Cameras - Introduction
If you are not experienced with cameras or videography, this section provides an introduction to camera basics.
Resolution of Sensors
Cameras supported by Motif can reach 25MP and more.
The resolution and the physical size of the sensor determine which lens can be used. Larger and higher resolution sensors usually need more expensive lenses than smaller and lower resolution sensors.
Sensor and Pixel Size
The size of a camera sensor is given by the size if the individual pixels (usually squares with a size of 1 to 5 µm) and the resolution of the sensor.
Grayscale and Color Sensors
Grayscale sensors acquire a black and white image with different shades of gray for each pixel. Most cameras are sensitive to light from ~400nm to ~1000nm. The use of a band pass or long pass filter is recommended for reducing spectral aberration.
Individual pixels of a color camera have either a red, green, or blue filter attached, in other words, each pixel is sensitive to only one of these colors. This is called the Bayer Pattern. In a processing step the missing two color values (e.g. green and red for a pixel that is only sensitive to blue light) are computed from the respective surrounding pixels (Bayer color interpolation).
Compared to the matching grayscale camera (same resolution etc.), color cameras are
- ~30% less light sensitive (up to 75%)
- have 30% less effective resolution due to interpolation
- can not be used whit IR illumination (except some specific models)
IR Cut Filter
Most color cameras have a built in IR cut filter that blocks light with a wavelength of >700nm (please read your camera manual for the exact specifications). The filter is mounted in close proximity to the sensor.
Color cameras not featuring, or with a removed IR cut filter are sensitive to IR light. Using white light (not containing any IR component) will result in a perfect color image. Using IR light for illumination will result in a false color grayscale image with red being dominant. The sensitivity is significantly reduced compared to a similar grayscale camera. Using IR and white light at the same time will result in a color image with a strong reddish cast.
Bayer pattern images represent the raw pixel values (grayscale) and need to be computed into color images. Each pixel of a color camera is equipped with either a red, green or blue filter. The pattern of the filters is called "Bayer Pattern". In a processing step the missing color values must be interpolated from the respective surrounding pixels (e.g. the blue and the red value must be interpolated for each green pixel). This is called de-Bayering and is a resource intensive processing task.
Due to this pattern, the effective resolution of color cameras are 30% less compared to the same resolution grayscale camera. Also the sensitivity of the pixels is reduced.
Global shutter cameras read all pixels at the same point in time. This has advantages for 3D tracking systems and for systems where the light source is flashed.
Rolling shutter cameras read the sensors line by line. This can result in artifacts if an object is moving fast with or against the direction the sensor is read. Cameras with a rolling shutter are often more cost-efficient compared to cameras with the same resolution and a global shutter.
Each camera model has different specifications. Please refer to the data sheet for your camera model for exact numbers.
If not mentioned differently, cameras are specified with IP30 and a maximal ambient temperature of 30°C (30 °F). A maximal housing temperature during operation of 50 °C (122 °F) is allowed, additional cooling (artificial airflow: mounting to heat conductive mount) might be required for operation above ~30°C. Humidity during operation 20–80 %, relative, non-condensing. Motif recording units are specified for conditions of 0 °C to 35 °C (32°F to 95°F) and humidity of 10 % to 80 % (non-condensing).
It is always remanded to mount the camera to a heat dispersing source and in addition to equip the camera with heat sinks. Ask Loopbio support for the metal knuckle-head and heat sinks for your Motif cameras (email@example.com).
The sensitivity of the camera sensor is influenced by the following factors:
Sensitivity of the camera sensor (EMVA Quantum Efficiency). Usually the larger the physical size of the individual pixels of a camera sensor is more sensitive.
Wavelength or spectral composition of the illumination. Most camera sensors are most sensitive to green and red light (500mn to 600nm). For color cameras this is balanced and most color cameras are not sensitive to IR light (> ~750nm). Grayscale cameras are sensitive to a much wider spectrum including near infrared (NIR or non-visual, non-thermal IR light of 800nm - 950nm). However, most camera sensors are less than half as sensitive to light with ~850nm as compared to green light and only 15% as sensitive to light whit a wavelength of 94nm.
Continue reading here on how to get a brighter image
Using the GPIO port of a camera and configuring the camera properly allows triggering the acquisition of images. This can be used for synchronizing multiple cameras for 3D applications, for synchronizing a single or multiple cameras with other devices or for synchronizing image acquisition with light flashes.
The trigger signal is a square wave (usually 0 to +5V), each pulse triggers the acquisition of a image (considering the trigger frequency is not exceeding the maximal frame rate of the camera and the exposure time of the camera is smaller than 1/trigger_frequency seconds).
The Loopbio Tigger Box is a perfect solution for camera synchronization.
Certain settings or combination of settings might not be supported by all camera models.
The frame rate is defined as the number of images taken per second (fps = frames per second).
The maximal frame rate is defined by:
- setting it with the frame rate slider
- the set exposure time has to be smaller than
- specifications of your Motif system (camera frame rate, recording unit)
Also called shutter speed, is the time the camera integrates over when taking an image. The longer the exposure time is, the longer the period of time that light is gathered for that image. Hence, long exposure times need less light but fast movements resulting in so called "motion blur".
Auto exposure automatically adjusts the exposure time within specified limits until a target brightness has been reached. This can be set to be done continuously or once.
Gain makes the image brighter by amplifying the whole image including the background noise. Increasing the gain does not make the camera more sensitive. It can be thought of, as similar to increasing the brightness in software.
Gain should only be used as a last resort if lighting is already at the brightest, the aperture is as open as possible, and exposure was set to the maximal value for a given frame rate.
Auto gain automatically adjusts the gain within specified limits until a target brightness has been reached. This can be set to be done continuously or once.
Gamma optimizes the visual appearance of the brightness of an image, it does not improve the image quality.
Pixels of multiple columns and rows are combined in a specific binning patter to reduce the overall size of the image without changing the field of view.
Different binning patterns can be used:
- 2x2 binning: combining 2 horizontal and 2 vertical pixels into one output pixel (results in 1/4 the resolution of the sensor)
- 3x3 binning: combining 3 horizontal and 3 vertical pixels into one output pixel (results in 1/9 the resolution of the sensor)
- 4x4 binning: combining 4 horizontal and 4 vertical pixels into one output pixel (results in 1/16 the resolution of the sensor)
Example: applying a 2x2 binning pattern on a 2448 x 2048 pixel image would result in a image with 1224 x 1024 pixels.
The binning mode can be set to, sum, average, or a mixture of both.
- Sum: sums up all raw pixel values of the pattern. The resulting image is much brighter, but also more noisy as the noise is also summed up.
- Average: averages all pixels values of the pattern. The resulting image will have the same visual brightness as the original image but noise will be reduced.
- Mixture: It could be specified that e.g. the columns are summed up and then averaged, resulting in a bringer less noisy image.
Not all cameras support binning (most color cameras don't), and not all cameras support all or more or different binning modes or pattern than mentioned here.
Other more advanced methods for sub-sampling, such as vertical and horizontal decimation, may be available on certain camera models
Region of Interest (Crop)
Defining a Region of Interest (ROI) allows to specify a region on the sensor that is used for reacquiring the image. By default, no ROI is set and the whole sensor is read. Setting a ROI reduces the pixels that need to be read and therefore will reduce the data the needs to be transferred to the Motif Recording Unit and substantially be stored. \
For some cameras setting an ROI can increase the maximal frame rate. Often this is linked to one axis (the height) of the image only.
Example for Motif a USB3 Camera
- 2048x2048pix -> 90fps (native resolution and frame rate)
- 2048x1500pix -> 122fps
- 2048x1000pix -> 182ps
- 2048x900pix -> 202fps
- 900x900pix -> 202fps (only reducing the height increases the frame rate)
After specifying a ROI an Offset for X and Y can be specified for moving the ROI to a specific area of the sensor.
Only applicable for color cameras.
Changing the balance between the Red, Green and Blue channel for making a white object appear white in the camera image is called White Balance. Depending on lighting and and the camera sensor, a white object might not be white in the camera image (color shift). The human eye will correct for this automatically but for an imaging system the white balance would have to be adjusted.
The Green channel is used as the reference while the gain for the Red and Blue are adjusted.
The format in which the image data is transmitted to the Motif Recording Unit can be specified. These are the most common options:
- Mono8: 8bit grayscale image - Available and default on grayscale cameras. Most color cameras support this format, the conversion between color and grayscale is done within the camera.
- BayerBG8: 8bit Bayer image - Available on most color cameras, read more. This mode is the least resource demanding (hat producing) mode on a color camera.
- YUV: 8bit YUV image - Available on most color cameras, de-bayering is dome on the camera resulting in less CPU utilization on the Motif Recording Unit. Reduces the maximal frame rate as more bandwidth is required compared to Bayer8.
- RGB: 8bit RGB image - de-Bayering is done on the cameras and full RBG images are transferred to the Motif Recording Unit. This reduces the maximal frame rate of the camera to ~1/3 of its frame rate at Bayer8 as more bandwidth is required.