Camera system for colour shade sorting and black spot inspection, for a recycled paper sheet manufacturer

How the brief came about

This system uses a camera to sort recycled paper sheets by colour shade and to assess sheet cleanliness from black spots on the surface. The task was to prove, before investing, that a camera could do these 2 tasks on real workpieces. The recycled paper sheet manufacturer wanted to separate sheets with unsuitable brightness from good ones, and to measure cleanliness from the number and size of black spots or foreign matter on the paper surface.

Colour and cleanliness are what the end customer sees immediately on receiving the paper sheets. When the sorting criteria depend on an inspector's eye, it is hard to state as a repeatable criterion which sheets pass and which do not. So we began by testing a prototype with the customer's sample workpieces in 2016.

Requirements and constraints

The prototype had to meet the customer's 2 requirements and be tested with sample workpieces on a test cabinet. The main requirements and constraints are as follows.

• Inspect the sheet's brightness (colour shade) to separate sheets with suitable brightness from unsuitable ones
• Assess sheet cleanliness from the number and size of black spots and plastic marks on the surface
• Use the customer's sample workpieces in 2 groups, each group with its own prototype workpiece (Master)
• Test on a test cabinet with controllable lighting; lighting conditions and the installation position in the factory must be surveyed separately before installation

How does this system sort colour shades and inspect black spots?

This prototype captures an image of each paper sheet, one at a time, under brightness-controlled lighting, then uses image processing to compare the sheet's colour shade against the group's prototype workpiece (Master) and to count black spots and plastic marks on the surface. The result for each sheet is shown as OK (pass) or NG (fail).

1. 1

   Control the lighting

   2 bar lights illuminate from either side of the lens and workpiece. The Lighting Control unit adjusts the light brightness to the set value.

2. 2

   Capture images of the paper sheet

   A Cube 5MP camera mounted on the test cabinet captures images of the recycled paper sheet placed below.

3. 3

   Set up and select the prototype workpiece

   The user sets sharpness, colour tone and brightness on the program's Setting page, then assigns a Master for each workpiece group as the reference.

4. 4

   Compare shades and inspect black spots

   The Medusa processing unit compares the sheet's shade with the Master and shows OK in green or NG in red, and also checks the number and size of black spots and plastic marks on the sheet.

We chose to compare against each group's Master because the customer's task was to separate sheets that differ from that group's reference sheet. In colour shade comparison, the brightness the camera sees also depends on the lighting, so the test rig controls the bar lights with a Lighting Control unit.

What we built and tested

We built a test rig and a prototype program, then tested them with the customer's sample workpieces. The details are as follows.

• Test rig: Cube 5MP camera on a test cabinet, 2 bar lights with a Lighting Control unit, and a Medusa processing unit
• Prototype program with Login, Process and Setting pages
• Testing with the customer's sample workpieces, covering both colour shade comparison and inspection of black spots and plastic marks
• Test report and the design of the full system: a report sorting good and defective work, and Manual and Automatic modes

Details of test results on real workpieces

In the test with the customer's 2 groups of sample workpieces, the program reported that 1 sheet's shade did not match the group's Master, while the other sheets were reported as matching their own group's Master. The cleanliness inspection found many plastic marks and black spots on sheets in the same group. This result comes from a test cabinet with controlled lighting. The documents do not separately confirm that the sheet's shade was actually wrong, and results on a production line must be measured again under on-site lighting.

The design target for inspection time is about 3 seconds per sheet in Manual mode (the user controls the inspection) and about 1 second per sheet in Automatic mode (the system inspects automatically). Both values must be measured again against the actual production line speed.

What we learned

This work gives 3 principles for camera-based colour shade and cleanliness inspection of flat sheet workpieces.

• Control the light before comparing colour. The brightness the camera sees also depends on the lighting, so a shade comparison can be trusted when the lighting is stable and adjustable, not merely when the camera has high resolution.
• Set criteria from an agreed reference workpiece. Each workpiece group should have its own Master, and before actual use the Master should be a sheet the customer accepts, so that an OK or NG result means the same thing every time.
• A result in a test cabinet is not yet a result on a production line. Before actual installation, the lighting conditions and installation position in the factory must be surveyed, and inspection time per sheet must be measured against the actual line speed.

How this approach can apply to your factory

If your factory needs to sort flat sheet workpieces by colour shade or brightness, or wants cleanliness criteria based on black spots instead of visual judgement, you can use the same approach as this prototype: control the lighting, compare against each group's Master, and count spots on the surface. What needs adjusting is the lighting and the decision criteria, when your workpiece or site differs from this work.

• Prepare sample workpieces, both good and defective sheets, along with a Master for each group, for testing and setting criteria
• Choose between a mode the user controls and an automatic mode, then compare inspection time per sheet with the production line speed
• Plan how sorting results will be passed on, for example using a report for people to remove sheets, or connecting to the production line's sorting mechanism