KR100933164B1 - Optimal Trace Detection of Linear Laser Beam Image based on Dynamic Programming for Accurate 3D Contour Information of Objects - Google Patents

Abstract

The present invention relates to a method for detecting the optimal linear laser beam using dynamic programming for accurate three-dimensional contour detection of objects. A method commonly used for detecting three-dimensional contours of an object is a method of calculating the image position of a laser line obtained by projecting a linear laser beam onto a surface of an object by a triangulation method. However, when the surface of an object absorbs or totally reflects a laser beam, it is difficult to calculate the three-dimensional contour information because it is difficult to obtain the exact position of the laser line from the image. Dynamic programming, on the other hand, is a very effective way to find the path that minimizes the total cost from the starting point to the destination by using partial cost values on the path. The present invention uses a trace of a linear laser image as a partial cost value of a path, and sets a target line and a start line of each dynamic programming method at the top and bottom of the image, starting from the target line and starting from the target line. Accumulate and propagate to the starting line, and find the exact position of the laser trace along the path of the smallest cumulative cost value starting from the starting line to find the exact location of the laser trace, thereby finding the correct three-dimensional contour of the object. It is about how to get information.

Dynamic Programming, Linear Laser, Trace Detection of Linear Laser Image, 3D Contour Detection of Objects

Description

Dynamic Programming-based Optimal Trace Detection of Line Type Laser Light Image for Obtaining Correct 3 Dimensional Shape Information}

The present invention relates to a method for detecting the optimal linear laser beam using dynamic programming for accurate three-dimensional contour detection of objects. The methods used to detect three-dimensional contours of an object include a method using an ultrasonic wave, a method using a stereo vision, and a method using a linear laser. The method of using ultrasonic waves is a method of measuring the flight time of ultrasonic waves and converting them into distances. However, since the precision is low, the ultrasonic waves are only used for the application of checking the existence of an object. The stereovision method is a method of calculating the distance to an object by using a triangulation method using a pixel position at which an object is projected by two cameras. Although this method is more accurate than using ultrasound, it is difficult to use for precise three-dimensional information detection because it is difficult to find a matching pair at the surface of an object without singularity. In the method using the linear laser, while projecting the linear laser beam on the surface of the object, the image of the laser beam is captured by the camera and the distance information is calculated using the distortion information of the linear laser image. This method can create an artificial laser image pattern on the surface of a plain object without a pattern, so it is used frequently because only the pixel position of the laser image can be precisely calculated. However, when the surface of an object absorbs or totally reflects a laser beam, it is difficult to calculate the three-dimensional contour information because it is difficult to obtain the exact position of the laser line from the image.

According to the present invention, in the conventional three-dimensional information calculation method using a linear laser beam, when the surface of an object absorbs or totally reflects the laser beam, it is difficult to obtain the exact position of the laser line from the image. In order to solve the problem of difficulty, the inverse of the linear laser image pixel value is used as the partial cost value of the path, and the three-dimensional contour of the object is found by applying the dynamic programming method that is effective for the optimal path detection. It is about how to get information.

 The present invention is to calculate the accurate three-dimensional information of the object, Figure 1 is a phenomenon in which the laser image is distorted according to the contour of the object in the image captured by the camera when the linear laser is projected on the object and the distance using the same An example of a measuring device is shown. The common feature of devices that can obtain three-dimensional information about object contours through such devices is that they can calculate distance, height and direction from the camera to the object by using the pixel position of the linear laser light in the image. .

As described above, the present invention relates to a method for detecting an optimal linear laser beam using dynamic programming for accurate three-dimensional contour detection of objects. The method of using a linear laser has the effect of creating an artificial laser image pattern on a plain object surface without a pattern, and it is frequently used because precise distance calculation is possible only by calculating the pixel position of the laser image. However, when the surface of an object absorbs or totally reflects a laser beam, it is difficult to calculate the three-dimensional contour information because it is difficult to obtain the exact position of the laser line from the image. Using the present invention, it is possible to find the optimal position of the laser line by utilizing the optimization function of the dynamic programming method, which is very effective in obtaining three-dimensional information of the object.

 The present invention is to calculate the accurate three-dimensional information of the object, Figure 1 is a phenomenon in which the laser image is distorted according to the contour of the object in the image captured by the camera when the linear laser is projected on the object and the distance using the same An example of a measuring device is shown. The common feature of devices that can obtain three-dimensional information about object contours through such devices is that they can calculate distance, height and direction from the camera to the object by using the pixel position of the linear laser light in the image. .

However, when the surface of the object absorbs or totally reflects the laser beam, it is difficult to obtain the exact position of the laser line from the image, which makes it difficult to calculate three-dimensional contour information. In order to solve this problem, the present invention separates a laser image from a background image, and applies dynamic programming to the separated laser image. In this case, in order to separate the laser image from the background image well, dark illumination is used, or an infrared laser and an infrared filter are used.

2A shows an example of a laser image after removing a background image through a method using an infrared filter. In FIG. 2A, the bending of the laser image means that the surface of the object is curved, and when the line is cut up and down, the laser image does not appear due to a problem of the surface of the object or lighting. An object of the present invention is to connect a line cut up and down as shown in FIG. 2A along an optimal line as shown in FIG. 2B using a dynamic planning method described below.

The dynamic programming method is an efficient calculation algorithm that calculates the overall optimal path through the local minimum calculation at each node as shown in FIG. 3. If D (i, j) Let D (k, l) be the minimum distance from each node (i, j) and node (k, l) to the destination. In addition, assuming that d _{ij , kl} is the local minimum distance from node (i, j) to node (k, l), D (i, j) is D (k, l) and d _{ij , It} is expressed as Equation 1 using _kl .

Where R (i, j) is the set of all nodes around node (i, j). If an operation such as Equation 1 is repeated at all nodes, cost values between each node are accumulated and propagated to all nodes. The calculation and propagation of this cost value is continued for a sufficient time and repeated until all nodes have no change in output value.

In addition, the determination of the optimal path is performed through the backward tracking using the output values of nodes starting from the starting point. That is, if (i, j) is an arbitrary node, the next point on the optimal path to the target point may be determined as the point (k, l) that satisfies the expression (2).

와 같은 과정을 반복하여 목적지 셀까지의 최소 거리의 경로를 구할 수 있다.

By repeating the above process, the path of the minimum distance to the destination cell can be obtained.

Optimal Trace Detection of Linear Laser Beam Images Using Dynamic Programming

In order to use the dynamic programming method for the optimal trace detection of the linear laser beam image, the linear laser beam trace in the linear laser beam image is installed as the cost value of the dynamic programming method, and then the dynamics are respectively displayed at the top and bottom of the image as shown in FIG. Set the goal line and start line of the programming method. Then, starting from the target line, the cost value is accumulated by the dynamic programming method and propagated to the starting line. Then, starting from the starting line, find the optimal path of the laser beam trace along the path of the lowest cumulative cost value.

More specifically, the dynamic programming calculation uses the inverse of the edge values in the edge image as the cost values d _{ij , kl} between two adjacent nodes (i, j) and (k, l) of the dynamic programming;

Apply a start cost value of 0 to the nodes corresponding to the target line;

Accumulate and propagate the cost value to all cells by performing the operation of Equation 1 on all nodes;

After the accumulated cost value reaches the start line, it is repeated until there is no change in the cost accumulation value in the entire image.

Thereafter, among the neighboring nodes of each node on the start line of the image, the node (k, l) satisfying Equation 2 is found and set as a starting point;

A position (k, l) of a node satisfying Equation y by checking the adjacent eight node regions R (i, j) of (i, j) as the starting point (k, l) as a starting point (i, j) Is the next position (i, j) on the optimal path;

This process is repeated starting from the starting point at the bottom of the image until reaching the destination line set at the top to determine the optimal linear laser ray image trace.

As described above, although the present invention has been described with reference to the above embodiments, the present invention is not limited thereto, and various modified embodiments may be included in the present invention without departing from the scope and spirit of the present invention.

FIG. 4 shows that a faint and cut laser image as shown in FIG. 4A as a real image thereof results in a continuous laser image as shown in FIG. 4B by using the method of the present invention. When such a laser line image is obtained for the whole object, three-dimensional information of the whole object can be obtained. In addition, FIG. 5 illustrates the restoration ability of the whole object using the present invention. For the real objects of FIG. 5A, the restoration image is incomplete as shown in FIG. 5B when the method of the present invention is not used. In this case, complete 3D contour information can be obtained as shown in FIG. 5C.

1 is an illustration of an apparatus for obtaining three-dimensional information of an object using linear laser light;

FIG. 2A is a diagram illustrating the frequent cutting of a laser image in a general linear laser beam image; FIG.

FIG. 2B is an exemplary view illustrating the effect of connecting and reconstructing the cut linear laser beam image of FIG. 2A by installing a target line and a starting line and applying a dynamic programming method according to the present invention.

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

210: truncated linear laser beam image

            220: target line

230: Laser beam image connected by the present invention

            240: starting line

3 is a conceptual diagram for explaining the principle of the dynamic planning method.

4A is an illustration of an actual image when a linear laser beam is projected onto an object;

4B is an exemplary view of detecting an optimal linear laser ray trace using the present invention.

Also, 5A is a photograph illustrating real objects used to verify the effect of the present invention;

FIG. 5B is an exemplary view of the result of reconstructing the three-dimensional information about the objects of FIG. 5A without using the method of the present invention; FIG.

5C is an exemplary view of a result of reconstructing three-dimensional information about the objects of FIG. 5A when using the method of the present invention.

Claims (3)

In extracting three-dimensional information about an object using a linear laser beam image, Setting a target line and a start line at upper and lower ends of the linear laser beam image; Accumulating cost values by dynamic programming for a plurality of paths connecting the target line and the starting line; And Finding a path of a laser ray trace, using a path having the least accumulated cost value among the plurality of paths; Optimal linear laser beam image trace detection method comprising a. The method of claim 1, Accumulating the cost value by the dynamic programming method,

(; D (i, j) is the minimum distance it takes to reach the target line from node (i, j), ; D (k, l) is the minimum distance it takes to reach the target line from node (k, l), ; d _{ij, kl} is the local minimum distance from node (i, j) to node (k, l), ; R (i, j) is a collection of nodes around node (i, j).) Accumulating cost values starting from the target line by performing the calculation of the operation, and repeating the operation until the cost value accumulated up to the starting line does not change in the entire image. Optimal linear laser beam image trace detection method comprising a. The method of claim 2, Finding the path of the laser beam trace, Among the neighboring nodes of each node on the starting line,

(; y (k, l) is the output of adjacent nodes in R (i, j), ; y (i, j) is the output of node (i, j), ; d _{ij, kl} is the local minimum distance from node (i, j) to node (k, l)., Determining a node (k, l) that satisfies as a starting point; The starting point (k, l) is a starting point (i, j), and among the R (i, j) adjacent to the starting point (i, j),

A first step of determining a position (k, l) of a node that satisfies the next starting point (i, j) on the path; And Repeating the first step until reaching the target line and finding the found path line as the path of the laser ray trace; Optimal linear laser beam image trace detection method comprising a.

Images