Using 3-D Machine Vision to Solve Issues Faced by Production Sites

Robots play an essential role in the manufacturing industry, such as in production lines for automobiles and automobile components. Some tasks, however, prove too difficult even for robots to perform. One such task is the selection of individual parts from randomly piled parts in a box or pallet.

The industrial sector often makes use of relatively large components that are randomly piled. Because workers must first reposition each part for use by robots, efforts to streamline and automate production lines reached a bottleneck.

Canon's 3-D machine vision system can help to resolve this issue. Machine vision refers to the use of industrial image sensors. The most common form of machine vision currently in use is 2-D machine vision, which has difficulty in identifying the positioning and orientation of randomly piled parts. As such, Canon developed the RV1100, a machine vision system capable of the high-speed, high-accuracy three-dimensional recognition of objects. By enabling a robotic arm to pick up individual items which have been randomly piled, the system enables the automation of parts supply on production lines, a task that conventionally had to be performed manually.

3-D Machine Vision System Workflow

• (1)
  Various patterns are projected onto randomly piled parts
• (2)
  The distance between the randomly piled parts and the sensor is measured
• (3)
  A pre-registered pattern dictionary and 3D CAD models are used to recognize part positioning and orientation
• (4)
  The system determines if the robot hand can grasp the part without making contact with other parts
• (5)
  Data is sent to the robot controller unit

Easy Installation, Enhanced Productivity

Enabling the three-dimensional recognition of objects, the RV1100 projects image-recognition patterns on to parts and then analyzes images of the resulting patterns. The system's three-dimensional recognition capabilities are made possible by using multiple projection patterns to analyze the images.
Conventional 3-D machine vision systems often prove difficult to install because of the importance of the positional relationship between the pattern projector and image sensor, which must be carefully calibrated to achieve high levels of precision. Canon's 3-D machine vision system, however, incorporates both the pattern projector and imaging sensor in a single unit, enabling installation without the need for difficult calibration adjustments. The lightweight, compact design enables easy installation without having to modify or move production lines. In addition, it features a dust- and water-resistant body design, making the system easy to maintain once installed.

High-Speed, High-Precision 3-D Recognition of Various Parts

Canon's 3-D machine vision system delivers unrivaled recognition precision. Users can easily register part data into the system by inputting CAD data for components with curved surfaces, components with few distinguishing features, or those with complex shapes, and by capturing images of the parts randomly assembled in a pile. Using a new approach that matches CAD data with not only distance measurement data but also gray-scale images, the system is able to recognize a wide range of parts and enables users to easily register new parts or modify existing parts without the need for complex programming.

The RV1100 combines a 3-D machine vision head with recognition software to perform three-dimensional object recognition, and sends this control data to the robotic arm. Robots produced by major global robotic manufacturers support this control data, delivering a high level of flexibility for integrating the system into existing facilities. The RV1100 conforms to regulations in Japan, the EU and EFTA (European Free Trade Association), the U.S., South Korea, and China, allowing manufacturers to employ the system at their overseas operations as well.

Canon is working to promote sales for the system and has already received inquiries from manufacturers in not only the automobile industry, but also the electronics, metal products, resin and chemical industries. As a leader in the 3-D machine vision market, the company will continue to actively pursue product development, aiming to achieve further improvements that enable the handling of parts that are both smaller and larger in scale than those currently handled by the system.

The RV1100 takes only approximately 2.5 seconds to recognize randomly piled parts and transmit data to the robotic arm's controller unit. The system helps to accelerate the parts supply process