Video Capture

We are often asked for advice on recording videos for automated analysis.

This article brings together our 9 tips for how to setup and film trials to ensure they are best suited to automated analysis of lab animal behavior. As you can see there is a lot to consider, so check out our set-up service and we will ensure that you are properly set-up to record good quality data.

1. Get the right hardware:

There is a bewildering number of cameras available ranging from consumer web-cams and camcorders to professional machine vision cameras. Digital video cameras are common and are often the best choice due the flexibility of digital video. It is important that the camera and lenses used allow a clear in-focus view of the whole arena. Both resolution and frame rate are important, and the required level depends on what behavior you want to identify. Where color will be used within an trial (e.g. to identify different animals) the camera should be able to clearly distinguish the different colors (many low cost cameras have very poor color quality).

2. Get the right software:

Before you start make sure it is clear how the camera will communicate with your recording software, many consumer cameras come with software to allow video to be recorded on a commodity PC, often these tools are not of high quality or particularly flexible. Professional cameras often require specialised hardware or software to control the camera but allow sophisticated control of the entire recording process such as triggering start / stop recording on the basis of another signal such as a movable door opening.

3. Select the best resolution:

When setting up the camera it is important to set the optimal resolution: too high and the data will be unnecessarily large and difficult to store; too low and the video may not capture enough information of what you are interested in. There are two aspects to resolution which should be considered: temporal and spatial. Temporal (number of frames per second) should be increased when identifying behaviours such as rapid walking or limb movements. Spatial (pixels per inch) resolution should be increased when attempting to identify fine details, for example the detailed position of limbs. For basic tracking each arena should be no smaller than 320x240, and 10 fps, for more advanced behaviour detection at least 640x480, and 30fps is recommended.

4. Position the camera:

No matter how high-end the camera is, if the camera isn't well positioned the quality of the data will be impacted. Typically the camera should be directly above, or to the side of the arena facing totally face on to the area of interest. If the camera is not aligned correctly the accuracy of tracking data will be significantly reduced. The camera should be sturdily fixed so that it does not move during or between trials.

5. Choose the right file format:

Even if the video you record is of good quality it must be stored in the correct format so that this quality is not lost. Video formats (e.g. .avi, .wmv) compress video to store it which loses information. These formats are designed to compress video so it still looks good to the human eye, however a computer notices the loss of information more and this can seriously affect analysis quality. MPEG2 offers a good compromise of computing power required, quality, and file size.

6. Get the lighting right:

Ideally shadows and reflections should be removed using matt surfaces for arenas, and diffuse lighting from directly above or below (if using a translucent or transparent floor). LED lighting is better than fluorescent lighting which can operate at frequencies that interfere with acquisition hardware. This will allow a clear outline of the animal to be obtained and lead to less error-prone analysis. If using multiple arenas make sure all arenas are identically lit and isolated from one-and-other.

7. Distinct colors between subject and arena:

The arena should have a uniform, non-reflective surface that contrasts with the subjects' color and texture i.e. black on white, and with other objects of interest within the arena. If there are multiple objects (or subjects) that look alike e.g. same color, size and shape it will be difficult for the computer to tell them apart.

8. Stay consistent over trials:

Keeping the environment consistent across a study is not only vital for scientific integrity but also for capturing good quality video. Specifically there should be no motion within the arena except for the objects (and subjects) to be analysed, the researcher should remain out of shot as much as possible, and avoid reaching into the environment until the trial is complete. For example, if the researcher moves around the arena, or a door is opened, then even this slight movement will change the lighting in the arena and affect the recorded videos.

9. Store the data:

Once the video has been recorded it must be stored appropriately. You should think about whether backups are necessary (reduces risk of data loss, but increases cost), the medium on which to store the data (DVDs, local hard drives, or internet based storage), and how the videos will be used in the future. How accessible does the data need to be? What data needs to be stored with the video to make sure it can be used in the future? Do you want to share data with peers and colleagues? Are you required to make the data available?

Bring in the Experts

At Actual we have the expertise to help you get the optimum setup for your lab. From our lab setup service to tailored solutions we're committed to ensuring the analysis of animal behavior is as efficient as possible. Contact us for further details and a discussion on the best options for you.