Vegetation management system and vegetation management method

Abstract

A system according to the present invention specifies a vegetation boundary in a target region on the basis of a remote sensing image and DEM data of the target region, and generates a height map indicating the height of each position in the target region. The system detects, from the height map, an edge treetop that is a treetop within a buffer distance from an asset in the target region. The system estimates a tree crown size corresponding to a height indicated by the height map using a plant allometry model which is a domain knowledge base model that receives heights as input and outputs tree crown sizes. The system estimates a tree crown area from the tree crown size corresponding to the edge treetop, estimates an azimuth angle on the basis of the relationship between the tree crown area and a vegetation boundary, and generates a 3D bounding box of a tree having the edge treetop on the basis of the height of the edge treetop, the tree crown size corresponding to the edge treetop, and the azimuth angle.

Description

Vegetation management system and vegetation management method

[0001] The present invention relates generally to vegetation management.

[0002] When trees encroach on assets (for example, power lines or houses), there is a possibility that the assets may be damaged or a fire (for example, a forest fire) may occur. Therefore, a vegetation management method disclosed in Patent Document 1, for example, is known.

[0003] US11270111

[0004] The method disclosed in Patent Document 1 detects tree canopies by image processing based on similarities and discontinuities in the spectral values ​​of pixels in an image. However, this method does not provide high accuracy or robustness in detecting tree canopies. In particular, in dense forests where tree canopies overlap, many unnecessary tree canopies (e.g., canopies of trees that cannot invade an asset) may be detected.

[0005] The system uses a semantic segmentation model to identify vegetation boundaries in a target area based on remote sensing images, which are satellite or aerial photographs of the target area, and DEM (Digital Elevation Model) data of the target area. The system generates a height map, which is data representing the height of each location in the target area, based on the remote sensing image and the DEM data. The system detects edge treetops, which are treetops that satisfy specified height conditions and are located within a buffer distance from assets in the target area, from the height map. The system estimates tree crown sizes corresponding to the heights represented in the height map using a plant allometry model, which is a domain knowledge-based model that inputs height and outputs tree crown sizes. The system estimates canopy areas from the detected canopy sizes corresponding to the edge treetops, estimates azimuth angles based on the relationship between the canopy area and the vegetation boundaries, and generates 3D bounding boxes for the edge treetops, which are trees with edge treetops, based on the height of the edge treetops, the detected canopy sizes corresponding to the edge treetops, and the azimuth angles.

[0006] According to the present invention, the accuracy and robustness of tree crown detection in vegetation management is improved.

[0007] 1 shows an example of the configuration of a vegetation management system according to an embodiment; 2 shows an example of the configuration of the entire system including the vegetation management system; 3 shows a DB (database) stored in the vegetation management system; 4 shows the overall flow of processing performed by the vegetation management system; 5 shows the flow of a canopy detection process; 6 shows the flow of a crown size estimation process; 7 shows the flow of a bounding box generation process; 8 shows the flow of a risk analysis process; 9 shows a schematic overview of an example of a height estimation process; 10 shows a schematic overview of an example of a boundary identification process; 11 shows a schematic overview of an example of a canopy detection process; 12 shows a schematic overview of a parameter value detection related to a 3D bounding box; 13 shows a schematic overview of a 3D bounding box; 14 shows a schematic overview of a bounding box construction for each edge tree; 15 shows a schematic overview of a risk analysis.

[0008] In the following description, an "interface device" may be one or more interface devices. The one or more interface devices may be at least one of the following: - An I / O interface device that is one or more I / O (Input / Output) interface devices. The I / O (Input / Output) interface device is an interface device for at least one of an I / O device and a remote display computer. The I / O interface device for the display computer may be a communication interface device. The at least one I / O device may be a user interface device, for example, either an input device such as a keyboard and a pointing device, or an output device such as a display device. - A communication interface device that is one or more communication interface devices. The one or more communication interface devices may be one or more homogeneous communication interface devices (e.g., one or more NICs (Network Interface Cards)) or two or more heterogeneous communication interface devices (e.g., a NIC and an HBA (Host Bus Adapter)).

[0009] In the following description, "memory" refers to one or more memory devices, which are an example of one or more storage devices, and may typically be a primary storage device. At least one memory device in the memory may be a volatile memory device or a non-volatile memory device.

[0010] In the following description, a "persistent storage device" may refer to one or more persistent storage devices, which are an example of one or more storage devices. A persistent storage device may typically be a non-volatile storage device (e.g., an auxiliary storage device), and more specifically, may be, for example, a hard disk drive (HDD), a solid state drive (SSD), a non-volatile memory express (NVME) drive, or a storage class memory (SCM).

[0011] Also, in the following description, "storage device" may be at least one of memory and persistent storage device.

[0012] In the following description, a "processor" may refer to one or more processor devices. The at least one processor device may typically be a microprocessor device such as a CPU (Central Processing Unit), but may also be another type of processor device such as a GPU (Graphics Processing Unit). The at least one processor device may be a single-core or multi-core. The at least one processor device may also be a processor core. At least one processor device may be a processor device in a broad sense, such as a circuit that is a collection of gate arrays written in a hardware description language that performs some or all of the processing (for example, an FPGA (Field-Programmable Gate Array), a CPLD (Complex Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit)).

[0013] In the following description, functions may be described using expressions such as "yyy unit." However, the functions may be implemented by one or more computer programs executed by a processor, by one or more hardware circuits (e.g., FPGAs or ASICs), or by a combination thereof. When a function is implemented by a program executed by a processor, the specified processing is performed using a storage device and / or an interface device, etc., as appropriate, and therefore the function may be considered to be at least a part of the processor. Processing described using a function as the subject may be processing performed by a processor or a device having the processor. A program may be installed from a program source. The program source may be, for example, a program distribution computer or a computer-readable storage medium (e.g., a non-transitory storage medium). The description of each function is merely an example; multiple functions may be combined into one function, or one function may be divided into multiple functions.

[0014] In the following description, data that produces an output in response to an input may be described using expressions such as "xxx DB." However, this data may have any structure (for example, structured data or unstructured data), or may be a learning model such as a neural network, genetic algorithm, or random forest that generates an output in response to an input. Therefore, "xxx DB" can be referred to as "xxx dataset." In the following description, the configuration of each DB is an example, and one DB may be divided into two or more DBs, or all or part of two or more DBs may be one DB. "DB" is an abbreviation for database.

[0015] In the following description, when elements of the same type are described without distinction, common reference symbols are used, and when elements of the same type are described with distinction, reference symbols are used.

[0016] Fig. 1 shows an example of the configuration of a vegetation management system according to an embodiment, and Fig. 2 shows an example of the configuration of the entire system including the vegetation management system.

[0017] In this embodiment, the vegetation management system 100 is a physical computer system (one or more physical computers), but it may also be a logical computer system (e.g., a virtual machine or a cloud computing service) based on a physical computer system (e.g., a cloud platform). The vegetation management system 100 includes an interface device 211, a storage device 212, and a processor 213 connected to them. The vegetation management system 100 is, for example, a server system.

[0018] The interface device 211 can communicate with a UI (User Interface) device 250 and a remote sensor 201 via a communication network (e.g., the Internet or a WAN (Wide Area Network)). The UI device 250 may be an input device or a display device, or a remote device such as a client computer equipped with them. A user of the UI device 250 may be, for example, an administrator. The remote sensor 201 is typically a camera mounted on a satellite or an aircraft. Therefore, the remote sensing image 202 output from the remote sensor 201 is typically image data of a satellite image or an aerial photograph. The interface device 211 receives the remote sensing image 202 output from the remote sensor 201, and a processor 213 (e.g., a data interface unit not shown) stores the remote sensing image 202 in a storage device 212. Although not shown, the interface device 211 receives weather data, DEM (Digital Elevation Model) data, and environmental data from an external data source (e.g., a server), and the processor 213 (e.g., a data interface unit not shown) stores the received data in the storage device 212.

[0019] Data and programs are stored in the storage device 212. For example, the storage device 212 stores a weather DB 101, an image DB 102, a DEM DB 103, and an environment DB 104. Programs are read from the storage device 212 and executed by the processor 213, thereby realizing functions such as a height estimation unit 111, a boundary identification unit 112, an edge treetop detection unit 113, a crown size estimation unit 114, a bounding box generation unit 115, and a risk analysis unit 116.

[0020] FIG. 3 shows an example of DBs 101 to 104 in the storage device 212.

[0021] The weather DB 101 has one or more pieces of weather data, such as daily precipitation, wind speed, and temperature. The image DB 102 includes one or more remote sensing images 202. The DEM DB 103 includes one or more pieces of DEM data 301. Note that the weather data, remote sensing images 202, and DEM data 301 may each be sorted by region, and each region may be associated with latitude and longitude. The source of these pieces of data may be, for example, a LANDFIRE dataset.

[0022] The environment DB 104 includes one or more pieces of environment data, such as soil data 311 and climate data 312. The soil data 311 represents the latitude and longitude, soil type, soil moisture, and soil contents for each location (e.g., region), while the climate data 312 represents the latitude and longitude and climate classification (e.g., Köppen climate classification) for each location (e.g., region).

[0023] Below, an example of the processing performed in this embodiment will be described, and each function of the vegetation management system 100 will be described as appropriate.

[0024] FIG. 4 shows the overall flow of the processing performed by the vegetation management system 100 .

[0025] The height estimation unit 111 acquires meteorological data for a region subject to vegetation management from the meteorological DB 101 using, for example, the latitude and longitude of the region as a key (S401), acquires DEM data 301 from the DEM DB 103 (S402), and acquires remote sensing images 202 from the image DB 102 (S403). The height estimation unit 111 performs height estimation processing based on the data acquired in S401 to S403 (S404). In S404, for example, the height estimation unit 111 generates a height map using the data acquired in S401 to S403 and a height estimation model. Note that the height estimation model may be a model (typically a machine learning model) that receives meteorological data, DEM data 301, and remote sensing images 202 as input and outputs a height map (an example of data representing height). In S401 to S403, in addition to data on the area subject to vegetation management, data for training the height estimation model may also be acquired, and the acquired data may be used to train the height estimation model in S404. The height map represents the estimated height for each region (e.g., one or more pixels) in the image represented by the remote sensing image 202. FIG. 9 shows a schematic overview of an example of height estimation processing. The remote sensing image 202A is input to the height estimation model, and a height map 902 is output. The height map 902 has a value for each pixel in the remote sensing image 202A that corresponds to the height of the position (latitude and longitude) corresponding to that pixel. Weather data does not necessarily have to be used for height estimation.

[0026] The boundary identification unit 112 performs steps S402 and S403 for the area subject to vegetation management, using, for example, the latitude and longitude of the area as a key. The boundary identification unit 112 performs boundary identification processing (S405). In S405, the boundary identification unit 112 finds the vegetation boundary by segmentation. Specifically, in S405, the boundary identification unit 112 identifies the vegetation boundary using the data acquired in S402 and S403 and a semantic segmentation model. Note that the semantic segmentation model may be a model that inputs the DEM data 301 and the remote sensing image 202 and outputs a vegetation boundary identification result. In S402 and S403, in addition to data about the area subject to vegetation management, data for training the semantic segmentation model may also be acquired, and the semantic segmentation model may be trained using the acquired data in S405. FIG. 10 schematically illustrates an example of the boundary identification processing. The remote sensing image 202B is input into the semantic segmentation model, and a vegetation boundary identification result 1002 is output. The vegetation boundary identification result 1002 is image data in which the vegetation area represented by the remote sensing image 202B is masked. Specifically, for example, the vegetation boundary identification result 1002 is image data in which each pixel is expressed in binary, with pixels in the vegetation area being expressed in black and pixels in areas other than the vegetation area being expressed in white. The vegetation boundary is the boundary of the generated mask of the vegetation area.

[0027] The edge treetop detection unit 113 performs treetop detection processing (S406). The crown size estimation unit 114 performs crown size estimation processing (S407). The bounding box generation unit 115 performs bounding box generation processing (S408). The risk analysis unit 116 performs risk analysis processing (S409).

[0028] Steps S406 to S409 will be explained in detail below.

[0029] FIG. 5 shows the flow of the treetop detection process.

[0030] The edge detection unit 113 acquires the height map estimated by the height estimation unit 111 (S501).

[0031] The edge treetop detection unit 113 finds local maxima in the height map that represent treetops (the tops of trees) (S502). Specifically, for example, the edge treetop detection unit 113 finds local maxima assumed to represent treetops in the height map by applying a filtering window (maximum filtering window) to the height map. Specifically, for example, as shown in FIG. 11 , many treetops (local maxima) 1101 are detected by using the height map of the remote sensing image 202C and the filtering window. Note that the remote sensing image 202C illustrated in FIG. 11 includes an asset region 1102. The asset region 1102 is an area defined to be occupied by an asset. In this embodiment, the asset is a power transmission line or a steel tower supporting the power transmission line.

[0032] Of the treetops detected in S502, the edge treetop detection unit 113 maintains edge treetops, which are treetops within a buffer distance from the asset (S503). Specifically, for example, as shown in FIG. 11 , the edge treetop detection unit 113 maintains (leaves) treetops 1101 from among multiple treetops 1101 that are within the buffer distance from the asset area 1102, and disables the other treetops. Treetops maintained in this way, i.e., treetops within the buffer distance from the asset, may be referred to as "edge treetops" hereinafter. Tree crowns with edge treetops may be referred to as "edge tree crowns." Trees with edge tree crowns may be referred to as "edge trees."

[0033] FIG. 6 shows the flow of the crown size estimation process.

[0034] The crown size estimation unit 114 trains a plant allometry model using a dataset including multiple height-crown pairs (S601). The plant allometry model is a domain knowledge-based model (typically a machine learning model) that takes tree height as input and tree crown size as output. A height-crown pair is a pair of height and crown size. The plant allometry model may be trained in advance, and therefore, S601 may not be included in the crown size estimation process.

[0035] The crown size estimation unit 114 acquires the height map estimated by the height estimation unit 111 (S602).

[0036] The crown size estimation unit 114 inputs the height map into a plant allometry model to estimate the crown sizes of all trees in the height map (S603). Note that, among the heights represented by the height map, heights equal to or greater than a threshold may be considered to be the height of the trees. Alternatively, the crown size may be output for each height represented by the height map.

[0037] FIG. 7 shows the flow of the bounding box generation process.

[0038] The bounding box generation unit 115 acquires the vegetation boundary (vegetation boundary identification result) identified by the boundary identification unit 112 (S701), acquires the height map estimated by the height estimation unit 111 (S702), and acquires data regarding each edge branch detected by the edge branch detection unit 113 (e.g., data including position data of each edge branch) (S703).

[0039] The bounding box generation unit 115 determines the height of the edge treetops from the data acquired in S703 (data related to each edge treetop) and the height map acquired in S702, and determines the crown size (crown size of the edge tree crown) corresponding to the height of the edge treetops from the estimation results by the crown size estimation unit 114 (crown sizes of all trees in the height map) (S704).

[0040] The bounding box generation unit 115 estimates the canopy area using the canopy radius (S705). The bounding box generation unit 115 identifies boundary segments, which are ranges delimited by the estimated canopy area, from the vegetation boundary acquired in S701 (S706).

[0041] The bounding box generation unit 115 generates a 3D bounding box (three-dimensional bounding box) of the tree based on the results of S704 to S706 (S707).

[0042] A specific example of generating a 3D bounding box will be described below with reference to FIGS.

[0043] 12, three edge crowns 1201A to 1201C (i.e., three edge trees) are arranged side by side, and each of the edge crowns 1201A to 1201C partially overlaps with the adjacent edge crown in plan view. Taking the edge crown 1201B as an example, we will explain how to generate a 3D bounding box for an edge tree having the edge crown 1201B.

[0044] Assume that the edge treetop obtained in S703 is edge treetop C. Assume also that the crown size identified in S704 is crown radius r. In this case, in S705, the bounding box generation unit 115 estimates the area surrounded by a circle of crown radius r centered on edge treetop C as the crown area 1250 of the edge treetop 1201B.

[0045] Assume that the vegetation boundary acquired in S701 is a vegetation boundary 1202. In S706, the bounding box generation unit 115 identifies a boundary segment AB, which is a range of the vegetation boundary 1202 that is bounded by the estimated canopy area 1250. The boundary segment AB is a line segment connecting two intersections A and B between the perimeter of the canopy area 1250 and the vegetation boundary 1202.

[0046] Then, in S707, the bounding box generation unit 115 generates a 3D bounding box of the edge tree having the edge crown 1201B as follows.

[0047] That is, the bounding box generation unit 115 translates a straight line including the boundary segment AB (line segment AB) so that it passes through edge treetop C (the center of the circle). This creates intersections D and E between the perimeter of the canopy area 1250 and the translated line passing through edge treetop C. There is also a vertical diameter XY (intersections X and Y between the perimeter of the canopy area 1250 and the vertical line passing through edge treetop C). As a result, an angle XCD is created. The angle XCD is the azimuth angle, and the magnitude of the azimuth angle is defined as "θ". Furthermore, the height of edge treetop C is defined as "h".

[0048] 13, the bounding box generator 115 constructs a 3D bounding box with a width and height of "2r" (twice r) and a height of "h", and then rotates the 3D bounding box horizontally by an azimuth angle θ around the edge treetop C (i.e., adapts the 3D bounding box to the orientation of the tree). The 3D bounding box rotated by the azimuth angle θ in this way is used as the 3D bounding box of the edge tree having the edge crown 1201B.

[0049] For each edge tree, the bounding box generator 115 generates a 3D bounding box of the edge tree having that edge tree, using the method described with reference to Figures 12 and 13. As a result, 3D bounding boxes 1300A to 1300C are constructed for the edge tree crowns 1201A to 1201C shown in Figure 12, as shown in Figure 14. For each edge tree, the bounding box generator 115 stores data related to the 3D bounding box of that edge tree in the storage device 212. For each edge tree, the data related to the 3D bounding box may include width, length, and height values, an azimuth angle value, and position data of the edge tree.

[0050] FIG. 8 shows the flow of the risk analysis process.

[0051] The risk analysis unit 116 acquires the 3D bounding box (data related to the 3D bounding box) of each edge tree from the storage device 212 (S801). Hereinafter, the 3D bounding box of an edge tree may be referred to as an "edge tree box."

[0052] The risk analysis unit 116 acquires the 3D bounding box of the asset (data related to the 3D bounding box) from the storage device 212 (S802). For example, according to the remote sensing image 202C shown in FIG. 11 , an asset area 1102 is identified from the remote sensing image 202C, and data related to the 3D bounding box of each asset (e.g., a steel tower) in the asset area 1102 is stored in the storage device 212. The 3D bounding box of each asset may be generated by the bounding box generation unit 115 based on the height map of the remote sensing image 202C. Hereinafter, the 3D bounding box of the asset may be referred to as an "asset box."

[0053] The risk analysis unit 116 acquires weather data from the weather DB 101 and environmental data from the environmental DB 104 for the area subject to vegetation management, for example, using the latitude and longitude of the area as a key (S803).

[0054] The risk analysis unit 116 simulates the collapse of edge trees (S804). Specifically, for example, the risk analysis unit 116 places the edge tree box 1300 of each edge tree at a position based on the position data of the edge tree, places the asset box of each asset at a position based on the position data of the asset, and then rotates each edge tree box 1300 in a three-dimensional direction, for example, by 0 to 90 degrees, as shown in FIG. 15. The risk analysis unit 116 determines for each edge tree whether the edge tree box comes into contact with the asset box.

[0055] The risk analysis unit 116 estimates the contact risk (S805). Specifically, for example, the risk analysis unit 116 may estimate the contact risk (e.g., calculate a numerical value representing the magnitude of the contact risk) based on the number of edge tree boxes contacting any asset box. The risk analysis unit 116 may calculate the probability and degree of edge tree collapse (e.g., blow-in or fall-in) occurring during the period represented by the acquired weather data (e.g., data representing wind speed, etc.) and environmental data (e.g., data representing the soil and climate zone at the location of the edge tree) acquired in S803, and use the probability and degree of collapse as the basis for estimating the contact risk. Furthermore, the risk analysis unit 116 may calculate the probability of an edge tree collapsing and contacting an asset based on the probability and degree of collapse and the rotation angle when the edge tree box contacts the asset box, and use the calculated contact probability as the basis for estimating the contact risk.

[0056] Various data acquired or generated in the vegetation management process may be stored in the storage device 212. For example, height maps, position data of detected edge treetops, canopy size data for each edge treetop, data representing vegetation boundaries, data related to 3D bounding boxes of edge treetops, and data representing results of the risk analysis process (e.g., the simulation results of S804 and the contact risk estimated in S805) may be stored in the storage device 212. The processor 213 (e.g., a data interface unit, not shown) may display information represented by the data stored in the storage device 212 on the UI device 250. The displayed information may be, for example, at least one of a height map, a remote sensing image on which objects representing edge treetops, vegetation boundaries, asset areas, etc. are superimposed, a 3D bounding box of an edge treetop, an animation of the rotation of the 3D bounding box of an edge treetop, a 3D bounding box of an asset, and results of the risk analysis process (e.g., the simulation results of S804 and the contact risk estimated in S805).

[0057] Although one embodiment has been described above, this is merely an example for explaining the present invention, and the scope of the present invention is not limited to this embodiment. The present invention can be implemented in various other forms.

[0058] The above description can be summarized as follows: The following summary may include supplementary explanations and explanations of variations of the above description.

[0059] The vegetation management system 100 has a boundary identification unit 112, a height estimation unit 111, an edge treetop detection unit 113, a crown size estimation unit 114, and a bounding box generation unit 115 (in other words, of the functions 111 to 116 shown in Figure 1, at least the risk analysis unit 116 is not required).

[0060] The boundary identification unit 112 identifies vegetation boundaries in the target area using a semantic segmentation model based on a remote sensing image 202 (typically a satellite image or aerial photograph image) of the target area and DEM data of the target area. The height estimation unit 111 generates a height map, which is data representing the height of each location in the target area, based on the remote sensing image and DEM data. The edge treetop detection unit 113 detects edge treetops 1101 from the height map, which are treetops that satisfy specified height conditions and are located within a buffer distance from assets in the target area. The crown size estimation unit 114 estimates the crown size corresponding to the height represented by the height map using a plant allometry model, which is a domain knowledge-based model that inputs height and outputs crown size. The bounding box generation unit 115 estimates the canopy area from the detected canopy size corresponding to the edge treetop 1101, estimates the azimuth angle based on the relationship between the canopy area and the vegetation boundary, and generates a 3D bounding box of the edge tree based on the height of the edge treetop, the detected canopy size corresponding to the edge treetop 1101, and the azimuth angle.

[0061] This will improve the accuracy and robustness of crown detection in vegetation management.

[0062] Specifically, the accuracy of tree crown detection is expected to improve because 3D bounding boxes are generated only for edge trees based on domain knowledge, rather than for all trees.More specifically, the method uses satellite or aerial photographic images and is based on vegetation boundaries and plant allometry models (domain knowledge), resulting in high interpretability.

[0063] In addition, the height of the 3D bounding box is the height of the edge tree as the top of the edge tree (the longest vertical distance), so coverage of the height of the edge tree is guaranteed. Furthermore, risk analysis can be performed using the 3D bounding box, making the method more robust.

[0064] The crown size estimation unit 114 may estimate the crown size corresponding to the height of each position represented by the height map, or may estimate the crown size corresponding to the height (and position) of only the edge treetops detected by the edge treetop detection unit 113.

[0065] The canopy area may be the canopy area 1250 as a circle with the edge treetop at its center C and the canopy radius based on the canopy size. The 3D bounding box of the edge tree may be the 3D bounding box 1300 shown in FIG. 13 , that is, a 3D bounding box with a width and length of 2r and a height of h. "r" may be the canopy radius. "h" may be the height of the edge treetop 1101. The azimuth angle may be the central angle XCD of the circle (e.g., 90 degrees or less). X may be the intersection of a line running vertically through the center C and the perimeter of the canopy area 1250, overlapping the diameter of the canopy area 1250. D may be the intersection of a line passing through the two intersections between the vegetation boundary and the canopy area perimeter, translated so that it passes through the center C, with the perimeter of the canopy area. In this way, a 3D bounding box appropriate for the overall size and orientation of the edge tree is expected to be generated.

[0066] The vegetation management system 100 may include a risk analysis unit 116. The risk analysis unit 116 may place an edge tree box (a 3D bounding box of the edge tree) based on the position of the edge tree, place an asset box (a 3D bounding box of the asset) based on the position of the asset, and estimate the contact risk based on whether the edge tree box would come into contact with the asset box when rotated in a three-dimensional direction. Because a 3D bounding box has width and length in addition to height, accurate estimation of the contact risk is expected. The risk analysis unit may calculate the probability that an edge tree will collapse and contact an asset based on at least one of meteorological data related to the weather in the target area (e.g., data representing a time series of wind speeds in the target area (e.g., forecast data)) and environmental data including data related to the soil of the edge tree (e.g., soil at or around the location of the edge tree), and use this probability as the basis for estimating the contact risk. This is expected to result in more accurate estimation of the contact risk. The risk analysis unit 116 may identify the rotation angle (rotation angle of the edge tree box) when the edge tree box comes into contact with the asset box, and based on that rotation angle and at least one of meteorological data and environmental data, calculate the probability that the edge tree will collapse to the extent of that rotation angle (the probability that the edge tree will collapse and come into contact with the asset).

[0067] 100: Vegetation management system

Claims

1. A boundary identification unit that identifies vegetation boundaries in a target area using a semantic segmentation model based on remote sensing images, which are satellite images or aerial photographs, of the target area and DEM (Digital Elevation Model) data of the target area; a height estimation unit that generates a height map, which is data representing the height of each position in the target area, based on the remote sensing images and the DEM data; an edge treetop detection unit that detects edge treetops, which are treetops that meet specified height conditions and are within a buffer distance from assets in the target area, from the height map; and a crown size estimation unit that estimates the crown size corresponding to the height represented by the height map using a plant allometry model, which is a domain knowledge-based model that takes height as input and outputs crown size. a bounding box generation unit that estimates a canopy area from the detected canopy size corresponding to the edge tree, estimates an azimuth angle based on the relationship between the canopy area and the vegetation boundary, and generates a 3D bounding box of an edge tree, which is a tree having the edge tree, based on the height of the edge tree, the detected canopy size corresponding to the edge tree, and the azimuth angle.

2. The vegetation management system of claim 1, further comprising a risk analysis unit that places an edge tree box, which is a 3D bounding box of the edge tree, based on the position of the edge tree, places an asset box, which is a 3D bounding box of the asset, based on the position of the asset, and estimates a contact risk based on whether the edge tree box may come into contact with the asset box when rotated in a three-dimensional direction.

3. A vegetation management system as described in claim 2, wherein the risk analysis unit calculates the probability that the edge tree will collapse and come into contact with the asset based on at least one of meteorological data regarding the weather in the target area and environmental data including data regarding the soil of the edge tree, and uses the probability as the basis for estimating the contact risk.

4. The vegetation management system of claim 1, wherein the canopy area is a circle with the edge treetop at its center C and a canopy radius based on the canopy size, the 3D bounding box has a width and length of 2r and a height of h, r is the canopy radius, h is the height of the edge treetop, the azimuth angle is the central angle XCD of the circle, X is the intersection of a straight line that overlaps the diameter of the canopy area and passes through the center C in a vertical direction with the perimeter of the canopy area, and D is the intersection of a straight line that is translated parallel to the center C and passes through the two intersections between the vegetation boundary and the perimeter of the canopy area, with the perimeter of the canopy area.

5. Using a semantic segmentation model, identify vegetation boundaries in the target area based on remote sensing images, which are satellite images or aerial photographs, of the target area and DEM (Digital Elevation Model) data of the target area; generate a height map, which is data representing the height of each position in the target area, based on the remote sensing images and the DEM data; detect edge treetops, which are treetops that meet specified height conditions and are within a buffer distance from assets in the target area, from the height map; and estimate the treetop size corresponding to the height represented by the height map using a plant allometry model, which is a domain knowledge-based model that takes height as input and outputs treetop size. A vegetation management method executed by a computer, comprising: estimating a canopy area from the detected canopy size corresponding to the edge treetop; estimating an azimuth angle based on a relationship between the canopy area and the vegetation boundary; and generating a 3D bounding box of an edge tree, which is a tree having the edge treetop, based on the height of the edge treetop, the detected canopy size corresponding to the edge treetop, and the azimuth angle.

6. Using a semantic segmentation model, identify vegetation boundaries in the target area based on remote sensing images, which are satellite images or aerial photographs, of the target area and DEM (Digital Elevation Model) data of the target area; generate a height map, which is data representing the height of each position in the target area, based on the remote sensing images and the DEM data; detect edge treetops, which are treetops that meet specified height conditions and are within a buffer distance from assets in the target area, from the height map; and estimate the treetop size corresponding to the height represented by the height map using a plant allometry model, which is a domain knowledge-based model that takes height as input and outputs treetop size. A computer program that causes a computer to execute the following steps: estimate a canopy area from the detected canopy size corresponding to the edge treetop; estimate an azimuth angle based on a relationship between the canopy area and the vegetation boundary; and generate a 3D bounding box of an edge tree, which is a tree that has the edge treetop, based on the height of the edge treetop, the detected canopy size corresponding to the edge treetop, and the azimuth angle.

Images