Feasibility Vision Solution

Vision inspection problems can be classified in 2 major classes:
- Easy and difficult-
Needless to say, this is an open door!

Easy vision inspection systems

inca311_white The easy class contains inspections:
- That are solvable with a back light (with or without flashlight);
- That are not very accurate; for example inspections concerning pixels accuracy;
- That have enough time to do the processing.
pictor Of course it is necessary to take care of all the inspection or measurement disturbance aspects e.g. vibration (flash light), indirect and unwanted light reflections, position inaccuracy of the product and bad reproducibility of the feature to detect or measure.
For these in principle easy vision problems a programmable camera e.g. Vision & Control
Sensor Partners BV or Philips AppTech can do the job.

In these cases a vision library with a minimum of mathematics is sufficient.
These programmable cameras can also communicate with production control systems as PLC’s and trigger the flash light.
These cameras are designed to be programmed by engineers with a minimum of vision knowledge.

Difficult vision inspection systems

The difficult class contains inspection were high accuracy and reliability are needed. This class also contains applications were a front light is used or a more complex situation with more than one light source.

Measure data The difficult class contains inspections:
- Where multiple illumination are necessary e.g. front and back light;
- Where a high accuracy or repeatability (less then one pixel) is required;
- Where more then one camera is used;
- Where there is an inaccurate positioning of the product, camera or light sources;
- In a rough environment.
In these situations, a feasibility study is necessary to prevents disasters.

The problem with vision systems is not the detection (measurement or inspection) of the feature but to prevent the vision system of getting unwanted disturbances of the measurement area and unforeseen changes in the product itself.

This indicates that mismeasurement, not the measurement itself is the problem.
Within a feasibility study, all possible and expectable disturbances that can cause a mismeasurement must be taken in account. When a high accuracy measurement is required the repeatability an reproducibility in a comparable situation has to be measured to prevent disasters. Of course it is difficult enough to create a comparable test situation. Therefore a lot of experience to build industrial vision systems is essential.

Accept and reject decision

Most customers want a detection system that accepts good product and rejects a bad one.
Therefore, they expect that a vision system takes in account the difference between good and bad.
But a vision system can not make the difference between good and bad, a vision system only measures!
Damaged paper 2 points Damaged paper 3 points A vision system measures by way of light (electromagnetic wave), so all other physical parameters, e.g. pressure and flow, can not directly be measured by a vision system.
This is only possible in an indirect way. A vision system measures the features which are visible and imaged on the camera sensor.
Length measurements are measured from the point where the intensity of light changes to the next change of light. The distance between these changes of light intensity is the length of an object.
The decision between good and bad is made after comparing the measured values with the rejection limits for that value. These rejection limits are defined by the operators, for the system itself can not decide between good and bad.

For example: when no holes or damages are allowed in a piece of paper, the first question of a vision expert is to define a hole.
Is a hole defined by the diameter? Or by the length and width? Or by the area or ratio between the length and the width? Or by a ratio between the area and the contour?
A sheet of paper is o.k. when a hole in the paper is (very) small. So the next question is: how small is small? What is the minimum acceptable hole diameter (or length or area)? This is very important for the resolution of the vision system and optics. What if there are several (very) small holes in the same area, which a human can interpret as a large hole?

Correlation Basically, a vision system only measures and if the measured value exceeds the (adjustable) limits it classifies the product as a bad product. Important is that the vision systems measures the same product with a small and known reproducibility.
By whom the decision is made, by the vision system itself, by operator or by machine control system (PLC), in terms of accepting or rejecting a product, is not of major interest.

Because a vision system is measuring equipment, a feasibility study is mainly based on measuring the product (or feature) in all relevant and acceptable situations, including expected environmental disturbances and product variations.
The reproducibility and repeatability, based on the collected measurement data, can be calculated and analyzed with all kind of statistical tools.
For a performance conclusion, still the knowledge and the experience of a vision expert is necessary: not only the data is decisive.

Do not just put the measurement data in a software analysis package and trust these results, but use your own common sense, vision knowledge, product knowledge and experience!

A feasibility study is based on a (sub)set of product which contains several good and several bad products. A major part of the set contains borderline products, because this is the zone where the measurement inaccuracy causes major unwanted rejects and slip-throughs.

Gage R&R When a vision system is running in a production environment the rejects are easy countable by means of a simple counter in the program or by an ejector unit.
Slip-throughs however, are more difficult to detect. In those cases every produced product has to be manually checked. In most cases, this is unfeasible. Therefore engineers found different ways to predict and check the throughput.

For this difficult class of vision inspections JAIMS always starts with a feasibility study. Not only to protect JAIMS from disasters but even more to protect the customer for a disaster.
With help of the results of a feasibility study it is clear what the customer can expect from the vision system and what the limits are.
During the feasibility study a lot of knowledge is gained from the product (in optic and measurable way) and the environment.
Using this knowledge the best performance of the final inspection system can be achieved, potential problems can caught away in time and the costumer will know what he can expect of the system. In this way, customers will be free of false promises and can count on a fair price.

In fact, an feasibility study is a win-win situation.

A feasibility study can also be used for a second opinion, to prevent false promises and disasters.