There is a wide spectrum of technical resources available to design automated vision systems. For particle inspection, SD technology is still the preferred choice by many; known for ease of recipe set-up, reliable performance, and unaffected by particulate color/reflectivity or exterior container influences.
Nevertheless, there are several use-cases where camera systems can be an alternative or an add-on feature to support automated vision systems delivering high quality particle inspection results.
When looking into camera technology for automated visual inspection, it is of prime importance to obtain good imaging quality in order to maximize the detection probability of particulates. This has to be achieved within the constraints of mechanical transport (limited degrees of freedom for observation) and available inspection time (machine speed vs observation time).
There is a variety of resources available acting as influencing factors for the images quality. Those include:
- Cameras: Area cameras based on 2D array, Line scan cameras based on 1D array; Color and monochrome versions, different resolution in pixel size, frame rates and transmission protocols
- Lenses: variety of view fields, focus depths, brightness level, image distortions, color aberrations
- Lighting types: LED, 2D panels in linear bars in round or dome shape, lighting color, intensity, luminance evenness
- Optical Filters: color filters; neutral density; polarizers
- Angles for image capture and for illumination
- Mechanical parameters such as container spinning speeds, acceleration and braking, spinning patterns, etc.
- Vision computer, including image processing software and dedicated detection algorithms to achieve desired performance within the limited time available for calculation and judgment.
As many of the factors influence one another, choosing the suitable camera hardware can be a key decision in designing the inspection system. The following illustration shows the different detection results, based on the applied camera hardware, field of view, container movement and imaging time.
However, those differences become more obvious, when applied to realized solutions, e.g. for the inspection of syringes.
Example 1: Enhancing detection of fibers by polarized lighting
While the container is rotated, one single shot image is captured and analyzed, using an area camera. The rotation of the container allows the fiber particle to elevate along the vortex. When using a polarized filter, the particle can be properly detected and differentiated from a scratch.
Example 2: Particulate detection in viscous solutions, filled in syringes
A challenge, well known to vision engineers is the complete elimination of bubbles before start of inspection in dense, high-viscous solutions. In syringes, there comes another complexity by particulate that could stick to the rubber stopper. This is particularly challenging, if bubbles are located in narrow angles at or nearby the stopper/wall point of contact. As there is also the probability of particles in this region of the syringe, a reliable differentiation between particle and bubbles is necessary.
In complex application cases, the appropriate design of the system is key. By adjusting the geometric angle to the inspection item and using a line camera instead of an area camera, the vision system can judge differential features easier. A texture analysis tool within the vision software enables the interpretation of image data such as texture, change rate of pixel surrounding the particles and grey scales within the particle. Combining those adjustments, the differentiation between a potentially contaminating particle and harmless bubbles can be achieved.
Product Manger, Japan
Marketing group, Japan