JP7703065B2 - Tree health assessment system - Google Patents

Description

The present invention relates to a tree health assessment system that assesses the health of trees such as roadside trees.

Traditionally, tree doctors regularly inspect each tree one by one to conduct tree diagnosis. Tree diagnosis involves visually inspecting the appearance of each tree to check for bark defects, decay, and other abnormalities.

However, this type of tree diagnosis requires highly skilled diagnosticians, and there are concerns that there will be a shortage of personnel and a decline in diagnostic capabilities in the future. In addition, regularly inspecting a huge number of trees is extremely costly. Furthermore, because the health of trees is constantly declining, tree diagnosis must be carried out in a timely manner.

As an alternative to such tree diagnosis by an arborist, a method for efficiently assessing the health of trees is known, as shown in Patent Document 1, in which a health database is created in advance that associates the spectral characteristics of leaves with the health of trees for each tree species, soil characteristics, and nutrients in the leaves, and the spectral characteristics of the leaves of trees are read using high-resolution satellite images, and the health database is then referenced from the soil characteristics and nutrient data in the leaves of the tree being evaluated to assess the health of each individual tree.

JP 2007-166967 A

However, the evaluation method described in Patent Document 1 requires collecting the soil in which the trees are growing and the leaves of the trees, and then measuring the nutrients in the leaves, which are closely related to the properties of the soil and the components of the soil, in advance, which could result in a huge amount of time and cost being required to evaluate the health of the trees.

Therefore, a technical problem arises that needs to be solved in order to easily determine the health of trees, and the present invention aims to solve this problem.

In order to achieve the above object, the tree health determination system of the present invention is a tree health determination system that determines the health of a tree, and includes an illumination unit that irradiates the tree with measurement light having a wavelength in the near-infrared region, a camera that receives light reflected from the tree and captures a near-infrared image, a calculation unit that calculates the reflection intensity of the reflected light at a first wavelength that is absorbed by the moisture of the tree based on the near-infrared image, and a determination unit that determines the health of the tree from the reflection intensity at the first wavelength calculated by the calculation unit based on a pre-stored relationship between the reflection intensity and the health of the tree.

In addition, in the tree health assessment system according to the present invention, it is preferable that the illumination unit irradiates the tree with measurement light having a wavelength in the visible light region, the camera receives light reflected from the tree from the measurement light to capture a visible light image, and the calculation unit detects position coordinates of a depression formed on the bark surface of the tree from the visible light image, and calculates the reflection intensity at the first wavelength at the position coordinates of the depression.

Furthermore, in the tree health assessment system according to the present invention, it is preferable that the calculation unit calculates the reflection intensity of the reflected light at second and third wavelengths that are less easily absorbed by the moisture of the tree compared to the first wavelength, detects the position coordinates of a measurement area in the near-infrared image where the reflection intensities at the second and third wavelengths are approximately equal, and calculates the reflection intensity at the first wavelength in the measurement area.

Furthermore, in the tree health assessment system according to the present invention, it is preferable that the assessment unit prestores a first threshold value set based on the reflection intensity of the reflected light acquired during a period when the photosynthesis of the tree is relatively inactive, and the assessment unit assesses that the health of the tree is high when the reflection intensity of the reflected light acquired during a period when the photosynthesis of the tree is relatively active is less than the first threshold value.

Furthermore, in the tree health assessment system according to the present invention, it is preferable that the assessment unit prestores a first threshold value set based on the reflection intensity of the reflected light acquired during a period when the photosynthesis of the tree is relatively inactive, and the assessment unit assesses that the health of the tree is low when the reflection intensity of the reflected light acquired during a period when the photosynthesis of the tree is relatively active is approximately equal to the first threshold value.

In addition, in the tree health assessment system according to the present invention, the assessment unit preferably stores in advance a second threshold value that is obtained when the photosynthesis of the tree is relatively active and is set based on the reflection intensity of the reflected light of the tree with high health, and the assessment unit determines that the tree has high health when the reflection intensity of the reflected light of the tree to be assessed is approximately equal to the second threshold value.

Furthermore, in the tree health assessment system according to the present invention, the assessment unit prestores a first threshold value set based on the reflection intensity of the reflected light obtained during a period when the tree's photosynthesis is relatively inactive, and a second threshold value set based on the reflection intensity of the reflected light obtained during a period when the tree's photosynthesis is relatively active and from the tree with a high level of health, and the assessment unit preferably determines that the closer the reflection intensity of the reflected light in the tree being assessed is to the first threshold value, the lower the level of health, and that the closer the reflection intensity is to the second threshold value, the higher the level of health.

This invention makes it possible to easily determine the health of trees without the need for highly skilled diagnosticians as in conventional tree diagnosis.

1 is a schematic diagram showing the configuration of a tree health determination system according to one embodiment of the present invention; FIG. 1 is a schematic diagram showing how the amount of light absorbed increases in proportion to the amount of water on the underside of a tree trunk. 1A is a near-infrared image of the trunk of a cherry tree diagnosed as an "unhealthy tree" among the cherry trees belonging to the first group in Experimental Example 1. 1B is a near-infrared image of the trunk of a cherry tree diagnosed as a "healthy tree" among the cherry trees belonging to the first group in Experimental Example 1. 1A is a near-infrared image of the trunk of a cherry tree that was diagnosed as an "unhealthy tree" among the cherry trees in the second group in Experimental Example 1. 1B and 1C are near-infrared images of the trunks of cherry trees that were diagnosed as "healthy trees" among the cherry trees in the second group in Experimental Example 1. 1A is a near-infrared image of a trunk of an oak tree diagnosed as an "unhealthy tree" in Experimental Example 1. FIG. 1B is a near-infrared image of a trunk of an oak tree diagnosed as a "healthy tree" in Experimental Example 1. 1A is a near-infrared image of the trunk of a Zelkova tree diagnosed as an "unhealthy tree" in Experimental Example 1. FIG. 1B is a near-infrared image of the trunk of a Zelkova tree diagnosed as a "healthy tree" in Experimental Example 1. 13 is a graph showing the average value of the reflection intensity of reflected light in a near-infrared image for each tree in Experimental Example 1. 1 is a table showing the results of tree vitality assessment by a tree surgeon in Experimental Example 2. 13 is a graph showing the average value and balance ratio of the reflection intensity of reflected light of dangerous trees and healthy trees in Experimental Example 2. 13 is a table showing balance indexes and balance ratios of dangerous trees and healthy trees in Experimental Example 2. 13 is a near-infrared image of Case 1 of Experimental Example 3 at a wavelength of about 1100 nm. 13 is a near-infrared image of Case 1 of Experimental Example 3 at a wavelength of about 1450 nm. 13 is a near-infrared image of Case 2 of Experimental Example 3 at a wavelength of about 1100 nm. 13 is a near-infrared image of Case 2 of Experimental Example 3 at a wavelength of about 1450 nm. 13 is an image showing the ratio of reflection intensities at wavelengths of 1000 nm and 1100 nm in Case 3 of Experimental Example 3. 13 is a graph showing the reflection intensity of reflected light in a wavelength band of about 900 to 1700 nm within a measurement area of a subject in cases 1 and 2 of Experimental Example 3. 13 is a graph showing the reflection intensity of reflected light in a wavelength band of about 900 to 1700 nm within a measurement area of a subject in cases 3 and 4 of Experimental Example 3. 13 is a graph showing differences in reflection intensity in cases 1 to 4 of Experimental Example 3. 13 is a table showing average reflection intensities of near-infrared images measured by tree species in winter in Experimental Example 4. 20 is a graph showing the average reflectance intensity of the trees in FIG. 19 . (a) to (c) are near-infrared images taken in summer of the trunk of a zelkova tree that is a "healthy tree" in Experimental Example 1, and (d) is a near-infrared image taken in summer of the trunk of a zelkova tree that is an "unhealthy tree." 22 is a graph showing the average reflection intensity of near-infrared images of the three zelkova trees that are “healthy trees” shown in FIG. 21 in Experimental Example 1, and a first threshold value for tree species group G1 to which zelkova trees belong in Experimental Example 4.

One embodiment of the present invention will be described with reference to the drawings. Note that in the following, when referring to the number, numerical value, amount, range, etc. of components, unless otherwise specified or when it is clearly limited to a specific number in principle, it is not limited to that specific number and may be more or less than the specific number.

In addition, when referring to the shape or positional relationship of components, etc., this includes things that are substantially similar or similar to that shape, etc., unless otherwise specified or considered in principle to be clearly different.

In addition, drawings may exaggerate characteristic parts to make the features easier to understand, and the dimensional ratios of components may not necessarily be the same as in reality.

Figures 1(a) and (b) are schematic diagrams showing the configuration of a tree health assessment system 1 according to one embodiment of the present invention. The tree health assessment system 1 according to this embodiment includes a lighting unit 2 and a camera 3.

The lighting unit 2 irradiates measurement light toward trees T, such as roadside trees. The measurement light has a wavelength in the visible light region (approximately 400-800 nm) and a wavelength in the near-infrared region (approximately 1000-1500 nm). The lighting unit 2 is, for example, a halogen lamp. The lighting unit 2 may be divided into one that irradiates measurement light having a wavelength in the visible light region and one that irradiates measurement light having a wavelength in the near-infrared region. The lighting unit 2 is positioned approximately 1 m away from the tree T. The distance from the tree T to the lighting unit 2 can be adjusted as desired.

Camera 3 is equipped with, for example, a SWIR camera that is sensitive to the near-infrared wavelength range. Camera 3 receives reflected light having a wavelength in the near-infrared range that is the measurement light reflected by tree T, and captures a near-infrared image. It is preferable to capture images with camera 3 at night to prevent reflected light from sunlight from appearing in the near-infrared image. It is also preferable to capture images with camera 3 on a clear day to prevent the bark surface and leaves from becoming wet due to rainfall.

In addition, at the same time as capturing the near-infrared image, camera 3 receives reflected light having a wavelength in the visible light region that is the measurement light reflected by tree T, and captures a visible light image that shows the trunk surface of tree T. Note that camera 3 may be divided into one that captures near-infrared images and one that captures visible light images.

The camera 3 is positioned at a distance of about 1 m from the tree T. The height of the camera 3 above the ground is set to, for example, about 0.9 m. The reflection angle formed by the optical path of the measurement light irradiated from the illumination unit 2 and the optical path of the reflected light reflected by the tree T and entering the camera 3 is set to about 45 degrees. The distance from the tree T to the camera 3, the height of the camera 3 above the ground, and the reflection angle of the reflected light can be adjusted as desired.

The operation of the tree health determination system 1 is controlled by the controller 4. The controller 4 controls each of the components that make up the tree health determination system 1. The controller 4 is composed of, for example, a CPU, a memory, etc. The functions of the controller 4 may be realized by controlling using software, or may be realized by operating using hardware. The controller 4 includes a calculation unit 41 and a determination unit 42.

The calculation unit 41 calculates the reflection intensity of the reflected light at a first wavelength (e.g., about 1450 nm) where absorption by moisture in the tree T is significant, based on the near-infrared image captured by the camera 3. As shown in FIG. 2, moisture is present on the underside of the bark of the tree T, which consists of xylem sap, which is moisture sucked up from the ground and sent from the roots to the entire tree T, and phloem sap, which is a nutrient produced in the leaves. The amount of absorption of the measurement light of the first wavelength irradiated from the illumination unit 2 increases in proportion to the moisture content of the underside of the bark. In other words, when the moisture content of the underside of the bark is high and the tree T is highly humid, much of the measurement light of the first wavelength is absorbed, and the near-infrared image appears darker (with less light). On the other hand, when the moisture content of the underside of the bark is low and the tree T is less humid, much of the measurement light of the first wavelength is reflected, and the near-infrared image appears whiter (with more light).

The reflection intensity of the reflected light calculated by the calculation unit 41 may be the average reflection intensity of the entire tree T captured in the near-infrared image, or it may be the local reflection intensity in any region of the tree T captured in the near-infrared image.

An arbitrary region of tree T in the near-infrared image is, for example, a recess formed on the bark surface of tree T that the calculation unit 41 extracts from the visible light image. In this way, in the case of a tree species in which protrusions that have become corky and dried and recesses with a high moisture content are formed on the bark surface, the actual moisture content of the tree can be accurately determined by excluding the protrusions from the calculation of the reflection intensity and calculating the reflection intensity in the recesses.

Alternatively, the calculation unit 41 may calculate the reflection intensity of reflected light at a second wavelength (e.g., about 1000 nm) and a third wavelength (e.g., about 1100 nm), which are hardly absorbed by the moisture in the tree T compared to the first wavelength, for the entire tree T captured in the near-infrared image, and set the measurement area extracted as an area of the tree T where the reflection intensities of the reflected light at the second and third wavelengths are approximately equal to any area of the tree T in the near-infrared image. This eliminates noise caused by light reflection and scattering (diffuse reflection), making it possible to obtain a highly reliable reflection intensity at the first wavelength.

The determination unit 42 determines the healthiness of the tree T from the reflection intensity at the first wavelength calculated by the calculation unit 41 based on the relationship between the reflection intensity of the reflected light and the healthiness of the tree T, which has been previously obtained and stored through experiments, etc.

Specifically, when a tree T performs photosynthesis, moisture for melting and absorbing carbon dioxide needs to be supplied from the roots of the tree T to the leaves through xylem, and it is presumed that a tree T with a high moisture content in the bark or underside of the leaves will grow by actively performing photosynthesis and respiration. For example, the determination unit 42 determines that a tree T whose reflection intensity in the near-infrared image calculated by the calculation unit 41 is lower than the threshold and appears black based on a threshold value of the reflection intensity of the reflected light set according to the tree species and measurement conditions is lively and has a high degree of health, and determines that a tree T whose reflection intensity in the near-infrared image calculated by the calculation unit 41 is higher than the threshold and appears white is low in health because photosynthesis and respiration are not performed sufficiently due to the progression of root decay, etc. Alternatively, the determination unit 42 may compare the reflection intensities at the first wavelength calculated by the calculation unit 41 for each of the multiple trees T and relatively determine the superiority or inferiority of the health of the multiple trees T.

The lighting unit 2, camera 3, and controller 4 may be mounted on a vehicle or the like, and the camera 3 may capture near-infrared images and visible light images while moving relative to the tree T. This allows efficient imaging of multiple trees T when multiple trees T are planted at intervals along a road, such as roadside trees.

In this way, the tree health determination system 1 according to this embodiment is a tree health determination system 1 that determines the health of a tree T, and is configured to include an illumination unit 2 that irradiates the tree T with measurement light having a wavelength in the near-infrared region, a camera 3 that receives light reflected from the tree T and captures a near-infrared image, a calculation unit 41 that calculates the reflection intensity of the reflected light at a first wavelength that is absorbed by the moisture of the tree T based on the near-infrared image, and a determination unit 42 that determines the health of the tree T from the reflection intensity at the first wavelength calculated by the calculation unit 41 based on the relationship between the reflection intensity and the health of the tree T that has been stored in advance.

This configuration makes it possible to determine the physiological health of trees in a timely, simple and low-cost manner, without the need for highly skilled diagnosticians as in conventional tree diagnosis.

<Experimental Example 1>
We will explain a comparative experiment of tree health assessment using the tree health assessment system 1 for cherry, zelkova, and oak trees and conventional tree diagnosis by an arborist.

Experimental procedure First, a conventional tree diagnosis was performed on cherry, zelkova, or oak trees by an arborist, and trees diagnosed as "healthy trees" and "unhealthy trees" in the tree diagnosis were selected. Note that in this experiment, trees diagnosed as "healthy trees" were selected that were close to trees diagnosed as "unhealthy trees." Note that "unhealthy trees" refer to trees that received a B2 rating (significant damage observed) in a road tree diagnosis based on the "Road Tree Diagnosis Manual (Tokyo Metropolitan Government Bureau of Construction)," and "healthy trees" refer to trees that received an A rating (healthy or close to healthy) or a B1 rating (noticeable damage observed).

Next, on a clear night, a halogen lamp was used to irradiate measurement light onto trees diagnosed as "healthy" or "unhealthy," and camera 3 was used to receive the light reflected from the trees and capture near-infrared images at a wavelength of approximately 1450 nm. A bandpass filter was provided on camera 3, and camera 3 captured near-infrared images corresponding to a wavelength of approximately 1450 nm that was transmitted through the bandpass filter. Next, calculation unit 41 calculated the average reflection intensity of the reflected light from the entire tree captured in each near-infrared image.

Experimental Results FIG. 3(a) is a near-infrared image of the trunk of the cherry tree 1 group (hereinafter referred to as "cherry tree 1 group") that was diagnosed as "unhealthy tree" by tree diagnosis among the cherry trees belonging to the first group, and FIG. 3(b) is a near-infrared image of the trunk of the cherry tree 1 group that was diagnosed as "healthy tree" by tree diagnosis. FIG. 4(a) is a near-infrared image of the trunk of the cherry tree 2 group (hereinafter referred to as "cherry tree 2 group") that was diagnosed as "unhealthy tree" by tree diagnosis among the cherry trees belonging to the second group, and FIGS. 4(b) and (c) are near-infrared images of the trunk of the cherry tree 2 group that was diagnosed as "healthy tree" by tree diagnosis. FIG. 5(a) is a near-infrared image of the trunk of a konara oak tree that was diagnosed as "unhealthy tree" by tree diagnosis, and FIG. 5(b) is a near-infrared image of the trunk of a konara oak tree that was diagnosed as "healthy tree" by tree diagnosis. Fig. 6(a) is a near-infrared image of the trunk of a zelkova tree diagnosed as "unhealthy" by tree diagnosis, and Fig. 6(b) is a near-infrared image of the trunk of a zelkova tree diagnosed as "healthy" by tree diagnosis. Fig. 7 is a graph showing the average value of the reflection intensity of reflected light in each near-infrared image. In this experiment, the image captured by the SWIR camera was divided into 256 gradations, with black being assigned a reflection intensity of 0 and white being assigned a reflection intensity of 256.

From Figure 7, it can be seen that the reflection intensity of "healthy trees" is lower than the reflection intensity of "unhealthy trees" in cherry group 1, cherry group 2, and zelkova. In other words, for tree species such as cherry and zelkova, a threshold value for the average reflection intensity of reflected light corresponding to "unhealthy trees" obtained in advance through experiments, etc. is set (e.g., cherry: 75, zelkova: 60), and if the average reflection intensity of the tree in the near-infrared image captured by camera 3 exceeds the threshold value, the tree is determined to be unhealthy, and if it is equal to or less than the threshold value, the tree is determined to be healthy.

On the other hand, in the case of Quercus serrata, it can be seen that the reflection intensity of "healthy trees" is higher than that of "unhealthy trees."

This is because in tree species such as oak, sawtooth oak, chestnut, and ginkgo, the healthier the tree, the more the older phloem turns into periderm, and the older periderm becomes bark, and as the annual rings grow, cracks appear and unevenness forms on the bark surface. The convex parts, which take up a large area of the bark surface, are old periderm that have been cut off from moisture and have turned into cork and are drying out, whereas the concave parts contain a newly formed cortex of periderm that contains a high moisture content, so the actual health of the tree cannot be accurately determined from the average reflection intensity of the reflected light over the entire tree.

Therefore, for tree species such as oak, chestnut, or ginkgo, the calculation unit 41 detects the position coordinates of the recess from the visible light image captured by the camera 3, calculates the reflection intensity of the reflected light at the recess in the near-infrared image, and sets a threshold value for the average value of the reflection intensity of the reflected light corresponding to an "unhealthy tree" obtained in advance through experiments, etc., and if the average value of the reflection intensity of the recess of the tree in the near-infrared image captured by the camera 3 exceeds the threshold value, the tree is judged to be unhealthy, and if it is equal to or less than the threshold value, the tree is judged to be healthy.

<Experimental Example 2>
A comparative experiment on determining the health of zelkova trees using the tree health determination system 1 and a conventional tree diagnosis by an arborist will be described.

- Experimental procedure First, five Zelkova trees that were close to each other were selected and a tree diagnosis was performed on them by a conventional arborist.

Next, on a clear night, a halogen lamp was used to irradiate each tree with measurement light, and a SWIR camera serving as camera 3 received the light reflected from the trees and captured near-infrared images at a wavelength of approximately 1450 nm. Next, the calculation unit 41 calculated the average reflection intensity of the light reflected from the entire tree captured in each near-infrared image.

To evaluate the balance between photosynthesis and tree body size to obtain nutrients, a balance index was used, calculated by dividing the crown projection area by the volume, and a balance ratio was used, calculated by multiplying the inverse of the balance index by 100. The crown projection area was obtained by the measurer looking up at the tree's crown and determining the edge of the crown from which the branches extend most in eight directions from the trunk, measuring the distance from the trunk to the edge of the crown with a tape measure in eight directions, and calculating the crown projection area using a radar chart, while the volume was calculated by estimating the trunk radius from the trunk circumference at 1.2 m above ground level. Trees with a high balance ratio are presumed to be in poor health because they are not receiving the amount of photosynthesis and nutrients appropriate to the tree's body.

Experimental results The results of tree vitality assessments by arborists using the "Tokyo Arborist Association Tree Survey Sheet" are shown in Figure 8. The trees diagnosed as "dangerous trees" were given a vitality rating of 1.09, giving them a diagnosis of "slightly poor" vitality category 2. The four trees diagnosed as "healthy trees" (healthy trees 1 to 4) were given vitality ratings of 0.45 to 0.55, giving them a diagnosis of "good" vitality category 1. The fact that only the "dangerous trees" were given a B2 rating is similar to the previous roadside tree health assessments explained in Experimental Example 1.

Figure 9 is a graph showing the average reflection intensity and balance ratio of reflected light in near-infrared images of five trees. In this experiment, as in Experimental Example 2, black was set to a reflection intensity of 0 and white was set to a reflection intensity of 256. Figure 10 is a table showing the balance index and balance ratio of dangerous trees and healthy trees 1-4. Figure 9 shows that the reflection intensity and balance ratio tendencies are similar for dangerous trees and healthy trees 1-4. In addition, when focusing on reflection intensity, it is clear that the bark of healthy tree 1 is the driest, and when focusing on the balance ratio, it is clear that the photosynthetic ability required to maintain the tree body of healthy tree 1 is inferior to the other trees.

Figures 8 and 9 show that healthy tree 1 is less healthy than the dangerous tree, and does not match the results of the previous street tree diagnosis described in Experimental Example 1 or the results of the tree diagnosis by an arborist. This is thought to be because in conventional street tree diagnosis health assessments, trees are diagnosed as extremely low in health due to reasons such as decay marks on the main branches caused by heavy pruning, poor bark condition, confirmed material decay, damage to the trunk, or poor wrapping after pruning. However, since there is almost no correlation between the presence or absence of decay marks due to heavy pruning and health, it is presumed that health assessment based on the reflection intensity of reflected light accurately evaluates the actual state of the tree by focusing on the moisture content on the underside of the trunk.

<Experimental Example 3>
A verification experiment of a method for accurately calculating the reflection intensity of reflected light from a near-infrared image will be described.

・Experimental procedure (Case 1)
Two near-infrared 500W lamps and a SWIR camera (e.g., Pika NIR-640 manufactured by RESONON) were prepared, and the average reflection intensity of the entire surface of the subject was measured based on the reflected light of a wavelength band of about 900 to 1700 nm reflected by the subject. Two blocks of cypress cut into a plate shape were prepared as the subject, one of which was dried and the other was moistened. In addition, the surface of the subject was covered with plastic wrap to prevent the block surface from drying due to the heat of the illumination and changing the degree of wetness. The angle (incident angle) of the measurement light with respect to the perpendicular line of the irradiated surface of the subject was set to about 30 degrees, and the distance between the subject and the SWIR camera was set to about 1 m. FIG. 11 shows a near-infrared image at a wavelength of about 1100 nm, and FIG. 12 shows a near-infrared image at a wavelength of about 1450 nm. The area surrounded by the dashed line in each image indicates the area where the average reflection intensity was evaluated.

・Experimental procedure (Case 2)
The same lighting, SWIR camera, and subject as in Case 1 were prepared, the angle of incidence of the measurement light was set to about 60 degrees, and the distance between the subject and the SWIR camera was set to about 1 m, and the average reflection intensity of the entire surface of the subject was measured based on the reflected light in the wavelength band of about 900 to 1700 nm reflected by the subject. Figure 13 shows a near-infrared image at a wavelength of about 1100 nm, and Figure 14 shows a near-infrared image at a wavelength of about 1450 nm. The area surrounded by the dashed line in each image indicates the area where the average reflection intensity was evaluated.

・Experimental procedure (Case 3)
The same illumination, SWIR camera, and subject were prepared as in case 1, and the same angle of incidence of the measurement light and the distance between the subject and the SWIR camera were set as in case 1. Then, the calculation unit 41 extracted the coordinates of the area (measurement area) where the reflection intensity was approximately constant in the near-infrared images at wavelengths of about 1000 nm and about 1100 nm captured by the SWIR camera. Next, the SWIR camera measured the reflection intensity of the reflected light in the wavelength band of about 900 to 1700 nm reflected in the measurement area of the subject. FIG. 15 shows an image showing the ratio (band ratio) of the reflection intensity at each wavelength of about 1000 nm and about 1100 nm in the subject in the dry state and the wet state. In the image of FIG. 15, the area where the reflection intensity does not change according to the wavelength change is shown in white, the area surrounded by the solid line in each image indicates the measurement area, and the area surrounded by the dashed line in each image indicates the area where the reflection intensity fluctuated particularly greatly.

・Experimental procedure (Case 4)
The same lighting, SWIR camera, and subject were prepared as in case 2, and the same angle of incidence of the measurement light and the distance between the subject and the SWIR camera were set as in case 2. Also, as in case 3, the calculation unit 41 extracted the coordinates of an area (measurement area) where the reflection intensity was approximately constant in the near-infrared images at wavelengths of approximately 1000 nm and approximately 1100 nm captured by the SWIR camera, and the SWIR camera measured the reflection intensity of the reflected light in the wavelength band of approximately 900 to 1700 nm reflected in the measurement area of the subject.

Experimental Results Fig. 16 is a graph showing the average reflection intensity of light in the wavelength band of about 900 to 1700 nm reflected from the entire surface of the subject in cases 1 and 2. Fig. 17 is a graph showing the reflection intensity of light in the wavelength band of about 900 to 1700 nm within the measurement area in cases 3 and 4. In this experiment, the reflection intensity was the measurement value of the SWIR camera (maximum value: about 65535).

Figure 16 shows that at a wavelength of about 1450 nm, which is absorbed by moisture, changes in reflection intensity are detected according to the degree of wetness. On the other hand, at a wavelength of about 1100 nm, which is not absorbed by moisture, the reflection intensity should be kept approximately constant regardless of the degree of wetness, but it can be seen that the reflection intensity changes.

This kind of change in reflection intensity (noise), which should not occur in reality, is presumed to be caused by light reflection or scattering (diffuse reflection) due to the angle of incidence of the measurement light on the subject, the angle of incidence of the reflected light on the SWIR camera, the distance from the subject to the SWIR camera, or the bending of the wrap covering the subject. And, even when photographing roadside trees, this kind of noise is presumed to occur depending on the lighting and camera conditions (position, incidence/reflection angle, illuminance, etc.) or the subject conditions (bark condition, trunk shape, etc.).

On the other hand, according to Figure 17, in cases 3 and 4, the reflection intensity is the same regardless of the degree of wetness in the entire wavelength band of approximately 1100 nm or less, and since the noise caused by the reflection and scattering (diffuse reflection) of light described above is corrected, it is inferred that the detection results of the reflection intensity according to the degree of wetness at a wavelength of 1450 nm, which is absorbed by moisture, are highly reliable.

Furthermore, as shown in FIG. 18, the difference in reflection intensity between subjects with different wetness at a wavelength of approximately 1450 nm in case 1 (dry: 5330, wet: 1693) was 3637, and the difference in reflection intensity between subjects with different wetness at a wavelength of approximately 1450 nm in case 2 (dry: 7247, wet: 3359) was 3888, whereas the difference in reflection intensity between subjects with different wetness at a wavelength of approximately 1450 nm in case 3 (dry: 5644, wet: 1883) was 3760, and the difference in reflection intensity between subjects with different wetness at a wavelength of approximately 1450 nm in case 4 (dry: 7876, wet: 4069) was 3807. This shows that the difference in reflection intensity is smaller in cases 3 and 4 than in cases 1 and 2, even between two cases with different angles of incidence and distances. This shows that by using the reflection intensity of light with a wavelength of approximately 1,450 nm in the measurement area within the subject, it is possible to accurately determine the health of trees even in groups with a mixture of bark conditions and trunk shapes.

<Experimental Example 4>
We will now describe a verification experiment of a method for determining the health of trees in summer using a threshold value (first threshold value) that is set based on the reflection intensity of reflected light obtained in winter.

- Experimental procedure First, in this experiment, trees were divided into the following six groups according to their species.
(1) Group 1 (G1): Deciduous broad-leaved trees such as Zelkova and Zelkova, which have thin outer bark and are thought to have low water retention and are prone to drying out, especially in dry outdoor conditions such as winter.
(2) Group 2 (G2): Deciduous broad-leaved trees such as autumn alder, Chinese hackberry, and Yoshino cherry, which have thick outer bark and are thought to have high water retention and are resistant to drying out, especially in dry outdoor conditions such as winter.
(3) Group 3 (G3): Deciduous broad-leaved trees such as tulip tree, persimmon, oak, ginkgo, etc., whose outer bark thickness is roughly intermediate between those in G1 and G2.
(4) Group 4 (G4): Evergreen broad-leaved trees such as Quercus myrsinae and Quercus sieboldii, with little leaf volume due to pruning of thick branches that are similar to the thickness of the trunk. The outer bark thickness of Group 4 is equivalent to that of Group 3.
(5) Group 5 (G5): Evergreen broad-leaved trees such as Quercus myrsinae and Quercus sieboldii, with thick branches that are similar to the thickness of the trunk and have a large amount of leaves. The outer bark thickness of Group 5 corresponds to that of Group 3.
(6) Group 6 (G6): Evergreen coniferous trees such as cedar.

Next, two near-infrared 36V 45W lights (e.g., HySiL1500-2 manufactured by KLS) and a SWIR camera (e.g., NIR640SN manufactured by Vision Sensing) were prepared, and the angle of the measurement light with respect to the perpendicular line of the irradiated surface of the subject (incident angle) was set to approximately 0 degrees, and the distance between the subject and the SWIR camera was set to approximately 1m. Then, for groups 1 to 6, the reflection intensity of the entire subject was measured based on the reflected light of approximately 1450 nm wavelength band reflected by the subject (tree) during the period when the photosynthesis of the trees is relatively inactive (winter).

Experimental Results FIG. 19 is a table showing the average reflection intensity of near-infrared images at a wavelength of about 1450 nm measured by tree species in winter. The average reflection intensity in FIG. 19 means the value when the reflection intensity is measured from one direction on the tree, and means the average value when the reflection intensity is measured from multiple directions on the tree. FIG. 20 is a graph showing the average reflection intensity by tree species shown in FIG. 19. FIGS. 21(a) to (c) are near-infrared images at a wavelength of about 1450 nm taken of each of the trunks of three zelkova trees diagnosed as healthy trees in Experimental Example 1 during a period when the photosynthesis of the trees is relatively active (summer), and FIG. 21(d) is a near-infrared image taken of the trunk of a zelkova tree diagnosed as unhealthy in Experimental Example 1 during summer. FIG. 22 is a graph showing the average reflection intensity of each near-infrared image of the three zelkova trees shown in FIG. 21 and the first threshold value described below for the tree species group G1 to which the zelkova trees belong. In this experiment, the reflection intensity was the measurement value of the SWIR camera (maximum value: approximately 65535).

19 and 20, it can be seen that there is no significant difference in the reflection intensity measured in winter between unhealthy trees with low health and healthy trees with high health within the same tree species. This is presumably because in winter, sunlight energy is weak, and deciduous trees lose their leaves, so photosynthesis is inactive and bark moisture is low regardless of the health of the tree. Therefore, the reflection intensity measured in winter is considered to correspond to trees in a dormant state, and a first threshold value used for health assessment was set for each of groups 1 to 6 based on the reflection intensity measured in winter. Note that a "dormant state" refers to a state in which the amount of photosynthesis and the amount of water absorbed by the roots are at a minimum due to leaf fall or reduced solar radiation energy, such as during winter, and growth activity is reduced.

Specifically, assuming that the reflection intensity in group 1 is about 30,000 and that a difference of about ±10% is not significant, the first threshold for group 1 was set to 28,000. Similarly, since the reflection intensity in group 2 is about 20,000, the first threshold for group 2 is set to 18,000, since the reflection intensity in group 3 is about 25,000, the first threshold for group 3 is set to 23,000, since the reflection intensity in group 4 is about 23,000, the first threshold for group 4 is set to 22,000, since the reflection intensity in group 5 is about 16,000, the first threshold for group 5 is set to 15,000, and since the reflection intensity in group 6 is about 28,000, the first threshold for group 6 is set to 28,000. The threshold values for groups 1 to 6 are not limited to the values mentioned above, and may vary depending on the lighting and camera conditions (position, incident and reflection angles, illuminance, etc.) or the subject conditions (bark condition, trunk shape, etc.).

As shown in FIG. 22, the reflection intensities of reflected light from three zelkova trees (healthy trees 1 to 3) measured in summer, corresponding to the images in FIG. 21(a) to (c), are 15,437, 17,784, and 16,900, respectively, which are significantly lower than the first threshold value of 28,000 for group 1. This is presumably because the energy of sunlight is strong in summer, and deciduous trees have leaves and photosynthesis is active, so the bark wetness is higher in summer than in winter. Therefore, these trees are determined to be healthy trees with a high level of health. The reflection intensity of reflected light from a zelkova tree measured in summer, corresponding to the image in FIG. 21(d), was 29,442.

In this way, if the reflection intensity of the reflected light from the tree being evaluated, measured in summer, is less than the first threshold value for the group to which the tree belongs, the tree can be determined to have a higher bark wetness and higher health than dormant trees. On the other hand, if the reflection intensity of the reflected light from the tree measured in summer is approximately equal to the first threshold value for the group to which the tree belongs, the tree can be determined to have a lower bark wetness and lower health, similar to dormant trees. Note that approximately equal to the first threshold value refers to, for example, a range (e.g., ±10% or less) within which it can be assumed that there is no significant difference based on the first threshold value.

In addition, the reflection intensities of the reflected light from healthy trees belonging to groups 1 to 6 measured in the summer can be regarded as corresponding to trees with high health, and a second threshold can be set based on these reflection intensities to determine whether the trees are healthy or not.

Specifically, when the second threshold value for group 1, which indicates trees with a high level of health, is set to 17,000 based on the reflection intensity of reflected light from healthy trees 1 to 3 in group 1 shown in FIG. 22, if the reflection intensity of reflected light from a tree belonging to group 1 is approximately equal to the second threshold value, the tree is determined to have a high bark moisture level, similar to healthy trees, and therefore a high level of health. Note that approximately equal to the second threshold value refers to, for example, a range (e.g., ±10% or less) within which it can be assumed that there is no significant difference based on the second threshold value. Note that, in this embodiment, the numerical value for group 1 is exemplified as the second threshold value, but the second threshold values for the other groups 2 to 6 are set separately.

In addition, a threshold value (first threshold value) set based on the reflection intensity of the reflected light from dormant trees belonging to each of groups 1 to 6 measured in winter and a threshold value (second threshold value) set based on the reflection intensity of healthy trees belonging to each of groups 1 to 6 measured in summer may be stored in advance, and the closer the reflection intensity of the reflected light from the tree being evaluated is to the first threshold value of the group 1 to 6 to which the tree belongs, the lower the healthiness may be determined, and the closer the reflection intensity is to the second threshold value of the group 1 to 6 to which the tree belongs, the higher the healthiness may be determined.

The present invention can be modified in various ways without departing from the spirit of the invention, and it goes without saying that the present invention extends to such modifications.

1: Tree health determination system 2: Lighting unit 3: Camera 4: Controller 41: Calculation unit 42: Determination unit T: Tree

Claims (9)

A tree health assessment system for assessing the health of trees, comprising:
an illumination unit that irradiates measurement light having a wavelength in the near-infrared region toward the trunk of the tree;
A camera that captures a near-infrared image by receiving light reflected by the bark surface of the trunk .
a calculation unit that calculates a reflection intensity of the reflected light at a first wavelength that is absorbed by water of the tree based on the near-infrared image;
a determination unit that determines a health level of the tree from the reflection intensity at the first wavelength calculated by the calculation unit based on a relationship between the reflection intensity and the health level of the tree that is stored in advance;
A tree health assessment system comprising: The illumination unit irradiates measurement light having a wavelength in the visible light region toward the tree trunk ,
The camera receives the measurement light reflected by the bark surface and captures a visible light image,
The tree health determination system of claim 1, characterized in that the calculation unit detects position coordinates of a depression formed on the bark surface from the visible light image and calculates the reflection intensity at the first wavelength at the position coordinates of the depression. The tree health assessment system of claim 1, characterized in that the calculation unit calculates the reflection intensity of the reflected light at second and third wavelengths that are less easily absorbed by the moisture of the tree than the first wavelength, detects position coordinates of a measurement area in the near-infrared image where the reflection intensities at the second and third wavelengths are approximately equal, and calculates the reflection intensity at the first wavelength in the measurement area. The determination unit preliminarily stores a first threshold value that is set based on the reflection intensity of the reflected light acquired during a period when photosynthesis of the tree is relatively inactive,
The tree health determination system of claim 1, characterized in that the determination unit determines that the tree is in good health when the reflection intensity of the reflected light from the tree to be determined, obtained during a period when the tree's photosynthesis is relatively active, is less than the first threshold value. The determination unit preliminarily stores a first threshold value that is set based on the reflection intensity of the reflected light acquired during a period when photosynthesis of the tree is relatively inactive,
The tree health determination system of claim 1, characterized in that the determination unit determines that the health of the tree is low when the reflection intensity of the reflected light from the tree to be determined, obtained during a period when the tree's photosynthesis is relatively active, is approximately equal to the first threshold value. The determination unit prestores a second threshold value that is acquired during a period when photosynthesis of the tree is relatively active and is set based on the reflection intensity of the reflected light of the tree having a high degree of health,
The tree health determination system of claim 1, characterized in that the determination unit determines that the tree is of high health when the reflection intensity of the reflected light from the tree to be determined is approximately equal to the second threshold value. The determination unit prestores a first threshold value set based on the reflection intensity of the reflected light acquired during a period when the photosynthesis of the tree is relatively inactive, and a second threshold value set based on the reflection intensity of the reflected light acquired during a period when the photosynthesis of the tree is relatively active and from the tree having a high degree of health,
The tree health determination system described in claim 1, characterized in that the determination unit determines that the closer the reflection intensity of the reflected light in the tree to be determined is to the first threshold value, the lower the healthiness, and that the closer the reflection intensity is to the second threshold value, the higher the healthiness. The illumination unit irradiates the measurement light at night,
The tree health determination system according to claim 1 , wherein the camera captures the near-infrared image at night. the lighting unit and the camera are mounted on a vehicle that is movable horizontally relative to the tree;
The illumination unit irradiates the measurement light while moving in a horizontal direction,
2. The tree health determination system according to claim 1, wherein the camera captures the near-infrared image while moving in a horizontal direction.