Tree measuring device

Abstract

This enables the estimation of the minimum diameter of trees with greater accuracy while further reducing the processing load. [Solution] The system includes a 3D image acquisition unit 101 that acquires a 3D image, which is a 3D point cloud or 3D model obtained by observing an area containing trees from multiple directions with sensors; a cross-section extraction unit 102 that extracts cross-sectional data, which is data of a cross section at a specified height from the ground surface, from the 3D image acquired by the 3D image acquisition unit 101; a cross-section conversion unit 103 that converts the cross-sectional data extracted by the cross-section extraction unit 102 into a 2D cross-sectional image viewed from directly above the tree; a tree recognition unit 104 that recognizes a tree cross-sectional image, which is a 2D cross-sectional image of the tree trunk, based on the 2D cross-sectional image converted by the cross-section conversion unit 103; and a diameter estimation unit 108 that estimates the diameter of the largest circle that fits in the tree cross-sectional image as the minimum diameter of the tree, based on the tree cross-sectional image recognized by the tree recognition unit 104.

Description

Technical Field

[0001] The present disclosure relates to a measuring device for trees.

Background Art

[0002] One method for quantifying the amount of carbon sequestration in a forest is to derive the amount of carbon sequestration in the forest from the tree ages in the forest. The tree age is estimated using the tree diameter, but manually measuring each tree in the forest one by one requires a great deal of labor. Also, in order to accurately derive the amount of carbon sequestration in the forest from the tree ages in the forest, it is necessary to estimate with high accuracy the minimum diameter, which is the so-called tip diameter of the cross-section of the tree. In contrast, Patent Document 1 discloses a technique for estimating the diameter of a single tree from continuous distance data for the same single tree at a plurality of measurement positions using a scanning laser rangefinder. In the technique of Patent Document 1, it is assumed that the cross-section of the tree is circular, and from the continuous distance data, the least squares method, the Hough transform, etc. are used to fit it to a circle and identify the radius.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the actual cross-section of a tree is not circular but has a distorted shape. Also, in the technique of Patent Document 1, a circle of the cross-section is estimated from only the distance data for a part of the cross-section of the tree. Therefore, in the technique of Patent Document 1, there is a possibility that it will be fitted to a circle whose size is significantly different from the circle from which the minimum radius of the cross-section of the tree can be obtained. Therefore, in the technique of Patent Document 1, there is a possibility that the minimum diameter of the tree cannot be accurately estimated. Also, when estimating the minimum diameter of the cross-section of a tree using a device, it is desirable to reduce the processing load on the device.

[0005] One objective of this disclosure is to provide a tree measuring device that enables the estimation of the minimum diameter of a tree with greater accuracy while further reducing the processing load. [Means for solving the problem]

[0006] The above objectives are achieved by a combination of features described in the independent claims, and the subordinate claims provide further advantageous specific examples of the disclosure. The reference numerals in parentheses in the claims indicate correspondences with specific means described in the embodiments described later as one aspect, and do not limit the technical scope of this disclosure.

[0007] To achieve the above objective, the tree measuring device of this disclosure includes a 3D image acquisition unit (101) that acquires a 3D image which is a 3D point cloud or 3D model obtained by observing an area containing a tree from multiple directions with a sensor (20), a cross section extraction unit (102) that extracts cross section data which is data of a cross section at a specified height from the ground surface from the 3D image acquired by the 3D image acquisition unit, a cross section conversion unit (103) that converts the cross section data extracted by the cross section extraction unit into a 2D cross section image viewed from directly above the tree, a tree recognition unit (104) that recognizes a tree cross section image which is a 2D cross section image of the tree trunk based on the 2D cross section image converted by the cross section conversion unit, and a diameter estimation unit (108) that estimates the diameter of the largest circle that fits in the tree cross section image as the minimum diameter of the tree based on the tree cross section image recognized by the tree recognition unit.

[0008] With the above configuration, the minimum diameter of the tree is estimated by extracting cross-sectional data at a specified height from the ground surface from a 3D image obtained by observing the area containing the tree from multiple directions using sensors. Therefore, the processing load is reduced because the processing target can be reduced compared to processing the entire 3D image. In addition, since the tree cross-sectional image, which is a 2D cross-sectional image of the tree trunk, is recognized based on a 2D cross-sectional image obtained by converting the cross-sectional data into an image viewed from directly above the tree, it is possible to suppress the estimation of the minimum diameter of the tree by targeting parts other than the tree trunk. Furthermore, since cross-sectional data is extracted from a 3D image obtained by observing from multiple directions and converted into a 2D cross-sectional image, it is easier to obtain a tree cross-sectional image that is closer to the actual cross-section of the tree. Therefore, the diameter of the largest circle that fits in the tree cross-sectional image will be closer to the actual minimum diameter of the tree. Consequently, it becomes possible to estimate the minimum diameter of the tree with greater accuracy in the diameter estimation unit. As a result, it becomes possible to estimate the minimum diameter of the tree with greater accuracy while further reducing the processing load. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows an example of a schematic configuration for a tree measurement system. [Figure 2] This figure shows an example of a general configuration of a measuring device for trees. [Figure 3] This diagram schematically represents the set of three-dimensional images before horizontal level projection. [Figure 4] This is an enlarged view of section A in Figure 3. [Figure 5] This diagram schematically represents the set of three-dimensional images after horizontal level projection. [Figure 6] This is an enlarged view of section B in Figure 5. [Figure 7] This is a schematic diagram illustrating the extraction of cross-sectional data at multiple heights. [Figure 8] This figure illustrates an example of converting cross-sectional data into a two-dimensional image. [Figure 9] This is a diagram illustrating one example of a method for determining the lean of a tree. [Figure 10] This diagram illustrates an example of tree cross-sectional image correction in the tilt correction unit. [Figure 11] This diagram illustrates an example of tree cross-sectional image correction in the tilt correction unit. [Figure 12] This diagram illustrates an example of multi-stem discrimination in the multi-stem discrimination unit. [Figure 13] This is a diagram illustrating the estimation of the minimum diameter of a tree. [Figure 14] This flowchart shows an example of the minimum diameter estimation process using a tree measuring device. [Modes for carrying out the invention]

[0010] (Embodiment 1) <Outline configuration of tree measurement system 1> Embodiments for disclosure will be described with reference to the drawings. Hereinafter, Embodiment 1 of this disclosure will be described with reference to the drawings. As an example, the tree measurement system 1 includes a tree measuring device 10, a sensor 20, and a 3D image generating device 30, as shown in Figure 1.

[0011] Sensor 20 is used to observe trees in a forest. Examples of sensor 20 include an imaging device and a search wave sensor. The imaging device images a predetermined range starting from itself. The search wave sensor transmits a search wave to a predetermined range starting from itself. The search wave sensor obtains sensing information based on the scanned signal obtained when it receives the reflected wave that has been reflected by the object. The search wave sensor observes the distance and shape of the object by transmitting and receiving the search wave. Examples of search wave sensors include Lidar (Light Detection and Ranging / Laser Imaging Detection and Ranging). The search wave sensor may also be a millimeter-wave radar, sonar, etc. Sensor 20 may be an imaging device or a search wave sensor. For example, Lidar may be used as the search wave sensor.

[0012] While moving manually or automatically within the forest, the sensor 20 observes the area containing each tree from multiple directions. When the sensor 20 is automatically moved within the forest, it may be realized by mounting the sensor 20 on an autonomous robot, drone, etc. In this embodiment, the sensor 20 shall collect observation results for a pre-planned section within the forest (hereinafter referred to as a planned section).

[0013] The three-dimensional image generation device 30 constructs a three-dimensional image from the observation results of the sensor 20. The three-dimensional image may be a three-dimensional point cloud or a three-dimensional model. In the case of a three-dimensional point cloud, it may be done as follows. When the sensor 20 is an exploration wave sensor, a three-dimensional point cloud may be constructed from the point cloud obtained by positioning. The three-dimensional point cloud is a point cloud on a three-dimensional coordinate. It is assumed that this three-dimensional coordinate corresponds to a coordinate in a geographical coordinate system. As an example, by positioning the observation position of the sensor 20 with a locator, it is assumed that the coordinate of the three-dimensional point cloud corresponds to the coordinate in the geographical coordinate system. The locator may be equipped with, for example, a GNSS (Global Navigation Satellite System) receiver and an inertial sensor. The GNSS receiver receives positioning signals from a plurality of positioning satellites. The inertial sensor includes, for example, a gyro sensor and an acceleration sensor. The locator sequentially positions the observation position of the sensor 20 by combining the positioning signal received by the GNSS receiver and the measurement result of the inertial sensor. The observation position may be coordinates of latitude, longitude, and altitude. The three-dimensional point cloud may also be configured to be used after being converted into coordinates in a geographical coordinate system.

[0014] When the 3D image generation device 30 constructs a 3D model, it may be done as follows. The 3D image generation device 30 may construct a 3D model from a plurality of captured images, for example, by SfM (Structure From Motion), 3D Gaussian Splatting, etc. Regarding the 3D model as well, it may be assumed that a correspondence with coordinates in the geographical coordinate system is established in the same manner as the 3D point cloud. The 3D model may also be configured to be converted into coordinates in the geographical coordinate system and used. In the present embodiment, the case where the 3D image generation device 30 constructs a 3D image by converting it into coordinates in the geographical coordinate system will be taken as an example, and the following description will continue.

[0015] The tree measurement device 10 is mainly composed of, for example, a computer including a processor, a volatile memory, a non-volatile memory, I / O, and a bus connecting these. The tree measurement device 10 executes processing related to estimating the diameter of a tree, etc., based on the 3D image acquired from the 3D image generation device 30. The tree measurement device 10 may be configured to be provided in a robot or the like equipped with the sensor 20 and the 3D image generation device 30, or may be configured to be provided in a server or the like. Details of the tree measurement device 10 will be described below.

[0016] <Schematic configuration of the tree measurement device 10> As shown in FIG. 2, the tree measurement device 10 includes a 3D image acquisition unit 101, a cross-section extraction unit 102, a cross-section conversion unit 103, a tree recognition unit 104, an inclination identification unit 105, an inclination correction unit 106, a multi-trunk discrimination unit 107, a diameter estimation unit 108, a tree number identification unit 109, a tree position identification unit 110, and a tree distance identification unit 111.

[0017] The 3D image acquisition unit 101 acquires the 3D image constructed by the 3D image generation device 30. That is, the 3D image acquisition unit 101 acquires a 3D image obtained by observing a region including a tree from a plurality of directions with the sensor 20.

[0018] The cross-section extraction unit 102 extracts cross-sectional data (hereinafter referred to as cross-sectional data) from the 3D image acquired by the 3D image acquisition unit 101, specifically the cross-sectional data at a specified height from the ground surface. The specified height can be the height commonly known as chest height. For example, the specified height can be 1.2m or 1.3m. In this embodiment, the specified height is 1.3m. As an example, the coordinate group of the layer intersecting the horizontal plane at the specified height in the 3D image can be extracted as cross-sectional data. In other words, the cross-sectional data corresponds to the data of a cross-section obtained by slicing the 3D image with a horizontal plane at the specified height.

[0019] The cross-section extraction unit 102 has a reference level projection unit 121 as a sub-functional block. The reference level projection unit 121 projects the highest part of the 3D image acquired by the 3D image acquisition unit 101, which is in contact with the ground surface, to a reference horizontal level. This projection will be referred to as horizontal level projection below. The reference horizontal level will be referred to as the reference horizontal level below. The part of the 3D image that is in contact with the ground surface (hereinafter referred to as the ground contact part) can be the bottom of the 3D image. The ground contact part can also be described as the part observed by the sensor 20 as being in contact with the ground surface. For a 3D image of a tree growing from a nearly horizontal ground, the coordinate of the height of the ground contact part for one 3D image will be uniformly the same value. On the other hand, for a 3D image of a tree growing from a slope, the coordinate of the height of the ground contact part for one 3D image will vary to different values ​​depending on the slope. The reference level projection unit 121 recognizes the coordinate of the highest point in the ground contact area of ​​the 3D image acquired by the 3D image acquisition unit 101 as the ground level of that 3D image. The highest point in the ground contact area will be referred to as the highest ground contact point below. The reference horizontal level can be the lowest ground level among the ground levels of the 3D image acquired by the 3D image acquisition unit 101. Alternatively, the reference horizontal level may be the coordinate of height "0" in the geographic coordinate system. The reference horizontal level can be set by the cross-section extraction unit 102.

[0020] Here, horizontal level projection will be explained using Figures 3 to 6. Figure 3 is a schematic representation of the set of three-dimensional images before horizontal level projection. Figure 4 is an enlarged view of part A in Figure 3. Figure 5 is a schematic representation of the set of three-dimensional images after horizontal level projection. Figure 6 is an enlarged view of part B in Figure 5. In Figures 3 to 6, HD indicates the height direction, and HOR indicates the horizontal direction for a given orientation. In Figures 3 to 6, 3DI shows the three-dimensional image. In Figures 3 to 6, SHL indicates the reference horizontal level. In Figures 3 to 4, GL indicates the ground level. In Figures 5 to 6, 3DIP shows the three-dimensional image after horizontal level projection. In Figures 4 and 6, HP indicates the highest point of contact with the ground.

[0021] When observing trees in an area including a slope with sensor 20, a 3D image 3DI is obtained in which the ground level GL differs from the reference horizontal level SHL, as shown in Figure 3. In particular, for trees growing from a slope, a 3D image 3DI is obtained in which the coordinates of the height of the ground contact point vary depending on the slope, as shown in Figure 4. In horizontal level projection, the height value of the highest ground contact point HP of such a 3D image 3DI is projected to match the height of the reference horizontal level SHL, as shown in Figure 6. As a result, as shown in Figure 5, 3D images 3DI with varying ground levels are projected onto a single reference horizontal level SHL.

[0022] The cross-section extraction unit 102 then performs horizontal level projection using the reference level projection unit 121, and then extracts cross-section data at a specified height from that ground surface, using the reference horizontal level as the ground surface. For trees growing on a slope, the coordinates of the height of the part touching the ground vary depending on the slope, so the height corresponding to breast height also varies depending on the orientation. Therefore, depending on the orientation, the cross-section extraction unit 102 may extract cross-section data at a height unsuitable for estimating the tree's diameter. In contrast, by extracting cross-section data after performing horizontal level projection, it becomes possible to suppress the extraction of cross-section data at a height unsuitable for estimating the tree's diameter, even for trees growing on a slope. Furthermore, in order to reduce unnecessary processing load, it is preferable for the reference level projection unit 121 to focus on 3D images where the ground level is above the reference horizontal level and perform horizontal level projection only.

[0023] The cross-section extraction unit 102 preferably extracts cross-sectional data by cutting out a three-dimensional image from horizontal planes at multiple heights, which are a specified height and heights different from the specified height, from the ground surface. The heights different from the specified height may be one type or multiple types. For example, the heights different from the specified height may be two types: a height higher than the specified height and a height lower than the specified height. In the example of this embodiment, the heights different from the specified height are 1.8m, which is 0.5m higher than the specified height of 1.3m, and 0.8m, which is 0.5m lower than the specified height of 1.3m. In other words, in the example of this embodiment, cross-sectional data is extracted at three types of heights: 1.8m, 1.3m, and 0.8m.

[0024] Here, we will explain the extraction of cross-sectional data at multiple heights using Figure 7. In Figure 7, we will explain using the example where the ground level of the 3D image 3DI has been projected onto the reference horizontal level SHL. As shown in Figure 7, cross-sectional data CSDH, CSDC, and CSDL are obtained by cutting out the 3D image 3DL from horizontal planes at multiple heights. Cross-sectional data CSDH is the cross-sectional data cut out at 1.8m, which is higher than the specified height. Cross-sectional data CSDC is the cross-sectional data cut out at the specified height of 1.3m. Cross-sectional data CSDL is the cross-sectional data cut out at 0.8m, which is lower than the specified height.

[0025] The section transformation unit 103 converts the section data extracted by the section extraction unit 102 into a two-dimensional section image viewed from directly above the tree. As shown in Figure 8, the section transformation unit 103 converts the section data CSDC into a two-dimensional section image 2DI, which is a two-dimensional image viewed from the axis of the height direction HD. Figure 8 is a diagram illustrating an example of converting section data into a two-dimensional image. In the example in Figure 8, an example is shown of converting section data CSDC into a two-dimensional section image 2DI viewed from the axis of the height direction HD. In Figure 8, Lat indicates the latitude direction and Lon indicates the longitude direction. Note that the axes of the two-dimensional coordinates of the two-dimensional section image 2DI do not have to be axes of the geographic coordinate system if georeferencing is performed later. For section data of different heights extracted by the section extraction unit 102, the section transformation unit 103 only needs to convert them into two-dimensional section images of different heights.

[0026] The tree recognition unit 104 recognizes a tree cross-sectional image, which is a two-dimensional cross-sectional image of a tree trunk, based on the two-dimensional cross-sectional image converted by the cross-sectional conversion unit 103. For example, this can be done as follows: The tree recognition unit 104 performs edge detection on the two-dimensional cross-sectional image and identifies the point cloud of edges in the two-dimensional cross-sectional image. The tree recognition unit 104 constructs a rounded contour shape (hereinafter referred to as a round shape) from the point cloud of adjacent edges. If the size of the constructed round shape is within a specified range, the tree recognition unit 104 recognizes the image of the constructed round shape as a tree cross-sectional image. The specified range can be set arbitrarily to a size corresponding to the trunk of the tree being observed. If the cross-sectional conversion unit 103 converts two-dimensional cross-sectional images for different heights, the tree recognition unit 104 can recognize tree cross-sectional images for different heights based on those two-dimensional cross-sectional images.

[0027] The tilt identification unit 105 identifies the tree's tilt based on the degree of horizontal displacement between tree cross-sectional images of different heights recognized by the tree recognition unit 104. Here, an example of how to identify the tree's tilt will be explained using Figure 9. In Figure 9, SL represents the tree's tilt estimated from the alignment of the cross-sectional data of the tree's 3D image at different heights. The cross-sectional data of the tree's 3D image at different heights will be shifted horizontally along the tree's tilt. Therefore, as shown in Figure 9, the tree's tilt SL can be estimated from the height of each cross-sectional data and the horizontal displacement of each cross-sectional data CSDH, CSDC, CSDL. Using this, the tilt identification unit 105 identifies the tree's tilt SL from the horizontal displacement between tree cross-sectional images of different heights recognized by the tree recognition unit 104 and the height corresponding to each tree cross-sectional image. As an example, the tree's tilt SL can be identified by the tilt of an approximate straight line, with each point being the centroid position of the tree cross-sectional images of different heights, with respect to the direction of the aforementioned horizontal displacement. Alternatively, instead of this approximation line, a line passing through the centroids of two tree cross-section images, ordered from highest to lowest height, such as in the cross-section data CSDH and cross-section data CSDC, may be used. The reason for using tree cross-section images instead of cross-section data is to prevent the incorrect identification of tree inclination using non-tree data.

[0028] The tilt correction unit 106 corrects the tree cross-section image at a specified height. Here, an example of tree cross-section image correction by the tilt correction unit 106 will be explained using Figures 10 to 11. As shown in Figure 10, the tilt correction unit 106 corrects the inclination of the plane from which the cross-section data CSDC is extracted at a specified height by the amount of inclination SL identified by the inclination identification unit 105, and then re-extracts the cross-section data. Subsequently, after re-extracting the cross-section data CSDC, as shown in Figure 11, tilt correction is performed to correct the inclination of the data of that cross-section with respect to the horizontal plane by the amount of inclination SL. The tilt correction unit 106 re-extracts the tilt-corrected data as cross-section data CSDC. Then, based on this re-extracted cross-section data CSDC, the tree cross-section image at a specified height is corrected. More specifically, this extracted corrected cross-section data CSDC is converted into a 2D cross-section image by the cross-section conversion unit 103, and the tree recognition unit 104 performs correction to re-recognize the tree cross-section image based on this 2D cross-section image.

[0029] Furthermore, the tilt correction unit 106 may be configured not to correct the tree cross-section image at a specified height if the tilt identified by the tilt identification unit 105 is below a threshold. The threshold here should be a value that does not affect the accuracy of the tree diameter estimation, and can be set arbitrarily.

[0030] The multi-stem discrimination unit 107 compares tree cross-sectional images of different heights recognized by the tree recognition unit 104 to determine whether multiple tree cross-sectional images of the same height belong to a single multi-stem tree or to multiple separate trees. This determination will be referred to as multi-stem discrimination below. A single multi-stem tree will be referred to as a multi-stem tree below. Here, an example of multi-stem discrimination by the multi-stem discrimination unit 107 will be explained using Figure 12. CSDHa and SCDHb in Figure 12 show cross-sectional data of different trunks of a multi-stem tree at a height of 1.8m, which is higher than the specified height. CSDCa and SCDCb in Figure 12 show cross-sectional data of different trunks of a multi-stem tree at a specified height of 1.3m. SLa and SLb in Figure 12 show the inclination of different trunks of a multi-stem tree. Inclination SLa shows the trunk inclination of cross-sectional data CSDHa and CSDCa. Inclination SLb shows the trunk inclination of cross-sectional data CSDHa and CSDCb.

[0031] When multiple tree cross-sectional images of the same height are recognized within a predetermined distance, the multi-stem discrimination unit 107 determines the inclination of the tree based on the degree of horizontal displacement between the tree cross-sectional images of different heights. This determination may also be performed by the inclination determination unit 105. The predetermined distance here should be a value that is estimated to be impossible as the distance between the multiple stems of a single tree, and can be set arbitrarily. When a tree has multiple stems, even if it is the same single tree, multiple cross-sectional data will be extracted at the same height, as shown in Figure 12. Here, if it is the trunk of a single tree, the straight lines of inclination indicated by the alignment of the cross-sectional data of different heights will intersect at a height higher than the reference horizontal level. The multi-stem discrimination unit 107 uses this point to determine the inclination of the tree for each of the multiple tree cross-sectional images of the same height, based on the horizontal displacement between the tree cross-sectional images of different heights and the height corresponding to each tree cross-sectional image. Furthermore, if the lines of each inclination intersect at a height higher than the reference horizontal level, it is identified as a multi-trunk tree. On the other hand, if the lines of each inclination do not intersect at a height higher than the reference horizontal level, they are identified as multiple separate trees.

[0032] The diameter estimation unit 108 estimates the minimum diameter of a tree based on the tree cross-section image recognized by the tree recognition unit 104, by determining the diameter of the largest circle that fits within the tree cross-section image. As shown in Figure 13, TSC is the tree cross-section image, and IC is the largest circle that fits within the tree cross-section image. Also, COG in Figure 13 is the centroid of this circle, and MD is the minimum diameter of this circle. The minimum diameter can be estimated as follows, for example. First, the centroid of the tree cross-section image is determined. Next, the distance between the two points that intersect the line passing through this centroid with the edge points of the tree cross-section image, and the distance between these two points is estimated as the minimum diameter. Alternatively, the largest circle that fits within the tree cross-section image can be determined, for example, by simulation, and the diameter of this circle can be estimated as the minimum diameter.

[0033] With the above configuration, the minimum diameter of the tree is estimated by extracting cross-sectional data at a specified height from the ground surface from a 3D image obtained by observing the area containing the tree from multiple directions using the sensor 20. Therefore, the processing load is reduced because the number of objects to be processed is reduced compared to processing the entire 3D image. In addition, since the tree cross-sectional image, which is a 2D cross-sectional image of the tree trunk, is recognized based on a 2D cross-sectional image obtained by converting the cross-sectional data into an image viewed from directly above the tree, it is possible to suppress the estimation of the minimum diameter of the tree by targeting parts other than the tree trunk. Furthermore, since cross-sectional data is extracted from a 3D image obtained by observing from multiple directions and converted into a 2D cross-sectional image, it is easier to obtain a tree cross-sectional image that is closer to the actual cross-section of the tree. Therefore, the diameter of the largest circle that fits in the tree cross-sectional image will be closer to the actual minimum diameter of the tree. Consequently, the diameter estimation unit 108 can estimate the minimum diameter of the tree with greater accuracy. As a result, it is possible to estimate the minimum diameter of the tree with greater accuracy while further reducing the processing load.

[0034] When the tilt correction unit 106 corrects the tree cross-section image at a specified height, the diameter estimation unit 108 should perform the following: Based on the tree cross-section image at a specified height corrected by the tilt correction unit 106, the diameter estimation unit 108 should estimate the diameter of the largest circle that fits within the tree cross-section image as the minimum diameter of the tree. When the trunk of a tree is tilted relative to the horizontal plane, the cross-sectional data used to extract a three-dimensional image from a horizontal plane may differ significantly in shape and size compared to when the cross-section of the trunk is not tilted relative to the horizontal plane. This difference may reduce the accuracy of the minimum diameter of the tree estimated based on the tree cross-section image. In contrast, by correcting the tree cross-section image at a specified height with the tilt correction unit 106, a tree cross-section image corrected for such differences can be obtained. Therefore, with the above configuration, it becomes possible to estimate the minimum diameter of the tree with greater accuracy even when the trunk of a tree is tilted relative to the horizontal plane.

[0035] When the multi-trunk discrimination unit 107 determines that a tree is a single tree with multiple trunks, the diameter estimation unit 108 should proceed as follows: The diameter estimation unit 108 should estimate the minimum diameter of the tree by using a tree cross-sectional image of the tree that is not divided into multiple tree cross-sectional images and is at a height lower than a specified height, and determining the diameter of the largest circle that fits within the tree cross-sectional image. A tree cross-sectional image of the tree that is not divided into multiple tree cross-sectional images and is at a height lower than a specified height will be referred to below as a common partial cross-sectional image. If a common partial cross-sectional image has already been obtained by the tree recognition unit 104, the minimum diameter of the tree should be estimated based on this common partial cross-sectional image. In the example of this embodiment, this applies when the tree cross-sectional image at a height of 0.8m, which is lower than the specified height, is a tree cross-sectional image that is not divided into multiple tree cross-sectional images. On the other hand, if a common partial cross-sectional image has not been obtained in advance by the tree recognition unit 104, the cross-sectional data at an even lower height should be extracted by the cross-sectional extraction unit 102 to obtain a common partial cross-sectional image.

[0036] With the above configuration, it is possible to avoid incorrectly estimating the minimum diameter by treating each trunk of a multi-trunk tree as a separate tree. Therefore, even when a tree has multiple trunks, it becomes possible to estimate the minimum diameter of the tree with greater accuracy. Here, "multiple trunks" refers to the presence of multiple trunks of a certain thickness or greater on a single tree. Here, "certain thickness" can be defined as the thickness that distinguishes branches from trunks, and can be set to any arbitrary value. Furthermore, with the above configuration, it is possible to prevent incorrect calculation of the number, density, etc., of trees by treating each multi-trunk tree as a separate tree.

[0037] Furthermore, if the multi-trunk discrimination unit 107 determines that the diameter estimation unit 108 belongs to multiple separate trees, it should proceed as follows: The diameter estimation unit 108 should estimate the minimum diameter of each separate tree based on the tree cross-sectional images that have been determined to belong to multiple separate trees.

[0038] The tree number identification unit 109 identifies the number of trees based on the number of trees whose minimum diameter is calculated by the diameter estimation unit 108. This makes it easy to identify the number of trees within the planned area. Therefore, it becomes easy to meet the need to identify the number of trees within a planned area. Alternatively, the system may also be configured to identify the density of trees within the planned area based on the number of trees relative to the area of ​​the planned area.

[0039] The tree location identification unit 110 uses tree cross-sectional images of different heights of trees whose minimum diameter has been calculated by the diameter estimation unit 108 to estimate the position of the tree's roots and identify its location. For example, the inclination identification unit 105 identifies the inclination of the tree. Next, for example, the position of the tree's roots can be estimated as the point on the line indicated by that inclination passing through the centroid of the tree cross-sectional image of a specified height, at the same height as the target tree's ground level. Then, the position of that point in the geographic coordinate system can be identified as the tree's location. This makes it easy to identify the tree's location in the geographic coordinate system. Therefore, it becomes easy to meet the need to identify the location of a tree in the geographic coordinate system.

[0040] The tree distance identification unit 111 identifies the distance between trees whose minimum diameter has been calculated by the diameter estimation unit 108. For example, the distance between the centroids of the tree cross-sectional images of the trees whose minimum diameter has been calculated can be calculated from the coordinates in the geographic coordinate system and identified as the distance between trees. This makes it easy to identify the distance between trees. Therefore, it becomes easy to meet the need to identify the distance between trees. In addition, the tree distance identification unit 111 may also be configured to calculate the average value of the distances between trees for multiple combinations.

[0041] Furthermore, the tree measuring device 10 may also identify the tree species if it is possible to do so. For example, the tree species may be identified by pattern matching of the 3D image acquired by the 3D image acquisition unit 101.

[0042] <Minimum diameter estimation process using tree measuring device 10> Here, using the flowchart in Figure 14, we will explain an example of the process related to estimating the minimum diameter of a tree using the tree measuring device 10 (hereinafter referred to as the minimum diameter estimation process). The flowchart in Figure 14 can be configured to start each time the 3D image acquisition unit 101 acquires a 3D image.

[0043] First, in step S1, the reference level projection unit 121 recognizes the ground level of the 3D image acquired by the 3D image acquisition unit 101. In step S2, the cross-section extraction unit 102 sets the reference horizontal level.

[0044] In step S3, if the ground level recognized in S1 is above the reference horizontal level set in S2 (YES in S3), the process proceeds to step S4. On the other hand, if the ground level recognized in S1 is not above the reference horizontal level set in S2 (NO in S3), the process proceeds to step S5. In step S4, the reference level projection unit 121 performs horizontal level projection on the 3D image acquired by the 3D image acquisition unit 101.

[0045] In step S5, the cross-section extraction unit 102 extracts cross-sectional data at multiple heights from the 3D image. In S5, if horizontal level projection was performed in S4, cross-sectional data is extracted from the 3D image that was subjected to horizontal level projection. In step S6, the cross-section conversion unit 103 converts the cross-sectional data extracted in S5 into a 2D cross-sectional image viewed from directly above the tree.

[0046] In step S7, the tree recognition unit 104 performs edge detection on the 2D cross-sectional image converted in S6 and identifies the point cloud of edges in the 2D cross-sectional image. Next, the tree recognition unit 104 constructs a rounded contour from the point cloud of adjacent edges. Then, if the size of the constructed round shape is within a specified range, the tree recognition unit 104 recognizes the image of the constructed round shape as a tree cross-sectional image. The flowchart in Figure 14 illustrates the case where the 2D cross-sectional image is recognized as a tree cross-sectional image as an example. If the 2D cross-sectional image is not recognized as a tree cross-sectional image, the minimum diameter estimation process can be terminated.

[0047] In step S8, the tilt identification unit 105 compares the positions of the tree cross-section images of different heights recognized in S7. In step S9, the tilt identification unit 105 identifies the tilt of the tree based on the degree of horizontal displacement between the tree cross-section images of different heights recognized in S7. In steps S8 and S9, if multiple tree cross-section images of the same height are recognized within a predetermined distance, the tilt of the tree is identified for each tree cross-section image based on the degree of horizontal displacement between the tree cross-section images of different heights.

[0048] In step S10, if the inclination identified by the inclination identification unit 105 is greater than or equal to a threshold (YES in S10), the process proceeds to step S11. On the other hand, if the inclination identified by the inclination identification unit 105 is less than a threshold (NO in S10), the process proceeds to step S14. In step S11, the multi-trunk discrimination unit 107 determines whether or not a tree is a multi-trunk tree by comparing the tree cross-section images of different heights recognized by the tree recognition unit 104. As mentioned above, the determination of whether or not a tree is a multi-trunk tree can be made by checking whether the straight lines of the tree's inclination intersect at a height higher than the reference horizontal level for each of the multiple tree cross-section images of the same height. If it is determined to be a multi-trunk tree (YES in S11), the process proceeds to step S12. On the other hand, if it is determined not to be a multi-trunk tree (NO in S11), the process proceeds to step S13.

[0049] In step S12, the diameter estimation unit 108 acquires the aforementioned common partial cross-sectional image and proceeds to step S14. In step S13, the tilt correction unit 106 corrects the tree cross-sectional image at a specified height according to the tilt identified in S9, as described above, and proceeds to step S14. In step S14, the diameter estimation unit 108 estimates the diameter of the largest circle that fits within the tree cross-sectional image as the minimum diameter of the tree, based on the tree cross-sectional image, and terminates the minimum diameter estimation process. In S14, if the common partial cross-sectional image was acquired in S12, the minimum diameter of the tree should be estimated based on the tree cross-sectional image as the common partial cross-sectional image. In S14, if the tree cross-sectional image was corrected in S13, the minimum diameter of the tree should be estimated based on this corrected tree cross-sectional image. Otherwise, the minimum diameter of the tree should be estimated based on the tree cross-sectional image at a specified height recognized in S7.

[0050] This disclosure is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of this disclosure. Furthermore, the control unit and method described in this disclosure may be implemented by a dedicated computer comprising a processor programmed to execute one or more functions embodied by a computer program. Alternatively, the apparatus and method described in this disclosure may be implemented by a dedicated hardware logic circuit. Alternatively, the apparatus and method described in this disclosure may be implemented by one or more dedicated computers comprising a combination of a processor that executes a computer program and one or more hardware logic circuits. Furthermore, the computer program may be stored as instructions executed by the computer on a computer-readable non-transitional tangible recording medium. [Explanation of Symbols]

[0051] 1 Tree measurement system, 10 Tree measurement device, 20 Sensor, 101 3D image acquisition unit, 102 Cross section extraction unit, 103 Cross section conversion unit, 104 Tree recognition unit, 105 Inclination identification unit, 106 Inclination correction unit, 107 Multi-trunk discrimination unit, 108 Diameter estimation unit, 109 Tree number identification unit, 110 Tree position identification unit, 111 Tree distance identification unit, 121 Reference level projection unit

Claims

1. A 3D image acquisition unit (101) acquires a 3D image, which is a 3D point cloud or 3D model obtained by observing an area containing trees from multiple directions with a sensor (20), A cross-sectional extraction unit (102) extracts cross-sectional data, which is data of a cross-section at a specified height from the ground surface, from the three-dimensional image acquired by the three-dimensional image acquisition unit, A cross-sectional conversion unit (103) converts the cross-sectional data extracted by the cross-sectional extraction unit into a two-dimensional cross-sectional image viewed from directly above the tree, A tree recognition unit (104) recognizes a tree cross-sectional image, which is a two-dimensional cross-sectional image of the tree trunk, based on the two-dimensional cross-sectional image converted by the cross-sectional conversion unit, A tree measuring device comprising a diameter estimation unit (108) that estimates the diameter of the largest circle that fits within the tree cross-section image as the minimum diameter of the tree, based on the tree cross-section image recognized by the tree recognition unit.

2. A measuring device for trees according to claim 1, The cross-sectional extraction unit is, The three-dimensional image acquired by the three-dimensional image acquisition unit has a reference level projection unit (121) that projects the highest part of the portion in contact with the ground surface to match a reference horizontal level. A tree measuring device that performs the projection on the reference level projection unit and then extracts the cross-sectional data with the horizontal level as the ground surface.

3. A measuring device for trees according to claim 1, The cross-section extraction unit extracts cross-sectional data obtained by cutting out the three-dimensional image from horizontal planes at multiple heights, which are defined as a specified height and heights different from the specified height, from the ground surface. The section conversion unit converts the height-specific section data extracted by the section extraction unit into height-specific two-dimensional section images. The tree recognition unit recognizes the tree cross-sectional images for each height based on the two-dimensional cross-sectional images for each height converted by the cross-sectional conversion unit. A tilt determination unit (105) determines the tilt of the tree based on the degree of horizontal displacement between the tree cross-sectional images of different heights recognized by the tree recognition unit, The system includes a tilt correction unit (106) that corrects the tree cross-sectional image at the specified height by correcting the tilt of the surface from which the cross-sectional data is extracted at the specified height by the amount of the tilt identified by the tilt identification unit, re-extracting the cross-sectional data, and then extracting the data with the tilt of the cross-sectional data relative to the horizontal plane corrected by the amount of the tilt, thereby correcting the tree cross-sectional image at the specified height. The diameter estimation unit is a tree measuring device that estimates the minimum diameter of the tree based on the tree cross-section image of the specified height corrected by the tilt correction unit, by determining the diameter of the largest circle that fits within the tree cross-section image.

4. A measuring device for trees according to claim 1, The cross-section extraction unit extracts cross-sectional data obtained by cutting out the three-dimensional image from horizontal planes at multiple heights, which are defined as a specified height and heights different from the specified height, from the ground surface. The section conversion unit converts the height-specific section data extracted by the section extraction unit into height-specific two-dimensional section images. The tree recognition unit recognizes the tree cross-sectional images for each height based on the two-dimensional cross-sectional images for each height converted by the cross-sectional conversion unit. The system includes a multi-trunk discrimination unit (107) that compares the tree cross-section images of different heights recognized by the tree recognition unit to determine whether multiple tree cross-section images of the same height belong to a single multi-trunk tree or to multiple separate trees. The diameter estimation unit, when the multi-stem discrimination unit determines that a tree is a single tree with multiple stems, estimates the minimum diameter of the tree based on the tree cross-sectional image of that tree, which is not divided into multiple tree cross-sectional images and is at a height lower than the specified height, by determining the diameter of the largest circle that fits within the tree cross-sectional image.

5. A measuring device for trees according to any one of claims 1 to 4, The cross-section extraction unit extracts cross-sectional data obtained by cutting out the three-dimensional image from horizontal planes at multiple heights, which are defined as a specified height and heights different from the specified height, from the ground surface. The section conversion unit converts the height-specific section data extracted by the section extraction unit into height-specific two-dimensional section images. The tree recognition unit recognizes the tree cross-sectional images for each height based on the two-dimensional cross-sectional images for each height converted by the cross-sectional conversion unit. A tree number identification unit (109) identifies the number of trees from the number of trees whose minimum diameter has been calculated by the diameter estimation unit, A tree measuring device comprising: a tree position identification unit (110) that estimates the position of the base of the tree and identifies the position of the tree using cross-sectional images of the tree at different heights, for which the minimum diameter has been calculated by the diameter estimation unit.

Images