Interactive human motion acquisition from video sequences

Biomechanical analysis of anterior cruciate ligament injury mechanisms: three‐dimensional motion reconstruction from video sequences

“…Although a considerable research effort has been put into automating tracking and reconstruction of human motion from video sequences, the challenges in handling single and multiple uncalibrated camera recordings without restrictions on camera motion, human motion, clothing, light setting, etc., still prevent a general solution to the problem (Moeslund & Granum, 2001; Moeslund, 2003). Furthermore, reaching adequate accuracy for biomechanical purposes, or simply estimating axial rotations of the segments seems to be difficult using automated (Cheung et al, 2005; Sminchisescu & Triggs, 2005) or even semi‐automated systems (Zheng et al, 2000; Barron & Kakadiaris, 2005). Thus, our model‐based image‐matching technique at present provides a suitable framework for utilizing the available information optimally in injury situations, where controlled experimental techniques cannot be used.…”

A model-based image-matching technique for three-dimensional reconstruction of human motion from uncalibrated video sequences

“…The advantage, on the other hand, is that manual assessment allows us to simultaneously utilize edge, surface, color, contrast, segment shape, texture and size as well as point landmark properties, to provide optimal motion estimates without the need to implement these complex criteria mathematically. In the computer vision literature, we also see that manual assessment of key frames is often needed to increase the accuracy of the tracking and reconstruction (Wilhelms et al, 2000;Yamamoto et al, 2000;Zheng et al, 2000;Perales and Torres, 1994;Taha et al, 1997). Furthermore, using a model is the key to reconstruct the poses of the body segments reliably.…”

“…In many indoor sports such as basketball and European team handball, there are several lines and markings on the floor, as well as the basket, goal posts and advertising boards. In situations where there is not sufficient calibration information in the field of view to reconstruct the camera parameters, estimates of the body configuration can still be obtained by using the person as reference for the camera location(s) similar to the approach of Zheng et al (2000). A drawback, however, is that the technique involves manual frame-by-frame matching by the operator, which means that the method is time consuming, and possibly biased by the operator's subjective judgement.…”

Human motion reconstruction from monocular images using genetic algorithms

“…Suppose the searching step is 0.1 , computational expenses for every joint are listed in Table 1 . The total cost of the 12 joints is 5.12838*10 11 . If the step value is less than 0.1 , the search space will be even larger, thus optimization methods have to be applied to solve the problem.…”