Method Of3D Measurement And Tracking For Dynamic Deformable Surfaces

There are lots of nature phenomena which can be considered as3D dynamic de-formable motion, such as flags waving in the wind, swimming sharks in the water, human faces, etc. On the one hand, it has scientific significance for revealing the move-ment of objects to study the these nature phenomena, on the other hand, it has practical value for dynamic deformable modeling in movies, animations, and games. Currently, there are no mature methods to capture the motions. If we can measure the3D mo-tion data of dynamic deformable objects accurately, which will be very meaningful for revealing the mechanism hidden behind the nature phenomena, at the same time, can boost related industry progress.In order to acquire the3D motion data of dynamic deformable objects, this paper presents three methods. The first method of this paper is a novel framework of3D measurement and tracking of dynamic surface based on joint optimization, in which a surface is divided into several blocks and block matching in stereo and temporal images is used instead of matching the whole surface. This method is a passive method of measurement, and we consider the stereo matching and temporal tracking at the same time. In the optimization, we assume there exist affine transformation between stereo image patches and shift transformation between temporal image patches in the stereo image sequence.However, the above method adopts joint optimization, in which we need consider more optimized parameters. Thus, this joint optimization method will bring difficul-ties to the solution. According to this reason, this paper presented the second method, namely, the method of3D measurement and tracking by expansion from the most reli-able region. This method is also a passive method to measure dynamic deformable3D surfaces from a calibrated stereo image sequence. This method adopts a patch-based expansion technique that performs3D measuring and tracking by computing the motion and disparity independently. Due to the separate consideration of stereo matching and temporal tracking, we can obtain more accurate results because of parameter reduction. We here adopt a patch-based expansion method, because patch-based methods have been proven to be very effective to cope with complex surfaces, and region expansion techniques have been validated to be very robust in lots of applications.Although the two methods can obtain good measurement results, they require there are rich textures on the measured objects, which is the limitation of passive methods. Due to the limitation, the passive methods will generate accumulated errors when we deal with long image sequences. To solve the limitation, we can use the active method. Thus, this paper presented the third method, namely3D measurement and tracking based on gray random triangle pattern, which belong to an active measurement tech-nique. To the best of our knowledge, this is the first study that presents the measure-ment of the dynamic deformable surface using the technique of random gray triangle pattern. The proposed method solves random triangle generating, detecting, matching, and tracking from stereo videos in a novel way. This method can measure the motion of the dynamic surface stably and accurately without error accumulation and the influence of local illumination change.According to experiments, each of the proposed methods has its own advantages, and can acquire3D motion data of dynamic deformable objects effectively, which can provide a useful method for scientific research and project application.