Hardware

Hardware Setup

Tracking box

Many of the challenges in achieving robust tracking over long time scales that would be very difficult to solve by adjusting parameters in the tracking can be entirely avoided by constructing a dedicated space for tracking. A tracking box has the combined benefit of isolating both the camera and the tracked individuals from external optical and mechanical perturbations. Although a tracking box is not strictly required, a box of the general configuration in the sample schematic below may serve as a reference for good tracking conditions before attempting to optimize other hardware or software parameters (fig 2.1). The only essential features of a tracking box are opaque walls, a camera mount, and a diffuse illumination source. The difference image calculation used by MARGO is very sensitive to any changes within the field of view of the camera. This is good because it means that the software is sensitive to very small movements of small objects, but it also means that the only movement visible to the camera should be the tracked objects. MARGO can automatically detect and adjust for spikes of noise in the image, but walls will drastically reduce the time MARGO spends attempting to correct noisy imaging and will increase the time spent tracking.

{width="50%"}

In addition to walls, use clear sanded plastic, diffuser film, or paper either between the illumination source and behavioral arenas or on the floor of the arenas themselves to diffuse the light source. Having a sanded or matte finish on all surfaces inside of the tracking box can also help reduce reflections and glares inside the box.

::: {#arenasection} Individual behavioral arenas

:::

MARGO ROI setting can operate in two different modes: automatic and grid modes. Choosing appropriate individual behavioral arenas will be dependent on the ROI detection mode used. Automatic detection mode works by finding bright regions of the image that are all roughly the same size. This means that ideal behavioral arenas used for automatic ROI detection are backlit, transparent areas separated by opaque areas in between. Sample arena construction for automatic detection is shown below in fig. 5.

{width="95%"}

Grid detection mode works only on the assumption that ROIs will be arranged in regularly spaced rows and columns. The user is prompted to draw and adjust the position of one or more grids of any arbitrary dimensions to specify ROIs. This means that behavioral arenas can be anything with stereotyped dimensions. Wellplates of any dimensions make ideal behavioral arenas for grid mode detection. Grid mode also works well with linear arrays of tunnels. Sample arena constructions for grid detection mode are shown below in fig. 6.

{width="50%"}

::: {#illumsection} Illumination

:::

Quality illumination is a prerequisite for high-quality tracking. An ideal light source will be evenly diffuse and sufficiently bright. Because MARGO uses a single tracking threshold for the entire field of view of the camera, even illumination is the top priority for an illumination source. MARGO can correct for uneven brightness, but having an even illumination source will ensure that the quality of tracking is equally good in all parts of the cameras field of view. Any illumination source bright enough to saturate the camera is sufficiently bright for tracking.

We perform all of our tracking with infrared light. Using an infrared light source and longpass filter on the camera has the dual benefit of reducing sensitivity of the tracking to perturbation by most external light sources (not sunlight) and allowing for independent control of visible light sources to deliver stimuli not visible to the camera. An affordable example of two color channel LED light panels configured with white and infrared LEDs can be found here. Alternatively, LCD backlight panels can provide a ready-made solution to bright, even illumination.

::: {#camconfig} Camera configuration

:::

MARGO has built-in tools to automatically detect and configure any camera visible to MATLAB. The appropriate configuration for your camera will largely depend on your particular experiment. Camera customization features are accessible under the "Hardware" menu bar. But before configuring the camera in MARGO, ensure that the camera and camera lens are properly setup for tracking. The following are good guidelines for configuring your camera:

• Aperture opened enough that ROIs are nearly saturation

• Lens maximally zoomed to reduce lens fisheye

• Lens focused on ROIs

• Camera rigidly fixed to a camera mount

::: {#camdetect}

camera detection

:::

Camera support in MARGO is built on MATLAB's Image Acquisition Toolbox. Before a camera can be detected by MARGO, the associated the associated MATLAB camera adaptors must first be installed. The appropriate adaptor to install will depend on the camera manufucturer. See MATLAB's tutorial for complete instructions on installing MATLAB camera support packages.

Once the camera adaptors are installed, a list of available cameras and camera modes should auto-populate upon launching MARGO. The list available cameras can be refreshed under the hardware menubar.

::: {#cammodes}

camera modes

:::

MARGO will automatically detect available camera modes for operating at different pixel resolutions and color bit depths. The tracking is designed to work with 8-bit color depth because most tracking operations are performed on binary images. It is also strongly recommended that monochromatic cameras are used. By default, MARGO only tracks the green color channel of RGB images.

::: {#camsettings}

camera settings

:::

Camera properties such as exposure time, shutter speed, gain, and frame rate can be adjusted through camera settings under the Hardware menu bar. Because the configurable settings are specific to each device, the properties that can be adjusted here will depend on your camera. Many cameras have modes to automatically adjust the camera exposure time, shutter speed, frame rate, gain or focus in real time. These modes are also often enabled by default, which may be good for many applications but is highly problematic for tracking. Constantly fluctuating images will introduce a lot of noise into the tracking due to the sensitivity of difference image to even minor changes from frame to frames.

To adjust camera settings, first initiate preview camera and leave it running to get feedback on the settings as they change. Select Hardware $>$ Camera $>$ camera settings to open the camera settings window. Keep in mind that the available settings and names of each property will be dependent on the camera. From the settings menu:

• Ensure that any automatically adjust fields are set to "off" or "manual"

• Adjust exposure and shutter speeds until the preview is just below saturation

• Set max frame rate (target acquisition rate can be set lower in tracking parameters)

• Close window to save settings