Loading...
 

Video cameras and mounts

The choice of video cameras can make a big difference in the quality of the footage. Viable systems range from high-end video cameras and DSLRs in expensive underwater housings down to action sports cameras like GoPros.

Focus features

Cameras best suited for videogrammetry allow manual focusing at repeatable distances. Less expensive, fixed-focus cameras like GoPros are also useful, as their non-adjustable focus is typically set at a distance that maximizes depth-of-field.

Camera recording formats and storage

Because most cameras support several recording options, these attributes are discussed on the camera settings page.

When purchasing cameras, consider the settings you want to use and make sure your model supports them. It's also useful to calculate the expected disk space occupied per hour of video, multiply that by the expected hours of video you'll collect, and make sure you have enough disk space on hand to manage your footage during data collection.

Recording formats

Preserving the finest details in video footage requires some attention not only on how the video was filmed, but also on how it is saved and encoded digitally. This process will vary among users. Files will play in VidSync as long as they're in any format compatible with Quicktime player on the same computer (.mov or .mp4 files using the H.264 codec are a good choice), but quality and file size depend on compression settings. Some cameras record to unusual, compressed formats like AVCHD, or in formats with little to no compression such as Apple ProRes. In both cases, conversion of the files (called transcoding) is required to make the videos playable on a computer, bring the files down to a reasonable size, or both. Final Cut Compressor is an excellent high-end video transcoder for Mac. Cheaper options, such as Wondershare Video Converter and many others, may be suitable for many users.

High-end transcoders such as Compressor allow you to control the bitrate of the converted file. For 1080p (full HD) videos, I've found that a bitrate of 4 mb/s (about 30 gb per camera for 2 hours of footage) presented a good tradeoff between file size and quality (erring on the size of larger files and higher quality, if anything).

External recorders

System design is also influenced by whether or not the operator requires a live view of the objects being filmed. A live view can help with verifying suitable lighting, focus, and aim. Self-contained camcorders in underwater housings do not offer such live views unless manually operated by a diver. An alternative is to use a closed-circuit video system, with cameras tethered to external recorders or computers. It is an advantage that these systems are not limited by the size of the battery and recording medium that fit inside a housing, but they may be more fragile and less portable than the self-contained system we described.

DSLRs and video

Digital SLRs have recently become popular for video as well as still photography, and they're a popular choice among people looking for fine manual control over exposure settings (although their manual options for video are not always as good as for still photography). But most of them have one very important limitation: they can only shoot video for 30 minutes at a time before recording stops. This goofy limitation is reportedly used by the camera manufacturers to avoid a video recording device classification that incurs an extra tariff on imports to the European Union. It poses a major problem for DSLRs with the kind of research we typically do with VidSync.

For some cameras, it is possible to circumvent this limitation with warranty-voiding firmware hacks. For other models, this is not possible. The Drift Model Project circumvents this limitation on Nikon D5300 DSLRs with a complicated combination of external recorders, hardware hacks, external power, heat sinks, and finicky operating procedures. Although we've been able to record high-quality data videos 5+ hours long, I would not advise others to try to beat the system with DSLRs in this manner. It was a lot of trouble.

Overheating can be a problem with both DSLRs and external records on very long recording sessions. Even sitting in extremely cold water, an underwater housing can insulate the air inside well enough that it heats from some combination of electrical activity inside the camera and a greenhouse effect if direct sunlight is shining into the housing. Painting the housing black (with Krylon Fusion spraypaint for plastic) solved this problem in our case, but then our Atomos Ninja Blade recorders were overheating on several-hour sessions. So we sandwiched those between plates of aluminum resting in a tub of shallow, cold water to serve as a heat sink.

Camera mounts

Typically, a camera system for VidSync consists of two side-by-side cameras mounted firmly to some platform or crossbar. The configuration and sturdiness of this mount are both important.

The configuration of the cameras on their mount can significantly affect the accuracy of measurements. Separation and angles of the cameras affect accuracy and precision, which depend on the angle of convergence of the lines of sight from the different cameras. When the subject is far from the cameras compared to their separation, the lines converge at a narrow angle, and small errors in line positions create relatively large errors in 3-D positions. Precision is maximized at the opposite extreme, when the cameras are separated so far (or the subject is so close) that the lines intersect at a right angle. This setup is ideal for many lab experiments. For the common “stereo pair” setup of two side-by-side cameras, Harvey and Shortis (1996) found (using calculation methods different from ours) that cameras should be separated by 1/3 to 1/5 of the camera-to-subject range for acceptable precision. We found suitable precision at a greater distance. In general, camera separation should be maximized subject to the constraints of subject distance, field of view, and field handling convenience for the cameras and the calibration frame (which must be larger as camera separation increases). Underwater work usually benefits from the use of wide-angle optics to increase the field of view, particularly because refraction at the water-glass-air interface “zooms in” the view by a factor of about 1.3. We found that a high-quality, wide-angle dome port minimized picture distortion with fewer lighting artifacts than a flat lens port; however, the opposite may be true for some systems (notably, GoPro ® cameras require a flat underwater housing port for optimal image quality).

The mount must hold the cameras completely steady relative to one another during the time between calibration and the end of data recording. This may include both the orientations of underwater housings and the positions of cameras within their housings; even if waterproof housings are fixed firmly in place, cameras may shift slightly within them. Care should be taken to secure all potentially movable parts, and to re-calibrate after each change to the system, such as removing cameras from their housings to change batteries. With extremely stable hardware, it may not be necessary to perform a new calibration for each new recording. Regardless of configuration, the calibration frame and chessboard should be filmed under similar conditions to the intended measurements, because the refractive index of water and the shape of the housings/ports may vary with large changes in temperature, depth, and salinity (Shortis et al. 2007a).

Desiccants to prevent fogging

Camera housings placed into cold water can fog up quickly as the moister, warmer air inside the housing is cooled against the lens port. Using a bit of electrical tape to stick a desiccant packet somewhere inside the housing can prevent this problem in all but the worst conditions. GoPro housings are too small for normal packets, but GoPro sells tiny desiccant wafers that can fit inside under the camera.

GoPros and VidSync

Many people have used VidSync with GoPro video cameras, and they are undoubtedly one of the cheapest, easiest systems to deploy and a very good choice for man projects. 

Advantages to GoPros:

  1. Cost.
  2. Ease of use. No need to make decisions about focus distance or exposure modes.
  3. Small size and lightweight. Fits easily into tight spaces, on the end of a long rod, etc.
  4. Pretty good quality.

Disadvantages:

  1. Their battery life isn't ideal for long recording sessions (multiple hours), although they make an accessory that will hold an extra battery.
  2. Older GoPros split video files into chunks you have to re-join manually using other software, and sometimes a frame is lost in the joining, which can affect synchronization. I heard the new ones don't have this problem, but I haven't verified that myself.
  3. Their sharpness suffers quite a bit underwater near the corners of the image. There are third-party underwater housings that alleviate this problem. It's also a problem that doesn't matter for many applications where you're pointing the cameras at your subject, but it can be important in behavioral studies where you fix the cameras in a steady position and have fish milling around out in front of them requirement measurement all over the screen.

Here are a few tips for using VidSync with GoPros:

  1. Make sure to disable the "HyperSmooth" stabilization feature, which may be enabled by default in some recent GoPro models. It's great for making action videos look smooth, but it can severely interfere with 3-D measurement calculations, especially if the cameras are in motion, as when being held by a diver.
  2. The black plastic screws GoPro uses with all their mounting hardware work fine to secure cameras in a sturdy orientation for VidSync, but if you hand-tighten them they'll easily get bumped out of alignment. Use a wrench (channellock tools work well) to crank them down more tightly.
  3. The default "Wide" exposure settings aren't ideal for every project. Because of the way GoPros use their sensors, you can effectively use the "Medium" or "Narrow" modes to emulate a longer focal length, i.e. zooming in on your subject, without any loss of quality. This is a great way to see better details if you don't need the wide field-of-view.
  4. The native aspect ratio of the GoPro sensor is 4:3. If you use a 3:2 mode such as the common 1080p used for most HD video, you'll be cropping off the top and bottom of the recordable image. The "Superview" mode on newer models will scrunch the information from the 4:3 sensor down into a 3:2 package instead of cropping it, but if you really need that extra information on the upper and lower parts of the screen it may be better to just use 1440p and record it in its native size.
  5. To get a feel for the field-of-view and other settings, you can use the GoPro phone app to see a live view of what your camera's seeing. It can't stay connected when the camera's underwater, unfortunately. If you require a live view then, there are moderately expensive third-party housings and viewing units that will let you watch a wired HDMI feed from the camera. This expense may be very worthwhile if it helps you make sure the cameras are pointed where they need to be to get your data.
  6. If you're using the GoPro in very shallow water, it might be more convenient to hang it upside-down from the camera mount than right-side-up. The cameras conveniently have a setting to determine which way is "up," and you can use this to film normal footage with the camera upside-down.