Resumen en español

One of the most important challenges when a movement analysis system is been developed is the three-dimensional reconstruction. This process involves several algorithms that must work synergistically in order to reconstruct the movement correctly and decreasing the user interventions. For this reason this paper describes in detail a new methodology to carry out the three-dimensional reconstruction of the human movement through the join of different algorithms decreasing the user input during the process (movement analysis system CineMED III). To analyze and evaluate the performance achieved using the methodology proposed, three experimental tests were carried out; a static marker test, bar with two markers in motion and using an error measurement device. CineMED III finally has a performance close to that of commercial systems, showing in the tests performed with the bar a lower average absolute error for the Vicon system, but higher with respect to the Elite system and testing the device with a value of RMS more about the Ariel system, but a smaller maximum error on the latter system., However, the proposed approach presents problems when markers are occluded for long periods of time and when the paths of two markers intersect. To solve these problems, it is proposed to add some cameras into the system to obtain redundant information from the motion capture.