JP2012098247A - Tree position detection device, tree position detection method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problem that it is not easy to accurately recognize positions of individual trees on the basis of tree crown shapes recognized from the sky above a forest.SOLUTION: Positions of trees are detected using three-dimensional point group data of a forest, which is acquired by airborne laser measurement which measures a reflected signal waveform of a laser pulse made to sweep from the sky above. Heights represented by the point group data are converted to substantial heights from the ground surface to generate normalized point group data (S52). A certain height range within the forest is set as a branch height layer on the basis of difference in distribution of the normalized point group data between a tree crown region and a branch height region under it in the forest (S54). Normalized point group data belonging to the branch height layer is extracted and is projected on a plane along the ground surface to obtain a two-dimensional frequency distribution (S56). Positions of high densities of the normalized point group data in the two-dimensional frequency distribution are detected on the basis of prescribed criteria and are defined as tree positions (S58).

Description

  The present invention relates to a tree position detection device, a tree position detection method, and a program for detecting a tree position in a forest.

  Evaluation of forest tree species, tree age, number of standing trees, tree height, etc. can be carried out by investigators visiting the site and making measurements with various measuring instruments and visual observations. However, if the forest to be surveyed is vast, it is not practical to conduct a field survey of the entire survey area because it requires a lot of cost and labor. Therefore, in the past, field surveys were conducted in some areas with typical forest features and topography within the survey area, and the entire survey area was evaluated based on the survey results.

  In recent years, global warming countermeasures are being studied in various countries around the world. Carbon dioxide absorption by forests is considered as a measure to prevent global warming, and it is necessary to ensure the accuracy of the above evaluation for forests in order to calculate the amount of absorption. However, even if, for example, the tree species and the age of the entire survey target area are the same, the growth state of the tree may differ depending on the location condition of the tree. As can be understood from this, it is extremely difficult to accurately evaluate the entire forest under study based on the results of field surveys in some areas.

  In this regard, forests are also evaluated by visually interpreting aerial photographs and analyzing aerial laser measurement data for the entire survey area.

JP 2007-198760 A

  For example, when forest management is not perfect, or when a broadleaf tree has the property of spreading laterally compared to conifers, the crowns of the trees tend to be close to each other and dense, or the crown apex tends to be unclear. In such a case, the shape of the crown of each tree is not clear in the aerial photograph, and visual interpretation of the position and number of the trees becomes difficult. For this reason, the interpretation result largely depends on the technical level of the interpreter, and there is a problem that accuracy and reliability may not be sufficient.

  In this regard, in the numerical surface model (Digital Surface Model: DSM) obtained by aerial laser measurement, the undulations on the surface of the canopy are expressed numerically, so it is possible to objectively grasp the crown shape compared to aerial photographs, When broad-leaved trees are dense as described above, it is not always easy to associate the undulations on the crown surface with the trees.

  The present invention has been made to solve the above problems, and provides a tree position detection device, a tree position detection method, and a program for accurately detecting the position of a tree without depending on a field survey. Objective.

  A tree position detection device according to the present invention is a device that detects the position of a tree using three-dimensional point cloud data of a forest obtained by aerial laser measurement that sweeps a laser pulse from above and measures a reflected signal waveform thereof. A normalization means for generating normalized point cloud data by converting the height represented by the point cloud data into a real height from the ground surface, and a canopy region of the forest and a sub-branch region below the normal region. Based on the difference in the distribution of the normalized point cloud data, the lower branch setting means for setting a constant height range in the forest as the lower branch, and the normalized point cloud data belonging to the lower branch is extracted. Plane projection means for obtaining a two-dimensional frequency distribution by projecting onto a plane along the ground surface, and a tree position by detecting a location where the normalized point cloud data gathers in the two-dimensional frequency distribution based on a predetermined criterion; And position detecting means for

  In the tree position detecting apparatus according to another aspect of the present invention, the lower branch setting means obtains a frequency distribution in the height direction of the normalized point cloud data, and the distribution of the normalized point cloud data in the lower branch region The lower branch layer is set in a range that is formed in the frequency distribution according to the density being lower than that of the canopy region and is lower in frequency than the canopy region side on the ground surface side.

  The tree position detecting apparatus according to another aspect of the present invention further includes a digital elevation model generating unit that generates a digital elevation model based on the point cloud data, and the normalizing unit is a height represented by the point cloud data. The real height is obtained by subtracting the height represented by the digital elevation model from

  The tree position detection device according to another aspect of the present invention further includes a crown image extraction means for extracting a crown image for each tree in an image obtained by photographing the forest from above, and the tree position obtained by the position detection means as the tree position. Collating means for confirming with a tree position of the tree crown image.

  According to a preferred aspect of the present invention, in the tree position detection apparatus according to the present invention, the tree position is obtained for each area that is pre-divided based on forest management information including tree species or tree age for the forest.

  The tree position detection method according to the present invention is a method for detecting the position of a tree using three-dimensional point cloud data of a forest obtained by aerial laser measurement that sweeps a laser pulse from above and measures a reflected signal waveform thereof. A normalization step for generating normalized point cloud data by converting the height represented by the point cloud data into a real height from the ground surface, and a canopy region of the forest and a sub-branch region below the normal region. Based on the difference in the distribution of the normalized point cloud data, a lower branch setting step for setting a constant height range in the forest as a lower branch layer, and extracting the normalized point cloud data belonging to the lower branch layer A plane projection step for obtaining a two-dimensional frequency distribution by projecting onto a plane along the ground surface, and a tree position by detecting a location where the normalized point cloud data gathers in the two-dimensional frequency distribution based on a predetermined criterion; Position detection step Has a flop, the.

  The program according to the present invention is a process for detecting the position of a tree using three-dimensional point cloud data of a forest acquired by an aerial laser measurement that sweeps a laser pulse from the sky and measures the reflected signal waveform in a computer. A normalizing means for generating normalized point cloud data by converting the height represented by the point cloud data into a real height from the ground surface, and a canopy region of the forest Based on the difference in distribution of the normalized point cloud data between the lower branch area and the lower branch area, the branch lower layer setting means for setting a constant height range in the forest as the branch lower layer, the branch belonging to the branch lower layer Normalized point cloud data is extracted and projected onto a plane along the ground surface to obtain a two-dimensional frequency distribution, and based on a predetermined criterion, the normalized point cloud data is extracted from the two-dimensional frequency distribution. Collection Position detecting means for location and detect and trees position that, to function as a.

  According to the present invention, it is possible to detect a tree position with improved accuracy without relying on an on-site survey using data of an aerial laser measurement.

It is a block diagram which shows the schematic structure of the tree position detection apparatus which concerns on embodiment of this invention. It is a general | schematic flowchart of the tree position detection process by the tree position detection apparatus of embodiment of this invention. It is a schematic diagram of the forest explaining a normalization process. It is a schematic diagram of the forest explaining a branch lower layer setting process. It is a schematic diagram explaining the frequency distribution of the height direction of normalized point cloud data. It is a schematic diagram of the two-dimensional frequency distribution obtained by projecting the normalized point cloud data of the lower branch layer on the horizontal plane.

  Hereinafter, a tree position detection apparatus 2 according to an embodiment of the present invention (hereinafter referred to as an embodiment) will be described with reference to the drawings. This device detects the position of the tree using the three-dimensional point cloud data of the forest acquired by the aviation laser measurement.

  The three-dimensional point cloud data is acquired using, for example, a laser measurement device mounted on an aircraft, a helicopter, or the like. The laser measuring device sweeps the laser pulse from the sky, receives the reflection signal, and obtains the distance from the time from pulse emission to reception to the reflection point.

  In the conventional laser measuring apparatus, only a predetermined number (for example, many apparatuses having four points) of high intensity among the peaks appearing in the reflected signal waveform are recorded. However, in recent years, full waveform measurement is possible. Laser measuring devices have been developed. In full waveform measurement, a reflected signal is continuously measured and recorded, whereby a waveform is represented by measurement data. In this apparatus, data by this full waveform measurement is used. For example, by a full waveform measurement, the reflected signal is sampled and recorded at 256 points, and as a result, the three-dimensional point cloud data used in this apparatus is reflected at an arbitrary height in the direction of laser pulse irradiation. Is included regardless of its strength.

  On flat ground, the laser measuring device is usually mounted in a direct view method, and the irradiation of the laser pulse is scanned in a direction perpendicular to the flight direction centering on the direct downward direction of an aircraft or the like. That is, the laser pulse strikes the tree more or less diagonally, except for vertical irradiation. In this regard, when measuring topography, in order to make it easier for the laser pulse to pass through the trees and hit the ground surface in the forest area, it is possible to set the scan angle small and irradiate the laser at an angle close to the ground surface. Is preferred. On the other hand, since this apparatus uses reflection at the trunk to detect the position of the tree, laser measurement data acquired by increasing the scan angle and making the laser pulse easily hit the trunk is more suitable. is there. It is also preferable to adjust the scan width overlap rate so that laser measurement data obtained by oblique irradiation can be obtained at each measurement point. In addition, it is also preferable to use laser measurement data acquired by mounting a laser measurement device in the oblique viewing method adopted when measuring a slope.

  While this apparatus can detect a tree position with good accuracy only by laser measurement data, it can also improve accuracy and reliability by using aerial photograph information together. Therefore, for example, an aerial photograph of the survey area is acquired in addition to the laser measurement data of the survey target area by installing a digital camera on the aircraft that performs laser measurement and taking a photograph at the same time as the laser measurement. To do.

  FIG. 1 is a block diagram illustrating a schematic configuration of the tree position detection apparatus 2. The system includes an arithmetic processing device 4, a storage device 6, an input device 8, and an output device 10. As the arithmetic processing device 4, it is possible to make dedicated hardware for performing various arithmetic processing of this system, but in this embodiment, the arithmetic processing device 4 uses a computer and a program executed on the computer. Built.

  A CPU (Central Processing Unit) of the computer constitutes an arithmetic processing unit 4, which will be described later, such as DTM generation means 20, normalization means 22, branch layer setting means 24, plane projection means 26, position detection means 28, and canopy image extraction means. 30 and the matching means 32.

  The storage device 6 is composed of a hard disk or the like built in the computer. The storage device 6 is a program for causing the arithmetic processing unit 4 to function as the DTM generation unit 20, the normalization unit 22, the lower layer setting unit 24, the plane projection unit 26, the position detection unit 28, the tree crown image extraction unit 30, and the matching unit 32. And other programs and various data necessary for processing of the system. For example, the laser measurement data acquired by the above-described laser measurement device for the forest that is the survey target area is the three-dimensional point cloud data 40 represented on the ground, for example, in an orthogonal coordinate system, and the processing target data is stored in the storage device 6. Is stored as The aerial photograph image data 42 taken by the digital camera is also stored in the storage device 6 as processing target data. The storage device 6 is also used to hold intermediate data in each process, and stores, for example, normalized point cloud data 44 described later.

  The input device 8 is a keyboard, a mouse, or the like, and is used for a user to operate the system.

  The output device 10 is a display, a printer, or the like, and is used for showing the tree position obtained by this system to the user by screen display, printing, or the like.

  FIG. 2 is a schematic flowchart of tree position detection processing by the tree position detection device 2. Each means of the arithmetic processing unit 4 will be described with reference to FIG.

  The DTM generation means 20 generates a digital elevation model (Digital Terrain Model: DTM) based on the point cloud data 40 (DTM generation processing S50). It can be expected that the last pulse obtained at the end of the reflection signal for each laser pulse emitted to the forest is due to reflection on the ground surface. Therefore, the DTM generating unit 20 generates a DTM based on the point corresponding to the last pulse in the point cloud data 40. At that time, filtering processing for removing features such as trees and buildings may be further performed on the three-dimensional shape represented by the last pulse. In the point cloud data 40 obtained by the full waveform measurement, more reflection points on the ground surface can be obtained than in the conventional measurement, so that the DTM can be obtained with high accuracy. By improving the accuracy of the ground surface elevation, the height of the root of the tree can be determined with high accuracy, and as a result, there is an advantage that a lower branch layer described later can be obtained with high accuracy. When a DTM obtained in advance by other means is used in the next normalization process, the process by the DTM generating means 20 can be omitted.

  The normalizing means 22 generates the normalized point cloud data 44 by converting the height of the point cloud data 40 represented by the Z value of the orthogonal coordinate system XYZ into a substantial height from the ground surface (normalization processing S52). . Specifically, the normalized point cloud data 44 is obtained by replacing the height (elevation) of the point cloud data 40 with the actual height obtained by subtracting the height (elevation) represented by the DTM from the height. Generated.

  FIG. 3 is a schematic diagram of the forest for explaining the normalization process S52. The horizontal direction corresponds to the horizontal direction (X and Y axis directions), and the vertical direction corresponds to the vertical direction (Z axis direction). 3A shows a state before the normalization process S52, and FIG. 3B shows a state after the normalization process S52.

  The forest tree 70 shown in FIG. In the drawing, the positions of the reflection points 74 constituting the point cloud data 40 are illustrated by “x” marks. Since the height of the reflection point 74 of the tree 70 is represented by an altitude in the point cloud data 40, the value is the altitude of the ground surface 72 where the tree 70 stands, and the actual height from the root of the tree 70 to the reflection point 74. Will be added. That is, as shown in FIG. 3A, when the ground surface 72 has undulations, the height range of the trunk portion 78 under the crown portion 76 where branches and leaves grow in each tree 70 varies depending on the altitude of the ground surface 72. To do.

  The normalization process S52 subtracts the DTM value representing the altitude of the ground surface 72 from the height of the reflection point 74 of the tree 70, so that the root of the trunk portion 78 has a height of 0 as shown in FIG. Thus, the extraction process of the trunk portion 78 in the process described later is facilitated.

  Based on the difference in the distribution of the normalized point cloud data 44 between the canopy region of the forest to be investigated and the sub-branch region below it, the branch lower layer setting means 24 calculates a constant height range within the forest. (Branch lower layer setting process S54).

  FIG. 4 is a schematic diagram of the forest for explaining the branch lower layer setting process S54, and corresponds to the state after the normalization process S52 shown in FIG. FIG. 4 shows a canopy region 80 made up of a crown portion 76 of each tree 70 and a sub-branch region 82 made up of a trunk portion 78 having no branches and leaves below. The lower end of the lower branch region 82 coincides with the ground surface 72b which is a plane, while the upper end may be different for each tree 70.

  The difference in distribution of the normalized point cloud data 44 between the tree crown region 80 and the under-branch region 82 appears in the frequency distribution in the height direction of the normalized point cloud data 44. FIG. 5 is a schematic diagram for explaining the frequency distribution. The left side of FIG. 5 shows the schematic diagram of the forest shown in FIG. 4, and the right side shows the frequency distribution 90 in the height direction of the forest. The frequency distribution 90 is displayed with the Z axis in common with the Z axis of the schematic diagram of the left forest. In the crown portion 76 of the tree 70, the branches and leaves that extend over the trunk extend in the horizontal direction from the trunk, and the branches and leaves can serve as the reflection points 74 of the laser pulse, so that basically there are more reflection points 74 than the trunk portion 78. Distribution density increases. On the other hand, since the trunk portion 78 is basically thinner than the crown portion 76, the laser pulse is less likely to hit as compared with the crown portion 76, and the number of reflection points 74 is small, so the distribution density is low. As a result, in the frequency distribution 90, the frequency in the height range Ra corresponding to the canopy region 80 is generally larger than the height range Rb corresponding to the under-branch region 82. The frequency distribution 90 has a height range Rb from a state having a relatively high average level in the height range Ra in the height range Rm in which the crown region 80 and the under-branch region 82 coexist in the height range Ra. Transition to a low frequency distribution at. That is, the frequency distribution 90 has a height range Rt whose frequency is lower on the ground surface 72b side than the height range Ra on the tree crown region 80 side.

  As shown in FIG. 5, the height range Rt can be extracted as a range lower than the threshold Th by setting a predetermined threshold Th lower than the average level A in the height range Ra of the frequency distribution 90, for example. . Here, the threshold Th is set to be larger than the frequency of the reflection points 74 in the height range Rb. However, since the frequency in the height range Rb changes according to the number of trees 70 in the forest, the threshold Th is also a fixed value. Instead, it is preferable to change the number according to the number of trees 70. For example, since the average value A of the frequency in the height range Ra also changes according to the number of trees 70, Th = αA can be set using a proportionality coefficient of 0 <α <1. α is set in advance by actually measuring several trees.

  The lower branch layer 100 is set in the height range Rt thus extracted. As shown in FIG. 5, the lower branch layer 100 is a flat region in the horizontal direction. The height range of the branch lower layer 100 can be set to coincide with the height range Rt, or can be set to occupy only a part of the height range Rt. For example, the region on the upper limit side of the height range Rt is located in the height range Rm of the coexistence region of the canopy region 80 and the under-branch region 82 and may partially include the reflection point 74 of the canopy region 80. Since the reflection point 74 of the tree crown region 80 can lower the accuracy of position detection of the trunk portion 78 described later, when the accuracy of position detection is desired to be more suitable, the height upper limit of the branch lower layer 100 is set to the height range. Set lower than the upper limit of Rt. Further, there may be a reflection point 74 caused by grass or the like near the ground surface 72b, and such a reflection point 74 can also reduce the accuracy of detecting the position of the trunk portion 78, so that the height lower limit of the branch lower layer 100 is increased. It may be set higher than the ground surface 72b which is the lower limit of the range Rt.

  In this way, the reflection points 74 of the crown portion 76 occupying a large proportion in the normalized point cloud data 44 are basically removed from the lower branch layer 100, and the reflection points 74 included in the lower branch layer 100 are those of the trunk portion 78. Become dominant.

  The plane projection unit 26 extracts the normalized point cloud data belonging to the lower branch layer 100 and projects it onto the plane along the ground surface 72b to obtain a two-dimensional frequency distribution (plane projection process S56). FIG. 6 is a schematic diagram of the two-dimensional frequency distribution 110, and the position of the projection point 112 of the normalized point cloud data 44 in the branch layer 100 on the XY plane which is a horizontal plane is indicated by “◯”. Most of the normalized point cloud data 44 is the reflection points 74 of the tree crown portion 76, and the reflection points 74 of the trunk portion 78 are few. Therefore, the reflection points 74 of the trunk portion 78 in the lower branch layer 100 are sparsely distributed three-dimensionally, and it is generally not easy to specify the position of the trunk portion 78 from the distribution. In this respect, since the trunk portion 78 has a feature extending roughly in the height direction, in the two-dimensional frequency distribution 110 obtained by projection in the height direction, the plurality of reflection points 74 of each trunk portion 78 are close to each other. Projected to position. That is, in the two-dimensional frequency distribution 110, the projection points 112 of the respective trunk portions 78 gather together at the position of the trunk to form a group 114, whereby the position of the trunk portion 78 can be easily specified.

  Based on a predetermined criterion, the position detection unit 28 detects a place where the normalized point cloud data 44 gathers and forms the group 114 in the two-dimensional frequency distribution 110 to obtain a tree position (position detection process S58). When the group 114 is composed of a large number of projection points 112 and a peak is formed in the two-dimensional frequency distribution 110, for example, the group is located at a place where a frequency equal to or higher than the threshold is obtained from the two-dimensional frequency distribution 110 by threshold determination. The presence of 114 is detected. Further, the peak position may be obtained by a peak detection method other than the threshold determination.

  When the number of projection points 112 of the trunk portion 78 is small, the value of the two-dimensional frequency distribution 110 is either “0” or “1” even in the region of the group 114, and the peak of the frequency distribution may not be formed. is assumed. In such a case, for example, filtering is performed to scan and smooth the two-dimensional frequency distribution 110 obtained by simple projection with a window having a thickness of a trunk, and a peak is formed at a place where the projection points 112 gather. What should I do?

  For example, the position detection unit 28 outputs the peak position as a tree position (Xd, Yd). Alternatively, the position of the center of gravity of the projection point 112 may be obtained for each group 114 to obtain the tree position (Xd, Yd). The tree position detection device 2 can output the tree position (Xd, Yd) detected by the position detection processing S58 to the output device 10, for example.

  Further, the tree position detection apparatus 2 can further improve the accuracy and reliability of the tree position using the aerial photograph image data 42. In this case, the position (Xd, Yd) obtained in the position detection process S58 is a tree position candidate.

  The crown image extraction means 30 analyzes the image data 42 and extracts a crown image for each tree (tree crown image extraction processing S60). For example, as shown in Japanese Patent Application Laid-Open No. 2003-344048, the tree crown image extraction means 30 extracts the tree crown shape from the image data 42 based on the watershed algorithm. Further, the tree image extracting means 30 obtains, for example, a point (Xh, Yh) corresponding to the center of the tree crown shape as a tree position. Note that the point (Xh, Yh) obtained by the operator reading the aerial photograph, extracting the tree crown image, and obtaining the tree position may be stored in the storage device 6 and used in the next matching unit 32.

  The collation means 32 corrects the tree position (Xd, Yd) by collation with the tree position (Xh, Yh) obtained from the tree crown image (collation processing S62). The collation process S62 checks whether or not a tree position (Xh, Yh) corresponding to an arbitrary candidate position (Xd, Yd) exists, and if it exists, the collation is established, and the position (Xd, Yd) is set as a tree position. If it does not exist, the candidate position (Xd, Yd) is determined not to be a tree position because the verification is not established. That is, the collation means 32 performs a correction to remove those that have not been collated from the group of tree positions obtained in the position detection process S58.

  Specifically, if the allowable distance of misalignment in collation is β and there is a position (Xh, Yh) of the crown image that satisfies both the following expressions (1) and (2), it is determined that the collation is established. For example, β can be set to be equal to or less than the radius of a circle having a size corresponding to the crown shape.

Xd−β ≦ Xh ≦ Xd + β (1)
Yd−β ≦ Yh ≦ Yd + β (2)

  The tree position detection device 2 can output, for example, the tree position (Xd, Yd) where the verification is established to the output device 10 when the verification processing S62 is performed.

  In the above, for the sake of simplicity, an example has been described in which forests in the survey area are handled in a lump. However, for example, when exploring a vast forest, there are differences in tree species, age, topography, etc. depending on the location, so the forest in the survey area is divided into multiple areas and It is preferable to execute all or part of the process. For example, the height of the lower branch region 82 varies depending on the tree type and age, and the topography and other growth environments also affect the height of the lower branch region 82. Therefore, if an area is set in a range in which those factors are common or not greatly different, the reflection points 74 of the crown portion 76 are preferably excluded, and the branch points such that many reflection points 74 of the trunk portion 78 are included. Since the lower layer 100 can be set, the position of the tree can appear more clearly in the two-dimensional frequency distribution 110.

  Most of the forest is managed and maintained by humans, and the tree species and age are almost the same for each place. The situation is often grasped as forest management information. Therefore, in this case, the area can be set based on the forest management information. Based on the topographic map, the area can be set in consideration of the topography. Moreover, even if the entire survey area is simply divided into meshes of a predetermined size, the lower branch layer 100 is suitably adjusted for each location, and an effect of improving the detection accuracy of the tree position can be expected.

  In another configuration for setting the branch lower layer, the branch lower layer 100 is temporarily set at the height on the ground surface 72 side, and the two-dimensional frequency distribution 110 for the temporarily set branch lower layer 100 is obtained. In the desirable lower branch layer 100, the ratio of the reflection points 74 of the trunk portion 78 is large. As a result, the projection points 112 are concentrated at positions corresponding to the trunks as shown in FIG. do not do. On the other hand, if the provisional lower branch layer 100 includes a reflection point 74 such as a tree portion 76 or grass near the ground surface 72b, the projection points 112 are distributed in addition to the position of the trunk, and the degree of concentration is low. Such distribution differences can be quantified and compared using the dispersion index in the same manner as the distribution pattern in biology. Therefore, the branch lower layer setting means 24 obtains the dispersion index in the two-dimensional frequency distribution 110 by changing the upper limit height and the lower limit height of the temporarily set branch lower layer 100, and the height range in which the projection points 112 are preferably concentrated. Can be determined as the lower branch layer 100.

  The present invention has been described above with a focus on the detection of the tree position in the forest. However, if the tree position is obtained, the number of trees can be easily counted by the tree position detection device 2, and the tree The density can also be determined.

  2 Tree position detection device, 4 arithmetic processing device, 6 storage device, 8 input device, 10 output device, 20 DTM generation means, 22 normalization means, 24 branch lower layer setting means, 26 plane projection means, 28 position detection means, 30 Tree crown image extraction means, 32 collation means, 40 point cloud data, 42 image data, 44 normalized point cloud data, 70 tree, 72 ground surface, 74 reflection point, 76 tree crown part, 78 tree trunk part, 80 tree crown area, 82 under branch Area, 90 frequency distribution, 100 lower branch, 110 two-dimensional frequency distribution, 112 projection points, 114 groups.

Claims (7)

A tree position detection device that detects the position of a tree using three-dimensional point cloud data of a forest acquired by aerial laser measurement that sweeps a laser pulse from the sky and measures a reflected signal waveform thereof,
Normalizing means for generating normalized point cloud data by converting the height represented by the point cloud data into a substantial height from the ground surface;
Based on the difference in the distribution of the normalized point cloud data in the forest crown area and the lower branch area below the forest, a lower branch setting means for setting a constant height range in the forest as a lower branch,
A plane projection means for extracting the normalized point cloud data belonging to the lower branch layer and projecting it on a plane along the ground surface to obtain a two-dimensional frequency distribution;
Based on a predetermined criterion, a position detection means for detecting a location where the normalized point cloud data gathers in the two-dimensional frequency distribution and setting it as a tree position;
A tree position detecting device characterized by comprising: In the tree position detection apparatus according to claim 1,
The lower branch setting means obtains a frequency distribution in the height direction of the normalized point cloud data, and according to the distribution density of the normalized point cloud data in the lower branch region being lower than the crown region A tree position detection apparatus, characterized in that the lower branch layer is set in a range formed in the frequency distribution and having a lower frequency on the ground surface side than on the crown area side. In the tree position detection apparatus according to claim 1 or 2,
A digital elevation model generating means for generating a digital elevation model based on the point cloud data;
The normalization means obtains the substantial height by subtracting the height represented by the digital elevation model from the height represented by the point cloud data;
Tree position detecting device characterized by the above. In the tree position detection apparatus according to claim 1,
A crown image extraction means for extracting a crown image for each tree in an image obtained by photographing the forest from above;
Collating means for determining the tree position obtained by the position detecting means by collating with the tree position of the crown image;
A tree position detecting device characterized by comprising: In the tree position detection device according to claim 1,
A tree position detection apparatus, wherein the tree position is obtained for each area that has been divided in advance based on forest management information including tree species or tree age for the forest. A method of detecting the position of a tree using three-dimensional point cloud data of a forest obtained by aerial laser measurement that sweeps a laser pulse from above and measures its reflected signal waveform,
A normalization step of generating normalized point cloud data by converting the height represented by the point cloud data into a substantial height from the ground surface;
Based on the difference in distribution of the normalized point cloud data between the forest crown area and the lower branch area below the branch area, a lower branch setting step for setting a constant height range in the forest as a lower branch,
A plane projection step of extracting the normalized point cloud data belonging to the lower branch layer and projecting it on a plane along the ground surface to obtain a two-dimensional frequency distribution;
Based on a predetermined criterion, a position detection step of detecting a location where the normalized point cloud data gathers in the two-dimensional frequency distribution and setting it as a tree position;
A tree position detecting method characterized by comprising: A program for causing a computer to detect the position of a tree using three-dimensional point cloud data of a forest acquired by aerial laser measurement that sweeps a laser pulse from the sky and measures the reflected signal waveform And the computer
Normalizing means for generating normalized point cloud data by converting the height represented by the point cloud data into a substantial height from the ground surface,
Based on the difference in the distribution of the normalized point cloud data between the canopy region of the forest and the sub-branch region below the branch region, a branch layer setting means for setting a constant height range in the forest as a branch layer,
A plane projection means for extracting the normalized point cloud data belonging to the lower branch layer and projecting it on a plane along the ground surface to obtain a two-dimensional frequency distribution; and
A program that functions as a position detection unit that detects a location where the normalized point cloud data gathers in the two-dimensional frequency distribution based on a predetermined criterion and sets it as a tree position.

Images