KR101844328B1 - Occlusion and rotation invariant object recognition system and method in factory automation - Google Patents

Abstract

The present invention relates to a factory automation object recognizing system and method robust against cloaking and rotation, and more particularly, to a system and method for recognizing a factory automation object, comprising an input unit for photographing an object placed on a factory automation line and acquiring a scene image; A region of interest setting unit for setting a region of interest by comparing the scene image with a previously input background image; A ring projection unit for extracting a candidate group corresponding to a value having a correlation of at least a predetermined first threshold value using a ring projection for a set ROI; A radiation projection unit for extracting a projection vector from a template image, dividing a scene image from the extracted candidate group into quadrants and extracting a projection vector from each divided region; And an object determination unit for calculating a correlation through the extracted template image and the radiation projection vector of each region, and for comparing the correlations and recognizing the object.

Description

BACKGROUND OF THE INVENTION Field of the Invention [0001] The present invention relates to a factory automation object recognizing system and method that are robust against obstacles and rotations,

The present invention relates to an object recognition system and method, and more particularly, to an object recognition system and a recognition method for visually recognizing a part to be applied to an assembly during assembly among factory automation techniques applied at a production site.

In an object recognition system used in a general factory automation environment, objects are moved by a robot through a conveyor belt. At this time, a camera is installed on the conveyor belt to photograph an object on the conveyor belt. The input image is analyzed and the robot on the conveyor belt informs the position of the object in the house.

In the vision recognition system using existing object block based matching, there is a problem in that the recognition rate significantly decreases or is not recognized when the registered image is rotated on the actually photographed image. In reality, however, moving objects on a conveyor belt are often not perfectly aligned. This problem must be overcome.

In the conventional vision recognition system described above, there is a problem that the recognition rate is remarkably deteriorated or is not recognized when the object is robust against rotation and the object is covered with another object. Also, in an actual automation environment, objects moving on a conveyor belt may have a complex mixture of objects. In this case, it is very difficult to recognize if the desired object is rotated and covered by other objects.

In addition, since the conventional factory automation object recognition system and method can be recognized only when the object to be recognized by the automation line is correctly aligned or not covered by other objects, it is necessary to add a task of sorting the object through the preprocessing process .

Korean Patent Publication No. 10-2016-0075135 (Published date: June 29, 2016) Korean Patent Laid-Open Publication No. 10-2014-0067604 (Published date: June 05, 2014)

The factory automation object recognizing system and method robust against obstruction and rotation according to the present invention have the following problems.

First, the present invention aims to provide a system and method for recognizing a target object in a factory automation line even when the object is rotated or partially obscured.

Second, the object of the present invention is to provide an object recognition system and method capable of enhancing recognition performance so as to accurately recognize various environments (object position, rotation, and masking) without additional apparatuses.

The present invention has been made in view of the above problems, and it is an object of the present invention to provide an apparatus and method for controlling the same.

According to a first aspect of the present invention, there is provided a factory automation object recognition system robust against cloaking and rotation, comprising: an input unit for capturing an image of a target object placed on a factory automation line and acquiring a scene image; A region of interest setting unit for setting a region of interest by comparing the scene image with a previously input background image; A ring projection unit for extracting a candidate group corresponding to a value having a correlation of at least a predetermined first threshold value using a ring projection for a set ROI; A radiation projection unit for extracting a projection vector from a template image, dividing a scene image from the extracted candidate group into quadrants and extracting a projection vector from each divided region; And an object determination unit for calculating a correlation through the extracted template image and the radiation projection vector of each region, and for comparing the correlations and recognizing the object.

Here, the region of interest setting unit may include a reference image storage unit for previously registering and storing a background image and a template image of a target object; And a setting unit configured to compare the stored background image with the scene image input from the input unit to set a region of interest in the target object region.

The ring projection unit may further include: a ring vector extraction unit that extracts a ring projection vector from the scene image and the template image; A correlation calculating unit for calculating a correlation between the scene image and the template image using the ring projection vector; And a candidate group extracting unit for extracting only candidate pixels having a calculated degree of correlation equal to or greater than a preset first threshold value as a candidate group.

In addition, the equation for calculating the ring projection vector is:

,

,

중 작은 값이고, α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀의 총 개수를 M이라 하고,

는 영상의 좌표에서의 픽셀 값을 나타내고, x 및 y는 픽셀의 좌표값이고, w 및 h는 템플릿 영상의 가로 및 세로의 크기이고, W 및 H는 장면 영상의 가로 및 세로의 크기이고, t는 반지름의 길이이다.)와 같은 식을 만족하는 것이 바람직하다.(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

W and h are the horizontal and vertical sizes of the template image, W and H are the horizontal and vertical sizes of the scene image, t and t are the pixel values in the coordinates of the image, x and y are the coordinate values of the pixel, Is the length of the radius) is preferably satisfied.

Further, the equation for calculating the degree of correlation,

는 벡터 x 요소들 평균값을 나타낸다.)과 같은 식을 만족하는 것이 바람직하다.(Where NCC denotes Covariance, denotes a covariance, a and b denote a vector,

Represents an average value of the vector x elements).

The equation for extracting the projection vector from the template image is as follows:

중 작은 값이고, α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀의 총 개수를 M이라 하고,

는 영상의 좌표에서의 픽셀 값을 나타내고, x 및 y는 픽셀의 좌표값이고, w 및 h는 템플릿 영상의 가로 및 세로의 크기이고, t는 반지름의 길이이다.)과 같은 식을 만족하는 것이 바람직하다.(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

Where x and y are the coordinate values of the pixel, w and h are the width and height of the template image, and t is the length of the radius). desirable.

The expression for extracting the radiation projection vector in each of the above-

중 작은 값이고, α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀의 총 개수를 M이라 하고,

는 영상의 좌표에서의 픽셀 값을 나타내고, x 및 y는 픽셀의 좌표값이고, w 및 h는 템플릿 영상의 가로 및 세로의 크기이고, W 및 H는 장면 영상의 가로 및 세로의 크기이고, t는 반지름의 길이이다.)과 같은 식을 만족하는 것이 바람직하다.(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

W and h are the horizontal and vertical sizes of the template image, W and H are the horizontal and vertical sizes of the scene image, t and t are the pixel values in the coordinates of the image, x and y are the coordinate values of the pixel, Is the length of the radius).

In addition, the object determining unit may include: a correlation calculating unit that calculates a correlation between the projection vector of the extracted template image and the projection vector of each divided region; And an object recognition unit for recognizing the calculated object value if the calculated correlation value of each area is larger than a predetermined second mapped value.

According to a second aspect of the present invention, there is provided a method for recognizing a factory automation object that is robust against cloaking and rotation, comprising the steps of: (a) capturing a scene image by capturing an object placed on a factory automation line of an input unit; (b) comparing the scene image with the background image pre-input by the ROI setting unit to set the ROI; (c) extracting a candidate group whose degree of correlation is equal to or greater than a predetermined first threshold value by using a ring projection for a region of interest set in the ring projecting unit; (d) extracting a radiation projection vector from the radiation projection part template image, dividing the scene image from the extracted candidate group, and extracting a radiation projection vector from each divided area; And (e) calculating a correlation through the template image extracted by the object determining unit and the radiation projection vector of each region, and comparing the correlations to determine whether the object is an object.

The step (c) includes the steps of: (c1) extracting a ring projection vector from the scene image and the template image; And (c2) calculating a correlation between the scene image and the template image using the ring projection vector. And (c3) extracting only the pixels having the calculated degree of correlation equal to or greater than a predetermined first threshold value as a candidate group.

The equation for calculating the ring projection vector in the step (c2)

,

,

중 작은 값이고,

는 영상의 좌표에서의 픽셀 값을 나타낸다.)와 같은 식을 만족하는 것이 바람직하다.(Where, w and h are the width and height of the template image, W and H are the width and height of the scene image, and R is the maximum radius,

Lt; / RTI >

Represents a pixel value in the coordinates of the image).

In addition, the equation for calculating the above-

는 벡터 x 요소들 평균값을 나타낸다.)과 같은 식을 만족하는 것이 바람직하다.(Where NCC denotes Covariance, denotes a covariance, a and b denote a vector,

Represents an average value of the vector x elements).

The equation for extracting the projection vector from the template image is as follows:

중 작은 값이고, α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀의 총 개수를 M이라 하고,

는 영상의 좌표에서의 픽셀 값을 나타내고, x 및 y는 픽셀의 좌표값이다.)과 같은 식을 만족하는 것이 바람직하다.(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

And x and y are coordinate values of the pixel), as shown in Fig.

Further, the expression for extracting the projection vector into each of the above-

중 작은 값이고, α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀의 총 개수를 M이라 하고,

는 영상의 좌표에서의 픽셀 값을 나타내고, x 및 y는 픽셀의 좌표값이다.)과 같은 식을 만족하는 것이 바람직하다.(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

And x and y are coordinate values of the pixel), as shown in Fig.

The step (e) may further include: calculating a degree of correlation between a projection vector of the extracted template image and a projection vector of each divided region; And recognizing the calculated object value as a target object when the calculated correlation value of each area is larger than a predetermined second mapped value.

And a third aspect of the present invention features a computer program stored in a computer-readable medium in combination with hardware to execute the above-described masking and rotation robust factory automation object recognition method.

The system and method for recognizing a factory automation object robust against obstruction and rotation according to the present invention have the following effects.

First, the object of the present invention is to provide an object recognition system and method capable of enhancing recognition performance so as to recognize accurately without using an additional device in various environments (object position, rotation, and masking) by using a ring projection and a radiation projection to provide.

Second, the present invention provides an object recognition system and method capable of increasing the efficiency of a factory automation line by recognizing an object that is rotated or obstructed, without going through a preprocessing process.

The effects of the present invention are not limited to those mentioned above, and other solutions not mentioned can be clearly understood by those skilled in the art from the following description.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a plant automation object recognizing system robust to obstruction and rotation according to an embodiment of the present invention; FIG.
FIG. 2 is a flowchart illustrating a method for recognizing factory automation objects, which is robust against cloaking and rotation according to an embodiment of the present invention.
3 is a schematic diagram of a scene image and a template image applied to a block-based matching algorithm.
FIG. 4 shows a matching process for a template image and a scene image.
Fig. 5 is a schematic diagram showing the geometrical meaning of the ring projection. Fig.
6 is a schematic diagram showing the geometrical meaning of Radial Projection.
FIG. 7 is a diagram illustrating an area-of-interest setting applied to a factory automation object recognizing method according to an embodiment of the present invention.
8 is an image showing an example of setting the first threshold value (Threshold 1).
FIG. 9 is a flowchart illustrating more specifically steps (d) and (e) in a method for recognizing a factory automation object robust against obstruction and rotation according to an embodiment of the present invention.
10 is a schematic diagram of dividing the pixels included in the maximum radius into four by a to d.
11 is an example of a template (FIG. 11 (a)) and a background image (FIG. 11 (b)) applied to the embodiment of the present invention.
12 is an illustration of an image in which a region of interest is extracted using a background image in a scene image according to an embodiment of the present invention.
13 is an illustration of an image obtained by extracting a candidate group using a ring projection in a region of interest according to an embodiment of the present invention.
FIG. 14 is an example of an image showing a result of ultimately determining an object using a radial projection in a candidate group according to an embodiment of the present invention.

Further objects, features and advantages of the present invention will become more apparent from the following detailed description and the accompanying drawings.

Before describing the present invention in detail, it is to be understood that the present invention is capable of various modifications and various embodiments, and the examples described below and illustrated in the drawings are intended to limit the invention to specific embodiments It is to be understood that the invention includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, the terms "comprises" or "having" and the like refer to the presence of stated features, integers, steps, operations, elements, components, or combinations thereof, But do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

Further, terms such as " part, "" unit," " module, "and the like described in the specification may mean a unit for processing at least one function or operation.

In the following description of the present invention with reference to the accompanying drawings, the same components are denoted by the same reference numerals regardless of the reference numerals, and redundant explanations thereof will be omitted. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail.

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

FIG. 1 is a block diagram of a factory automation object recognizing system which is robust against obstacles and rotation according to an embodiment of the present invention. FIG. 2 is a diagram showing a factory automation object recognizing method robust to obstacles and rotation according to an embodiment of the present invention Fig.

As shown in FIG. 1, the system for recognizing factory automation objects, which is robust against cloaking and rotation according to an embodiment of the present invention, includes an input unit 100 for shooting a target object placed on a factory automation line and acquiring a scene image, (200) for setting a region of interest by comparing the background image and the scene image with each other, and determining a correlation value between a predetermined threshold value and a predetermined threshold value by using Ring Projection for the set region of interest A ring projection unit 300 for extracting a corresponding candidate group, a radiation projection unit 400 for extracting a radiation projection vector from the template image, dividing the scene image from the extracted candidate group into four, An object determining unit 500 for calculating the correlation through the extracted template image and the radiation projection vector of each region, comparing the correlations, and recognizing the object, It is configured to include.

As shown in FIG. 2, the method for recognizing factory automation objects, which is robust against cloaking and rotation, according to an embodiment of the present invention, includes the steps of: (a) (B) setting a region of interest by comparing the scene image with a background image previously input by the region-of-interest setting unit 200, and (c) setting a region of interest Extracting a candidate group corresponding to a value whose correlation is greater than or equal to a preset first threshold value by using a ring projection for the region; (d) Extracting a projection image from the extracted candidate group, and extracting a projection vector from each of the divided regions; and (e) using the extracted template image and the projection vector of each region, Prize Calculating a degree, and compares each correlation is configured to include the step of determining whether the object.

As described above, the embodiment of the present invention is an object recognition system and method used in a factory automation environment. In the factory automation environment, recognition performance in various environments (object position, rotation, and masking) The constraints are reduced for use in a factory automation environment and the efficiency of the automation line can be increased by omitting the preprocessing process.

More specifically, as shown in FIG. 1, the region-of-interest setting unit 200 registers a background image and a template image of a target object in advance and stores the template image And a setting unit configured to set a region of interest in the target object region by comparing the stored background image with a scene image input from the input unit 100. In addition,

The ring projecting unit 300 includes a ring vector extracting unit for extracting a ring projection vector from a scene image and a template image, and a correlation calculator for calculating a correlation between the scene image and the template image using the ring projection vector And a candidate group extracting unit for extracting only candidate pixels having a calculated correlation value equal to or greater than a predetermined first threshold value as a candidate group.

The object determining unit 500 further includes a correlation calculating unit that calculates a correlation between the projection vector of the extracted template image and the projection vector of each of the divided regions, And an object recognition unit for recognizing the object as an object when the value is greater than a two-parameter value.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a method for recognizing a factory automation object robust to a cloaking and rotation according to an embodiment of the present invention will be described in detail.

First, a factory automation object recognizing method which is robust against cloaking and rotation according to an embodiment of the present invention uses a block-based matching algorithm among object recognition methods. The block matching algorithm is a method of finding a matching part with an image registered as a template in a scene image.

3 is a schematic diagram of a scene image and a template image applied to a block-based matching algorithm. FIG. 3 (a) shows a scene image taken by a camera, and FIG. 3 (b) shows a template image in which an object to be searched is registered.

FIG. 4 shows a matching process for a template image and a scene image. 4, when the template image is w * h and the size of the scene image is W * H, the template image is shifted pixel by pixel on the scene image, and the degree of similarity with the scene image is calculated where the template image is located .

[Equation 1] denotes a range in which a template image moves when it moves on a scene image in pixel units.

Normalized Cross-Correlation (NCC) is one of the methods of calculating the similarity of a template image while moving a template image in a scene image.

는 벡터 x 요소의 표준편차이다.[Equation 2] is a method of measuring the similarity between the vectors a and b, Covariance means covariance

Is the standard deviation of the vector x elements.

는 벡터 x 요소들 평균값 이다.If Equation (2) is solved, Equation (3) is obtained.

Is the mean value of the vector x elements.

The range of the NCC calculation value is as shown in [Equation 4]. If the two vectors are exactly the same, the result is 1; the higher the correlation, the closer to 1.

In order to compare the degree of correlation in Equation (2) above, a vector must be extracted from the template image and the scene image. There are various methods for extracting the vector, but among them, Ring Projection_ and Radial Projection are used.

크기의 템플릿 영상의 중심점에서부터 반지름 r을 점진적으로 증가시키면서 반지름 r에 위치한 픽셀 값들의 합을 나타낸다.Fig. 5 is a schematic diagram showing the geometrical meaning of the ring projection. Fig. As shown in Fig. 5,

Represents the sum of the pixel values located at the radius r while gradually increasing the radius r from the center point of the template image of the size.

중 작은 값이고, α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀의 총 개수를 M이라 하고,

는 영상의 좌표에서의 픽셀 값을 나타내고, x 및 y는 픽셀의 좌표값이고, w 및 h는 템플릿 영상의 가로 및 세로의 크기이고, W 및 H는 장면 영상의 가로 및 세로의 크기이고, t는 반지름의 길이이다.[Equation 5] is a formula for obtaining a vector in a template image and [Equation 6] is a formula for obtaining a vector in a scene image. At this time, λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

W and h are the horizontal and vertical sizes of the template image, W and H are the horizontal and vertical sizes of the scene image, t and t are the pixel values in the coordinates of the image, x and y are the coordinate values of the pixel, Is the length of the radius.

크기의 템플릿 영상의 중심에서 최대 원의 픽셀에 해당되는 픽셀과의 직선상에 있는 픽셀의 합이다.6 is a schematic diagram showing the geometrical meaning of Radial Projection.

Is the sum of the pixels on the straight line with the pixel corresponding to the pixel of the maximum circle at the center of the template image of the size.

중 작은 값이다. 이때 α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀을 잇는 선분과 x축과의 각도이다. α는 중심으로부터 반지름이 R인 위치에 해당하는 픽셀의 총 개수를 M이라 하면 2π를 M으로 샘플링을 하여

를 구한다.

는 영상의 좌표에서의 픽셀 값이다. x 및 y는 픽셀의 좌표값이고, w 및 h는 템플릿 영상의 가로 및 세로의 크기이고, W 및 H는 장면 영상의 가로 및 세로의 크기이고, t는 반지름의 길이이다.Equation (7) is a formula for obtaining a vector from a template image and Equation (8) is an equation for obtaining a vector from a scene image. λ is equal to the maximum radius R

Which is a small value. Where alpha is the angle between the x-axis and the line segment connecting the pixel corresponding to the radius R from the center. If the total number of pixels corresponding to the position where the radius is R from the center is M, 2? is sampled by M

.

Is the pixel value in the coordinates of the image. w and h are the width and height of the template image, W and H are the width and height of the scene image, and t is the length of the radius.

Embodiments of the present invention are robust recognition systems and methods among block-based object recognition methods used in factory automation environments. In the factory automation environment, the camera's viewpoint and lighting are fixed and only certain kinds of specific parts are recognized, so focusing on robust recognition algorithms that are more rotation and shadowing than scaling and brightness-based recognition algorithms.

2, step (a) is a step of photographing and inputting an image on a factory automation line using the camera by the input unit 100, and step (b) Since the background in which the object and the object to be recognized in the environment are not frequently changed is a step of registering the background image and the object to be recognized in advance and comparing the image with the inputted image to designate or set the place where the object exists as the region of interest .

FIG. 7 is a diagram illustrating an area-of-interest setting applied to a factory automation object recognizing method according to an embodiment of the present invention. FIG. 7A is an example image of a background screen when there is no object, and FIG. 7B is an example image in which a region of interest is designated using a background image in a scene image.

block-based matching is performed between the scene image and the template image after steps (a) and (b). That is, in step (c), the ring projecting unit 300 performs matching only on the region of interest, and using the ring projection vector {circumflex over And calculates the correlation using Equation (3). At this time, only the pixels having the calculated correlation value equal to or higher than the first threshold value (Threshold 1) are extracted as candidates.

As described above, in the embodiment of the present invention, there are various methods for extracting the vector, but the reason for extracting the vector using the ring projection is that the method is robust to rotation. The above method can recognize a high performance even if an object is rotated in a scene image.

If the first threshold Threshold 1 is set too high, it is not included in the candidate group if there is occlusion, so that we are not included in the candidate group even if there is occlusion. It is preferable to set the first threshold value (Threshold 1).

8 is an image showing an example of setting the first threshold value (Threshold 1). 8A is an example image in which the first threshold Threshold 1 is set too high. If there is no obstructed object, you can find the exact center of the object. It is an image that can not recognize the object when the object is covered.

8B is an example image in which the value of the first threshold value (Threshold 1) is appropriately set. Even if the object is covered, the center point of the object should be extracted as the candidate group. As shown in FIGS. 8A and 8B, when only a ring projection vector is used, the rotated object is correctly recognized. However, in order to recognize an obstructed object, many parts other than the center point of the object to be recognized are recognized. Therefore, it can be seen that it is inappropriate to perform recognition using only a ring projection vector.

(d) is a step of extracting a radiation projection vector from the radiation projection unit 400. In step (c), a radiation projection vector is extracted using a radial projection for a pixel extracted as a candidate group , step (e) is a step of calculating the correlation through the radial projection vector extracted in step (d) and determining an object. At this time, a vector obtained by dividing the radial projection vector into four parts in the scene image is obtained in order to increase the recognition rate for the obscured object. At this time, if one part is larger than the second threshold value (Threshold 2), it is judged as an object.

FIG. 9 is a flowchart illustrating more specifically steps (d) and (e) in a factory automation object recognizing method robust to obstruction and rotation according to an embodiment of the present invention, ~ d. < / RTI >

As shown in Fig. 9, first, a P_t vector, which is a radiation projection vector of the template image, is obtained by using [Equation 7], and the radiation projection vectors P_s_a, P_s_b, P_s_c, and P_s_d are obtained by dividing the pixels included in the maximum radius into four regions a to d to extract a radial projection vector.

, b 부분의 α 범위를

, c 부분의 α범위를

, d 부분의 α범위를

로 하여, P_s_a, P_s_b, P_s_c, P_s_d 벡터를 추출한다.For a scene image, a vector is extracted using [Equation 8]. At this time,

, the alpha range of the part b

, the alpha range of the c portion

, and the alpha range of the d portion

And extracts the P_s_a, P_s_b, P_s_c, and P_s_d vectors.

The initial values of Max_a, Max_b, Max_c, Max_d, and I are 0, and the value of m is the number of pixels included in the maximum radius in the template image. The object discrimination is determined as an object when any one of Max_a, Max_b, Max_c and Max_d is greater than the second threshold (Threshold 2).

11A and 11B illustrate an example of a template (FIG. 11A) and a background image (FIG. 11B) applied to an embodiment of the present invention. FIG. FIG. 13 is an example of an image obtained by extracting a candidate region using a ring projection in a region of interest according to an embodiment of the present invention. FIG. An example of an image showing the result of ultimate determination of an object using Radial Projection in the candidate group according to the example.

Table 1 shows the result of recognizing an object in Fig. The objects were tested with occlusion between 0 and 75% and all were correctly recognized. Also, it can be seen that both are closed and rotated at the same time.

And, as another embodiment of the present invention, a computer program stored in a computer-readable medium may be stored in a computer-readable medium in combination with hardware in order to execute the above-described occlusion and rotation robust factory automation object recognition methods.

That is, the apparatus according to the embodiment of the present invention can be implemented as a computer-readable code on a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the recording medium include ROM, RAM, optical disk, magnetic tape, floppy disk, hard disk, nonvolatile memory and the like. The computer-readable recording medium may also be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner.

The embodiments and the accompanying drawings described in the present specification are merely illustrative of some of the technical ideas included in the present invention. Accordingly, the embodiments disclosed herein are for the purpose of describing rather than limiting the technical spirit of the present invention, and it is apparent that the scope of the technical idea of the present invention is not limited by these embodiments. It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

100: Input unit 200: Interest area setting unit
300: Ring projecting part 400: Radial projection part
500: Object judging unit

Claims (16)

An input unit for photographing an object placed on a factory automation line and acquiring a scene image; A region of interest setting unit for setting a region of interest by comparing the scene image with a previously input background image; A ring projection unit for extracting a candidate group corresponding to a value having a correlation of at least a predetermined first threshold value using a ring projection for a set ROI; A radiation projection unit for extracting a projection vector from a template image, dividing a scene image from the extracted candidate group into quadrants and extracting a projection vector from each divided region; And an object determination unit for calculating a correlation through the extracted template image and the radiation projection vector of each region, and for comparing the correlations and recognizing the object,
The formula for calculating the degree of correlation,

(Where NCC denotes Covariance, denotes a covariance, a and b denote vectors, and E (a) and E (b) denote the average values of the elements a and b).
And the rotation angle of the factory automation object recognition system. The method according to claim 1,
Wherein the ROI setting unit comprises:
A reference image storage unit for previously registering and storing a background image and a template image of a target object; And
And a setting unit configured to set a region of interest in the target object region by comparing the stored background image with a scene image input from the input unit. The method according to claim 1,
The ring-
A ring vector extracting unit for extracting a ring projection vector from the scene image and the template image;
A correlation calculating unit for calculating a correlation between the scene image and the template image using the ring projection vector; And
And a candidate group extracting unit for extracting only candidate pixels having a calculated correlation degree equal to or greater than a predetermined first threshold value as candidate groups. The method of claim 3,
The equation for calculating the ring projection vector is:

,

(Where, w and h are the width and height of the template image, W and H are the width and height of the scene image, and R is the maximum radius,

Lt; / RTI >

And a pixel value in a coordinate of the image is expressed by the following equation: < EMI ID = 1.0 > delete The method according to claim 1,
The expression for extracting the projection vector from the template image is:

(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

Where x and y are the coordinate values of the pixel, w and h are the width and height of the template image, and t is the length of the radius). A factory automation object recognition system that features robustness against obstruction and rotation. The method according to claim 1,
The expression for extracting the projection vector into each of the above-

(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

W and h are the horizontal and vertical sizes of the template image, W and H are the horizontal and vertical sizes of the scene image, t and t are the pixel values in the coordinates of the image, x and y are the coordinate values of the pixel, Is a length of a radius of the object to be inspected. The method according to claim 1,
The object determining unit
A correlation degree calculation unit for calculating a correlation between the radiation projection vector of the extracted template image and the radiation projection vector in each divided region; And
And an object recognition unit for recognizing the object as an object when the calculated correlation value of each area is larger than a preset second mterature value. (a) capturing and inputting a scene image of a target object placed on a factory automation line of an input unit; (b) comparing the scene image with the background image pre-input by the ROI setting unit to set the ROI; (c) extracting a candidate group whose degree of correlation is equal to or greater than a predetermined first threshold value by using a ring projection for a region of interest set in the ring projecting unit; (d) extracting a radiation projection vector from the radiation projection part template image, dividing the scene image from the extracted candidate group, and extracting a radiation projection vector from each divided area; And (e) calculating a degree of correlation through a template image extracted by the object determining unit and a projection vector of each region, and comparing the degrees of correlation to determine whether the object is an object,
The formula for calculating the degree of correlation,

(Where NCC denotes a covariance, covariance denotes a covariance, a and b denote a vector, and E (a) and E (b) denote an average value of the elements a and b) A factory automation object recognition method that is robust to obscuration and rotation. The method of claim 9,
The step (c)
(c1) extracting a ring projection vector from the scene image and the template image; And
(c2) calculating a correlation between the scene image and the template image using the ring projection vector; And
(c3) extracting only the pixels having the calculated degree of correlation equal to or greater than a predetermined first threshold value as a candidate group. The method of claim 10,
The equation for calculating the ring projection vector in the step (c2)

,

(Where, w and h are the width and height of the template image, W and H are the width and height of the scene image, and R is the maximum radius,

Lt; / RTI >

And a pixel value in a coordinate of the image. The method of recognizing a factory automation object is robust against obstruction and rotation. delete The method of claim 9,
The expression for extracting the projection vector from the template image is:

(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

Where x and y are the coordinate values of the pixel, w and h are the width and height of the template image, and t is the length of the radius). Characterized by a tangible and rotationally robust factory automation object recognition method. The method of claim 9,
The expression for extracting the projection vector into each of the above-

(λ is equal to the maximum radius R

And a is a total number of pixels corresponding to a position where the radius is R from the center is M,

W and h are the horizontal and vertical sizes of the template image, W and H are the horizontal and vertical sizes of the scene image, t and t are the pixel values in the coordinates of the image, x and y are the coordinate values of the pixel, Is a length of a radius), and a rotation-resistant factory automation object recognition method. The method of claim 9,
The step (e)
Calculating a correlation between the projection vector of the extracted template image and the projection vector of each divided region; And
Recognizing the object as a target object when the calculated correlation value of each area is larger than a preset second mterter value. Combined with hardware,
A computer program stored on a computer readable medium for executing the method of claim 9 and a method for recognizing factory automation objects robust to rotation and rotation.

Images