FAQ

In the Graphics and Virtual Reality Lab we have several motion capture systems, including a 24-cameras PhaseSpace Impulse X2 motion capture system, a 12-cameras Optitrack Flex motion capture system, XSens suits, several Kinect sensors (including the new Azure), as well as other depth sensors, and high quality RGB-video cameras.

For the purposes of this project, we used a passive motion capture system of twelve Optitrack Flex 3 cameras (resolution of 640x480) to capture the 3D motion of our articulated subjects. The cameras operate at high frequency (in our experiments, we capture movements at 100Hz), each of which can capture the position of any number of bright spots from the reflective markers. Motions were captured in a working volume of approximately 2m x 3m.

In addition to that, we used the AXY-4 triaxial accelerometers of Technosmat to record the accelerations of the reptiles, that have been strategically positioned on the reptiles’ body. These accelerometers can record at high frame rates, and in our experiment we record at 100 frames per second.

In this project we used the following formats:

• Video format: Movement recording using an HD camera. Data saved in .MP4 (MPEG-4 part 10 H.264) or .FLV (flash) format.
• C3D format: stored 3D coordinate information, analog data and associated information as it is recorded from the motion capture system.
• Character (.FBX) format: A rigged and skinned virtual character baked with the animation. This format is compatible with Adobe MotionBuilder, 3Ds Max, Maya, Unity3D, Blender etc.
• BVH format: BioVision Hierarchy file format. Exported BVH files do not include individual marker data. Instead, a selected skeleton is exported using hierarchical segment relationships.
• CSV format: This format stores the xyz-acceleration readings taken from the AXY-4 triaxial accelerometers of Technosma.

We used four RGB vision cameras. One fixed top view (GoPro HERO 7), two fixed site views (Canon EOS 60D and Canon EOS 7D Mark II) and one movable hand camera (Canon PowerShot SX40 HS). All cameras were recording at 24-30 fps (depending on camera model) with a resolution of 1920×1080. In the case of in-the-field recording, only two RGB vision cameras were used (Canon EOS 60D and Canon EOS 7D Mark II) equipped with zoom lenses (Canon EF 100-400 mm).

Recording from optical cameras were synchronized using the “Multi-Camera Source Sequence” function of Adobe Premiere Pro. Please note that this function uses sound recorded by all cameras as a means of synchronizing the visual records. For better synchronization we were talking during recording and producing instants timestamping sounds (e.g., clapping once) when starting and before stopping the records.

Since each recording method (vision records and acceleration) was running independently, all the processed data had to be synchronized for analysis. This was very challenging and had to be conducted manually, to the nearest second, by timestamping the starting point of each record using a handheld chronometer. Knowing the starting time of each device allowed as to synchronize each other (i.e., the vision cameras with the accelerometers) during the post-process analysis.

Recording with Mo-cap is based on the cameras being able to identify the reflecting markers attached on the animal. During our experiments we faced numerus challenges. The main once that we consider extremely important to share to assist replication or our approach are the followings:

• For small animals like the Stellagama stellio  lizard we had to use small face markers (3mm diameter) not to interfere with animals’ motion. This small reflectance surface limited the working distance from cameras forcing us to monitor the movements of the lizard in an area of 1,5m X 1,5m. In the case of snakes where larger markers were used, this area was increased to approximately another 1 meter from each site.
• Adding natural obstacles although will provide a more hospitable environment for the animal and will enrich its motions will highly interfere with line of sight of the markers. In any point in time at least three cameras must “see” each marker. When the animal hides behind an object (e.g., a brick) the system will not be able to see all markers and the reconstruction of animal movement will be practically impossible. Although adding more cameras (we used 12 cameras) might assist it will not provide a solid solution. 
• Since each recording equipment (i.e., video cameras, mo-cap cameras, accelerometers) works independently, proper timestamping is of absolute importance for the post-process synchronization between the recording methods. We recommend using a neutral handheld digital chronometer as a reference instrument for this timestamping. In addition, we stress the importance of talking during filming and initiating each record by pronouncing the time and making a clear fast knock for timestamping the initiation second.

All individuals were collected from the field using time survey and funnel traps and were released back in the field after the end of the behavioural trials. In all cases handling and human interaction was kept as limited as possible to minimize stress to the animals.

A total of 20 recording efforts (8 for lizard and 12 for snake), approximately 10 minute each, took place over a period of four days to further minimize stress to the animals. Between recordings animals were left to relax either within the monitoring arena, or within their terrarium. All appropriate licenses had been acquired from the Department of Environment, the Veterinary Services and Cyprus Bioethics Committee prior to collection.

During recording, animals were left alone to move freely, thus capturing general movements. Small obstacles (e.g., a single log, a brick) were added to enrich animal’s motion, while in some cases they were induced to climb, attack, flee or jump as means to record more specific movements. Please not that adding extensive natural obstacles in the monitoring arena is not recommended due to their interference with line of sight of the markers. In any point in time at least three cameras must “see” each marker. When the animal hides behind an object (e.g., a brick) the system will not be able to see all markers and the reconstruction of animal movement will be practically impossible. Although adding more cameras (we used 12 cameras) might assist it will not provide a solid solution. 

Not at this point. Social interaction is a very interesting and extremely challenging field of study that we intent to explore in the near future. Since that was the first effort to record movement and behavior of reptiles using Mo-Cap technology, we only record general motions and behaviors. We will assess all pros and cons before proceeding in using mo-cap for recording multiple individuals (and their interaction) simultaneously.

Markers were attached on the animals using non epoxy glue. The "3M Vetbond Tissue Adhesive" was used for attaching facial markers (3 mm diameter) on the lizard while "Bostik smart adhesive" was use for attaching medium size markers (14 mm diameter) on the snake.

Since each recording method (optical motion capture and vision records) was running independently, all the processed data had to be synchronized for analysis. This was very challenging and had to be conducted manually, to the nearest second, by timestamping the starting point of each record using a handheld chronometer. Knowing the starting time of each device allowed as to synchronize each other during the post-process analysis.

A total of 20 recording efforts (8 for lizard and 12 for snake)  took place over a period of four days. Each recording had a duration of approximately 10 minutes (due to size limitations of the .c3d format).