What are the vision applications of construction robots?

With the improvement of the digital level of the construction industry, machine vision will have a wider range of applications in robot positioning, recognition, detection and other fields. In the near future, visual technology will play an increasingly important role not only in the production of building steel structures, but also in complex environments such as construction sites. Let's take a look at the application of machine vision of "fixed construction robot" in construction robots. 

1. Robot vision system based on construction robot

Application occasions:

(1) Building steel structure factory

Welding robot: sheet metal processing, sheet metal loading and unloading, workpiece assembly, welding seam tracking, welding seam quality inspection

Painting robot: surface treatment of welding parts, position identification of painting workpieces, painting quality inspection

(2) Building site construction

Wall plastering robot: recognition of plastering wall position, construction quality inspection

Floor paving robot: floor tile position detection

A large number of field applications of construction robots require precise positioning of processing or construction objects. However, existing industrial robots can only perform predetermined command actions in a strictly defined structured environment, and lack the ability to perceive and respond to the environment, which greatly limits The application of robots.

Robot vision technology greatly improves the actual work efficiency of the robot, reduces or even eliminates the link of teaching or offline programming of the robot's motion trajectory, thereby saving a lot of programming time and improving production efficiency and production quality.

2. Application of construction welding robot:

There are two just-needed applications of robot vision in the field of construction robot welding

(1) The existing construction robot welding requires manual programming and teaching to ensure the trajectory of the robot welding. There are many varieties of components and small batches. Each time the product is replaced, it is necessary to change the fixture, re-teaching, and the final programming plan can be determined through multiple programming, which seriously affects the production efficiency.

(2) During the welding process, due to the large dimensional tolerances of the workpiece, the dimensional error of the tooling itself, or the deformation caused by the welding thermal stress, the actual weld trajectory and the programmed trajectory will be different, and the welding robot cannot recognize and correct the difference, which leads to There is a difference in welding quality.

The R&D team for the industrialization of fixed buildings has developed a construction robot intelligent welding system for prefabricated building welding equipment, which solves the technical problem of manual programming of welding robots without significantly increasing costs, and opens up the last part of welding robot substitution. Kilometers.

The intelligent welding system of building welding robot is composed of the following three subsystems

(1) Initial welding position recognition and guidance subsystem

The system uses a vision sensor to take images of the weldment in the work space. Through image processing and stereo matching, it extracts the coordinates of the initial point of the weld in the three-dimensional space, and transmits the result to the central control computer. The server controls the robot’s The welding gun automatically moves to the initial welding position to prepare for welding.

(2) The 2D camera is mainly used to quickly identify the start and end of the weld

Based on 2D vision technology, the staff can quickly obtain the X/Y axis coordinates of the starting point and the end point through a graphical interface. Through algorithms such as deep learning, the system can automatically identify and highlight the actual welds for the operator to select. Automatic tracking uses the pre-set welding seam starting point and other information to start the welding seam tracking/tracking dual-purpose camera, and through algorithm control, guide the manipulator to carry the welding gun to the ups and downs that can accurately start the welding process. Next, the system comprehensively uses neural network prediction, high-efficiency filtering, noise elimination and shape adaptive control algorithms, and the robot asynchronously seeks and determines the spatial information of the seam to be welded, and gives it based on the empirical parameters that have been set in the program. The best welding angle and robot welding torch motion trajectory are determined. The automatic tracing process (without manual intervention) is completed, and the robot itself generates the spatial weld trajectory that can only be obtained by traditional manual teaching.

3. Seam tracking subsystem based on vision sensing

Follow the previous step of work, take the image information of the weld position, shape and direction of the workpiece to be welded, and then extract the weld shape and direction features through a specially designed image processing algorithm, and determine the next approach or direction of the welding gun according to the weld position. Correct the movement direction and displacement, and then start the welding seam tracking calculation program, and drive the robot body to move the welding gun endpoint to track the welding seam direction and position correction through the central control machine and the robot control. In this way, the welding path can be adjusted in real time to ensure welding quality.

4. Real-time control subsystem of weld penetration based on visual sensing

Using a camera installed at the rear of the robot's welding torch, under the irradiation of the welding arc, an image of the half of the molten pool in the direction of the robot's movement is obtained. The algorithm extracts the shape features of the molten pool, such as width, half-length, area, shape feature information, etc. Based on this information, the central control machine combines the corresponding process parameters and the pre-established welding pool dynamic process model to predict the welding quality parameters such as penetration depth, penetration, penetration width and reinforcement. Invoke appropriate control strategies to provide proper welding parameter adjustments and robot movement speed, posture, and wire feeder speed adjustment changes, which are executed by the welding power supply and robot body and other institutions to realize real-time monitoring of the dynamic characteristics of the welding pool, and Intelligent control of penetration and weld forming quality.

In summary, the "Intelligent Welding System for Construction Robots" developed based on vision technology uses a large number of computer vision images and artificial intelligence technology. Through the extraction and analysis of welding groove features, as well as the changes in the four dimensions of the welding seam in three-dimensional position space and time space, combined with different welding processes, the real-time pose and pose of the welding machine and robot during the welding process are automatically planned. Movement trajectory. It can completely replace the work of engineers teaching programming or offline programming. In actual use, it will exert huge economic benefits.