KR100894136B1 - An image detection equipment and the method which applied the KD-tree search algorithm of the non-stack type for ray tracing - Google Patents

Abstract

The present invention relates to a non-stack KD-Tree search algorithm for ray tracing, which is operated in a graphics processing unit (GPU), and more particularly. A non-stacked cad tree for ray tracing, which is applied to the graphics processing unit and improves the performance of the navigator algorithm by reducing the number of duplicated nodes in the cad tree search algorithm, a kind of binary tree. An image detection apparatus and method applying a search algorithm are provided.

An image detection apparatus using a non-stacked Cady Tree search algorithm for ray tracing according to an embodiment of the present invention divides an image into a binary tree format, uses a final segment as an end node, and includes an object included in the image using ray tracing. A device comprising a Cady Tree algorithm for detecting a signal, comprising: a ray generator for generating a ray to be transmitted to an image, and a triangle for crossing the object and the ray while moving in a direction of a terminal node of the binary tree After the search through the down-search unit and the down-search unit, an up-search unit for searching the undiscovered internal node and a shading / texture mapping unit for calculating the color of the object intersection point detected by the down search unit. There are technical features.

Ray Tracing, Cady Tree, Stack, Image Detection

Description

An image detection equipment and the method which applied the KD-tree search algorithm of the non-stack type for ray tracing}

1 is a block diagram showing image segmentation and light beams by a caddy tree using a conventional stack;

2 is a structural diagram showing a node movement in a Caddy tree using a conventional stack;

3 is a flowchart illustrating a stack-based search method using a CAD tree using a conventional stack;

4 is a block diagram showing image segmentation and light beams by a caddy tree using a conventional non-stack;

5 is a flow chart showing the up and down navigation by the Caddy tree using a conventional non-stack,

6 is a block diagram of an image detection apparatus to which a Cady Tree search algorithm using a non-stack according to the present invention;

7 is a configuration diagram of a search unit included in an image detection apparatus according to the present invention;

8 is a block diagram showing image segmentation and light beam by the Caddy tree according to the present invention;

9 is a block diagram showing a node movement in the Cady tree using a non-stack in accordance with the present invention,

10 is a flowchart illustrating a node searching method in a Cady tree using a non-stack according to the present invention.

<Explanation of symbols for the main parts of the drawings>

110: ray generator 120: search unit

130: downward navigation unit 131: ray-triangle cross test unit

132: first demultiplexer 140: upward search unit

141: node determination unit 142: downward navigation node determination unit

143: ray-box cross test unit 144: second demultiplexer

150: shading / texture mapping unit

The present invention relates to a non-stack KD-tree search algorithm for ray tracing that is operated in a graphics processing unit (GPU), and more specifically, A non-stacked cad tree for ray tracing, which is applied to the graphics processing unit and improves the performance of the navigator algorithm by reducing the number of duplicated nodes in the cad tree search algorithm, a kind of binary tree. An image detection apparatus and method applying a search algorithm are provided.

In general, the graphics card that handles computer image information, acceleration, signal conversion, and screen output depends on the video RAM and graphics chip. It was created to solve the bottleneck caused by the CPU's graphic work. Graphics card is also called Graphics Accelerator. The graphics processing unit handles transforms and lightings that were previously handled by the CPU, and was developed to use the central processing unit's cycle for other tasks, reduce the burden, and use it more freely. The Graphics Processing Unit (GPU) is an improvement on the 3D graphics acceleration chip that was originally used for applications in implementing 3D graphics. A semiconductor chip that performs graphics arithmetic processing, also called a core. Faster GPUs mean faster cores. Therefore, the throughput of the CPU is reduced, but as the throughput of the GPU continues to increase, a lot of heat is generated, and a heat sink or cooling fan is often installed like a CPU.

Thus, as computers become more versatile, faster, and higher in performance, the importance of graphics processing devices becomes more important. As the use of special effects filters and digital signal processing (DSP) processors gradually increases, the throughput of graphic processing devices is increasing.

Caddy tree, which is an algorithm included in the graphic processing device and detects an object included in an image, is a kind of spatial partitioning tree such as an octree. The octree creates eight sub-nodes by dividing the middle of each coordinate axis of the parent node.

1 is a block diagram showing image segmentation and light beams by a caddy tree using a conventional stack. As shown in FIG. 1, the Cady Tree iteratively divides two regions around a position where an evaluation function can be minimized for one coordinate axis selected by an arbitrary evaluation function until a specific limit value is reached. Divides the input image.

The Cady tree created in this way is used to quickly search for an object that intersects a ray.

2 is a structural diagram showing node movement in a Caddy tree using a conventional stack. As shown in FIG. 2, the binary tree shown is a tree structure of the image segmented in FIG. 1, and the search for the tree for detecting the image starts from the root node S0.

The intersection of the root node S0 and the ray is examined, and if the intersection of the ray and the object is performed, the child node S2 (second node) farthest from the ray start point among the two child nodes of the root node S0. ) Is placed on the stack, and the child node S1 (first node) closer to the ray viewpoint is recursively cross-tested as the parent node root node S0.

That is, if a cross-test of the first node S1, which is one of the child nodes of the root node S0, is performed and the subnode of the first node S1 includes an area intersecting the light beam, the first node S1 ) Inserts the child node n2 (second node) farthest from the ray start point out of the two child nodes included in the stack, and cross-tests the child node n1 (first node) of the adjacent one. The cross test of the light beam and the triangle included in the first node n1 is performed.

Subsequently, when the intersection test is not found as a result of the cross test of the first node n1, in designating a node to be tested next, the first node is a second node that is a node stored in the stack because no child node exists. Take out (n2) and perform the search again.

That is, in the tree shown in FIG. 2, nodes to be stored in the stack are S2, n2 and n4, and the visit order of all nodes is S0, S1, n1, n2, S2, n3, n4.

3 is a flowchart illustrating a stack-based search method using a CAD tree using a conventional stack. As shown in FIG. 3, when an intersection test of a root node is performed (S110), and when an intersection exists in a lower node of the root node, a second node among the root nodes of the root node is stored in a stack (S120) and searched. In order to move to the first node (S130). In this case, when the first node is the terminal node (S140), the intersection point of the triangle and the light beam included in the terminal node is detected (S150), and when the first node is not the terminal node (S140), the corresponding first node As the parent node, the second node among the child nodes of the parent node is stored in the stack (S120) and moved to the first node for searching (S130).

If the intersection of the triangle and the ray occurs in step S150 (S160), the search is terminated, and if the intersection of the triangle and the ray does not occur (S160), if the node stored by checking the node stored in the stack exists (S170) In step S180, the controller moves to the node stored at the top of the stack.

If there are no nodes stored in the stack (S170), it is determined that all nodes of the corresponding tree have been visited, and the search ends.

4 is a block diagram showing image segmentation and light beams by a caddy tree using a conventional non-stack. A method that does not use a stack on a GPU suggests two Cady Tree raytrace algorithms: Restart and Backtrack. These algorithms transform the intersection tolerance of the rays, thus avoiding rediscovering already visited nodes. As shown in FIG. 4, the image used for the Caddy tree is divided as described in FIG. 1, and since the value of t_min is ahead of the node n1, the node n1, which is a node close to the ray time point, is first searched, and n1 is After the search, t_min and t_max are adjusted as in (b).

Next, when searching for n2, since node n1 is not included between t_min and t_max, only n2 is selected to perform the search.

At this time, if the restart method does not find a triangle that intersects a ray at the terminal node of the tree, the t_min value is changed, and then the search is performed by searching downward again from the root node. Is used when the child node performs a search up to the parent node using a pointer to the parent node. The double restart algorithm is superior to backtrack in memory usage, and the backtrack is superior to restart in performance.

Among the two algorithms, the backtrack algorithm has a problem that up to three visits and boundary detection tests are required for internal nodes.

FIG. 5 is a flowchart illustrating up and down navigation by a Caddy tree using a conventional non-stack. As shown in FIG. 5, when the terminal node arrives through the downlink search, the intersection point between the ray and the triangle included in the terminal node is detected (S210). After the intersection is not found (S220), t_max is substituted for t_min, and global_max is substituted for t_max (S230). Subsequently, when moving to the parent node (S240), and among the child nodes included in the corresponding parent node, if there is a child node that has not been searched (S250), the mobile node moves to the corresponding child node (S260), and searches downward. In step S210 to perform (S270).

If the search for all nodes among the child nodes of the parent node is made (S250), the process moves to the parent node of the parent node and re-executes from step S250 (S240).

As shown in FIGS. 1 to 3, in the case of the Caddy tree using the stack, since a memory write operation must be performed, a memory conflict frequently occurs in a graphic processing apparatus that performs an operation using a multi-pipeline. There is a problem that causes bottlenecks.

In addition, as shown in FIGS. 4 to 5, restart and backtrack always require a boundary detection test for a parent node to select a next visit node after a terminal node visit. Therefore, up to two more visit and border detection tests should be performed on the node that was visited during the first down search. Therefore, there is a problem that a performance degradation of up to about 3 times compared to an algorithm using a stack occurs.

The present invention is to solve the problems of the prior art, and an object of the present invention is to provide an image detection apparatus that applies a Cady tree search algorithm that does not require a stack and memory write function.

Another object of the present invention is to provide a cady tree search method in which the same node is not visited more than 1.5 times on average in order to minimize duplicate operations during restart and backtrack search.

The above object of the present invention is to divide the image into a binary tree format, the final divided region as a terminal node, and includes a Cady tree algorithm for detecting an object included in the image using ray tracing, After searching through the downlink search unit and the downlink search unit for searching the triangle crossing the object and the light beam while moving in the direction of the light generating unit and the terminal node of the binary tree, It is achieved by an uplink searcher for searching for an undiscovered internal node and a shading / texture mapping unit for calculating the color of the object intersection point detected by the downsearcher.

In addition, another object of the present invention is to divide the image into a binary tree format, the final segmented region as a terminal node, an image detection method using a Cady tree algorithm for detecting an object included in the image using ray tracing. In an embodiment, a first step of searching a first node of a lower node from a root node and a second step of detecting an intersection point of the light beam and a triangle in the terminal node when the terminal node is found through the down-searching method A third step of detecting the position of the sibling node of the current node if the intersection point is not found in the step; and a fourth step of moving to the sibling node of the parent node if the sibling node has already searched in the third step. And a fifth step of determining an address of a sibling node of the parent node and a corresponding node if the sibling node of the parent node has not been searched. From among the lower nodes of the sixth step and the sixth step of detecting the point of intersection with the light beam is achieved by a seventh step for performing a search in the case where the intersection of the rays detected from the first step.

 Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Prior to this, terms or words used in the specification and claims should not be construed as having a conventional or dictionary meaning, and the inventors should properly explain the concept of terms in order to best explain their own invention. Based on the principles that can be defined, it should be interpreted as meanings and concepts corresponding to the technical spirit of the present invention.

6 is a block diagram of an image detection apparatus to which the Cady Tree algorithm using the non-stack according to the present invention is applied. As shown in FIG. 6, the image detecting apparatus includes a light generating unit 110 for generating a light beam to be transmitted to an image, a searching unit 120 for searching a node for a divided image, and a light beam and a terminal node. It includes a shading / texture mapping unit 150 for analyzing the intersection where the triangle is intersected.

In this case, the searcher 120 includes a downlink search unit 130 for searching from a root node to a terminal node and an uplink search unit 140 for searching from a terminal node to a parent node.

7 is a block diagram of a search unit included in the image detection apparatus according to the present invention. As illustrated in FIG. 7, the search unit 120 is divided into a down search unit 130 and an up search unit 140, and a signal between each component includes beam information and node information. In addition, when the node is a terminal node, it includes a triangle list.

When the down-search unit 130 arrives at the terminal node while searching for the intersection point from the root node to the lower node direction, the down search unit 130 searches for the intersection point between the ray and the triangle using information of the beam and triangle information included in the terminal node. And a triangle intersection test unit 131. When the ray and triangle cross each other in the ray-triangle intersection test unit 131, the shading / texture mapping unit 150 sets a transmission path of the ray and node information. And one demultiplexer 132.

The uplink search unit 140 searches for the intersection point of the triangle and the ray existing in the terminal node in the downlink search unit 130, and if there is no intersection point, the node determining unit 141 for determining a node to perform the next search. ).

In addition, when the node to perform the next search is determined as the sibling node of the parent node, the nodes determined by the downlink node determining unit 142 and the downlink node determining unit 142 that find the node to be downlinked are the first node. In the case of a (near node), the ray-box cross test unit 143 for detecting a boundary line between a light ray and an image is included.

In addition, when the ray-box cross-test unit 143 intersects the ray and the boundary line, the second demultiplexer 144 sets a transmission path for transmitting the ray information and the node information to search downward from the corresponding node. It may include.

When the down-search unit 130 arrives at the terminal node through the down-search, the ray-triangle cross test unit 131 included in the down-search unit 130 uses the triangle information and the light information to transmit the light at the terminal node. The intersection point between the triangle and the triangle is searched, and when the intersection point is found, the intersection information and the ray information are transmitted to the shading / texture mapping unit 150 through the first demultiplexer 132.

If the intersection point is not detected by the ray-triangle intersection test unit 131, the beam information and the node information are transmitted to the node determination unit 141 included in the uplink search unit 140 through the first demultiplexer 132. send.

In determining the node to perform the next search, the node determining unit 141 determines that the next node to be processed is a sibling node of the corresponding node. Is transmitted to the ray-triangle intersection test unit 141 to search for the intersection of the ray and the triangle.

If the node determined by the node determination unit 141 is a sibling node of the parent node, the downlink discovery node determination unit 142 determines the sibling node of the parent node to perform the down search, and the downlink discovery node determination unit 142 If the downlink search node determined in FIG. 2 is a second node (far node), the sibling node of the parent node of the determined node is searched, and if the determined downlink node is the first node (near node), the ray-box cross test unit ( In step 143, the intersection search is performed.

When the intersection point is found as a result of the ray-box cross-test unit 143, the ray-triangle cross-test unit 131 included in the down search unit 130 passes through the second demultiplexer 144. Information and node information is transmitted to perform downlink search, and if no intersection point is found, the cross-test result information and light information are transmitted to the downlink node determination unit 142 through the second demultiplexer 146. For example, the sibling node of the parent node for performing the next search is determined based on the currently located node.

8 is a block diagram illustrating image segmentation and light beams by the Caddy Tree according to the present invention. As shown in FIG. 8, the image is divided into a binary tree format by the Cady Tree algorithm, and the light beam is irradiated from the light beam time point. In this case, the root node corresponds to S0 including the entire image, and the terminal nodes are n1 to n16 which are not divided anymore.

9 is a diagram illustrating a node movement in a Cady tree using a non-stack in accordance with the present invention. FIG. 9 is a representation of the divided image illustrated in FIG. 8 in a binary tree structure. Each arrow indicates a sequence for visiting a lower node of a root node when all terminal nodes include triangles.

At this time, a node located near the ray point of view among the child nodes of the root node is called a first node, and a node located far from the ray point of view is called a second node. That is, any node among the nodes other than the terminal node includes a child node, and a node near the ray point of view is called a first node (near node), and a node far from the ray point of view. Is called a second node (far node).

The search from the root node to the child node is called down search, and the search from the terminal node to the sibling node of the parent is called up search.

10 is a flowchart illustrating a node searching method in a Cady tree using a non-stack according to the present invention. As shown in FIG. 10, the node is first downlinked from the root node to the light ray through the downlink searcher (S305). Then, when arriving at the terminal node, the downlink search unit searches for the intersection point between the ray and the triangle in the ray-triangle intersection test unit using the information of the ray and the triangle information included in the terminal node (S310).

If an intersection is found at the terminal node (S315), the corresponding cross test result information and the ray information are transmitted to the shading / texture mapping unit using the first demultiplexer (S320), and the intersection node is not found at the terminal node. If it is (S315), it is detected whether the corresponding node exists in the outermost point from the light source viewpoint (S325).

In step S325, if the corresponding node is the outermost node, it is determined that the node to be searched no longer exists, and the search ends. If the node is not the outermost node, the search proceeds from the sibling node of the corresponding node. (S335).

If it is already necessary to complete the search and do not need to perform again (S330), to move to the sibling node of the parent node to perform the search (S340). In this case, when the address of the sibling node of the parent node is odd (that is, as a child node of the root node, which corresponds to the first node) (S345), the search is made up to the child node of the sibling node of the parent node already. Since it has been completed, it moves to the sibling node of the parent node of the node to be moved in order to perform again from step S335. Thereafter, the search is performed again from step S335.

If the address of the sibling node of the parent node is an even address (i.e., it is a node below the root node of the root node and corresponds to the second node) (S345), a node that has not yet been searched for intersection search is performed. As a search for whether an intersection point with a ray exists in a lower node of the node (S350), and if an intersection point is detected in a lower node of the node, the process moves to step S305 to re-search for detecting an intersection point of a ray and a triangle. .

Although the present invention has been shown and described with reference to the preferred embodiments as described above, it is not limited to the above embodiments and those skilled in the art without departing from the spirit of the present invention. Various changes and modifications will be possible.

Therefore, in the image detection apparatus using the non-stacked Cady Tree algorithm for ray tracing according to the present invention, in searching for a node using a Cady Tree search algorithm that does not require a stack and memory write function, a duplicate search and calculation execution time are required. The effect of reducing the time for image detection is reduced.

Claims (10)

An apparatus comprising a Cady Tree algorithm for dividing an image into a binary tree format, using the last divided region as a terminal node, and detecting an object included in the image by using ray tracing. A ray generator for generating a ray to be transmitted to the image; A downward search unit for searching for a triangle in which the object and the ray intersect while moving in a direction of a terminal node of the binary tree; An upward search unit for searching for an internal node that is not searched after the search through the down search unit; And Shading / texture mapping unit for calculating the color of the object intersection point detected by the downward search unit Apparatus for detecting a non-stacked Cady tree search algorithm for ray tracing comprising a. delete delete delete delete delete delete delete delete delete

Images