JP6044052B2 - Fisheye image data creation program and LAI calculation program - Google Patents

Description

  The present invention relates to a fisheye image data creation program, an LAI (Leaf Area Index) calculation program, and an LAI calculation apparatus.

  As a standard for evaluating the forest environment, a leaf area index (hereinafter referred to as “LAI”) is widely known. LAI refers to the area of leaves per unit area, and can be used as an index for measuring the amount of light in the forest. In other words, if LAI can be calculated at any location in the forest, the amount of light that can be increased by reducing the amount of leaves at those locations, etc. The amount of felling can be performed quantitatively.

  In general, the LAI is obtained by calculating the amount of light incident on the forest from an image taken from directly above the ground in the forest using a fisheye camera. Specifically, the image taken by the fisheye camera is binarized, and the amount of light on the image is obtained based on the number of shadow pixels. Then, assuming that the amount of light in the forest that reaches the forest floor is a Poisson distribution, an LAI is obtained using an estimation formula (for example, see Non-Patent Document 1 below).

  As for conventional leaf area calculation methods, for example, the following patent document 1 discloses an aircraft laser data analysis technique, the following patent document 2 uses a wide-angle lens and an electronic imaging device, and the following patent document 3 describes an optical technique. A method for measuring with an artificial vegetation index sensor is disclosed.

  Non-Patent Document 2 below discloses a study of calculating LAI using a ground laser.

JP 2008-1111725 A JP 2007-171033 A JP 2011-133451 A

MacAuthhur & Horn, Ecology, 50 (5), 1969. Seidel et al. , 2012, Agricultural and Forest Metalogy, 154-155, pp. 1-8

  The analysis technique described in Patent Literature 1 uses aircraft laser data (laser data emitted from the sky) and uses a method of calculating the leaf area from the “reflection intensity” of the laser. However, even if data is acquired at high density with an aircraft laser, the laser transmittance into the forest is low, and there is no guarantee that the exact situation inside the forest can be grasped. In addition, since the “reflection intensity” varies depending on the irradiation angle, the reflection intensity varies depending on the position of the aircraft when the laser is irradiated and the position of the target forest, and there is a limit to estimation based only on the reflection intensity.

  In addition, the technique described in Patent Document 2 described above is an “electronic image sensor” in order to overcome the problem that the weather when photographing with a fish-eye camera or the like greatly affects the result of estimating the LAI in the conventional method. Is to introduce. However, the technique described in Patent Document 2 has a problem that an “electronic image sensor” must be newly added as a device configuration. In addition, since actual shooting is performed using a fisheye camera or the like, there is a problem in that subjective judgment is added in the binarization process depending on the weather and time, resulting in a large difference.

  The technique described in Patent Document 3 is an original development of a “sensor for calculating an optical vegetation index” by further advancing the technique described in Patent Document 2. However, like the above-mentioned Patent Document 2, there is a problem that the unique sensor must be used. In addition, since actual shooting is performed using a fisheye camera or the like, there is a problem in that subjective judgment is added in the binarization process depending on the weather and time, resulting in a large difference.

  Non-Patent Document 2 is certainly a technique for calculating LAI using a terrestrial laser. However, since LAI is calculated after converting all point clouds to Voxel for calculation, one image is obtained. Since it takes about 4 hours to obtain the temperature, it is very slow and cannot be said to be a simple method. Further, the details of the image creation procedure are hardly clarified, and it cannot be said that it has been disclosed to the extent that it can be carried out.

  Therefore, in view of the above problems, the present invention provides a fisheye image data creation method and a fisheye image data creation program that are simpler, faster, more precise and less affected by the weather, and based on this, more easily, faster, and more accurate. An object of the present invention is to provide an LAI calculation method and a LAI calculation program.

  A fisheye image data creation program according to an aspect of the present invention for solving the above-described problem is a step of reading three-dimensional forest point cloud data into a computer, a step of creating three-dimensional reference point data, and reference point data Obtaining at least two declination data of the forest point data constituting the forest point cloud data with reference to, and creating two-dimensional fisheye image data based on the two declination data of the forest point data , Execute.

  A fisheye image data creation program according to another aspect of the present invention includes a step of reading a three-dimensional forest point cloud data into a computer, a step of creating three-dimensional reference point data, and a reference point data as a reference. Obtaining at least any two declination data of the forest point data constituting the forest point cloud data, creating two-dimensional fisheye image data based on the two declination data of the forest point data, fisheye A step of obtaining an LAI based on the image data is executed.

  As described above, the present invention provides a fish-eye image data creation method and a fish-eye image data creation program that are simpler, faster, more accurate and less affected by the weather, and based on this, a more simple, faster, and more accurate LAI calculation method and An LAI calculation program can be provided.

It is a figure which shows the outline of the step in the LAI calculation method which concerns on embodiment. It is an image figure of the acquisition method of forest point cloud data. It is an image figure of the acquired forest point cloud data. It is an image figure of the virtual space used when creating reference point data. It is an image figure of calculation at the time of calculating | requiring declination data. It is a figure which shows the image of a projective transformation process. It is a figure which shows the image which arrange | positions all the forest point clouds on the surface of a sphere. It is explanatory drawing in the case of determining the area | region where forest point cloud data exist, and the area | region where the forest point cloud is not arrange | positioned. It is a figure which shows the image of the produced fisheye image data. It is an image figure of the process which extracts only the circular area | region of the predetermined range from the center, and makes it the calculation object of LAI. It is a figure which shows the result of having compared LAI by AccuPAR and LAI by a fisheye lens. It is a figure which shows the result of having compared LAI calculated | required using the forest point cloud data by a laser sensor, and LAI by AccuPAR. It is a figure which shows the relationship between LAI defined using the fisheye lens, and LAI defined using the forest point cloud data calculated | required with the laser sensor.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. However, the present invention can be implemented in many different forms, and is not limited to specific examples of the embodiments shown below.

  The present embodiment relates to an LAI calculation method. The LAI calculation method according to the present embodiment (hereinafter referred to as “the present method”) includes (1) a step of reading three-dimensional forest point cloud data, (2) a step of generating three-dimensional reference point data, and (3) a reference. Obtaining at least any two declination data of the forest point data constituting the forest point cloud data on the basis of the point data; (4) two-dimensional fisheye based on the two declination data of the forest point data; Creating image data; and (5) obtaining LAI based on the fisheye image data. FIG. 1 shows an outline of the steps according to the present method.

  Although this method is not limited, it is preferable because it is easy to realize by recording a program capable of executing the above steps on a recording medium such as a hard disk of a computer and executing the program. The following description is based on data processing by a computer.

  First, the step (1) of reading the three-dimensional forest point cloud data of this method is literally the step of reading the three-dimensional forest point cloud data. This step is a typical mode in which, for example, recording is performed in advance on a recording medium such as a hard disk of a computer, and three-dimensional forest point cloud data is read into a recording medium such as a memory as necessary. By doing so, the processing described later can be started. Here, “three-dimensional forest point cloud data” refers to point data (forest points) including coordinate information representing the presence of trees, etc. in an area where trees such as forests or forests exist (hereinafter referred to as “forests”). Data). That is, it means that there are tree trunks, leaves, and the like at positions where forest point data exists. The three-dimensional forest point cloud data is not limited as long as it can be acquired, but it is preferable to acquire the three-dimensional forest point cloud data by a more detailed method. Can be obtained by placing and operating a device capable of doing so. An image diagram of the forest point cloud data acquisition method in this case is shown in FIG. 2, and an image diagram of the acquired forest point cloud data is shown in FIG. Here, the three-dimensional forest point cloud data is not limited, but is preferably binary data including information on the x coordinate, the y coordinate, and the z coordinate. By doing in this way, a simple process can be performed in a later step.

  The step (2) of creating three-dimensional reference point data in this method is a step of determining a reference point that is the center when creating fisheye image data. That is, the reference point data is a point that becomes the center of fish-eye image data created later.

  Here, the reference point data is preferably expressed in the same coordinate system as the three-dimensional forest point cloud data, and includes, but is not limited to, information on the x coordinate, y coordinate, and z coordinate. It is preferably binary data.

  Also, in this step, when considering the height from the ground surface when calculating the LAI, a new DTM (Digital Terrain Model) has been added to specify the height of the reference point from the ground surface. It is preferable to create.

The point cloud data is not classified in advance, such as reflection from the ground surface and reflection from other points. Therefore, the ground surface DTM is created. DTM is ground surface data created from forest point cloud data, and is processed to be smooth. By obtaining this data, the height of the reference point from the ground surface can be accurately determined. This DTM divides the xy coordinate plane in the forest point cloud data into a plurality of areas, finds the minimum value in each area (cell), compares it with the cells adjacent to each area, and only those that are the minimum value Create the remaining surface.
On the created surface, if there is a lower cell value in the adjacent cell centered around each cell (eg 3x3 or 5x5 range), substitute the lower adjacent cell value into the central cell (extract the minimum value) Image filter). This process is performed throughout. By comparing the DTM that has been subjected to the minimum value filter and the original DTM, only the cells that do not vary (the minimum value is not substituted from the adjacent cells) are determined. A TIN (Triangular Irregular Network) is created only from a place where there is no change, and the TIN is converted to a raster. Similar processing is applied to the DSM. In DSM, the maximum value filter for comparing the value with the adjacent cell and substituting the maximum value into the central cell is applied. By comparing the maximum value filter and the original DSM, only cells that do not change are determined. The determined cell represents the top of the tree. The tree vertices and positions obtained from the DSM are excellent information for determining the reference point because they are excellent in simply calculating the density of standing trees.

  Although not limited, when creating the reference point data, that is, when determining the reference point, for example, the virtual space is displayed on the display device connected to the information processing device, and the user can select the desired virtual space. The reference point data can be created by selecting the position and causing the position to be used as a reference point. In this way, if the forest point cloud data is acquired, fisheye image data at an arbitrary position can be created, and the LAI on the spot can be obtained. An image diagram in this case is shown in FIG.

  Further, the step (3) of obtaining at least any two declination data of the forest point data constituting the forest point cloud data on the basis of the reference point data in the present method is an important step in the present method. This process makes it easy to create a two-dimensional fisheye image based on a three-dimensional forest point cloud. This process can also be said to be a coordinate conversion process for converting to a polar coordinate system, more specifically, a spherical coordinate system.

  The two declination data obtained in this step is not limited, but the projection line to the xy plane of the line connecting the reference point and the forest point to be processed is the x axis (or y axis) on the xy plane. ) And horizontal angle data corresponding to the horizontal angle θ, and elevation data corresponding to an elevation angle φ that is an angle between the projection line and the z axis (or xy plane). By doing in this way, it becomes an expression in polar coordinates, more specifically in spherical coordinates, and processing corresponding to a circle which is the shape of a fisheye image becomes easy.

  In this step, the two declination angles may be obtained by any method, but it is preferable to obtain the distance D between the reference point and the forest store when obtaining the two declination angles. By obtaining the distance D, it is possible to easily obtain the declination and to accurately represent spherical coordinates. An image of calculation in this case is shown in FIG.

  In this step, after obtaining the two declination data by the above process, (4) two-dimensional fisheye image data is created based on the two declination data of the forest point data. In this step, first, projective transformation processing is performed to create two-dimensional projected image data. Specifically, projective transformation processing is performed on each forest point data of the forest point cloud data using two declination data. This image diagram is shown in FIG. In this figure, orthographic projection conversion processing is shown from the viewpoint of explanation, but as other projection transformation, equisolid angle projection and equidistant projection may be used.

  In addition, in this projective transformation process, or before that, it is preferable to process the distance D between the reference point and each forest point data so as to have the same value (for example, 1). By doing so, it becomes possible to project all the forest point groups to be processed within a range of a predetermined radius (for example, 1), which is different from a fish-eye image actually using a lens and becomes projection image data. . When the distance D from the reference point is set to the same value (for example, 1) before the projective transformation process, the image should be regarded as the process of arranging all the forest point groups on the surface of the sphere having the value. Can do. An image diagram in this case is shown in FIG.

  In this step, the projected image data can be used as fisheye image data as it is, but it is preferable to further adjust the roughness of the projected image data. Specifically, the projected image is cut into a plurality of regions, and when the projected forest point cloud is arranged in each region, the region where the forest point cloud data exists, the forest point cloud is placed. If not, the region is determined to be a region where no forest point cloud data exists. FIG. 8 shows an image in this case, and FIG. 9 shows an image of the created fisheye image data. As a result, fisheye image data can be created. When this processing is performed, it is possible to freely obtain an LAI having a desired accuracy, and it is possible to achieve a balance between the reduction of processing time and accuracy. Note that the fisheye image data is preferably one in which each pixel data is one of white and black binary values of 0 and 1, from the viewpoint of more easily obtaining the LAI, but the distance is more For the sake of clarity, it is also preferable to perform weighting according to distance, for example, before calculating LAI (for example, forests that are darker and farther from the reference point are darker when representing forest points that are closer to the reference point). When expressing a point, it is also preferable to make it brighter).

  In addition, according to the above-described roughness adjustment process, it is possible to remove errors that occur when acquiring forest point cloud data. This point will be specifically described. Forest point cloud data acquisition is preferably acquired by a three-dimensional position measurement device using a laser sensor or the like. In this case, the device transmits laser light at a certain angular interval ( For example, forest point cloud data is acquired while rotating at α). That is, when r is the distance from the laser light source to the measurement point, the value obtained by multiplying the distance r and the rotation angle interval α is the limit of resolution. The roughness needs to be coarser than this resolution limit. More specifically, the roughness of the projection image data needs to be larger than the two-dimensional projection component at the rotation angle interval α, and more preferably at least twice the two-dimensional projection component at the rotation angle interval α. It is. By doing so, errors can be reduced.

  As described above, two-dimensional fisheye image data can be obtained by this step.

  In this step, the process for determining the presence / absence of forest points may be limited to a necessary range. Specifically, at least one of a predetermined distance and a predetermined angle may be determined, and only the forest point cloud data within the predetermined range may be processed. In this way, it is possible not only to avoid unnecessary processing of the forest point cloud data, but also by matching the LAI value measured by the device with the range that other devices can measure. It is possible to compare the LAI calculated in accordance with the same range.

  As a specific example of the above processing, for example, when the distance is within a predetermined distance range, distance range data including information on the predetermined distance is created, and point data that is equal to or greater than the predetermined distance from the reference point It is preferable that the processing is excluded from the processing target.

  If the angle is within a predetermined angle range, angle range data including information on the predetermined angle is created, and point data whose elevation angle is greater than or equal to the predetermined angle from the reference point is excluded from the processing target. It is preferable.

  The method further includes the step of (5) obtaining LAI based on the fisheye image data. The fisheye image data created in the above steps is two-dimensional circular image data, and the forest point data in this circle is binarized to calculate the area occupied by the forest points in the circle. To calculate the LAI. The LAI calculation method can be realized by a known method and is not limited. For example, the LAI calculation method can be performed based on Non-Patent Document 1.

  In this step, it is also preferable to perform a process for making only a predetermined region of the fisheye image data the LAI calculation target. More specifically, it is preferable to extract only a circular region within a predetermined range from the center and make it an LAI calculation target. In this way, it is possible to calculate the LAI in a region of a predetermined elevation angle, and to compare the results with other LAI measurement devices having a limitation on the elevation angle. An image diagram in this case is shown in FIG.

  As described above, this method provides a fisheye image data creation method and a fisheye image data creation program that are simpler, faster, more accurate and less affected by the weather. It is possible to provide a difficult LAI calculation method and LAI calculation program.

  A demonstration experiment was conducted to verify the above method. Specifically, at Hagley Park in Christchurch, New Zealand, which is composed of various tree species, a fisheye image is actually taken using a fisheye lens, the fisheye image is read and an LAI is obtained, while a laser sensor device Forest point cloud data was acquired using, two-dimensional fisheye image data was created according to the method described in the above embodiment, and LAI was obtained.

  In addition, as the data for field verification, using AccuPAR (manufactured by Dragon Device), the data was acquired simultaneously at the two locations of the same measured location and the opened location, the light intensity corrected for the weather on the measurement date was acquired, and the LAI Calculated.

  First, FIG. 11 shows the result of comparing LAI with AccuPAR and LAI with fisheye lens. According to this result, when LAI was obtained using a fisheye lens, although the correlation coefficient R = 0.79, the LAI range was narrow and there was a tendency to underestimate the AccuPAR LAI evaluation.

  On the other hand, when the LAI obtained by using the forest point cloud data by the laser sensor according to the present embodiment and the LAI obtained by AccuPAR are compared, the LAI obtained by this method has a wide range and can capture various situations of the LAI. Confirmed that it was possible. It was also confirmed that the correlation coefficient R was excellent at 0.81. The result is shown in FIG.

  FIG. 13 shows the relationship between the LAI determined using the fisheye lens and the LAI determined using the forest point cloud data obtained by the laser sensor. As a result, it was confirmed that although the value tends to saturate in the evaluation using the fisheye lens in the place where the LAI is high, the calculation can be performed accurately even in the place where the LAI is high. This is thought to be due to factors such as the subjectivity of the measurer being included in the binarization criteria when the LAI is high.

  In addition, in the method according to the present embodiment, about 3 million points of forest point cloud data are targeted for processing, but it has been confirmed that this processing can be completed in about 45 seconds and can be performed in a very short time. . (Comment: Can you mention the actual number of processing points and processing time?)

  As described above, according to the present embodiment, a fisheye image data creation method and a fisheye image data creation program that are simpler, faster, more accurate and less affected by the weather are provided. It was confirmed that an LAI calculation method and an LAI calculation program that are difficult to be provided can be provided.

  INDUSTRIAL APPLICABILITY The present invention has industrial applicability as a fisheye image data creation method, LAI calculation method, fisheye image data creation program, and LAI calculation program.

Claims (2)

On the computer,
Reading 3D forest point cloud data;
Creating 3D reference point data;
Obtaining at least any two declination data of the forest point data constituting the forest point cloud data on the basis of the reference point data;
A fisheye image data creation program for executing the step of creating two-dimensional fisheye image data based on two declination data of the forest point data. On the computer,
Reading 3D forest point cloud data;
Creating 3D reference point data;
Obtaining at least any two declination data of the forest point data constituting the forest point cloud data on the basis of the reference point data;
A fisheye image data creation program for executing a step of creating two-dimensional fisheye image data based on two declination data of the forest point data and a step of obtaining an LAI based on the fisheye image data.