Machine Vision Applications Courtesey of Multipix
- TECH TIP: What is sensitivity and why are sensitivity statements often misleading?
- Analogue Goes Digital - Blood Fraction Identification Uses Basler ace GigE
- Machine vision gives Optometrists a clear view
TECH TIP: What is sensitivity and why are sensitivity statements often misleading?
Sensitivity
The response curve for a light sensitive sensor can be divided into three parts: the dark area, the linear area and the saturation area. A typical response curve is shown in the graph below.
The dark area of the response curve shows the sensor's response to very low light. The output of the sensor in the dark area is very low, is noisy and is unpredictable. As you gradually increase the light falling on a sensor, you will find a point where the output of the sensor begins to increase predictably as the amount of light increases. This point is called the Noise Equivalent Exposure (NEE).
After the NEE point is reached, the output of the sensor becomes linear. The output remains linear until a point called the Saturation Equivalent Exposure (SEE) is reached. At this point, increasing the light intensity results in a nonlinear increase in the sensor output.
The gradient of the linear portion of the sensor's response curve is commonly referred to as sensitivity and is usually measured in V/µJ/cm2. The higher a sensor's output voltage is for a given amount of light, the higher its sensitivity.
But when you are discussing sensors, talking about sensitivity alone does not make sense. For one thing, NEE is also very important. Since a sensor with a high NEE will be blind at low light levels, NEE should be as low as possible.
Another point to consider is that a digital camera is a system and that sensor sensitivity is just one of the factors involved in the output signal from the camera. Electronic devices in the camera such as Analogue to Digital converters and amplifiers also influence the output signal. It could be said that a camera's "responsivity" is a better measure of performance. Also that digital cameras responsivity should be stated as DN/µJ/cm2 (DN stands for digital number). The graph below shows a responsivity curve.
If a camera provides a gain feature as most of them do, responsivity will depend on the gain setting. And responsivity really only makes sense when it is stated in combination with a measurement of the camera's noise such as peak-to-peak, signal-to-noise ratio.
Let's consider an example. Suppose that you are comparing two cameras and that they have the following specifications:
• Camera One: Responsivity = 1 DN/µJ/cm2 Noise = 2 DN (peak-to-peak)
• Camera Two: Responsivity = 2 DN/µJ/cm2 Noise = 5 DN (peak-to-peak)
At first glance, camera two seems better than camera one because its responsivity is higher. However if camera one has a gain feature, we can adjust the gain and increase the responsivity to two. Keep in mind that if we adjust the gain to double the responsivity from one to two, we will also double the noise. Now we have this situation:
• Camera One: Responsivity = 2 DN/µJ/cm2 Noise = 4 DN (peak-to-peak)
• Camera Two: Responsivity = 2 DN/µJ/cm2 Noise = 5 DN (peak-to-peak)
Which camera is better? They now both have the same responsivity, but camera one has lower noise. Camera one would be the better choice.
The lesson to be learned from all of this is that sensor sensitivity alone does not tell the entire story and that we should be sure to use similar measuring criteria when we are comparing cameras.
To learn more contact Multipix Imaging (www.multipix.com). Information courtsey of Basler VC.
Analogue Goes Digital - Blood Fraction Identification Uses Basler ace GigE
RTS (www.rts-group.com) is a major supplier of automated products and solutions for scientific applications including biological sample management and biobanking. The company has knowledge of robot vision which allows them to develop high end solutions to challenging inspection task. One example of this is blood fractionation.
The Application
Whole blood fractionation is a time consuming process when performed manually, but blood samples are highly variable in nature and this makes automation extremely difficult. As part of the UK Biobank project (www.ukbiobank.ac.uk), RTS developed a number of high throughput automated blood fractionation systems. These systems were needed to handle the application’s high throughput and quality requirements while managing the inherent variability of volume, viscosity, and turbidity in blood samples. Without these systems, a small army of technicians would be required to perform the fractionation task, and they would face the
serious health and safety concerns inherent in working with unscreened blood.
Solution and Benefits
A unique patented vision system that is used to assess each individual blood sample and measure the individual fractions is key to the RTS solutions for automating the fractionation
of blood. The acquired information is linked to a barcode on each input container and is stored in a central database.
The stored fraction heights are used to calculate the volume of each fraction using known collection vessel dimensions. This information is then used by a liquid handling system
to accurately aspirate off each fraction, before dispensing them into cryovials in user specified aliquots for storage or downstream processing. Automating blood fractionation saves time and money and also reduces the risk of a potentially dangerous job. Technicians who would otherwise be performing the task are free to engage in more challenging and productive tasks.
RTS approached Multipix Imaging (www.multipix.com) to discuss ways to upgrade the existing design and in particular the camera technology. This resulted in RTS deploying a Basler ace GigE Vision camera (model acA1300-30gm) to replace an existing Sony XC-HR70 analogue camera. By doing this, RTS has been able to update to the latest technology thus achieving:
• significant cost savings in both camera and peripheral hardware.
• Gain an increase in camera resolution with no additional cost and potentially achieve an increase in accuracy of 25% for the detection of the fraction levels.
• Maintain the same sensor size and thus maintain the same FOV – a key requirement for the system.
• Avoid any costly changes to the design of the vision unit because the Basler ace GigE has the same mounting footprint as the XC-HR70.
• Avoid any software changes.
Multipix Imaging is a company which specialises in machine vision components. They welcome customer questions and are very happy to offer advice. For further information visit www.multipix.com
Machine vision gives Optometrists a clear view
Being fitted for your new spectacles, you go to the optometrist and without time consuming adaptations and measuring you look at yourself in the mirror, your chosen frame on your nose… job done!
The international glasses manufacturer Rodenstock, Germany, has developed an optometrist service terminal called ImpressionIST® that enables the adaptation of glasses regarding the individual parameters and center data in an absolutely unstrained atmosphere. Innovative 3D images determine the facial measurements, comfortably and accurately giving the optometrist the information they require and the customer a view of themselves in their new glasses.
This system has been successfully distributed over 17 countries. Due to its great success Rodenstock are now developing the second generation of this terminal, the ImpressionIST® Avantgarde.
The ImpressionIST is promoted as having a four-in-one functionality, comprising:
- A two-camera 3D video centre system, which also measures back vertex distance, pantoscopic tilt and frame wrap angle.
- A consulting program to assist in the selection of frames and added-value lenses.
- The ability to capture an image of the patient and demonstrate their appearance with various eyewear options.
- An interactive product information terminal for the patient.
It has been designed so that the customer takes center stage and the experience is comfortable and enjoyable.
The 3D video center system represents the technical heart of the device. The customer looks into a mirror from a distance of around 75cm. This puts the customers in a relaxed position and gives the optimal posture for the measuring.
Hidden from the customer’s view, the imaging system is mounted behind a semitransparent mirror.
Image Processing
Two images are simultaneously acquired by two machine vision cameras. One image views the customer’s face frontally (center camera), another one from below and beside (side camera).
For the next step, the images are processed by imaging software based on the software library HALCON by MVTec.
While video systems for spectacle fitting have been available for some time, only the ImpressionIST solution makes it possible to truly measure the spectacles in all three dimensions and to transfer these measurements 1:1 in optimum fashion to the later position in front of the eyes.
The 3D coordinates created by HALCON imaging software are computed for calibration. In a log printout, all results are documented with illustrating figures. These details go directly to the workshop to create the spectacles prescription.
The pupils can be found by the image processing software. Crucial criteria are the three reflections of the illumination around the pupil. After the pupils are found, the position crosses are set by the software.
Multipix Imaging is the UK distributor for HALCON software, with a great deal of experience and training in the use of this powerful software.
For further information please contact: sales@multpix.com Tel: 01730 233332 www.multipix.com
