A brief analysis of industrial robot visual positioning technology

Robot integrates electronic technology, sensing technology and intelligent control technology. It is a machine device that can automatically perform work tasks. It can accept human command and can also act independently in accordance with the principles and programs formulated by artificial intelligence technology. It has been applied in many fields.

At present, industrial robots can only perform predetermined instructions in a strictly defined structured environment and lack the ability to perceive and adapt to the environment, which greatly limits the application of robots. The workpiece visual positioning method, combined with dedicated image processing software, utilizes the robot's visual control without the need for pre-teaching or offline programming of the industrial robot's motion trajectory. It can achieve reliable positioning of the industrial robot's visual system and play an active role in improving workpiece positioning accuracy and processing effects. It can save a lot of programming time and improve production efficiency and processing quality. In China, this aspect is mainly used in welding robot tracking of weld seams.

The robot visual positioning system is composed of (as shown in the figure below). A spraying tool and a single camera are installed at the end of the articulated robot so that the workpiece can completely appear in the camera's image. The system includes camera system and control system:

1. (1) Camera system:

   It consists of a single camera, a computer and a collection system (including an image capture card), responsible for the collection of visual images and machine vision algorithms; it is recommended to use a digital camera for this system, and the extraction accuracy will be higher than that of a general camera.

2. (2) Control system:

   composed of a computer and a control box, used to control the actual position of the robot end;

The work area is photographed by a CCD camera, and the computer extracts tracking features through image recognition methods for data recognition and calculation.

The American TEO brand digital camera TM-C6597E is used, equipped with a TM-C520HP image acquisition system. It is designed with a dedicated camera, balanced transmission line, and image acquisition card to input the video signal into the calculator and process it quickly. First, select a local image of the object being tracked. This step is equivalent to the process of offline learning, establishing a coordinate system in the image and training the system to find the tracking object. After learning, the image card continuously collects images, extracts tracking features, performs data recognition and calculation, solves the given position values ​​of each joint of the robot through inverse kinematics, and finally controls the high-precision end actuator to adjust the position of the robot.

In this way, the visual positioning system combines area-based matching and shape feature recognition for data recognition and calculation, and can quickly and accurately identify the boundaries and centers of object features. The robot control system obtains the angle error of each joint position of the robot through inverse kinematics solution, and finally controls the high-precision end actuator to adjust the robot's posture to eliminate this error. This solves the problem that the actual position of the robot end is far from the desired position, and improves the positioning accuracy of traditional robots.

The contrast between the workpiece on the workbench and the background of the workbench forms a big difference in color, that is, the workpiece is identified as black, and the center line of the black image is extracted. The acquisition system uses this information as an important feature for identifying the workpiece.

When a general vision system acquires a workpiece image, it cannot be used directly in the vision system due to various conditions and noise interference. Image preprocessing such as grayscale correction and noise filtering must be performed through image analysis and recognition.