JP2014182624A - Data management device and method thereof and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To swiftly acquire a piece of information in a specified space range.SOLUTION: A space search tree storage stores a space search tree; each of terminal nodes is associated with a piece of attribute information of an object; and a host node is associated with a piece of aggregation attribute information which is generated according to the attribute information of all of lower terminal nodes. A search controller performs recursive search from a root node toward lower nodes of the space search tree on the basis of a search range, in which a part of the space is specified, to search a node to which an external envelope rectangular solid overlapping with the search range is allotted. When the node detected by the search controller is a first node that does not satisfy prescribed search conditions, a pruning section determines not to perform the search of nodes lower than the first node. The output section outputs a search result on the basis of the aggregation attribute information of the first node and the attribute information of the terminal node detected by the search controller.

Description

  Embodiments described herein relate generally to a data management apparatus and method, and a program.

  In a geographic information system, it is important to handle a collection of objects (called a scene) with various degrees of detail from a power generation device to neighboring features, and to aggregate information within a specified range accurately and at high speed.

  If there were a small number of objects within the specified range, all the objects in the target range could be simply extracted and tabulated using existing spatial index technology. However, when a range over which a scene is looked down is specified (overhead tabulation), a large number of objects are included in the range, and if all the objects are enumerated, high speed is sacrificed.

  When counting in 2D (dimensions), the degree of detail is determined for each set of objects of the same type (buildings, topography, etc.), and this problem is avoided by omitting detailed object counting in a bird's-eye view ( This method is called “decruttering”. However, in the case of 3D, even if the same kind of object is used, whether it is the target of aggregation changes depending on the altitude, or objects near the viewpoint need to be aggregated in detail, and distant objects need to be aggregated roughly.

  The existing 3D method prepares multiple levels of detail for each scene, and uses the data according to the distance from the viewpoint (uses detailed data for aggregation near the viewpoint, etc., and coarse data for distance). . Since this method requires a large amount of data, the data size is reduced at the expense of the accuracy of the object, and the visual accuracy is increased by the texture. However, since the total information requires numerical accuracy rather than visual accuracy, this conventional method cannot be used.

  Therefore, it is possible to perform extraction and aggregation processing for a set of objects with various levels of detail accurately and at high speed, and find the total population of a specified range in a geographic information system, or the total power generation of a specified generation system It has been desired that the amount can be obtained and the scene of the designated field of view can be displayed at high speed.

Patent registration No. 3267590 JP 2011-191430 A Patent registration No.4171884

  An object of an embodiment of the present invention is to obtain information on an object within a specified range in a space at high speed.

  A data management device as an embodiment of the present invention includes a first storage unit, a second storage unit, a search control unit, a pruning unit, and an output unit.

  The first storage unit includes an outer rectangular parallelepiped including an outer rectangular parallelepiped in which each outer rectangular parallelepiped of a plurality of objects arranged in a space is assigned to a terminal node and an upper node is assigned to each lower node immediately below the upper node. The space search tree to which is assigned is stored.

  The second storage unit stores the attribute information of the object as the attribute information of the end node, and the sum total determined according to the attribute information of all the end nodes existing below the upper node for each upper node Stores attribute information.

  The search control unit recursively searches from a root node of the spatial search tree toward a lower node based on a search range that specifies a part of the space, and a node to which an outer cuboid that overlaps the search range is assigned. To detect.

  The pruning unit determines not to perform a search lower than the first node when the node detected by the search control unit is a first node that does not satisfy a predetermined search condition. .

  The output unit outputs a search result based on the combined attribute information of the first node and the attribute information of the terminal node detected by the search control unit.

1 is a block diagram of a data management apparatus according to an embodiment of the present invention. The block diagram which added the process part for registering an object with respect to the apparatus of FIG. The flowchart of an object search. FIG. 4 is a detailed flowchart of a search performed in step S103 in FIG. The flowchart of an object registration process. FIG. 6 is a detailed flowchart of the search performed in step S302 of FIG. Shows a set of 15 regions. FIG. 8 shows a spatial search tree generated for the regional set of FIG. The state where the region P having a population of 10,000 is additionally registered in the region set of FIG. 7 is shown. FIG. 10 shows a spatial search tree generated for the regional set of FIG. A set of power plants is shown. FIG. 12 shows a space search tree generated for the power plant set of FIG. FIG. 12 is a diagram showing another example of a space search tree generated for the power plant set of FIG. A set of panels A to H is shown. FIG. 15 shows a spatial search tree generated for the panel set of FIG. Shows the rendering result.

  First, the technical background of the embodiment of the present invention will be described.

  In the search of a spatial index tree in the prior art, all objects that satisfy a condition (called a query) are listed. Examples of queries include "Extract objects that overlap with a specified range", "Extract objects that are included in a specified range", "Extract objects that completely include a specified range", etc. . This was realized as follows.

  That is, if the minimum bounded box (MBB) recorded for each node of the spatial search tree does not overlap with the range specified in the condition, the search for that child node is pruned (that is, its child node). However, whenever there is an overlap, the child node is recursively searched. This has sacrificed high speed.

  In the embodiment of the present invention, the total attribute information is determined for each node in advance based on the attribute information of all the terminal nodes below it. When the minimum envelope cuboid (MBB) recorded for each node has an overlap with the range specified in the query, it is determined whether or not the search condition specified separately is satisfied. If not satisfied, the search is terminated, and the child node of the node is not searched. Instead, the summation attribute information of the node is used as an alternative to the search below the child nodes from which the search is omitted, thereby supplementing the accuracy of the search. In the present embodiment as well, as in the prior art, the search below the nodes that do not overlap with the range specified by the query is not performed.

  Embodiments of the present invention will be described below with reference to the drawings.

(Example A)
FIG. 1 is a block diagram of a 3D data management apparatus according to an embodiment of the present invention.

The space search tree storage device with combined attribute information 106 stores a space search tree with combined attribute information constructed manually or automatically. The terminal node (leaf node) of the spatial search tree is assigned the smallest outer cuboid of the object, and the upper node, which is a node other than the terminal, specifies the smallest outer cuboid of each lower node connected to the upper node. The smallest enclosing cuboid is assigned. Such a space search tree can be generated by obtaining a minimum outer cuboid of an object and calculating a minimum outer cuboid (MBB) by collecting objects that are spatially close to each other. Note that a rectangular parallelepiped other than the smallest enclosing rectangular parallelepiped or a shape other than the rectangular parallelepiped may be used.
Attribute information is assigned to all nodes in the space search tree. In particular, the attribute information given to the intermediate or root node is called summed attribute information. Of the spatial search tree storage device with combined attribute information 106, the part storing the space search tree main body corresponds to the first storage unit, and the part storing the attribute information of each node corresponds to the second storage unit.

  Based on an instruction from the object search control unit 102, the node reading unit 104 receives data of the node of the spatial search tree from the spatial search tree storage device with combined attribute information 106, and attribute information / total attribute information attached to the node. Is read. The node reading unit 104 sends the read data to the object search control unit 102 and the combined attribute information pruning unit 103.

  The target range / search condition input unit 101 is a means for inputting a search range (target range) designating a part of the space and a search condition (target condition).

  The object search control unit 102 recursively searches a node having an overlap with the target range from the nodes in the space search tree as long as the search condition is satisfied. When the leaf of the node is reached, the attribute information of the object attached to the leaf node (object) is read out and summed up with the total results so far. When the combined attribute information pruning unit 103 detects that a node (root node or intermediate node) that does not satisfy the search condition has been reached, the object search control unit 102 is instructed to terminate the search below the child node of the node. The object search control unit 102 reads the total attribute information of the node via the node reading unit 104, and adds the total count result so far. When all the above search processes are completed, the search result output unit 105 finally outputs the total result (search result).

  FIG. 2 is a block diagram in which a processing unit for registering an object is added to the apparatus of FIG.

  The object input unit 201 inputs information on an object to be registered.

  The object registration control unit 202 performs control for registering an object. The node creation / update unit 204 creates and adds a terminal node at an appropriate position in the space search tree, and sets the attribute information of the node, thereby registering the object in the same manner as the conventional space search tree technique. . In addition, the total attribute information update unit 203 updates the total attribute information of related nodes (for example, each node from the registered node to the root node).

  Hereinafter, the operation of this embodiment will be described using a specific example.

  In this embodiment, a state in which 15 regions shown in FIG. 7 are registered is considered as an initial state. Each region has a rectangular shape, which is itself a minimum outer rectangular parallelepiped. Population (all 10,000 people in this example) is given as attribute information of each region (object). At this time, the spatial search tree storage device with combined attribute information stores a spatial search tree as shown in FIG.

  Here, each node is assumed to have at most four child nodes, but this number may be freely changed. Each terminal node is associated with a region (that is, an object). In this example, the object has two dimensions, but the same can be implemented in three dimensions.

  Here, as shown in FIG. 9, it is considered to additionally register a region P having a population of 10,000. The flow of object registration processing at this time will be described. The registration process is performed by each processing unit shown on the right side of FIG. FIG. 5 shows a flowchart of this process.

  First, a registered object is input (S301). Here, the data of the name “Region P” and the population “10,000 people” are input.

  Next, the position for registering the node is determined. This is performed by the processing of the sub-flowchart shown in FIG. This position determination is the same operation as the object registration of the conventional space search tree. That is, with respect to the spatial search tree of FIG. 8, first, for the four child nodes of the top node, the child node that requires the smallest expansion of the minimum outer cuboid of the node when the region P is additionally registered is selected. (S401). Here, the rightmost node is selected. When the same operation is repeated and the node reaches one node above the leaf node, the node is set as a registered node and the process returns to the flowchart of FIG.

  It is checked whether there is an empty slot in the registered node (S303). This time, each node can have 4 child nodes, and this registered node currently has only 3 child nodes, so it is considered that there is an empty slot. Therefore, the object is registered as a child node of the registration node (S306), and the spatial search tree shown in FIG. 10 is obtained (however, the combined attribute information of the intermediate node and the root node is not updated at this point). Then, the tree is traced back from the registered node to the root node, and the total attribute information is updated (S307).

  In this example, a simple sum of attribute information of child nodes is adopted as the summation routine. That is, the total population (40,000 people) of the regions M to P is updated as the total attribute information of the registered nodes. Move to the upper node and perform the same operation to rewrite the total attribute information of 150,000 people in the root node to 160,000 people. This completes the registration operation. If it is determined in step S303 that there is no empty slot, the registered node is divided to generate a new node (S304), and the new node and the divided node, and the higher intermediate node and root node are added together. Attribute information is calculated (S305).

  Here, using the target range indicated by a circle in FIG. 7, let us consider obtaining the sum of populations in an area having an overlap in the target range as an aggregation process.

  The search condition is “the minimum outer cuboid of the node is not completely included in the target range (only a part is included)”.

  FIG. 3 shows a flowchart of object search. First, data representing the information shown above is input as a target range and a search condition (S101, S102). The space search tree shown in FIG. 10 is searched from the root node (S103).

  FIG. 4 shows a detailed flowchart of the search performed in step S103. As can be seen from FIG. 7, the minimum enveloping cuboid of the root node (the cuboid that includes everything from region A to region P) overlaps the target range (YES in S201), and the root node is not an object (not a leaf node). (“Intermediate node” in S202), it is determined whether the root node satisfies the search condition (S203).

  Since the root node is obviously not completely included in the target range, the search condition is satisfied (YES in S204), and the search continues (S205).

  Next, the search is advanced to the leftmost child node (having the object as a child) among the child nodes of the root node, the same processing is repeated (S201 to S205), and the search is advanced to the end node (object) (S202). "object"). Region A and region C are not subject to aggregation because they do not overlap with the target range. On the other hand, region B and region D are subject to aggregation because they overlap with the target range. Therefore, the sum of the population of region B and region D (20,000 people) is acquired as total information so far (S208, S209).

  It moves to search of the 2nd child node from the left among the child nodes of a root node, and this node determines whether the area | regions E-H are completely included (S201, S202, S203). Since the regions E to H are completely included in the target range, the search condition is not satisfied (S204). Therefore, pruning occurs. That is, the nodes (regions E to H) below this are not individually inspected and searched. Instead, the total attribute information (40,000 people) possessed by this node is counted (S206, S207). The current total information is 60,000.

  In the related technology, this information (40,000 people) was obtained by searching and summing the areas E to H individually, but in this example pruning, the accuracy of the information is not impaired, It can be seen that the processing speed can be increased.

  Further, the third and fourth child nodes from the left of the root node are searched in the same manner as the first child node, and the regions J, M, and N are determined to be counted. As a result, the total information is 90,000. This completes all searches and returns a total calculation result of 90,000 people.

(Example B)
In the present embodiment, a case where an average power generation amount of a power plant within a specified range is obtained as a totaling process is shown.

  The basic operation of this embodiment is the same as that of the embodiment A. The difference is that the combined attribute information is information (a set of sum and number) for calculating the average.

  In the present embodiment and later-described embodiment C, a set of power plants A to H shown in FIG. 11 will be described as a theme.

  These are expressed as a space search tree in FIG. Here, 2 is set as the upper limit of the number of child nodes of each node, but this value can be freely changed.

  Each node has the total sum and the number of power generation amounts of the power plants belonging to the node or lower as the combined attribute information. This is expressed as a sum / number.

  Here, using the target range indicated by an ellipse in FIG. 11, the average power generation amount of the power plant having an overlap with this range is totaled.

  As in the case of Example A, the search condition is “the minimum outer cuboid of the node is not completely included in the target range”.

  The child node on the left side of the root node has the minimum enveloping cuboid of plants A to D, and this is completely included in the target range, and therefore does not satisfy the search condition (NO in step 204 in FIG. 4). Therefore, the search is terminated, and the current total information is 10/4 (S206, S207).

  Move on to search for a child node on the right side of the root node. Since this node satisfies the search condition (YES in S204), further child nodes are searched. A child node having attribute information of 11/2 has a minimum enclosing rectangular parallelepiped of plants E and F, and does not overlap at all with the target range, so it is not subject to aggregation.

  Next, the child node having attribute information of 15/2 has the minimum outer cuboid of plants G and H, which overlaps the target range, but also satisfies the search condition (YES in S204), and further child nodes Explore.

  Since the plant G is an aggregation target, 7/1 is added as the attribute information of the plant G to the aggregation result, and the current aggregation information is 17/5 (“object” in S202, S208, S209). Plant H does not overlap with the target range and is not a target for aggregation.

  Note that although objects (plants) that overlap with the target range even a little are targeted for aggregation, it is also possible to include only objects that are completely included in the target range.

  With the above, all the aggregations are over, and the aggregation information is 17/5. Divide this sum by this number and output an average of about 3.4 as the aggregated result.

(Example C)
In the present embodiment, a case is shown in which the top three plants with the largest amount of power generation among the power generation plants within the specified range are obtained as the aggregation process.

  The basic operation of this embodiment is the same as that of the embodiment B. The difference is that the combined attribute information is the names (identification information) of the top three power plants.

  Similar to Example B, a set of power plants A to H shown in FIG. These are expressed as a space search tree in FIG. Here, 2 is set as the upper limit of the number of child nodes of each node, but this value can be freely changed.

  The root node and each intermediate node have the names of the top three plants with the largest amount of power generation among the power generation plants belonging to the node or lower as the combined attribute information. If there are fewer than three power plants belonging to the node or lower, all plants are in descending order of power generation.

  Here, using the target range indicated by an ellipse in FIG. 11, among the power generation plants having an overlap with this range, the top three plants with the largest power generation amount are aggregated.

  As in the case of Example A, the search condition is “the minimum outer cuboid of the node is not completely included in the target range”.

  The child node on the left side of the root node has the minimum enveloping cuboid of plants A to D, and this is completely included in the target range, and therefore does not satisfy the search condition (NO in step 204 in FIG. 4). Therefore, the search is terminated, and the current total information becomes D, C, and B (S206, S207).

  Move on to search for a child node on the right side of the root node. Since this node satisfies the search condition (YES in S204), further child nodes are searched. The child node having the attribute information of F and E has the minimum enclosing rectangular parallelepiped of plants E and F, and does not overlap at all with the target range, so that it is not an aggregation target.

  Next, the child node having attribute information of H and G has the minimum outer cuboid of plants G and H, which overlaps the target range but satisfies the search condition (YES in S204), and further searches for child nodes. To do.

  Since the plant G is an aggregation target, G is added to the aggregation result as attribute information of the plant G (“object” in S202, S208, S209). At this time, only the top 3 plants with the largest amount of power generation are retained, and the names of plants that are 4th or lower are discarded. As a specific method, simple comparison and merge sort are appropriate when the number of cases is large. As a result, the current total information becomes G, D, and C. Plant H does not overlap with the target range and is not a target for aggregation.

  With the above, all tabulation is completed, and tabulation information is G, D, and C.

(Example D)
Finally, Example D shows a case where a set of panels arranged in 3D space is rendered and displayed as an image. In addition, although the panel of photovoltaic power generation is assumed as an entity of this panel, it is not restricted to this.

  A set of panels A to H shown in FIG. 14 will be described as a theme. Although the panels A to H have a three-dimensional shape, they are displayed in a two-dimensional minimum outer rectangular parallelepiped in FIG. 14 for simplicity. Only panel A is red (# FF0000 for RGB) and other panels are blue (# 0000FF for RGB).

  These are represented as a space search tree in FIG. In the root node and each intermediate node, the average color of the child nodes is recorded as the total attribute information. For example, the average color of panel A (# FF0000) and panel B (# 0000FF) is # 7F007F.

  Here, 2 is set as the upper limit of the number of child nodes of each node, but this value can be freely changed.

  The above set of panels is rendered from the camera direction shown in FIG. The field of view of this camera is the target range.

  The search condition is “the minimum outer cuboid of the node does not have a size that fits in the minimum pixel unit when rendered on the screen”. The minimum pixel unit is an example, and more generally, “the minimum outer cuboid of a node does not have a size that can be accommodated in a pixel area unit of a predetermined size when rendered on the screen”.

  The spatial search tree shown in FIG. 15 is searched from the root node, and the found object (or node) is set as a rendering target. As a specific rendering method, various methods such as a method using a CPU and a method using a GPU can be considered.

  First, it is checked whether the left node (having # 3F00BE as total attribute information) among the child nodes of the root node satisfies the search condition (S203). Here, the search condition is satisfied (YES in S204 of FIG. 4).

  Next, it is further checked whether the child node (having # 7F007F as the total attribute information) satisfies the search condition (S203). Here, it is assumed that the search condition is not satisfied (NO in S204). Pruning occurs and panels A and B are not rendered individually, but instead the minimum outer cuboid of the node is rendered on the screen using the color recorded as the combined attribute information (S206, S207). ).

  Originally, panels A and B should be drawn individually, so rendering will be inaccurate, but this node does not satisfy the search conditions, so even if panels A and B are drawn correctly, they will overlap in the minimum pixel unit. Therefore, a large change does not occur in the generated image.

  Render other node objects. In the determination of the search conditions for other objects, it is assumed that the search conditions are all satisfied (YES in S204).

  Thus, the rendering result shown in FIG. 16 is obtained as the total information. Note that panels A and B are collapsed and displayed as the smallest pixel unit. In this way, detailed drawing of the panel that does not affect many of the rendering results can be omitted, so that rendering can be performed at high speed.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

Claims (14)

Spatial search in which each enveloping cuboid of a plurality of objects arranged in space is assigned to a terminal node, and an enclosing cuboid including an enclosing cuboid assigned to each lower node immediately below the upper node is assigned to an upper node. A first storage for storing the tree;
Storing the attribute information of the object as attribute information of the end node, and storing the sum attribute information determined according to the attribute information of all the end nodes existing below the upper node for each upper node A second storage unit;
A search control unit that recursively searches from a root node of the spatial search tree toward a lower node based on a search range that specifies a part of the space, and searches for a node to which an outer rectangular parallelepiped overlapping the search range is assigned; ,
When the node detected by the search control unit is a first node that does not satisfy a predetermined search condition, a pruning unit that determines not to perform a search lower than the first node;
Based on the summation attribute information of the first node and the attribute information of the terminal node detected by the search control unit, an output unit that outputs a search result;
A data management device. 2. The data management apparatus according to claim 1, wherein the combined attribute information of the upper node includes addition of attribute information of all terminal nodes existing under the upper node. 3. The data according to claim 2, wherein the output unit calculates a sum of the attribute information of the first node and the attribute information of the terminal node detected by the search control unit, and outputs a sum value as the search result. Management device. The combined attribute information of the upper node includes an addition of attribute information of all terminal nodes existing under the upper node and a total number of all terminal nodes existing under the upper node. Data management device. 5. The output unit calculates an average of attribute information of all end nodes existing under the first node and end nodes detected by the search control unit, and outputs the average as the search result. The data management device described in 1. The combined attribute information of the upper node includes identification information of objects corresponding to N terminal nodes selected according to the attribute information of the terminal node among all the terminal nodes existing under the upper node. The data management device according to claim 1. The output unit selects N objects based on the combined attribute information of the first node and the attribute information of the end node detected by the search control unit, and the identification information of the selected object is used as the search result. The data management device according to claim 6, wherein the data management device is output. 2. The data management apparatus according to claim 1, wherein the combined attribute information of the upper node includes an average of attribute information of all terminal nodes existing below the upper node. The data according to any one of claims 1 to 8, wherein the search condition given in advance is that an outer rectangular parallelepiped assigned to a node detected by the search is not completely included in the search range. Management device. The attribute information represents a color,
The output unit draws the minimum outer envelope of the first node in the color of the combined attribute information of the first node, and the outer rectangular parallelepiped or object of the end node detected by the search control unit is the attribute of the end node 9. The data management apparatus according to claim 8, wherein rendering data drawn with information colors is output as the search result. 11. The data management apparatus according to claim 10, wherein the search condition given in advance is that a minimum outer cuboid of a node detected by the search does not fit in a pixel area unit of a predetermined size when rendered on a screen. An object input section for inputting information of an object to be registered;
An object registration control unit that generates an outer cuboid of the object and adds a terminal node to which the outer cuboid is assigned to the spatial search tree;
12. An update unit that updates the total attribute information of each node from the upper node immediately above the added end node to the root node based on the attribute information of the object to be registered. The data management device according to claim 1. Spatial search in which each enveloping cuboid of a plurality of objects arranged in space is assigned to a terminal node, and an enclosing cuboid including an enclosing cuboid assigned to each lower node immediately below the upper node is assigned to an upper node. Reading data from a first storage for storing the tree;
Storing the attribute information of the object as the attribute information of the end node, and storing the sum attribute information determined according to the attribute information of all the end nodes existing below the upper node for each upper node Reading data from the storage unit;
A search control step for recursively searching from a root node of the spatial search tree toward a lower node based on a search range that specifies a part of the space, and searching for a node to which an outer rectangular parallelepiped overlapping the search range is assigned; ,
When the node detected by the search control step is the first node that does not satisfy the search condition given in advance, a pruning step for determining not to search lower than the first node;
Based on the combined attribute information of the first node and the attribute information of the terminal node detected by the search control step, an output step of outputting a search result;
Data management method with Spatial search in which each enveloping cuboid of a plurality of objects arranged in space is assigned to a terminal node, and an enclosing cuboid including an enclosing cuboid assigned to each lower node immediately below the upper node is assigned to an upper node. Reading data from a first storage for storing the tree;
Storing the attribute information of the object as the attribute information of the end node, and storing the sum attribute information determined according to the attribute information of all the end nodes existing below the upper node for each upper node Reading data from the storage unit;
A search control step for recursively searching from a root node of the spatial search tree toward a lower node based on a search range that specifies a part of the space, and searching for a node to which an outer rectangular parallelepiped overlapping the search range is assigned; ,
When the node detected by the search control step is the first node that does not satisfy the search condition given in advance, a pruning step for determining not to search lower than the first node;
Based on the combined attribute information of the first node and the attribute information of the terminal node detected by the search control step, an output step of outputting a search result;
A program that causes a computer to execute.

Images