Vision systems do more than capture images off a production line. Simple tasks include determining the presence of an object or a label on a bottle. More challenging chores check tubing sizes and inspect needles for small hooks and burs. Some medical companies consider vision systems part of their intellectual property. For instance, their system might look at a mixing level in a vat, or read the number of particulates in a vial and let the control system make a decision on that data. Companies build these proprietary clues into their manufacturing process and want to verify what they've done.

Building a vision system doesn't have to be a chore. Realizing what it must do and recognizing the core components of a system are the first steps. While drop-in systems are available, most applications require more tinkering. For example, some production parts are harder to image than others. Parts that are round or clear like glass are more difficult to image. Having to differentiate between two different clear parts, for example, gets even more complex.

Image quality

Imaging systems should create sufficient image quality to get information about the object from the image. What may be adequate image quality for one application may prove inadequate in another. A variety of factors contribute to the overall image quality, including resolution, image contrast, depth of field, perspective errors, and geometric errors (distortion).

Simply put, resolution is the amount of object detail reproduced by the imaging system. Object resolution can be calculated from the image resolution of the camera using the primary magnification of the imaging lens. It is the ratio between sensor size (mm) and field of vision (mm).

It should be noted that in many systems you will need to look beyond the camera resolution to determine the achievable system resolution. The lens that is chosen may or many not perform to the level of the cameras limiting resolution or to the resolution requirement. For this reason you must look at the camera and lenses capabilities in combination to determine if the system will perform as required.

Another equally important factor is contrast, or how effectively the differences between the object and the shades of gray in the background are reproduced in the image. For the image to appear defined, black object details must be seen as black and the white details as white, with all shades of gray in between accurately imaged. A black line on a white background is an example of 100% contrast.

The lens, sensor, and illumination all play key roles in determining image contrast. The lens contrast is typically defined in terms of the percentage of the object contrast that is reproduced. A sensor's ability to reproduce contrast is usually specified in terms of decibels (dB) in analog cameras and bits in digital cameras.

A lens' depth of field (DOF) is its ability to maintain a required amount of image quality as an object is positioned closer to or further from best focus. DOF also applies to objects with depth, because high-DOF lenses can image a whole object clearly. As the object is placed closer or farther than the working distance, it goes out of focus and both resolution and contrast suffer. For this reason, DOF only makes sense when it is defined with an associated resolution and contrast.

Distortion is an optical error, or aberration, caused by the lens which results in magnification differences between different points on the image. The points on the object are misplaced in the image relative to the center of the field. The difference between the actual (distorted image) and the predicted (non-distorted object) position can be expressed in terms of a percentage from the center of the field.

Although distortion exists in all lenses, it can often be fairly well corrected. It is more difficult to correct in short focal length (wide angle/fish eye) lenses. Distortion is a geometric aberration which means information about the object is not actually lost, but simply misplaced on the image plane. This can prove troublesome in some measurement applications. However, once distortion is accurately measured using a distortion target, it can be factored into measurement calculations. The distortion can be removed from the image using software.

Perspective error, also called parallax, is part of everyday human experience. In fact, parallax is what lets the brain interpret the 3D world. We expect closer objects to appear relatively larger than those positioned farther away. This phenomenon is also present in conventional imaging systems where magnification of the object changes with its distance from the lens. Lenses that are telecentric optically correct for this so objects remain the same perceived size, independent of their location in the space defined by the lens.

Parallax is most troublesome in measurement applications involving objects with depth or objects moving relative to the lens. While telecentric lenses do not inherently have more DOF than conventional designs, the telecentric image tends to blur symmetrically. The center of the blur corresponds to the center of the object, so no error is introduced when measuring the center-to-center separation of objects. This is true even if the image is not perfectly focused. Distortion is rarely a problem because it can effectively be corrected in telecentric lens designs.

The primary limitation of a telecentric design is that it cannot maintain its telecentric properties when imaging objects larger than the front-lens diameter. Although telecentric designs do not inherently have more depth of field than conventional lenses, depth of field is often incorporated into the design.

Electronics

The imaging sensor, as well as other electronic components, plays a significant role in the performance of an imaging system. Proper integration of all components, including camera, capture board, software, and cables result in optimum system performance.

Charge-Coupled Devices (CCDs) are the most common camera sensors in machine-vision applications. The CCD camera contains a silicon chip which consists of a matrix of light sensitive photosites or pixels. The CCD's wide use can be linked to its characteristically small size and light weight. Additionally, CCDs have an impressive dynamic range and yield a highly linear relationship between incoming energy and outgoing signal, making them ideal for metrology.

Analog CCD cameras can have rectangular pixels which are larger in the vertical dimension. This is a result of a limited number of scanning lines in the signal standards. Asymmetric pixels yield higher horizontal resolution than vertical. The required analog CCD cameras (with the same signal standard) usually have the same vertical resolution. For this reason, the industry standard is to specify resolution in terms of horizontal resolution. Digital cameras, on the other hand are not limited by vertical bandwidth and therefore can have either rectangular or square pixels.

The size of the sensor's active area is important in determining the system's field of view (FOV). Given a fixed primary magnification (determined by the lens), larger sensors yield greater FOVs. The standard CCD sensor sizes are 1/4, 1/3, 1/2, 2/3, and 1-in. Nomenclature for these standards dates back to the Vidicon vacuum tubes used for television, so it is important to note that the actual dimensions of the chips differ. All of these standards maintain a 4:3 (Horizontal:Vertical) dimensional aspect ratio. Another issue is the ability of the lens to support certain CCD chip sizes. If the chip is too large for the lens design, the resulting image may appear to fade away and degrade towards the edges because of vignetting (extinction of rays which pass through the outer edges of the lens). This is commonly referred to as the “tunnel” effect, since the edges of the field become dark. Smaller sizes do not yield such problems.

Illumination

The effect of illumination on image quality is often underestimated. A proper lighting scenario can increase the image contrast and resolution, improving the overall performance of the system. This can include the illumination setup as well as filtering and polarizing.

Engineers may often struggle with contrast and resolution problems in an imaging system, while underestimating the power of proper illumination. Often a required image quality is typically met by improving the illumination, rather than investing in higher resolution detectors, imaging lenses, and software. System integrators should keep in mind that proper light intensity in the final image depends on component selection.

Every component affects the amount of light incident on the sensor and, therefore, affects the image quality of the system. Obviously the aperture opening or f-stop of the imaging lens has a direct impact on the amount of light incident on the camera. Increase the illumination as the lens closes (higher f-stop). High-power lenses usually require more illumination because smaller viewing areas reflect less light back into the lens. The camera's minimum sensitivity is also important in determining the minimum amount of light required by the system. In addition, CCD-camera settings such as gain, and shutter speed affect the sensitivity of the sensor. Fiber-optic illumination usually involves an illuminator and a light guide, each of which should be integrated to optimize lighting at the object. Because illumination affects image quality, it is important to maintain constant lighting conditions for repeatable results.

The lens aperture, system magnification, camera settings, filtering, and other illumination parameters all affect the light incident on the sensor. These factors must be adjusted to accommodate objects with different characteristics such as profile and reflectivity.

Digital versus Analog

The debate over analog versus digital system was once a matter of evaluating the application. Analog works better in some situations and digital in others. To some extent this is still true. But the cost of digital products with analog capabilities has come down remarkably. While it was once expensive to put a digital image into a computer it's now inexpensive. It was also expensive to get a resolution higher than analog. Analog has a finite resolution as defined by TV standards. Because there was so much volume in this area, there wasn't demand for digital products. But in the last 8 to 10 years there has been a boom in the digital-camera industry and also in digital high-definition television sets. The price of digital vision systems are now two to three times what analog prices are but they give four to ten times the capability.

Digital versus Analog capabilities

Digital	Analog
Typically large cameras.	Size is typically smaller.
Vertical resolution is not limited, so digital cameras can offer higher resolution.	Vertical resolution is limited by the bandwidth of the analog signal.
With no bandwidth limit, these can offer higher numbers of pixels and larger CCD sensors, resulting in greater resolution.	Sensors are usually standard-size formats.
Computer and capture board required to display signal.	Computers and capture boards can be used for digitizing but are not necessary for display.
Signal can be compressed so user can transmit in lower bandwidth without loss.	Analog printing and recording can be easily incorporated into the system.
These typically have square pixels for identical horizontal and vertical resolution.	Usually have rectangular pixels (different horizontal and vertical resolution).
The output signal is digital, therefore little signal loss occurs during signal processing.	Analog signals are susceptible to noise and interference which cause signal loss.

Focusing on machine vision terms

Here are a few common terms to describe vision systems.

Depth of Field or DOF is the maximum object depth that can be maintained entirely in focus. DOF is also the amount of object movement (in and out of best focus) allowable while maintaining a required amount of focus.

Field of View or FOV is the viewable area of the object under inspection. In other words, this is the portion of the object that fills the camera's sensor.

The primary magnification or PMAG of the lens is the ratio between the sensor size and the FOV.

Resolution is the minimum feature size of the object that can be distinguished by the imaging system.

Sensor size is the size of a camera sensor's active area, typically specified in the horizontal dimension. This parameter is important in determining the proper lens magnification required to obtain a desired field of view.

Working distance is the distance from the front of the lens to the object under inspection.