 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
|
|
||
 |
||
|
|
|
|
||
|
|
||
 |
|
105 Technical Definitions of High-Speed Imaging |
Frame Rate
Frame rate, sample rate, capture rate and image (or camera) speed are interchangeable terms. Measured in frames per second, the imager’s speed is one of the most important considerations in motion capture analysis. The frame rate is determined after considering the speed of the event, the size of the area under study, the number of images needed to obtain all the event’s essential information, and the frame rates that are available from the motion analyzer being used. For example, at 1,000 fps a picture is taken once every millisecond. If an event takes place in 15 milliseconds, the imager will capture 15 frames of that event. If the frame rate is set too low, the imager will not capture enough images to provide meaningful analysis. If the frame rate is set higher than necessary, the file size may too large to handle. In other instances, too high of a frame rate sacrifices the area of coverage. This happens when an imager’s frame rate is set higher than its ability to provide full-frame coverage. In most of the new generation of motion analyzers, the imagers provide optional frame rates at reduced resolutions. At these higher rates, the height or width of the images is sacrificed but in return, the frame rate can be as much as sixty times the imager’s full resolution frame rate. Currently, the fastest motion analyzer provides 7000 full resolution frames per second (1024 by 1024) and up to 1.3 million reduced resolution frames per second.
When considering the frame rate performance of a motion analyzer be specific about your requirements. Look closely at a manufacture’s specification sheet to see what the true resolution is at any given frame rate. Some lower frame rate motion analyzer’s are using a technique called line doubling to increase their full frame rate performance. However, the true resolution at the stated frame rate is actually lower and upon display, the pixels are interpolated to fill out the image (4:3 aspect ratio). If no analysis is intended for the images this presents no problem. However, if measurements are to be made, it is important to know the true frame size (resolution) so that measurements can be corrected in the calculations. Typically, for these type of motion analyzers the imaging sensor was designed for standard video. By using this type of sensor the cost is less than a sensor designed for high frame rates. The sensor is being pushed to a higher frame rate. To achieve a higher frame rate beyond its original specification, the amount of image data read out of the sensor must be reduced (lower resolution). Therefore, make sure the frame rate performance matches the motion analyzer’s capability.
Sensor Dimensions
The size of the image sensor in a high-speed camera is important to know. Some common size sensors include 1/2 inch, 2/3 inch and 1 inch. The 1 inch sensor has an effective width of 12.8 millimeters, while the 2/3-inch sensor has an effective width of 8.8 millimeters. A lens that works properly on a camera having a small sensor may not produce a large enough image to work correctly on a camera having a large sensor. This is due to the distortion in the fringe areas of the lens. Some of the newer cameras available now have even larger sensors and may even require custom lenses to prevent image distortion to the edges of the image.
Exposure
Many factors influence the amount of light required to produce the best image possible. Without sufficient light, the image may be:
- under-exposed, detail is lost in dark
- unbalanced, poor color reproduction
- blurred, due to the lack of depth-of-field
The time that light is exposed to the imaging sensor depends on several factors. These factors include, lens f-stop, frame rate, shutter setting, amount of available light, reflectance of surrounding material, imaging sensor’s well capacity, and the sensor’s signal-to-noise (SNR) ratio. All of these factors can significantly impact the image quality. An often overlooked factor is the exposure time, also known as the shutter time.
The exposure time, shutter rate, shutter angle are interchangeable terms. The exposure time for mechanical shutters is set in terms of number of degrees that it is open. The exposure time for electronic sensors is either the inverse of the frame rate if no electronic shutter exists or the time that an electronic shuttered sensor is exposed in microseconds. Shown below are the relationships for defining the exposure time.
mechanical shutter =( revolutions per second x angle/360)
no shutter = 1/frame rate
electronic shutter = period of time that the sensor is exposed
The exposure time determines how sharp or blur free an image is—regardless of the frame rate. The exposure time needed to avoid blur depends on the subject’s velocity and direction, the amount of lens magnification, the shutter speed or frame rate (which ever is faster) and the resolution of the imaging system.
A high velocity subject may be blurred in an image if the velocity is too high during the integration of light on the sensor. If a sharp edge of an object is imaged, and the object moves more than 2 pixels or a line pair, within one frame, the object may be blurred. This is due to the fact that multiple pixels are imaging an averaged value of the edge. This creates a smear, or blur effect on the edge. To get good picture quality, the shutter rate should be 10x that of the subject’s velocity. For example, if you are recording at 1000 fps, the shutter setting should be 1/10,000 of a second (100 microseconds).
The lens magnification can influence the relative velocity of the subject being imaged. The velocity of an object moving across a magnified field-of-view (FOV) is increased linearly according to the magnification level. Instinctually, if an object is viewed far away, the relative velocity in the FOV is less than that viewed next to the object.
Motion analyzers use electronic or mechanical shutters that operate at or beyond 1 micro- seconds (1/1,000,000 of a second), which is fast enough to provide blur-free images of most high-speed events. The shutter controls the amount of light that is exposed to the sensor by the cycle rate of the shutter and the time that the shutter is open. The cycle time is set by the frame rate. The shutter then determines the exposure time. If no independent shutter capability exists for the imaging sensor, then the frame rate will be the effective exposure time. Therefore, for a high velocity object, higher frame rates are required. The shutter is synchronized to the sensor timing. Multiple cameras can be synchronized if the shutters can be controlled in unison. Shown in Table below are subjects (applications) and the estimated minimum frame rate and shutter setting (exposure time) that is expected to be required to obtain usual images of these events.
SUBJECT Min. Frame Rate Exposure (µSec) Money sorting machine (single bill time) 500 100 Flame pattern test (fuel combustion) 3000 20 Wire bonding (one cycle) 1000 50 Surface mount (one placement cycle, no pickup) 1000 100 Food—crackers on process line (three samples) 250 1000 Potato chips being bagged (one cycle) 250 1000 Tire testing, front and rear over glass plate 500 100 Hot glue applied to film box flap 500 500 Blood stream (one cell motion across screen) 1000 20 High voltage circuit breaker (one cycle) 1000 1000 Label pickup (one label) 250 1000 Golf ball impact and flight (club, a.k.a Biomechanics) 1000 20 Composite material fracture 1000 100 Car crash test (impact) 1000 100 Air Bag Inflation 3000 70 A proper shutter speed may be calculated as follows.
Exposure (shutter rate) 2X Pixel Size / Vr
where:
Vr = sensor dimension x (field-of-view / object’s velocity )
Pixel Size = pixel dimension / total pixels
Note: pixel dimension should correspond to the dimension used for the total pixel count.
If the object’s velocity, the field-of-view, the imaging sensor’s dimensions and pixel count are known, the shutter speed required to produce a sharp image can be calculated. The relative velocity (Vr) at the sensor can be calculated by reducing the subject’s velocity by the optical reduction at the sensor. The pixel size must be calculated by dividing the sensor size in the dimension of interest (x or y). Knowing that a relative velocity at the sensor plane that is less than 2 pixels or a line pair will produce a good image, we multiply the pixel size by two. Therefore, the shutter speed is calculated by dividing the 2X pixel size by the relative velocity (Vr). The inverse yields the minimum shutter speed or in the case of an imaging system without a shutter, it is the minimum frame rate for sharp images.
Depth of Field
Depth-of-field (DOF) is the range in which an object would be in focus within a scene. The largest DOF is when a lens is set to infinity. The smaller the f-stop the smaller the DOF. If the object is moved closer to the lens, the DOF also decreases. Lenses of different focal lengths will not have the same DOF for a given f-stop.
Sensitivity
Most modern image sensors have a sensitivity that is equivalent to a film Exposure Index value of between 125 ISO and 480 ISO in color and up to 3200 ISO in monochrome. The sensitivity is a very important factor for obtaining clear images. An inexperienced user may confuse motion blur with a poor depth-of-field. If the sensitivity of the camera is not high enough for imaging an object for a given scene, the lens aperture must be opened up. This reduces the depth-of-field for the object to remain in focus. As the object moves, it could take a path outside the area that is in focus. This would then give the appearance of an object with motion blur. However, in reality, it is out of focus.
In practice, a single 600-watt incandescent lamp placed four feet from a typical subject provides sufficient illumination to make recordings at 1,000 fps with an exposure of one millisecond (1/1,000 of a second) a f/4. This level of performance is fine for many applications, although some demanding high-speed events have characteristics where greater light sensitivity may be preferred.
Record Time
The recording time of a high-speed video system is dependent on the frame rate selected and the amount of storage medium available. The continuing technological advances in DRAM cards make higher storage levels affordable, but DRAM is still a limiting factor. However, as the following chart shows, most high-speed events occur in such short duration that 2000 frames is usually more than enough to capture the event. As memory chips get denser, the storage capacity will increase in motion analyzers. The table below provides average event times for various applications. The event times were measured from actual imaging data. The definition of an event time is the duration of event that produced significant information for motion analysis.
SUBJECT EVENT TIME (sec) FRAMES (1K fps) Money sorting machine (single bill time) 1.2 1,200 Flame pattern test (fuel combustion) 0.7 700 Wire bonding (one cycle) 0.8 800 Surface mount (one placement cycle, no pickup) 0.3 300 Food—crackers on process line (three samples) 0.3 300 Potato chips being bagged (one cycle) 1.1 1,100 Tire testing, front and rear over glass plate 0.4
400 Hot glue applied to film box flap 0.2 200 Blood stream (one cell motion across screen) 0.8 800 High voltage circuit breaker (one cycle) 0.2 200 Label pickup (one label) 0.6 600 Golf ball impact and flight (club) 0.6 600 Composite material fracture 0.1 100 Car crash test (impact) 0.3 300 Air Bag Inflation 0.035 35
Resolution
Resolution of a motion analyzer is general expressed in terms of the number of pixels in the horizontal and vertical dimension. A pixel is defined as a the smallest unit of a picture that can be individually addressed and read. At the present, high-speed-camera resolutions range from 64 x 16 to 2048 x 2048 pixels. Generally, the limiting resolution of the imaging system is the imaging sensor.
A rule of thumb for capturing high-speed events is that the smallest object or displacement to be detected by the camera should not be less than 2 pixels within the camera’s horizontal field of view.
The sensor resolution may be expressed also in terms of line pairs per millimeter (lp/mm). The meaning of line pairs per millimeter is an expression of how many transitions from black to white (lines) can be resolved in one millimeter. To calculate a sensor’s theoretical limiting resolution in lp/mm, take the inverse of two times the pixel size. Shown below is the limiting resolution of a sensor with a 16 micron pixel.
Theoretical Limiting Resolution = ( 1/ (2 x pixel size)) x 1000
= 1/(2 x 16) x 1000
= 31.25 lp/mm
Record Modes
Motion analyzer’s have various methods of recoding. The variety of recording methods is one of the most distinguishing features of high-speed imaging systems. Most of these recording methods cannot be matched by high-speed film cameras. The motion analyzer’s most useful recording method is when the camera is placed in record mode and continues to record waiting on a trigger. While in this continuous record mode (sometimes called End Trigger or Center Trigger) the camera records continuously, replacing its oldest images with the newest image until the event occurs and the camera is triggered and stops recording. Further flexibility allows the operator to program exactly how many images before and after an event is captured. For engineers and technicians trying to record something unpredictable or intermittent, these special triggering programs are the only feasible method of capturing the event.
Photron’s camera systems offers several different triggering programs that provides very flexible triggering techniques: Start, End, Center, Manual, Random, Random Reset, Random Center, Random Manual, and Dual Speed Recording. If you would like more information on these Recording modes, please reference the appropriate Equipment Operators Manual.
Time Magnification
The goal in using a high-speed camera is to obtain a sequence of pictures, that are observable in slow motion capture of a high-speed event. Time magnification describes the degree of "slowing down" of motion that occurs during the playback of an event. To determine the amount of time magnification, divided the recording rate by the replay rate. For example, a recording made at 1,000 fps and replayed at 30 fps will show a time magnification of 33:1. One second of real time will last for 30 seconds on the video monitor or computer monitor. If the same recording was replayed at only 1 fps, that one second event would take more than 16 minutes to play back. Most systems allow replay in forward or reverse with variable playback speeds. Therefore, it is important to capture only the information that is necessary otherwise, long recordings can take hours to playback. Some examples are shown below.
Record Rate
Time (sec)
Frames Recorded
Playback @ 30 fps
Playback @ 1 fps
250
20
5000
167 sec
83 min
500
50
30000
1000 sec
500 min
1000
2
1500
50 sec
25 min
4500
0.11
500
17 sec
8 min
30000
0.5
15000
500 sec
250 min
Lighting Techniques
Lighting an application properly can produce dynamic results over poor light management. There are four fundamental directions for lighting high speed video subjects; front, side, fill and backlight. Placing a light behind or adjacent to a lens is the most common method of illuminating a subject. However, some fill lighting or side lighting may be needed to eliminate the shadows produced by the front lighting. It is advisable to have the light behind the lens to avoid specular reflections off the lens. Side lighting is the next most common lighting technique. As the name implies, the light is at an angle from the side. This can produce a very pleasing illumination. In fact, for low contrast subjects, a low incident lighting angle from the side can enhance detail. Fill lighting may be used to remove shadows or other dark areas. Fill lighting may also be used to lessen the flicker from lamps that have poor uniformity. Fill is from the side or top of a scene. Backlighting may be used to illuminate a translucent subject from behind. It is not used that frequently in high speed video. However, certain applications such as microscopy, web analysis or flow visualization may be well suited for backlighting. All of these techniques are important for getting a high quality image.
Lighting Sources
There are a number of lighting sources available for high speed video. Some care must be taken in lighting selection due to the several factors. The areas that need to be considered included the type of light, the uniformity of the light source, the intensity of the light, the color temperature, the amount of flicker, the size of the light, the beam focus and the handling requirements. All of these factors are important in matching the light to the application.
Type of Lighting
Lighting types can be identified by two characteristics; physical design and the method of producing the light. The physical characteristics include lens, the reflector, packaging and the bulb design. The method of producing light includes tungsten, carbon arc, fluorescent and HMI.
Tungsten
Tungsten lighting is also referred to as incandescent lamps. Tungsten color temperature is 3200K. A type of tungsten lamp is called halogen. Halogen is a hotter lamp since the bulb must heat the regenerative tungsten. The tungsten lamps are efficient in their light output.
Carbon Arcs
This type of lamp forms an arc between two carbon electrodes. The arc produces a gas that fuels a bright flame that burns from one electrode to the other. In time, this consumes the carbon.
Gas Discharge
The fluorescent tube is one type of gas discharge lamp. At the end of each tube are electrodes. The tube is normally filled with argon and some mercury. As current is applied at the electrodes, the mercury is vaporized by the argon gas. The mercury emits an ultraviolet emission. This then strikes the side of the tube that is coated with a phosphor. The phosphor then transforms the ultraviolet to visible light. Most fluorescent lamps emit a dominant green hue which is not very suitable for a balanced light source. Additional, the discharge produces a non-uniform light that is easily detected as a 60 cycle flicker when playing images back from a high-speed motion analyzer.
Arc Discharge
HMI (mercury medium-arc iodide) is the most common lamp in this class of lighting. As current is passed through the HMI electrodes, an arc is generated and the gas in the lamp is excited to a light emitting state. The spectrum of light emitted includes visible as well as ultraviolet. This light source typically has a UV filter to block the harmful emissions. The HMI light is a balanced light source. It generates an intense white light. If a switching ballast is used with the HMI, it produces a uniform light with very low flicker. Other types of ballast are not as well regulated.
Color
Understanding color is difficult but necessary even for monochrome imaging. The color of light is determined by its wavelength. The longer wavelengths are hotter in color (red). The shorter wavelengths are cooler (blue).
Color perception is a function of the human eye. The surface of an object either reflects or absorbs different light wavelengths. The light that the human eye perceives is unique in that it produces a physiological effect in our brain. What is red to one person may have a slight difference of perception by another person. Terms that further describe the color of an object is hue, saturation and brightness. Hue is the base color such as red, blue violet, yellow and others. Saturation is the shades that vary from a basic color to that of a different shade. An example of a hue would be green and a saturated color would be lime (light green). Brightness also known as luminance is the intensity of the light. The subject of color would take an entire book to fully explain the science. However, studying a color chart can give the user some insight into the composition a color scene.
Color temperature is a common way of describing a light source. Color temperature originally derived it’s meaning from the heating of a theoretical black body to a temperature that caused the body to give off varying colors that ranged from red hot to white hot. This term was developed by Lord Kelvin and his name was associated with the unit measure.
Color versus Monochrome
Most of the early high-speed film was black-and-white. Once color film became available, the use of black and white declined. The use of high-speed color film set the format standard that video has attempted to meet. Over the years, monochrome images have been all that could be recorded on most motion analyzers. Today’s motion analyzers can produce images that replace color film for some high speed applications. Full 24-bit color images are now possible from motion analyzers. To understand the strengths and weaknesses of both color and monochrome in varying high speed video applications, some background must be discussed.
There are various methods of producing color in high speed video. The three the most widely used techniques are color wheel, beam splitter, and color filter arrays. The color wheel is used in still imaging. The subject does not move but, the wheel rotates to a position with a primary color filter and an image is taken. Then the wheel moves to the next filter and an image is taken. Finally, the last filter is in position and an image is taken. All three images taken with the primary filters are built into a three color plane image (RGB). This technique is not suitable for high speed video due to the motion differences between each successive image. Using three imaging sensors with stationary color filters and a beam splitter, true color reproduction is possible. True color means that the primary colors and all the saturations are possible. This technique is costly since all the electronics is tripled with the need for three imaging sensors. The alignment of the three sensors must be very precise. Otherwise, misregistration will occur on the colors. The last technique is a cost saving compromise. Color Filter Arrays (CFA) provide a more cost affective means for producing color (only one imaging device). There are individual color filters deposited on the surface of each pixel. There is some combination of Red, Blue and Green or a complimentary color scheme. Each pixel is isolated to a certain color spectrum. Although the pixels are filtered, the raw data must be interpolated for solving the missing pixels in each color plane.
Now that the main methods for producing color have been discussed, we need to review why you would want to image in color and not monochrome. Generally, monochrome images are better in image quality. Monochrome cameras are more sensitive due to the lack of color filtering. The resolving capability is better than CFA imaging sensors. This is due to the fact that there is no interpolation involved. The disadvantage of a monochrome image is the loss of color differentiation. The subtle change in gray levels is harder to observe than a change in hue or saturation. Color is valuable for differentiating shades. It also produces a bridge from color film to color video.
Continue to High-Speed University 201 - High-Speed Imaging Applications
