BotCam Stereo-Video System

The BotCam's stereo-video system is based on the principles of stereo-photogrammetry. Stereo-photogrammetry allows objects in images (photographs or frame grabs of video) to be measured with high precision and accuracy. All measurements require a minimum of two images of an object (moving objects must be synchronized in time) from two distinct locations with known relative camera orientations. This information combined with the internal geometry of the cameras allows three-dimensional space to be defined.

The image above shows the VMS PC interface. Stereo-video images from a BotCam deployment at Penguin Banks, Hawaii, are shown. A 
        306 mm Ehu (Etelis carbunculus) has been measured. Other species present include Dogfish Sharks 
        (Squalus mitsukurii) and Antigonia (Antigonia eos). Water depth is 250 meters.
The image above shows the VMS PC interface. Stereo-video images from a BotCam deployment at Penguin Banks, Hawaii, are shown. A 306 mm Ehu (Etelis carbunculus) has been measured. Other species present include Dogfish Sharks (Squalus mitsukurii) and Antigonia (Antigonia eos). Water depth is 250 meters.
VMS interface showing a Southern Bluefin Tuna (from Harvey et. al. 2003.).
VMS interface showing a Southern Bluefin Tuna (from Harvey et. al. 2003.).

The BotCam's stereo-video images are processed using a software package called Vision Metrology System (VMS) from Geomsoft (www.Geomsoft.com). An example of the VMS interface is shown. Objects (e.g. fish) that can be seen in both the left and right images are measured by clicking on the same points in both images and then choosing the object's name from a pull-down menu. The length of the object is projected on both images. The objects distance and angle from the cameras, the frame number (time) and the X, Y, Z coordinates in space are all reported in a data file that can be uploaded to alternative data management programs (e.g. Microsoft Excel, Matlab).

Forklength
(Sample size: 54 mm)
Error (mm) Absolute Error (mm) Relative Error (%) Relative Absolute Error (%)
Mean 1.72 6.06 0.16 0.56
1 Std. Dev. 8.13 5.62 0.76 0.54
1 Std. Error 1.11 0.77 0.10 0.07
Max. Body Depth
(Sample size: 47 mm)
Error (mm) Absolute Error (mm) Relative Error (%) Relative Absolute Error (%)
Mean 1.37 3.93 0.51 1.37
1 Std. Dev. 5.06 3.43 1.78 1.24
1 Std. Error 0.74 0.50 0.26 0.18
Error: Stereo Video Observed Value - Measured Value
Absolute Error: Absolute Value of Error
Relative Error: (Error / Measured Value) X 100
Relative Absolute Error: (Absolute Value of Relative Error) X 100
Measured Forklength Range: 830 - 1412 mm
Measured Maximum Body Depth Range: 228 - 365 mm
Table 1. Errors associated with stereo-video estimates of length and width measurements of southern bluefin tuna. (Adapted from Harvey et. al. 2003).

VMS has been tested in several underwater applications by Dr. Euan Harvey from the University of Western Australia and others. Results from one report by Harvey et al. (2003) using caged bluefin tuna are shown. The results compare stereo-video measurements of bluefin tuna to physical measurements of the fish taken with calipers after the fish were removed from the cages.