Mechatronics, Industrial Control & Instrumentation

Vision Sensors: Features, Concepts & Applications in Manufacturing

Contents

1 Vision Sensors
- 1.1 Vision Systems Concepts
  - 1.1.1 Vision Data Processing Techniques
- 1.2 Application of Vision Sensors in Manufacturing

Vision Sensors

A vision sensor is basically a TV camera connected to a computer, usually a PC. Special vision software analyzes the video image and makes specific recommendation to the work area.

Vision systems require computing power to process thousands of pixels of information continuously in order to arrive at a go/no-go decision about what is seen. Some of the technologies used include those employing powerful microprocessors such as ones using Pentium with MMX technology and the adoption of the PCI expansion bus for PCs. PCI cards can communicate with the microprocessor at much higher data rate than the legacy ISA cards. In most factory operations, vision guided systems have replaced the inflexible multi-sensor systems.

Vision Systems Concepts

The hardware for a vision system essentially consists of an electronic camera connected to a special video processing PCI card referred to as frame grabber. The PCI card is plugged into a Pentium-based PC that is running special vision recognition software. Normally the camera is rigidly mounted so that it is looking down upon the area of interest.

The visual information is passed from the camera to the frame grabber card. The card converts the visual information into digital format which is usually a 640 by 480 pixel frame as illustrated below:

Figure 1.1: A frame (one still picture), usually made up of 640 x 480 pixels.

Each pixel is represented by an 8-bit gray-scale number called the intensity, meaning that each pixel is assigned one of the 255 shades between black and white. Thereafter the entire frame is conveyed to the PC (through the PC’s built-in PCI bus) up to 30 times a second. The PC runs software that continuously analyzes the visual data based on some criterion and then sends the signals back to the work area either to reject a part that has failed inspection or to direct a pick and place robot to move to a certain location to pick up a box.

Vision Data Processing Techniques

There are various ways used to process the visual data. Any one of the technique applied depends on the particular application. Some of the vision data processing techniques include:

Edge detection.
Pattern recognition

Edge Detection Technique

This is one of the most used techniques for processing visual data. With edge detection technique, the computer attempts to define the edges of the object, an example of this process is illustrated below:

Figure 1.2: Computer uses edge detection to form an outline of the wrench.

The computer scans the entire frame looking for edges; it detects them on the basis of changes in the gray-scale intensity. An edge is where there is a sharp intensity discontinuity. It is detected on the basis of three parameters: contrast, width and steepness as demonstrated below:

Figure 1.3: The computer defines the edge on the basis of contrast, width and steepness.

Contrast is the numeric differences in intensity between two adjacent areas of the frame; steepness is how quickly the intensity changes between the areas; and width is the minimum size of the uniform intensity area on either side of the edge. The actual edge is defined as the steepest place in the intensity curve. The threshold values of contrast, steepness and width that the computer utilizes to attempt to identify the edges can be fine-tuned to get the best performance in each specific application. Furthermore, bright lighting and contrasting backgrounds can be used to highlight the edges as much as possible.

Once the outline of the object is formed within the computer, the dimensions can be checked on the visual image to determine whether the part is acceptable.

Pattern Recognition Technique

This technique is employed when an object needs to be identified. Case in point, in a system being used to sort parts into different bins, the computer maintains a library of the expected part shapes and by using various ‘best-fit’ algorithms, attempts to find the best match for each specific shape. Alternatively, the task of the computer may be to scan the entire visual frame looking for a specific shape, such as a hole pattern, in order to guide a robot to mount a component.

Don’t miss out on key updates, join our newsletter List

Application of Vision Sensors in Manufacturing

Vision sensors are used for instance, in inspections of parts to determine whether parts have been made or assembled correctly; vision systems may be used to discard damaged mangoes from a fruit processing line; also a vision system may be used to guide pick and place robots.

Related articles:

John Mulindi

John Mulindi is an Industrial Instrumentation and Control Professional with a wide range of experience in electrical and electronics, process measurement, control systems and automation. In free time he spends time reading, taking adventure walks and watching football.