Machine Vision & Automatic Optical Inspection

Where human vision is most effective for qualitative interpretation of complex, unstructured scenes, machine vision excels at quantitative measurements of a structured scene because it can process visual information faster, more accurately, and repeatably. For example, on a production line, a machine vision system can inspect hundreds or even thousands of parts with multiple features and attributes every minute. A machine vision system built with a suitable camera resolution and optics can easily inspect object details too small to be seen by the human eye.

Our automatic inspection systems are configured based on a combination of the following components to a complete system to cater for the requirements of specific applications:

1.     Optics including lighting and lenses

We can configure the optical system from a diverse variety of lenses including fixed and variable focus lenses, telecentric lenses, and zoom lenses. Telecentric lenses are particularly suitable for a variety of applications where measurement of critical dimensions of the target is required. We combine the lens system with a variety of lighting options including back lighting, front lighting, and through the lens lighting. In our labs, we develop the optimal optical solutions for an inspection task first and then integrated the optical system with cameras, and where required, robotic or gantry units. We work with a number of suppliers to obtain the lens and lighting systems which include Sill Optics and CCS America.

2.     Cameras

As for cameras we rely on our partners Dalsa, Sony  and Basler. Therefore, you can rest assured that the components that we use in our machine vision and automatic inspection systems are of the highest quality and are supported by a first class worldwide network.

3.     Software

Our software is based on open source libraries and our proprietary algorithms that we use together to quickly develop programs that will extract image patterns, features, and attributes and compare them with known good ones. The program is optimized to deliver a high through put. As a result, defective parts, features, and attributes can be identified reliably during production for operator action. 

Machine Vision Benefits Include:

  • Higher quality of inspection
  • Accurate measurements & gauging,
  • Verification of Assemblies
  • Increased productivity in Repetitive tasks
  • Higher efficiency of operation
  • Lower production costs
  • Early detection of flaws and defects in the process

Inspection using Machine Vision

  • Identifying defects and irregularities
  • Automatic testing of LCD displays
  • The Challenge of checking for dead pixels on LCD displays:
    • Cycling through all the red, green, and blue pixels followed by white, while checking to spot dead pixels
    • The human eye soon tires of such a task and dead pixels are then missed

Machine Vision Color and Pattern Matching

We can quickly create custom programs to identify defects, erroneous colors, in LED arrays

In the opposite picture, the pattern matches the expectation, so it is a PASS!

In the opposite picture, the color and the pattern of the LEDs does not match the expectation. So it is a FAIL!

Machine Vision Color and Pattern Matching

A Sample pseudo-code for Inspecting Automotive Instrument Cluster

The following steps represent the pseudo-code for executing the inspection tasks.

1.Find the location and orientation of the odometer display. Use edge detection tools to find the left edge (1) and the lower edge (2) of the odometer display area. Depending on the cluster, pattern matching can be used in place of using the edge detection tools. The choice will depend on the type of fiducials available for identification and location.

2.Create a coordinate reference frame based on step 1. This allows subsequent images to relocate the defined Regions Of Interest at the correct locations.


3.Check for the presence of the seatbelt icon (3) and the PS icon (4) using Pattern Matching. Check for their orientation and the distance between them to ensure correct placement of the cluster in its nest.

4.Read the tachometer gauge needle. Find both the edges (5) and (6). Compute the bisector of the two edges to find the center of the needle (7) (Figure 3). The inspection is based on the coordinate reference frame created earlier in step 2. Find the angle of the bisector with respect to the start position of the gauge.

5.Read the speedoometer (7, 8, 9) in the same way as in step 4.

6.Check for the status of the rest of the icons (ON/OFF) (10, 11, 12, 13) by comparing the mean intensity of these against expected values. The location of these is based on the coordinate reference frame created earlier in step 2.