KR20230108905A - Method of automatically write block shape and dimension information for robot OLP - Google Patents

Abstract

The shape information is created by learning the photo information of the block (U-cell), which is the work object acquired through the 3D image input device, by deep learning, and the dimension information is created using the coordinate information of the feature points of the shape. It relates to a method for automatically writing block shape and dimension information for a robot OLP that enables automatic writing of shape information and dimension information into robot OLP (Off-Line Programming), Acquiring image information, extracting learning data for shape information generation from the acquired image information, generating shape information by learning the extracted learning data through deep learning, and extracting feature point coordinates, which are location information, from image information of a work object and, based on the extracted location information, generate dimension information of the weld line, automatically write the shape information and dimension information of the created work object into the OLP, and allow the user to manually transfer the block information to the robot OLP using a teaching manipulator or the like. Reduce human error by minimizing input with .

Description

Method of automatically write block shape and dimension information for robot OLP {Method of automatically write block shape and dimension information for robot OLP}

The present invention relates to a method for automatically writing block shape and dimension information for a robot OLP. In particular, shape information is obtained by learning photo information of a block (U-cell), which is a work object, obtained through a 3D image input device by deep learning. block shape for robot OLP, which can be automatically written into robot OLP (Off-Line Programming) after creating dimension information using coordinate information on feature points It relates to a method of automatically entering dimension information.

Welding work, which accounts for most of the production in the shipbuilding field, is mainly performed in straight-line welding sections, but as the volume of complex hull structures increases recently, new technologies for automatic welding are required.

In particular, in the case of automatically welding a double hull or curved block (U-cell) with a robot, it can be classified into about 9 types, but drawing information must be provided to the robot (robot OLP) according to the factory situation.

As a method of providing block information (shape information, dimension information) to the robot OLP, an operator inputs dimension information generated from CAD information or block information generated by a user measuring a drawing or real object through a teaching manipulator. .

In this case, a case in which an accurate welding line may not be found due to a mismatch of CAD information or a user's mistake may occur, which may result in a welding operation defect.

Therefore, when a worker creates block information through drawing or physical measurement, it solves the problem of inaccuracy of information due to inaccurate tracking of welding lines and minimizes the work of manually inputting block information using a teaching manipulator, etc., resulting in dormant errors. We need a way to reduce

Korean Registered Patent No. 10-2127951 (registered on June 23, 2020) (multi-process based real-time welding robot control system and its control method) Republic of Korea Patent Publication 10-2017-0142379 (published on December 28, 2017) (welding line recognition device for welding robot using image processing)

Therefore, the present invention is proposed to solve various problems that occur when the above song block information is input to the robot OLP, and the photo information of the block (U-cell), which is a work object, obtained through a 3D image input device. After learning by deep learning to create shape information, and using coordinate information for feature points of the shape to create dimension information, the created shape and dimension information can be automatically written into the robot OLP (Off-Line Programming). The purpose of this study is to provide a method for automatically writing block shape and dimension information for a robot OLP.

In order to achieve the above object, the "method for automatically writing block shape and dimension information for robot OLP" according to the present invention,

(a) acquiring image information of a work object using a 3D image input device;

(b) extracting learning data for generating shape information from the acquired image information and generating shape information by learning the learning data through deep learning;

(c) extracting coordinates of feature points, which are location information, from the image information of the work object, and generating dimensional information of a weld line based on the extracted location information; and

(d) automatically writing the generated shape information and dimension information of the work object into the OLP.

Step (a) above is characterized in that image information of the work object is obtained using a 3D vision camera or a 3D scanner.

In the above step (b),

It is characterized in that shape information is extracted by applying an image classification technique to the obtained image information of the work object.

In the above step (b),

It is characterized in that the extracted shape information is set as learning data, and shape information is generated by learning the set learning data by deep learning.

In the above step (c),

Characterized in that feature points of the shape of the work object are extracted using a 3D camera, and coordinates of the extracted feature points are acquired.

In the above step (c),

Using the coordinates of the acquired feature points as location information, the shape including at least one of Longi width, Longi height, flange width, flange thickness, scallop height, trans thickness, and scallop size of each weld line part It is characterized in that the dimension information is generated by calculating the length.

According to the present invention, shape information is generated by learning image information of a block (U-cell), which is a work object, by deep learning, and dimension information is generated using coordinate information of feature points of the shape, and then the generated shape information and dimensions By automatically writing information into the robot OLP (Off-Line Programming), there is an effect of reducing human errors by minimizing the user's manual input of block information into the robot OLP using a teaching manipulator, etc. .

In addition, by automatically generating block shape information and dimension information, there is an effect of solving the problem of inconsistency of CAD information and the problem of not finding an accurate welding line due to user's mistake.

1 is a schematic configuration diagram of a welding robot control system to which the present invention is applied;
2 is a flowchart showing a method of automatically writing block shape and dimension information for a robot OLP according to the present invention;
3 is an example of generating block formation information by deep learning in the present invention;
4 is an exemplary view of generating block dimension information in the present invention;
5 is an exemplary view of automatically writing block shape information and dimension information generated in the present invention into a robot OLP.

Hereinafter, a method of automatically writing block shape and dimension information for a robot OLP according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

The terms or words used in the present invention described below should not be construed as being limited to a conventional or dictionary meaning, and the inventor may appropriately define the concept of the term in order to explain his/her invention in the best way. It should be interpreted as a meaning and concept consistent with the technical spirit of the present invention based on the principle that it can be.

Therefore, the embodiments described in this specification and the configurations shown in the drawings are only preferred embodiments of the present invention, and do not represent all of the technical spirit of the present invention, so various equivalents and equivalents that can replace them at the time of the present application It should be understood that variations may exist.

1 is a schematic configuration diagram of a welding robot control system to which the present invention is applied.

As shown in this, it is composed of a CPU and a network card, and is connected to the OLP server 1 and the monitor PC 2 through a network to receive a user's welding robot control command and control the welding robot (3). ), a system I/O module, and a servo driver, and is connected to the upper robot controller 3 to control the multi-joint welding robot 8 under the control of the upper controller 3 and transfer the control status information of the welding robot to the upper level. It includes a robot controller (5) consisting of a lower controller (4) communicating with the controller (3).

Here, the upper controller 3, the lower controller 4, and each module of the lower controller 4 are connected through a network.

The upper controller 3 and the teaching manipulator are connected through Ethernet and receive user input.

In addition, OLP data required for welding work is supplied through a separate OLP server (1) through a computer network, and the operation status of the robot is configured to be checked on an office PC through a monitoring PC (2).

The lower controller 4 is connected to the welding machine 7, the touch sensor 6, and the articulated welding robot 8, transmits a welding command to the welder 7, and sends an operation command to the articulated welding robot 7. will be conveyed

OLP data required for welding and for robot control are automatically generated in the same manner as shown in FIG. 2 .

The OLP data generated in this way can be generated in any part, such as an upper controller, a lower controller, an OLP server, a monitor PC, or a robot, by using the algorithm shown in FIG. 2 .

Therefore, the method of automatically writing block shape and dimension information for the robot OLP of the present invention can be implemented in any part that implements the welding robot control system, and is not specifically limited to any part in the present invention. OLP data can be created in the optimized part by the operator who operates the welding robot control system.

Figure 2 is a flowchart showing the "method for automatically writing block shape and dimension information for robot OLP" according to the present invention, (a) obtaining image information of a work object using a 3D image input device (S10); (b) extracting learning data for generating shape information from the acquired image information and generating shape information by learning the learning data through deep learning (S20 - S40); (c) the work object Extracting feature point coordinates, which are location information, from image information of and generating dimension information of the weld line based on the extracted location information (S50-S60), and (d) shape information and dimension information of the created work object. It may include automatically writing in the OLP (S70).

The detailed description of the "method for automatically writing block shape and dimension information for robot OLP" according to the present invention configured as described above is as follows.

First, image information (photo information) of the work object is obtained by photographing the work object (u-cell) using a 3D vision camera, a 3D scanner, etc. (S10).

Here, it is preferable to obtain image information by photographing the work object from various angles.

Subsequently, learning data for generating shape information is extracted from the obtained image information (S20). That is, shape information is extracted by applying an image classification technique to the acquired image information of the work object (S20).

Here, image classification is a known image classification technique that makes it easy to understand the content of an image by analyzing the characteristics of each pixel and grouping them into one of several types. As a numerical analysis of a numerical image including one or more spatial and visual spectral band and frequency band relationships including a category of information on a particular characteristic, it is a well-known technique in the image processing field, so a detailed description thereof will be omitted. do.

Next, the extracted shape information is set as learning data, and shape information is generated by learning with a deep learning algorithm (S30 to S40).

3 is an example of generating shape information data of a U-Cell, that is, a right-hand room, a scallop, a left-hand room, or a scallop, by learning the extracted shape information with a CNN algorithm, which is a deep learning algorithm.

Here, the CNN algorithm, which is one of the deep learning algorithms, classifies the input image in the convolutional layer to classify the image, processes the classified image information in the fully connected layer, and outputs the shape information data of the work object (U-Cell).

Using this deep learning algorithm, the shape information of the work object is automatically generated.

In addition, feature point coordinates, which are position information, are extracted from the image information of the work object captured by the 3D camera, etc., and dimensional information of the weld line is automatically generated based on the extracted position information (S50 to S60).

4 obtains coordinates for feature points based on location information from image information of a work object photographed by a 3D camera, and calculates dimension information for each weld line using the obtained coordinate information of feature points to obtain dimension information. generate Here, the dimension information about the shape is dimensional information by calculating the length of the shape including at least one of Longi width, Longi height, flange width, flange thickness, scallop height, trans thickness, and scallop size. generate

Subsequently, the generated shape information and dimension information of the work object are automatically written in the OLP (S70). 5 is an example of automatically writing generated shape information and dimension information into an OLP. That is, the generated shape information is automatically inserted into the shape information entry field, and each generated dimension information is automatically inserted into the dimension information input field. To this end, it is possible to search for an entry field by using item character recognition or location recognition of an OLP data sheet, and automatically fill in dimension information determined and generated in the searched entry field.

According to the present invention described above, shape information is generated by learning image information of a block (U-cell), which is a work object, by deep learning, and dimension information is generated using coordinate information of feature points of the shape. By automatically writing information and dimension information into the robot OLP (Off-Line Programming), it is possible to reduce human errors by minimizing the need for a user to manually input block information into the robot OLP using a teaching manipulator or the like. .

In addition, by automatically generating block shape information and dimension information, it is possible to solve the problem of inconsistency of CAD information and the problem of not finding an accurate welding line due to user's mistake.

Although the invention made by the present inventors has been specifically described according to the above embodiments, the present invention is not limited to the above embodiments, and it is common knowledge in the art that various changes can be made without departing from the gist of the present invention. It is self-evident to those who have

1: OLP server
2: Monitor PC
3: Upper robot controller
4: Sub controller
5: Robot controller
6: touch sensor
7: Welder
8: articulated welding robot

Claims (6)

(a) acquiring image information of a work object using a 3D image input device;
(b) extracting learning data for generating shape information from the obtained image information and generating shape information by learning the extracted learning data through deep learning;
(c) extracting coordinates of feature points, which are location information, from the image information of the work object, and generating dimensional information of a weld line based on the extracted location information; and
(d) Automatically writing block shape and dimension information for a robot OLP, comprising the step of automatically writing the generated shape information and dimension information of the work object into the OLP.
The method of claim 1, wherein the step (a) acquires image information of a work object using a 3D vision camera or a 3D scanner.
In claim 1, the step (b),
A method of automatically writing block shape and dimension information for a robot OLP, characterized in that for extracting shape information by applying an image classification technique to the obtained image information of the work object.
In claim 3, the step (b),
A method for automatically writing block shape and dimension information for a robot OLP, characterized in that the extracted shape information is set as learning data, and shape information is generated by learning the set learning data by deep learning.
In claim 1, the step (c),
A method for automatically writing block shape and dimension information for a robot OLP, characterized by extracting feature points for the shape of a work object using a three-dimensional camera and acquiring coordinates of the extracted feature points.
In claim 5, the step (c),
A shape including at least one of the Longi width, Longi height, flange width, flange thickness, scallop height, trans thickness, and scallop size of each weld line using the coordinates of the acquired feature points as location information A block shape and dimensional information automatic writing method for a robot OLP, characterized in that for generating dimensional information by calculating the length of.

Images