The following view is not just prevalent amongst experts: more than 2/3 of the success of image processing solutions is due to suitable lighting. The aim is quite simple: create contrast. It is the prerequisite for simple, fast and robust test programs. Without it, the most efficient processing software cannot operate reliably.
There are numerous direct and indirect reciprocal effects between lighting and the environment: test objects, ambient light, lenses, cameras and machine environment, as well as image processing hardware and software all have an effect on the success or failure of lighting.

But be careful. The expression "Yeah, I can see it" does not often provide much help. Light recognition in humans and machines is too different. On the one hand, you can in several instances follow your sense of sight in selecting lighting, but in many cases it leads you up the wrong track.

The composition of light should fundamentally match the sensitivity of the image sensor, whereby the width of the spectral range of light is normally restricted to 300 nm for UV light and 1,400 nm for NIR. The test object, lens and image sensor must be coordinated in this range.

Depending on technology and design, image sensors, including the human eye, have different sensitivity to different colours of light. Conclusions about whether a suitable lighting wavelength for one image sensor would also be suitable for all other image sensors are therefore not easily drawn in a comparative way. Particularly in the field of invisible light (UV and infrared) calculable, but not directly comparable, contexts apply to light recognition.

Human and machine recognition differs even during goal formulation – contrast-rich lighting. Unbeknown to us, differences in brightness described as contrast by humans include the existing maximum and minimum brightness in the entire visual field as part of contrast imaging.

Globally relative values are depicted. In the case of mechanical image processing only locally absolute differences in brightness are used and analysed. The results for human and machine are far removed and should therefore always be objectively calculated using image processing technology aids.

Distance is an essential factor for successful lighting. The photometric law of distance gives the following contexts for incident light:

If the distance from the lighting component to the test object is doubled, only a quarter of the original illuminance and irradiance (given in lux or W/m²) remains; if the distance is increased by three times, only 1/9 remains. Unfortunately, here too, the human eye, with its logarithmical recognition of brightness, is not a good guide as regards loss of brightness and leads to errors.
The image processing system's brightness value (grey value) should therefore be consulted for objective proof of brightness fall-off.

An important conclusion for incident light illumination is thus: place the lighting component as close to the test object as possible.
The conditions are different for back light. The factor that determines the brightness for lighting components is therefore luminance or ray density (given in candela/m² or in W/sr). It remains constant independent of the distance from the viewer/camera. Changes in distance thus have no influence on the lighting brightness. Back light illumination can therefore be placed at a greater distance from the test object without loss of brightness.

Different implementation scenarios require different components. Thus, on the one hand, lighting is required that allows illumination with the greatest possible homogeneity (e.g. for back light illumination). On the other hand, other forms of lighting are often designed in such a way that they create a targeted lighting profile that is determined by their individual shape (e.g. for dark field lighting).

Information about the absolute brightness, homogeneity or lighting profile provides illumination indicatrices for incident light and luminance indicatrices for back light.

Light colour, which is created by targeted wavelength radiation is a considerable feature of lighting that can be externally recognised by humans. This light colour causes a reaction on the test object (absorption, reflection and transmission). Light that corresponds to the colour of the test object under white light is reflected particularly well and is therefore particularly effective. The complementary colour, on the other hand, is particularly effectively eliminated and therefore appears particularly dark.
The targeted use of colour and complementary colour constitutes a principle for selecting light wavelengths.

Objects that appear achromatic under white light (e.g. metal surfaces) point to the fact that they reflect all wavelengths (of visible light) more or less evenly. Selection of lighting colour plays a lesser role for these objects.
However, not all white light is the same. In order that there is even brightness on the test object, the composition of spectrum used in the white light must be known. This is particularly important for colour image processing, during which differences in light composition can be corrected using white balance.
Human description is not effective for UV and infrared light, whose wavelengths lie beyond the visible field. No prediction can be made for the interaction between the light "colour" and the test object. They must be established by experiments.

Unfortunately, when lighting the test object, targeted illumination using the lighting components implemented does not always prevail. Changes to the position of the image processing test facility (relocation of the machine) or variable daylight radiation during the course of the day can lead to the implemented lighting not being able to deliver constant contrast, thus endangering the stability and reliability of the image processing.

The following examples can be used as countermeasures for extraneous and interfering light:

• Using considerably more efficient lighting so that targeted lighting is stronger than interference
• Short-term lighting in combination with flash lighting
• Use of light filters in combination with suitable light colour
• Light-proof protective housing around the image processing machine

If the specific machine conditions (lack of space and light, light colour not suitable for filtering or limited operation) mean that not all countermeasures can be taken, the problem of extraneous and interfering light remains a considerable destabilising factor for image processing.

Due to the large amounts of light energy radiated when using lasers and LEDs as light sources, precautions must be observed. There is also the fact that part of the light used is radiated in the invisible field of UV and infrared light, which is particularly dangerous as it cannot be directly recognised by humans.
The danger concerns both the eyes and skin (particularly with UV). However, some simple types of operation and behaviour can reduce the dangers caused by these light sources:

• Do not look directly into the light
• Only switch the light on when it is required for image processing (pulse or flash – not continuous operation)

Long gone are the times in which lighting in image processing was understood as "light to brighten the image". Modern lighting components have developed into high-tech products in which state-of-the-art knowledge from lighting, electronics, thermodynamics, materials engineering and production technology are united.

The variety of light sources used in image processing has been greatly reduced in recent years due to LEDs. However, numerous other light sources are still used in image processing:

Halogen lamps:
bright, even spectrum, for continuous operation only, short-life, slow-switching, sensitive to vibration, low efficiency, usually integrated into cold-light sources

Metal halide lamps:
very bright, for continuous operation only, non-universal spectrum, expensive

Xenon lamps:
good for flash (including short sequence), very bright, high-voltage operation, problematic with regard to electromagnetic tolerance, elaborate controls

Fluorescent lamps:
cost-effective (including illumination of large areas), operation only with high-frequency ballast, inflexible shape, highly dependent upon temperature

Laser:
many setup options for structured lighting, monochrome light, differences in intensity (speckles) make analysis more difficult, protective measures necessary!

LEDs:
becoming standard lighting in image processing, as they are:
highly proven in industry, can be controlled quickly and in a targeted manner using a variety of electrical options, long-life, almost any preferred construction possible, increasingly more efficient, increasingly more wavelengths/light colours/colour combinations, no maintenance or repair

The light output in the light sources used by the lighting components is determined by their internal structure. For the various applications in image processing, the direction of the light output must be manipulated in a targeted manner using optical components within the lighting components in order to achieve the desired lighting effect. Depending on use, the light is formed as:

• Diffuse lighting with the most even distribution of light radiation in different directions
• Directed light, which features a pronounced preferential direction
• Telecentric lighting, which is particularly strongly directed and in which parallel main beams provide most of the light
• Structured lighting, which features a local light distribution structure as an additional lighting property.

Lighting control is an essential part of lighting components of the interface to the controls or the image processing system. The age of faults caused by LEDs being soldered together and which are connected to the PLC supply voltage without further measures has passed. Effective, powerful and reliable image processing can only be achieved by defining lighting scenarios, as these are achieved using control circuits that are specially adapted to the light sources.

Due to their operating conditions, some light sources, such as halogen lamps, metal halide lamps and fluorescent lamps can only be switched on and off. AC-operated lighting in which 50 Hz supply frequency affects the temporal brightness distribution cannot be used due to surges during picture capture (variations in image brightness); high frequency pre-switch systems can be used, if necessary.
If a time-defined, brightness-defined or location-defined lighting scenario is required, more complex circuits in the form of a lighting control are necessary, as powerful light sources such as LEDs require defined operating conditions as regards voltage, current and temperature.
A large input power supply (typically 10 to 30 VDC) is an essential quality for connecting unregulated supply networks (PLC supply voltage).
Furthermore, a constant-current source should be integrated. This counteracts the ageing effect on LEDs in the long term and can provide uniform brightness in the event of short-term switching operations.

LEDs only remain powerful light sources if they are operated in such a targeted way that there are no permanent changes in temperature within the LED chips. Even short-term overheating of chips leads to extreme and irreversible aging. Losses in brightness (> 50%) and increased failure rates that appear only after years in cool operation can occur within a few hours. Temperature management is therefore an absolute must for LED lighting with constantly increasing levels of performance. Targeted and design-based heat dissipation also guarantees reliable operation without overheating even at high ambient temperatures. Low-cost lighting in the form of plastic-moulded or thermally isolated airborne LED fields cannot fulfil these conditions in any way and are best suited to pulse operation.
As well as temporally stable lighting operation, the ability to control its brightness is an important feature. This can be achieved in various ways:

• manually via potentiometer
• via analogue control voltage (typically 0 to 10 VDC)
• via digital interface using a controller

Adaptive lighting allows temporal and local brightness to be set for entire LED fields and adjusted to the geometry of the test object and environment in real time via ethernet.

The circuitry principles for the various types of operation with which the LED lighting can be operated are determined in the dimensioning of the lighting controls:

The simplest mode of operation is constant illumination. Once switched on, the lights provide illumination for a long period of time. It is important to turn on the timer, which elapses once the final brightness (noted in the datasheet) is achieved. For LED lighting, this amounts to times of up to several milliseconds depending on the integrated stabilisation circuit. It is therefore unsuitable to turn on this lighting via the supply voltage.

Lights that do not provide constant illumination are operated as pulse lighting. They feature a fast switch input, through which the lights can be switched on and off using a PLC gauge with a delay time of < 1 µs. This quick switch option is always used when complex lighting scenarios require that the camera captures several consecutive images under different lighting conditions.

Lights that are only intended for constant lighting are not suitable for pulse operation. The extensive capacitive wiring of the stabilisation circuit means that there are long delays in switching on and off that would prevent the necessary quick reaction.

Increasingly faster processes and production machines require flash lighting either so that motion blur can be avoided (motion "freezing") or so that extraneous and interfering light can be suppressed. The prerequisite for this is that, between the image processing system and the hardware, there is a quick reaction to a trigger signal and/or quick activation of flash lighting via digital input and output. Likewise, the image processing software must be able to synchronise the start of the flash, the image capture and the control the shutter speed.

TTL and PLC gauges are typical gauges for triggering flash components. It is therefore possible to trigger the flash from either the image processing system or the machine controls (PLC).

Once the trigger signal is sent there is a delay time until the flash is triggered (and therefore until image capture). The delay time must be very short so the test object does not unduly move out of the camera's visual field. Typical delay times are around 500 ns.

Maximum flash frequency denotes the largest number of flashes that a flash component can conduct, at constant light energy (without reductions in brightness). It is very dependent on the quality of the control circuit for the flash lighting.

Flash time describes the time in which the flash can deliver constant brightness. Flash times are usually between 1 and 100 µs. It must be possible to set flash time and brightness independently of each other.
The flash time needed for a specific application is dependent on a number of factors:

• Speed of the test object
• Visual field size
• Maximum permissible motion blur of the image
• Synchronisation
• Flash lighting output, etc.

Image: It must be possible to synchronise the various processes during flash lighting using hardware and software.

The life of lights depends on both the type of light source and on a number of other environmental and operational conditions. It can be between 300 hours (powerful halogen lights) and > 50,000 hours for LED lights.
The oft-cited MTBF (Mean Time Between Failure) for LED lighting of 100,000 hours is for a single red LED. For other light wave lengths (blue and UV) it is lower. Furthermore, MTBF for a lighting component means the product of the MTBF for all components used and is therefore always shorter than a single LED.

In the case of LED flash lighting, it can be assumed that several tens of millions of flashes can be realised at maximum illumination and without ageing and thus loss of brightness if certain principles of switch technology are complied with during construction.

The lighting that is technically the best is rendered useless if it is not defined and cannot be permanently and stably fixed and combined.
Standardised perforated grids for fastening and standardised connection gauges are required for lighting to be suitable for series production in mechanical engineering work. Connection cables with unprotected low-voltage wires cannot be used as the lighting connection cable can bend (tensile resistance) and twist (torsional resistance) when machines and robots move. It is not uncommon for aggressive media to affect connection cables; additional oil resistance is therefore often required.
Like all image processing components, lighting is also subject to oscillations and vibrations. As a light source, LEDs are extremely shock-resistant, however the interface to the printed circuit board or the LED mount is often not capable of coping with the force that occurs. At high acceleration, as components that are moved by robot arms are subject to, this leads to malfunctions due to components and connection cables breaking. In order to avoid this, precautions that can be proven by conforming to the corresponding standards must be taken at the design stage.

Lighting selection and configuration can never be reduced to just one suitable way of doing things – there are always several options, each of which has its own advantages and disadvantages. To discover and, if necessary, quantify these is often not a simple process.

This is because the geometric properties (edge shapes and machining marks), optical properties (colour, reflection, gloss level, scattering) and even chemical properties (condensation, oxidation, oil, coolant) of the test object are unknown, not tolerated or subject to major fluctuations. All these properties have a considerable influence on the options for lighting configuration and selection.

The brightness of the property to be highlighted can be influenced using the directional properties of the light emerging from the lighting:

• Diffuse lighting is suitable for incident light and back light illumination where smaller surface faults in the structure of the test object are to be made good

• Directed lighting due to its strong directionality, can ensure that pronounced directional properties on the test object are particularly well highlighted. Directed lighting is often used in incident light.

• Telecentric lighting finds its essential use in combination with → telecentric lenses for contrast-rich depiction of test objects in → back light.

• Structured lighting is used because of the light structures used, e.g. for 3D image processing tasks in incident light.

In general, adjustment of the lighting to the geometric-optical properties of the test object must be carried out during lighting selection and configuration. The top priority is to know what is to be accurately highlighted and how. If there are many characteristics to inspect, this can lead to several completely different lighting scenarios/components for a single test object.

Incident light illumination is suitable for tasks in which test characteristics on the surface of the test object are to be met.

Transparent materials can be inspected particularly well. If the test objects are partially or fully transparent, the top and underside cannot be separated because the components will then be illuminated from the inside.

Difficulties with incident light generally occur on the edges of the test objects as the edge properties influence the interplay between light and test object considerably. For this reason, incident light illumination is not suitable for metric measurements.

Many image processing applications have to fall back on incident lighting because the parameters of the industrial environment do not allow back light, e.g.

• Pick & place robots
• Inspection of components on conveyors
• Automatic assembly machines

Back lighting allows for the most contrast-rich depiction and sharp definition of silhouettes and contours of non-transparent components. Many transparent and partially transparent components can also be inspected in telecentric back light. The same is also true for applications that rely on precision, such as metric measurements.

Back lighting is generally independent of the lighting distance to the test object and is therefore particularly suitable for high-speed image processing applications.
Despite its advantages, opportunities to apply back lighting are usually restricted by the fact that they rely heavily on the design of the machine and therefore cannot always be implemented.

With this kind of lighting, the illuminated surface without the test object appears light by the fact that the lighting is on the optical axis of the camera and lens above (incident light) or under (back light) the test object. The camera "looks into the light" (back light) or the lights illuminate from the camera's observation direction (incident light). The details to be recognised are generally seen as dark on light.

Without the test object, the camera's visual field appears dark in dark field illumination because the lighting is outside the camera's visual field. Only when the test object is in the visual field does the lighting on the edge and surface elements work in such a way that the details to be recognised are visible as light on dark.

• Application in colour or black & white
  (affects choice of wavelength/light colours)
• Speed of test objects
  (affects light output and flash)
• Size of field to be lit
  (effect on size of lighting and type of light source)
• Maintenance costs
  (affects selection/omission of certain light sources)
• Environmental conditions
  (affects size of lighting, wavelength, light colour, etc.)

Moreover, a number of other factors influence the choice of lighting. Therefore, having experts select proper lighting is a service that makes sense in terms of money and time.