Dynamic accuracy survey of the new "single plane--single frame 3-D calibration" technique for use in biomedical...

This study tested the accuracy of a new 3-D calibration technique under dynamic situations. The technique was firstly introduced in 1998 for biomechanical human tests and calibrates 3-D volumes in an easy way. It revealed superior in static tests to others. In order to disclose dynamic accuracy two different tests were performed. With this technique it does not matter whether redundant information from multiple camera views is available or not. The mean error for distances measured at 0.018% for redundant information and at 0.012% for the non-redundant test in contrast to other procedures found in literature, which attain values of 0.09% and 0.04% respectively. The maximum error ranged there between 5.5% and 17.9%, whereas the presented data reached values of 0.33% and 0.48%. The more important angle error was at maximum 0.055% (9 times less than the most accurate in literature) and nearly zero for the mean error value. The level of noise was the same in the test with redundancy and 7.4 times lower in the present study than other commercial available systems for non-redundant video information. The new procedure revealed as a stable and very accurate 3-D reconstruction technique for a variety of application not limited to biomedical applications.