater than Db (Dc) are Psx, Psy, and Psz, respectively. The age As values ​​of the potted tree T when it is repotted, corresponding to the pixel position Psx, Psy, and Psz values, are Asx, Asy, and Asz, respectively. In this case, the relationship Asx ≤ Asy ≤ Asz holds. That is, the minimum value of these three age As values ​​is Asx, and the maximum value is Asz. Therefore, the expression Asx ≤ As ≤ Asz, which shows the range of the age As values, is determined as the determination result by the determination unit 108.

[0085] Figure 6 illustrates a learning model M that generates an estimated evaluation value Moe for a potted tree T. The learning model M shown in Figure 6 has been trained using training data by the learning unit 204 of the evaluation device 20. The evaluation estimation unit 208 of the evaluation device 20 uses the feature quantities Mic and the appearance image Mia of the tree T as input data Mi, and the learning model M which has been trained using training data, to generate an estimated evaluation value Moe for the tree T as output data Mo. The estimated evaluation value Moe is, for example, the estimated value of the tree T.

[0086] The learning model M can be constructed, for example, by an ensemble of LightGBM (Light Gradient Boosting Machine), TabNet (Tabular Network), CNN (Convolutional Neural Network), etc.

[0087] Figure 7A is a flowchart illustrating the procedure for determining the age At and potting period Qp of a potted tree T. This procedure is performed by the calculation unit 100 of the determination device 10 executing a program stored in the storage unit 150. When this procedure is started, in step S1, the trunk internal image acquisition unit 102 acquires an internal image Im of the inside of the trunk Tt of tree T from an external storage device or the like.

[0088] In step S2, the contrast adjustment unit 104 performs contrast adjustment to change the contrast of the internal image Im acquired in step S1. In step S3, the peak detection unit 106 detects the peak of pixel values ​​within the internal image Im that has undergone contrast adjustment in step S2.

[0089] In step S4, the determination unit 108 determines the age At of the tree T based on the number of peak pixel values ​​Np corresponding to the tree rings G captured in the internal image Im, as detected in step S3. The determination unit 108 further determines the potting period Qp during which the tree T was potted by subtracting the age Af, at which the spacing Ga of the tree rings G captured in the internal image Im changed, from the current age At. The determination unit 108 may further determine the age As at which the tree T was repotted, after the time of age Af, at which the spacing Ga of the tree rings G captured in the internal image Im changed. Once the processing in step S4 is completed, this processing procedure ends.

[0090] Figure 7B is a flowchart illustrating the processing procedure during the learning phase of the learning model M. This processing procedure is performed by the calculation unit 200 of the evaluation device 20 executing a program stored in the storage unit 250. When this processing procedure is started, in step S11, the training data acquisition unit 202 acquires training data in which the feature quantity Mic of tree T, which includes the judgment result from the judgment device 10, the appearance image Mia of tree T, and the evaluation value assigned to tree T are associated.

[0091] In step S12, the learning unit 204 trains the learning model M using the training data acquired in step S11. Once the processing in step S12 is complete, this processing procedure ends.

[0092] Figure 7C is a flowchart illustrating the processing procedure in the estimation stage using the learning model M. This processing procedure is performed by the calculation unit 200 of the evaluation device 20 executing a program stored in the storage unit 250. When this processing procedure is started, in step S21, the input data acquisition unit 206 acquires the input data Mi of the learning model M, which has been trained using the training data. The feature quantities Mic of the tree T, which include the judgment result from the judgment device 10, and the appearance image Mia of the tree T are acquired as the input data Mi.

[0093] In step S22, the evaluation estimation unit 208 uses the input data Mi obtained in step S21 and the learning model M which has been trained using the training data to generate an estimated evaluation value Moe for the potted tree T as output data Mo. Once the processing in step S22 is complete, this processing procedure is terminated.

[0094] With regard to the embodiments described above, the following additional information is disclosed.

[0095] (Note 1) The determination device (10) of this disclosure includes a tree trunk internal image acquisition unit (102) that acquires an internal image (Im) of the inside of a potted tree (T) obtained by irradiating the trunk (Tt) of the tree trunk (Tt) with electromagnetic waves (W) in an irradiation direction (B) intersecting the height direction (H) of the tree, and a determination unit (108) that determines the age (At) of the tree based on the number (Ng) of annual rings (G) formed inside the trunk, which are captured in the internal image. With this configuration, the age of a potted tree can be determined with high accuracy.

[0096] (Note 2) The determination device described in Appendix 1 may determine the period (Qp) during which the tree was in a pot based on the change in the spacing (Ga) of the annual rings captured in the internal image. With such a configuration, the period during which the tree was in a pot can be determined with high accuracy.

[0097] (Note 3) The determination device described in Appendix 2 further comprises a contrast adjustment unit (104) that performs contrast adjustment to change the contrast of the internal image, and the determination unit calculates the tree age value (Atx, Aty, Atz) and the potted planting period value (Qpx, Qpy, Qpz) based on each of the multiple internal images obtained by performing the contrast adjustment on a single internal image with different limiting parameter values ​​(x, y, z) that limit the change in contrast, and the determination unit determines the range of the tree age value calculated in multiple ways corresponding to the multiple internal images as the determination result for the tree age, and determines the range of the potted planting period value calculated in multiple ways corresponding to the multiple internal images as the determination result for the potted planting period. With such a configuration, a more realistic determination result can be obtained that takes into account the uncertainty of the internal image.

[0098] (Note 4) The evaluation device (20) of this disclosure includes a training data acquisition unit (202) that acquires training data associated with tree feature quantities (Mic) including the judgment result of the judgment device described in any one of the appendices 1 to 3, an appearance image (Mia) obtained by capturing the appearance of the tree with a camera, and an evaluation value assigned in advance as an evaluation of the tree, and a learning unit (204) that uses the training data to train a learning model (M) in which the tree feature quantities and the appearance image are input data (Mi) and an estimated value of the evaluation of the tree (Moe) is output data (Mo). With this configuration, a learning model that accurately estimates the evaluation of a tree can be obtained.

[0099] (Note 5) The evaluation device described in Appendix 4 may further include an input data acquisition unit (206) that acquires the input data, and an evaluation estimation unit (208) that uses the tree's feature quantities and appearance image as input data, and the learning model to generate the estimated value of the evaluation of the tree as output data. With such a configuration, the estimated value of the evaluation of the tree can be generated accurately and stably.

[0100] While this disclosure has been described in detail, it is not limited to the individual embodiments described above. These embodiments can be added, replaced, modified, partially deleted, etc., in any way that does not depart from the gist of this disclosure or from the spirit of this disclosure derived from the claims and their equivalents. These embodiments can also be implemented in combination. For example, the order of operations and processes in the embodiments described above are given as examples only and are not limited thereto. The same applies when numerical values ​​or mathematical formulas are used in the description of the embodiments described above. [Explanation of Symbols]

[0101] 10: Judgment device 20: Evaluation device 102:Tree trunk interior image acquisition unit 104: Contrast adjustment section 108: Judgment section 202: Training Data Acquisition Department 204: Learning Department 206: Input Data Acquisition Unit 208: Evaluation and Estimation Department At: Tree age Atx, Aty, Atz: Values ​​of tree age B: Irradiation direction G: Tree rings Ga: Spacing between tree rings H: Height direction Im: Internal image M: Learning Model Mi: Input data Mia: Exterior image Mic: Tree features Mo: Output data Moe: Estimated Value Ng: Number of tree rings Qp: Potting period Qpx, Qpy, Qpz: Values ​​for the potting period T: Tree Tt:Trunk x, y, z: Values ​​of the constraint parameters W: Electromagnetic waves

Claims

1. A tree trunk internal image acquisition unit acquires an internal image of the inside of a potted tree trunk by irradiating the trunk with electromagnetic waves in an irradiation direction intersecting the height direction of the tree, A determination unit that determines the age of the tree based on the number of annual rings formed inside the tree trunk captured in the internal image, A determination device equipped with the following features.

2. A determination device according to claim 1, The determination unit is a determination device that determines the period during which the tree was grown in a pot based on the change in the spacing of the annual rings captured in the internal image.

3. A determination device according to claim 2, The system further includes a contrast adjustment unit that performs contrast adjustment to change the contrast of the internal image, Based on each of the multiple internal images obtained by performing the contrast adjustment on a single internal image, each of which has a different limit parameter value that limits the change in contrast, the determination unit calculates the value of the tree age and the value of the potted planting period. The determination unit determines the range of the tree age values ​​calculated in relation to the multiple internal images as the determination result for the tree age, and determines the range of the potted planting period values ​​calculated in relation to the multiple internal images as the determination result for the potted planting period.

4. A training data acquisition unit that acquires training data in which the characteristic quantities of the tree, including the determination result by the determination device described in any one of claims 1 to 3, an appearance image obtained by capturing the appearance of the tree with a camera, and an evaluation value assigned in advance as an evaluation of the tree are associated; A learning unit that uses the aforementioned training data to train a learning model that takes the characteristics of the trees and the appearance images as input data and outputs estimated values ​​of the evaluation of the trees as output data. An evaluation device equipped with the following features.

5. An evaluation apparatus according to claim 4, An input data acquisition unit that acquires the aforementioned input data, An evaluation estimation unit that generates an estimated value of the evaluation of the tree as output data, using the tree feature quantities and the appearance image as input data and the learning model, An evaluation device that is further equipped with these features.

Images